FIIIIIIIIIIIIIII|III I I fill:

“45898

g
1

r-A
@0395é99

 

This is to certify that the
dissertation entitled

NATURAL AND ANTHROPOGENIC FACTORS INFLUENCING
SPATIAL AND TEMPORAL PATTERNS OF METAL
ACCUMULATION IN INLAND LAKES

presented by
SHARON SIMPSON YOHN

has been accepted towards fulﬁllment
of the requirements for the

Doctoral degree in Environmental Geosciences

 

 

WT mé)

Majon’ Professor’s Signature

[Sena/4:107

Date

MSU is an Affirmative Action/Equal Opportunity Institution

 

LIBRARY
Michigan State
University

 

 

 

PLACE IN RETURN Box to remove this checkout from your record.
TO AVOID FINES return on or before date due.
MAY BE RECALLED with earlier due date if requested.

 

DATE DUE DATE DUE DATE DUE

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

6/01 cJCIRC/DateDuo.p65-p.15

 

AllYIIIII,

NATURAL AND ANTHROPOGENIC FACTORS INFLUENCING SPATIAL AND
TEMPORAL PATTERNS OF METAL ACCUMULATION IN INLAND LAKES

By

Sharon Simpson Yohn

A DISSERTATION
Submitted to
Michigan State University
in partial fulﬁllment of the requirements
for the degree of
DOCTOR OF PHILOSOPHY
Department of Geological Sciences

2004

ABSTRACT

NATURAL AND ANTHROPOGENIC FACTORS INFLUENCING SPATIAL AND
TEMPORAL PATTERNS OF METAL ACCUMULATION IN INLAND LAKES

By
Sharon Simpson Yohn

The natural cycles of trace metals in the environment have been signiﬁcantly
inﬂuenced by humans, resulting in toxicological concerns for both humans and
wildlife. Although environmental legislation reduced loading of some metals, humans
continue to alter the environment in ever increasing ways. In order to reduce the input
of metals to the environment, it is necessary to understand both current and historical
sources and contaminant pathways. To help determine this, sediments were collected
ﬁ'om 17 lakes in Michigan, USA, sediment proﬁles were examined and interpreted,
and anthropogenic accumulation rates were correlated to characteristics of the
watershed during two time periods (19703 and 19903) to understand both natural and
anthropogenic factors inﬂuencing contaminant accumulation in the sediment. Lakes
record the input of contaminants ﬁ'om both the local watershed and from atmospheric
deposition over time, and are therefore ideal for identifying sources and evaluating
temporal changes.

Elements within each lake were grouped using cluster analysis based on the
similarity of depth proﬁles, with each group representing a unique source and/or
process, including terrestrial, diagenetic, anthropogenic and carbonate groups.
Representative elements were chosen for each group, and these elements were used to

help quantify anthropogenic accumulation rates of cadmium, copper, lead and zinc.

Cadmium, zinc and lead anthropogenic accumulation rates decreased from the
19703 to the 19903 in most lakes, probably in response to environmental legislation.
Copper accumulation rates remained the same or increased over that time period for
many of the study lakes, indicating that historical sources have not been reduced by
environmental legislation, or that new sources have emerged.

Multivariate regression models using a maximum of two watershed
characteristics to predict anthropogenic accumulation rates for each metal were
determined for both time periods to provide insight into the dominant sources and
pathways of these metals in the environment. Cadmium and lead accumulation rates
were best predicted by characteristics that seem to represent atmospheric transport and
deposition in the 19703, including sulfate deposition rates and either average slope of
the watershed or slope within a 100 m buffer of the lake. The reason for a negative
correlation between slope and lead and cadmium accumulation rates is unclear, but the
average slope of the watershed has a regional gradient similar to that of the overall
population distribution in the state and may represent local atmospheric deposition.
Copper and zinc are best predicted by population density in the 19703. Percentage
urban land cover in the watershed is the best predictor of all four metals in the 19903,
indicating that local sources are currently more important that regional atmospheric
sources. Future work can now focus on identifying the speciﬁc sources of these

metals that are represented by urban land cover.

DEDICATION
This dissertation is dedicated to my husband, Chuck Yohn,

for his constant encouragement, support and patience.

iv

ACKNOWLEDGEMENTS

Special thanks to my advisor, David Long, for his encouragement, guidance,
and insight throughout this project. Special thanks also to Lina Patino, who was both
a committee member and a mentor, for her many hours discussing ICP-MS results and
providing constant support. My committee, Gary Weissmann, Catherine Yansa, Lina
Patino, and Bryan Pij anowski were always helpful and encouraging, and have my
grateful appreciation. Thanks to Julie Winkler as well for her statistical insight and
signiﬁcant editing of Chapter 2. Many fellow graduate students assisted with
sampling, lab techniques, or moral support, including Lydia Scholle, Joel Fett,
Merideth Lindeman, Nathanial Saladine, Karen Wayland, Jennifer McGuire and many
others. Thanks especially to Matthew Parsons for competently taking over this
project. Thanks also to Hasand Gandhi for his help with freeze drying. This work
would not have been possible without the help of Jackie Bennett, Cathy Caswell,
Loretta Knutson, and Diane Baclawski, who always had solutions to my problems.

Many thanks to the people at the Michigan Department of Environmental
Quality, who provided both funding and support. Special thanks to Bill Taft, Neal
Godby, Sarah Walsh, and Rick Lundgren for their help both in the ﬁeld and the ofﬁce.
Thanks also to Paul Wilkinson of the Freshwater Institute, Manitoba, Canada forszb
dating and interpretations. Three years of my graduate education were funded by an
Environmental Protection Agency Science To Achieve Results (STAR) fellowship.

Finally, 1 would like to thank my husband, Chuck, and my parents, Joanne and

Mark, for their encouragement and support.

TABLE OF CONTENTS

LIST OF TABLES ......................................................................................................... x
LIST OF FIGURES ..................................................................................................... xiii
CHAPTER 1. INTRODUCTION ................................................................................... l
Hypothesis and approach ............................................................................................ 7
References ................................................................................................................ 12

CHAPTER 2. THE USE OF MU LTTVARIATE STATISTICS FOR THE

INTERPRETATION OF SEDIMENT CHRONOLOGIES ......................................... 17
Introduction .............................................................................................................. 17
Methods and Procedures ........................................................................................... 20

Geochemical analyses .......................................................................................... 21
Outlier analysis ..................................................................................................... 23
Factor analysis ...................................................................................................... 25
Cluster Analysis .................................................................................................... 29
Assessment ............................................................................................................... 31
Factor Analysis ..................................................................................................... 32
Cluster analysis ..................................................................................................... 48
Discussion ................................................................................................................. 53
Comments and recommendations ............................................................................. 55
References ................................................................................................................ 56

CHAPTER 3. THE MPORTANCE OF THE MULTI-ELEMENT APPROACH
FOR INTERPRETING SEDIMENT HISTORIES ...................................................... 60

Abstract ..................................................................................................................... 60

vi

Introduction .............................................................................................................. 61

Methods .................................................................................................................... 63
Results and discussion .............................................................................................. 65
2'on and dating .................................................................................................... 65
Outlier analysis ..................................................................................................... 67
Element grouping ................................................................................................. 68
Identiﬁcation and quantiﬁcation of human inputs ................................................ 73
Conclusions .............................................................................................................. 79
References ................................................................................................................ 80

CHAPTER 4. ASSESSING ENVIRONMENTAL CHANGE THROUGH

CHEMICAL SEDIMENT CHRONOLOGIES FROM INLAND LAKES ................. 84
Abstract ..................................................................................................................... 84
Introduction .............................................................................................................. 85
Methods .................................................................................................................... 89
Results and Discussion ............................................................................................. 92

Sedimentation rates and focusing ......................................................................... 92
Chemical sediment chronologies .......................................................................... 96
Anthropogenic inventories ................................................................................. 105
Temporal and spatial patterns of accumulation .................................................. 108
Population effects on lead and copper accumulation ......................................... 1 10
Summary ................................................................................................................. 1 14
Acknowledgements ................................................................................................ 1 16
References .............................................................................................................. 1 l7

vii

CHAPTER 5. REGIONAL VERSUS LOCAL INFLUENCES ON LEAD AND

CADMIUM LOADING TO THE GREAT LAKES REGION .................................. 121
Abstract ................................................................................................................... 121
Introduction ............................................................................................................ 122
Methodology ........................................................................................................... 126
Results and discussion ............................................................................................ 130

mph results ........................................................................................................ 130
Temporal patterns ............................................................................................... 134
Spatial patterns ................................................................................................... 138
Population density .............................................................................................. 147
Production records .............................................................................................. 155
Conclusions ............................................................................................................ 158
Acknowledgements ................................................................................................ 159
References .............................................................................................................. 160

CHAPTER 6. LEAD IN THE ENVIRONMENT: EVALUATING SOURCES

AND PATHWAYS OVER TIME USING LAKE SEDIMENTS ............................. 165
Introduction ............................................................................................................ 165
Methods .................................................................................................................. 170
Results and discussion ............................................................................................ 176

Patterns of lead deposition .................................................................................. 176
Lead inputs - watershed characteristics relationships ........................................ 178
Multivariate regression ....................................................................................... 183
References ............................ ................................................................................. 188

viii

CHAPTER 7. IDENTIFYING CHANGES IN SOURCES FOR CADMIUM,

COPPER AND ZINC TO INLAND LAKES OVER THREE DECADES ............... 193
Introduction ............................................................................................................ 193
Methods .................................................................................................................. 199
Results .................................................................................................................... 207

Sediment chronologies ....................................................................................... 207
Spatial patterns and atmospheric deposition ...................................................... 213
Watershed characteristics ................................................................................... 222
Correlations ........................................................................................................ 225
Multivariate regression ....................................................................................... 231
Summary and Conclusions ..................................................................................... 240
References .............................................................................................................. 243

CHAPTER 8. SUMMARY AND RECOMMENDATIONS ..................................... 248
Summary ................................................................................................................. 248
Recommendations .................................................................................................. 254
References .............................................................................................................. 258

APPENDIX A. SEDIMENT CHEMICAL AND WATERSHED

CHARACTERISTICS DATA .................................................................................... 262

APPENDIX B. METHODS FOR ACQUISITION OF WATERSHED

CHARACTERISTICS DATA .................................................................................... 273

APPENDIX C. INDUCTIVELY COUPLED MASS SPECTROMETER

ANALYTICAL METHODS ...................................................................................... 297

BIBLIOGRAPHY ...................................................................................................... 312

ix

Chapter 2

Table 1

Table 2
Table 3
Table 4
Table 5

Table 6

Chapter 3

Table 1

Table 2

Chapter 4

Table 1

Table 2

LIST OF TABLES

Communalities of each element for the four study lakes and

Elk Lake with outliers removed .......................................... 29
Factor analysis results for Gratiot Lake ................................. 34
Factor analysis results for Gull Lake .................................... 37
Factor analysis results for Paw Paw Lake .............................. 41
Factor analysis results for Elk Lake ..................................... 43

Final groupings for each element in the four study lakes,
and for Elk Lake with outliers removed (Elk OR) ..................... 50

Characteristics of 12 Michigan study lakes and selected data
determined from 2”Pb analysis, including sampling year, lake

surface area, lake depth (depth of the water column above the

coring location), model used for dating, the approximate mixed

depth, the average sedimentation rate, the focusing factor (F F),

and the oldest section in the core .......................................... 66

Classiﬁcation of elements into terrestrial (T), carbonate (C),
diagenetic (D, D1,D2),and anthropogenic (A, A1, A2) ............... 69

Date of the oldest section (date), focusing factors (FF) and
average sedimentation rates (rate) from sediment cores in ﬁve
lakes in Michigan, USA .................................................... 93

Areas, populations, and population densities for watersheds
of the ﬁve study lakes in Michigan ....................................... 111

Chapter 5
Table 1

Table 2

Table 3

Chapter 6
Table 1

Table 2

Chapter 7

Table 1

Table 2

Table 3

Table 4

Table 5

Table 6

Characteristics ofthe 12 Michigan study lakes

Selected data from 210Ph analysis, including the model used

for dating, the approximate mixed depth, the average
sedimentation rate, the focusing factor (FF) and the oldest

section in the core .........................................................

Summary of lead and cadmium accumulation rates in this
study (average 1985-1995) and accumulation rates measured
in the sediment, atmospheric deposition, and snowpack ﬁ'om

previous studies ............................................................

Correlation coefﬁcients (r) between anthropogenic lead
accumulation rates from 1970-1980 (19703) and

1990-2000 (19903) and watershed characteristics ....................

Correlation coefﬁcients (r) between anthropogenic lead
accumulation rates ﬁ'om 1970—1980 (19703) and

1990-2000 (19903) and land cover variables ..........................

Level 1 land cover groupings from two time periods .................

Accumulation rates of cadmium, copper, and lead (mg m'2 y")

during the 19703 ...........................................................

Accumulation rates (mg rn‘2 y") of cadmium, copper, and zinc

during the 19903 ...........................................................

Pearson product moment correlations (r) between log
anthropogenic cadmium accumulation rates and watershed

characteristics ..............................................................

Pearson product moment correlations (r) between log
anthropogenic cadmium accumulation rates and land cover

variables .....................................................................

Pearson product moment correlations (r) between log
anthropogenic copper accumulation rates and watershed

characteristics ..............................................................

xi

127

133

143

179

183

.. 205

.. 216

.. 218

227

227

228

Table 7

Table 8

Table 9

Appendix A

Table 1

Table 2

Table 3

Table 4

Table 5

Table 6

Appendix B

Table 1

Table 2

Appendix C

Table 1

Table 2

Table 3

Table 4

Pearson product moment correlations (r) between log
anthropogenic copper accumulation rates and land cover
variables ....................................................................... 228

Pearson product moment correlations (r) between log
anthropogenic zinc accumulation rates and watershed
characteristics ................................................................ 229

Pearson product moment correlations (r) between log
anthropogenic zinc accumulation rates and land cover

variables ...................................................................... 229
Focusing corrected anthropogenic inventories (ug/cmz) of

cadmium, copper, lead and zinc in ﬁfteen Michigan lakes.. .. .. 263
Focusing corrected anthropogenic accumulation rates (pg/mz/y)

of cadmium, copper, lead and zinc averaged from 1970 — 1980 ...... 263
Focusing corrected anthropogenic accumulation rates (pg/mzly)

of cadmium, copper, lead and zinc averaged ﬁom 1990 — 2000 ...... 264
Watershed characteristics that do not vary over time ................... 264
Watershed characteristics for the 19703 (1970-1980) .................. 265
Watershed characteristics for the 19903 (1990-2000) .................. 268
MIRIS land cover data codes ............................................... 286
2001 IFMAP land cover data codes ....................................... 287
Final concentrations of standards in ppb ................................. 298
Dilutions to create solutions for ICP-MS standards ..................... 303
Volume of solution added in uL to create ICP-MS standards ......... 304
Final concentrations of ICP-MS standards .............................. 304

xii

LIST OF FIGURES

Chapter 1
Figure 1 Flow chart of research approach .......................................... 8
Figure 2 Location of lakes sampled in Michgan from 1999-2001 ............. 9
Chapter 2

Figure l Sediment chemical concentration (mg/kg) proﬁles for
Gratiot Lake ................................................................... 35

Figure 2 Sediment chemical concentration (mg/kg) proﬁles for
Gull Lake ...................................................................... 38

Figure 3 Standardized concentrations of aluminum and lead in
sediments of Gull Lake .................................................... 39

Figure 4 Sediment chemical concentration (mg/kg) proﬁles for
Paw Paw Lake ............................................................... 42

Figure 5 Sediment chemical concentration (mg/kg) proﬁles for
Elk Lake ..................................................................... 44

Figure 6 Unmodiﬁed concentrations (mg/kg) and ranked values of lead in
sediments of Elk Lake ..................................................... 46

Figure 7 Cluster trees for sediment chronologies for four Michigan lakes. . . 49
Chapter 3
Figure 1 Location of eleven study lakes in Michigan ............................ 63
Figure 2 Sediment aluminum concentration proﬁles (mg/kg) of four
lakes representing the diversity of concentrations and proﬁles

found among the study lakes .............................................. 70

Figure 3 Sediment aluminum and iron concentration proﬁles (mg/kg) in
Elk Lake ...................................................................... 73

Figure 4 Arsenic and manganese sediment concentration proﬁles in
Elk Lake ...................................................................... 75

xiii

Figure 5

Chapter 4
Figure 1

Figure2

Figure 3

Figure 4

Figure 5

Figure 6

Figure 7

Figure 8

Figure 9

Chapter 5

Figure 1

Figure 2

Figure 3

Figure4

Figure 5

Focusing corrected anthropogenic inventories of (a) copper and
(b) lead calculated using the constant background (white) and

watershed correction technique (black) ...................................

Location of lakes sampled in Michigan, USA ........................

Log 2")Pb activity (Bq/g) versus accumulated dry
mass (g/cm2 )in Elk Lake .................................................

Copper accumulation rates; total, pre-1800 background values,
and watershed (aluminum corrected) background values for

Gull Lake .....................................................................

Aluminum accumulation rates over time in ﬁve inland lakes .........

Graphs showing the metal accumulation rates for study lakes
3. lead, b. copper ............................................................

Graphs showing the metal normalized rates for study lakes ...........

Graphs showing the metal inventories as a function of background

correction for study lakes: a. lead, b. copper ............................

Graphs showing decadal metal accumulation rates for study

lakes; a. lead, b. copper ......................................................

Graphs showing the metal accumulation rates versus watershed
population densities for study lakes; a. lead with Cass Lake,

b. lead without Cass Lake, c. copper, d. copper without Cass Lake.

Lakes sampled in Michigan, USA .......................................

2|on activity (Bq/g) proﬁles for Gratiot (21), Elk (b), Crystal M (c)
and Littleﬁeld (d) lakes ...................................................

Lead concentrations (mg/kg) over time in 12 Michigan lakes .......

Lead concentrations normalized to peak concentration in 12
Michigan lakes .............................................................

Cadmium concentrations (mg/kg) over time in 12 Michigan lakes..

xiv

78

89

93

97

98

101

103

107

109

113

127

130

135

135

137

Figure6

Figure 7

Figure8

Figure9

Cadmium concentrations normalized to peak concentration in
12 Michigan lakes .......................................................... 137

Average lead focusing-corrected anthropogenic accumulation
rates (ug/mZ/y) for: (a) 1925-1935, (b) 1965-1975, (c) 1985-1995... 141

Average cadmium focusing-corrected anthropogenic accumulation
rates (pg/mz/y) for: (a) 1925-1935, (b) 1965-1975, (c) 1985-1995... 145

Watershed population densities (people/kmz) over time ............... 148

Figure 10 Average focusing-corrected lead anthropogenic accumulation rates

(pg/m /y) versus watershed population densities for (a) 1925-1935,
(b) 1965-1975, (c) 1985-1995 .............................................. 150

Figure 11 Average focusing-corrected cadmium anthropogenic accumulation

rates (pg/mz/y) versus watershed population densities for
(a) 1925-1935, (b) 1965-1975, (c) 1985-1995 ........................... 154

Figure 12 Production of leaded petroleum, and concentration proﬁles of Elk

and Gull Lakes normalized to the peak values ........................... 156

Figure 13 Total production of cadmium and smelter production of cadmium

Chapter 6
Figure 1

Figure 2

Figure 3

Figure 4

in the United States, and concentration proﬁles of Gull and
Paw Paw Lakes normalized to peak values ............................... 157

Location of lakes sampled in Michigan, USA ............................ 171

Anthropogenic accumulation rates of lead (1.1g m'2 y") in
16 Michigan lakes ............................................................ 177

Log of average anthropogenic lead accumulation rates from the
period 1990 — 2000 plotted versus average rates from 1970 — 1980... 177

Log lead anthropogenic accumulation rates (11g rn'2 y") for the

19703 and 19903 versus the two variables for each time period that
together explains the greatest variance in lead accumulation rates. . . 185

XV

Chapter 7
Figure 1

Figure 2

Figure 3

Figure4

Figure 5

Figure6

Figure 7

Figure 8

Figure 9

Location of study lakes in Michigan, USA .............................. 200

Sediment anthropogenic accumulation rates (mg rn'2 y") of
cadmium in sixteen Michigan lakes ....................................... 207

Average cadmium sediment log anthropogenic accumulation

rates (11g m'2 y 1') 1n the 19903 plotted versus average values for

the 19703 ....................................................................... 208
Sediment copper anthropogenic accumulation rates (mg m'2 y") ...... 210

Average copper sediment log anthropogenic accumulation rates

(11g m y I') in the 19903 plotted versus average values for the
19703 ........................................................................... 211
Sediment zinc anthropogenic accumulation rates (mg m'2 y"). .. . 212

AverageI zinc sediment log anthropogenic accumulation rates
(11g m y 1') 1n the 19903 plotted versus average values for the
19703 ........................................................................... 212

Average log cadmium anthropogenic accumulation (11g m'2 y")
rates in sediments of Michigan lakes ....................................... 215

Average log copper anthropogenic accumulation (ug rn'2 y")
rates in sediments of Michigan lakes ....................................... 219

Figure 10 Average log zinc anthropogenic accumulation (11g m'2 y") rates

in sediments of Michigan lakes ............................................. 221

Figure 11 Estimated rates of sulfate deposition rates (kg ha'1 y") to Michigan

study lakes ..................................................................... 223

Figure 12 Urban land cover in the Cass Lake watershed in 1975 and 1995 ...... 224

Figure 13 Average slope (log degrees) of 2study lake watersheds and

population density (people km 2) distribution in Michigan ............. 230

Figure 14 Average log anthropogenic accumulation rates (mg m'2 y") of

cadmium in sediments deposited in the 19703 plotted versus log
average slope 01f the watersheds (degrees) and sulfate deposition
rates (kg ha'1 y ) .............................................................. 233

xvi

Figure 15 Average log anthropogenic accumulation rates (mg in2 y") of
cadmium in sediments deposited in the 19903 plotted versus log
percent urban landcover of the watersheds .......................................... 234

Figure 16 Average log anthropogenic accumulation rates (mg m'2 y") of
copper in sediments deposited in the 19703 plotted versus log
population density (people km'z), and log slope corrected k-factor. . .. 236

Figure 17 Average log anthropogenic accumulation rates (mg m'2 y") of
copper in sediments deposited in the 19903 plotted versus log
percent urban landcover of the watersheds ................................ 237

Figure 18 Average log anthropogenic accumulation rates (mg rn'2 y") of
zinc in sediments deposited in the 19703 plotted versus log
population density (people km'z) ........................................... 238

Figure 19 Average log anthropogenic accumulation rates (mg m'2 y") of
zinc in sediments deposited in the 19903 plotted versus log
percentage urban in the watershed, and log slope within a 100 m
buffer of the lake (degrees) .................................................. 239

Images in this dissertation are presented in color.

xvii

CHAPTER 1
INTRODUCTION

Humans have signiﬁcantly modiﬁed the natural cycle of several metals (e.g.,
Cd, Cu, Pb, Zn), resulting in enrichment of these metals in the environment and
toxicological concerns for both wildlife and humans (Nriagu, 1978; Nriagu, 1989;
Eisenreich and Strachan, 1992; Catallo et al., 1995; Sanudo-Wilhelmy and Gill, 1999).
Environmental legislation, such as the Clean Air and Water Acts, reduced inputs of
lead to the environment (Graney et al., 1995; Long et al., 1995; Callender and
vanMetre, 1997; Callender and Rice, 2000; Simonetti et al., 2000; Mecray et al.,
2001), but concentrations in the water, air and sediment remain elevated (Graney et
al., 1995; Long et al., 1995; Callender and vanMetre, 1997; Mecray et al., 2001).
Humans continue to alter the environment in ever increasing ways, including
population growth and urban sprawl, which may result in new and emerging
contaminants and sources. In order to continue reducing the loading of toxic trace
metals to the environment it is necessary to understand current sources for these
metals, and their pathways through the environment. Additionally, by understanding
the history of contaminant inputs, it may be possible to determine how and if sources
have changed over time. One technique for understating the history of trace metal
inputs to a region is through the examination of sediment chronologies from multiple
lakes.

An accurate record of historical chemical inputs to the lake is needed to be able
to explore which processes and sources are inﬂuencing chemical loading to the

environment, and how they have changed over time. Ideally, this history would be

recorded in such a way that the different sources and/or processes inﬂuencing the
chemical inputs could be differentiated. Such a record exists within sediment
chronologies ﬁ'om inland lakes (Callender and Rice, 2000), which act as collectors of
metals from both the local watershed and regional atmospheric deposition. The
disadvantage of sediment chronologies is that terrestrial inputs from the watershed,
anthropogenic inputs from the watershed and atmosphere, and autochthonous
materials are deposited together in one mixed record, making it difﬁcult to
differentiate the relative importance of each different source (Balogh et al., 1999).
However, the presented approach has four aspects that facilitate the differentiation of
changes in sources and processes: data are collected for multiple elements rather than
only contaminants, data are represented ﬁ'om large time intervals (>200 y), data are
collected over a large spatial area, and data are compared to watershed characteristics
that represent different sources and pathways. Although all of these approaches have
been used individually (Bruland et al., 1974; Dillion and Evans, 1982; Birch et al.,
1996; Heyvaert et al., 2000), this work is the ﬁrst to incorporate all four.

It is necessary to quantify the rate of chemical input to a lake from
anthropogenic sources to determine possible sources and pathways of contaminants
using sediment chronologies. The multi—element approach allows the differentiation
between natural inputs (inputs of metals from naturally occurring erosion or
atmospheric deposition), secondary human inputs (increased inputs due to erosion
from human activities, such as clear cutting), and direct human inputs (selective
enrichment of certain elements due to human activities, such as lead enrichment due to

the use of leaded gasoline) (Bruland et al., 1974; Johnson and Nicholls, 1988;

Rarnanathan et al., 1996; Heyvaert et al., 2000; Kang et al., 2000). The examination
of multiple elements also gives insight into the effects of diagenesis (Harrington et al.,
1998; Brown et al., 2000), and identifying those elements that have been mobilized
and represent geochemical processes rather than a history of inputs.

Sediment chronologies from multiple lakes provide information over both
space and time, and provide useful insight into the history of contaminant inputs.
Spatial patterns can be used to help determine if sources are controlled regionally (a
spatial gradient is observed), or at the watershed scale (no spatial gradient), but
provide little information about speciﬁc sources. Temporal patterns of metal
accumulation can be compared to a qualitative (e.g., presence of smelters, tanneries)
(Catallo et al., 1995; Spiethoff and Hemond, 1996) or a quantitative (e.g., copper
mining production records) (Kerfoot et al., 1994) history of the region to understand
the input of metals to the lake. Unfortunately, while in some cases the actual ﬂux
from a source is known (e.g., atmospheric ﬂux or sewer outfall) (Bruland etal., 1974;
Evans and Dillon, 1982; Johnson and Nicholls, 1988; Balogh et al., 1999), generally
there is not a historical record of ﬂux ﬁ-om a source.

Because the exact ﬂux from a particular source is often unknown (e.g., input
from mining), a quantiﬁable watershed characteristic that acts as an indicator of
source(s) may be used instead (e. g., copper mining production records as an indicator
of input from mining) (Kerfoot et al., 1994). Although this watershed characteristic
does not provide the exact proportion of the input that comes ﬁom that source, use of
watershed characteristics can indicate if a particular source inﬂuences chemical input

into the lake at different time periods. This study will compare anthropogenic

accumulation rates from multiple lakes to watershed characteristics for two time
periods (1970-1980, 1990-2000) to identify sources and pathways of metals through
the environment, and evaluate changes over time. These two time periods were
chosen to represent the period of intense industrial activity (19703) and a more current
time period after signiﬁcant environmental legislation (19903).

The watershed characteristics that will be examined in this study represent
physical, transport, and source characteristics. The physical characteristics include
lake surface area, watershed area, and watershed to lake area ratio. These parameters
may give insight into the importance of atmospheric deposition; a lake with a small
watershed to lake area ratio may be dominated by atmospheric deposition, whereas a
lake with a large ratio may be more inﬂuenced by inputs from the watershed (Dillion
and Evans, 1982).

The transport characteristics represent the pathway of metals to a lake. The
presence of a source of metals will not be recorded in the lake sediments unless those
metals are transported to the lake. The average slope of the watershed and K factor
(measure of erodibility) were included to represent the erodibility of soils within the
watershed (Detenbeck et al., 1993). Any metal deposited on the land from sources
such as pesticides or fertilizers (Muhlbaier and Tisue, 1981; Rice, 1999) must be
eroded before reaching the lake, and may be dependant on soil erodibility. Sulfate
deposition rates were estimated to evaluate the importance of atmospheric deposition.
The atmosphere can be an important pathway of metals to lakes (Winchester and
Nifong, 1971; Gatz, 1975; Gatz et al., 1989; Nriagu, 1989; Callender and vanMetre,

1997; Sweet et al., 1998; Rice, 1999; Callender and Rice, 2000; Simonetti et al., 2000;

Goodarzi et al., 2001; Souch et al., 2003), but accurate estimations of deposition rates
of metals for each watershed are difficult to determine. Sulfate deposition, however,
has been monitored throughout Michigan (National Atmospheric Deposition Program
(NRSP-3)/National Trends Network, 2003), and can be used to estimate the relative
distribution of atmospheric deposition due to coal and oil burning. Both sulfate and
trace metals are released with the burning of fossil fuels (Nriagu, 1989; Berner and
Bemer, 1996). Sulfate deposition rates will not, however, represent atmospheric
deposition ﬁom other sources (e.g., copper from copper smelting, lead from the
burning of leaded gasoline).

The source characteristics include Toxic Release Inventory data, population
density, and land cover. Releases of toxic chemicals are required to be reported by the
Emergency Planning and Community Right-to-Know Act of 1986 (EPCRA),
including the location and quantity of release. These data represents point source
inputs that may be unique to that lake. Watershed population density and urban land
cover both represent sources related to human development, including: releases from
the wear of automobile parts (Rice, 1999), automobile emissions (Callender and
vanMetre, 1997; Rice, 1999), road dust (Shahin et al., 2000), wastewater, and
household and yard waste (Callender and Rice, 2000). Agricultural land use
represents pesticide and fertilizer use (Muhlbaier and Tisue, 1981; Rice, 1999). Forest
and wetland land cover correspond to areas that should have no anthropogenic
sources.

All of the characteristics are quantiﬁed on the watershed scale, but it is

conceivable that sources that are closer to the lake will be more important than those

further away (Comeleo et al., 1996; Richards et al., 1996; Basnyat et al., 1999).
Previous work on water quality parameters and aquatic sediments provides mixed
results, with some variables predicted best with whole watershed data, and others
when proximity to the waterbody was taken into account (Comeleo et al., 1996;
Richards et al., 1996; Johnson et al., 1997; Basnyat et al., 1999). To address the issue
of proximity, land cover, K factor, and slope were quantiﬁed in a 100 m buffer around
the lake (Richards et al., 1996; Johnson et al., 1997), and land cover was weighted by
the inverse of the distance to the lake along the ﬂow path (Comeleo et al., 1996).

Previous studies found relationships between: lead and zinc concentrations in
stream and reservoir sediments and population density (Callender and Rice, 2000); the
sum of concentrations of mercury, copper, lead and zinc in stream sediments and
population density (Rice, 1999); and metal concentrations in estuary sediments and
percent urban land, point source efﬂuent volumes, metals inputs (positive
relationships) and percent area of non-forested wetlands (negative relationship) (Paul
et al., 2002). These studies examine only recent relationships, rather than examining
the change in patterns over time. This study will use the approach of examining
spatial patterns and relationships to watershed characteristics, but also examine the
changes in these relationships over time.

This study will combine the multi-element, temporal and spatial aspects of
different studies preformed in the past to develop a more complete understanding of
what is inﬂuencing the chemical ﬂuxes and geochemistry of lake sediments. This
study is unique because it will combine all three aspects (spatial, temporal, multi-

element) for metal accumulation rates and both geochemical and anthropogenic

factors landscape variables such as land use, population density, and soil erodibility.
This approach will lend insight both into the factors that inﬂuence the geochemistry of
lake systems, and to what degree a system is inﬂuenced by anthropogenic activities. It

will also help identify sources and pathways of toxic metals in the environment.

Hypothesis and approach
The goal of this work is to evaluate the sources for metals and their pathways
through the ecosystem, and how they have changed over time by assessing the relative
importance of factors inﬂuencing the rate of chemical input to lakes for different time
periods. The working hypothesis is: sources and pathways controlling metals in the
environment have changed over time, and these changes should be reflected both
in the spatial patterns of metal accumulation rates in lake sediments and the
changing correlations between metal accumulation rates and characteristics of
the lake watersheds.
A three-step approach is used to test this hypothesis (Figure 1) and these are to:
1. Identify elements that are representative of classes. Classes are groups of
elements that are inﬂuenced by the same sources and processes, and
therefore can be represented by one element. Using this information,
identify elements that are strongly inﬂuenced by direct human inputs and
quantify anthropogenic accumulation rates.
2. Determine the inﬂuence of different watershed characteristics on
accumulation rates of each of the elements shown to be dominantly

inﬂuenced by anthropogenic actions at each time interval. At each time

interval, the inﬂuence of various characteristics (e.g., land use, population

density) will be determined for each of these elements.

3. Examine the temporal changes in the importance of different factors.
Pb Zn Cd

 

Identify elements influenced
by similar sources/processes,
determine rates of human

inputs
1 Concentration

Time

 

 

 

Determine influence of
different factors on
anthropogenic elements at
each time interval

 

 

 

 

 

 

 

 

 

 

 

l Watershed
Determine how the relative characteristic
importance of different
factors changes over time 9 3
‘3 a ,
a o
E g y a o
g m . .1 a .
Watershed
characteristic

Figure 1. Flow chart of research approach.

This hypothesis was investigated through the sampling of sediment from
seventeen Michigan lakes (Figure 2), analysis and interpretation of metals in the
sediment, and of data comparison to watershed characteristics during two time
intervals (19703 and 19903). This dissertation presents a broad scale approach to
determine regional patterns, sources, and pathways of metals. It is beyond the scope

of this work to describe each lake individually, but detailed descriptions of results and

additional data analysis for each of the lakes can be found in four year end reports

(Simpson et al., 2000; Yohn et al., 2001; Yohn et al., 2002b; Yohn et al., 2003).

 

Torch‘ "

Elk 0 .1?i
Crystal B -—- J. lL—IT‘E 'Higgins

isHoughton

Littleﬁeld

Cadillac . ITQT
I- ., .

    

Crystal M

 

Paw Paw

Figure 2. Location of lakes sampled in Michigan ﬁ'om 1999-2002 (0 1999,

O 2000, I 2001, A 2002).

Chapter 2 describes the advantages and disadvantages of using factor or cluster
analysis for grouping elements with similar depth proﬁles into classes (Yohn et al., in
preparation-a). This chapter focuses primarily on the statistical tests using data from
four representative lakes, while Chapter 3 provides the resultant groupings from the

eleven lakes sampled ﬁom 1999-2001, excluding Cass Lake (Yohn and Long, in

preparation). The geochemistry and possible sources and processes inﬂuencing each
class are discussed, and elements to represent each class are determined. The
importance of the multi-element approach in the calculation of anthropogenic
accumulation rates is also described.

Chapters 4 and 5 provide a proof of concept for steps two and three and
preliminary data analysis. In Chapter 4, temporal patterns and changing spatial
patterns in ﬁve Michigan lakes (sampled during 1999, Figure 2) are used to assess the
importance of local and regional sources for copper and lead (Y ohn et al., 20023).
Additionally, accumulation rates of lead and copper were compared to watershed
population densities to further evaluate the importance of local sources. In Chapter 5
the changing source of lead and cadmium were evaluated by examining the similarity
of sediment chronologies among 12 lakes (sampled during 1999-2001, Figure 2),
spatial patterns of lead and cadmium accumulation for three time periods (1925-1935,
1965-1975, and 1985-1995), and the relationship between lead and cadmium
accumulation rates in the sediment and watershed population densities for those three
time periods (Yohn et al., 2004).

Lead inputs to inland lakes are the focus of Chapter 6, as well as the
relationship between lead anthropogenic accumulation rates and characteristics of the
watersheds (Y ohn et al., in preparation-b). The accumulation rates of lead to 16 lakes
(excluding Witch Lake) were compared to the characteristics of the watershed (e.g.,
population density, land cover), and multiple regression was used to determine which
parameters best predict the rate of metal inputs. This chapter includes a comparison of

two time periods, 1970-1980 and 1990-2000. These time periods differ from Chapter

10

5 because of the availability of land cover data. Witch Lake was excluded due to the
local inﬂuence of mining on this lake.

Chapter 7 includes an analysis of copper, cadmium, and zinc. These elements
have been shown to be primarily controlled by human actions (Chapters 1 and 2) and
are discussed in a similar manner as lead in Chapter 6. This section is separated from
the previous chapter in order to keep Chapter 4 at publishable length, but contains
similar methods of data interpretation, and compares the same two time periods.

A summary and synthesis of the study is presented in Chapter 8. Additionally,
this section discusses recommendations for the approach and direction of future work,

and the transferability of this approach to other locations.

11

References

Balogh, S.J., Engstrom, D.R., Alemendinger, J .E., Meyer, ML. and Johnson, K.D.,
1999. History of mercury loading in the Upper Mississippi River reconstructed
from sediments of Lake Pepin. Environ. Sci. Technol. 33, 3297-3302.

Basnyat, P., Teeter, L.D., Flynn, K.M. and Lockaby, B.G., 1999. Relationships
between landscape characteristics and nonpoint source pollution inputs to
coastal estuaries. Environmental Management 23, 539-549.

Berner, E.K. and Berner, R.A., 1996. Global environment : water, air, and
geochemical cycles. Prentice Hall, Upper Saddle River, N.J., xiv, 376 pp.

Birch, L., Hanselrnann, K.W. and Bachofen, R., 1996. Heavy metal conservation in
Lake Cadagno sediments: historical records of anthropogenic emissions in a
meromictic alpine lake. Water Resources 30, 679-687.

Brown, E.T., Callonnec, LL. and German, CR, 2000. Geochemical cycling of redox-
sensitive metals in sediments from Lake Malawi: A diagnostic paleotracer for
episodic changes in mixing depth. Geochim. Cosmochim. Acta 64, 3515-3523.

Bruland, K.W., Bertine, K., Koide, M. and Goldberg, ED, 1974. History of metal
pollution in Southern California costal zone. Environ. Sci. Technol. 8, 425-
432.

Callender, E. and Rice, K.C., 2000. The urban environmental gradient: Anthropogenic
inﬂuences on the spatial and temporal distributions of lead and zinc in the
environment. Environ. Sci. Technol. 24, 232-238.

Callender, E. and vanMetre, RC, 1997. Reservoir sediment cores show US lead
declines. Environ. Sci. Technol. 31, A424-A428.

Catallo, WJ., Schlenker, M., Gambrell, RP. and Shane, B.S., 1995. Toxic chemicals
and trace metals from urban and rural Louisiana lakes: recent historical proﬁles
and toxicological signiﬁcance. Environ. Sci. Technol. 29, 1439-1446.

Comeleo, R.L., Paul, J .F ., August, P.V., C0peland, J., Baker, C., Hale, SS. and
Latimer, R.W., 1996. Relationships between watershed stressors and sediment
contamination in Chesapeake Bay estuaries. Landscape Ecology 11, 307-319.

Detenbeck, N.E., Johnston, CA. and Niemi, G.J., 1993. Wetland Effects on Lake

Water-Quality in the Minneapolis St-Paul Metropolitan-Area Landscape
Ecology 8, 39-61.

12

Dillion, PJ. and Evans, RD, 1982. Whole-lake lead burdens in sediments of southern
Ontario, Canada Hydrobiologia 91, 121-130.

Eisenreich, SJ. and Strachan, W.M.J., 1992. Estimating atmospheric deposition of
toxic substances in the Great Lakes - an update, Canada Centre for Inland
Waters, Burlington, Ontario, January 31-Febuary 2, 1992.

Evans, RD. and Dillon, P.J., 1982. Historical changes in anthropogenic lead fallout in
southern Ontario, Canada. Hydrobiologia 91, 131-137.

Gatz, DR, 1975. Pollutant aerosol deposition into southern Lake Michigan. Water Air
Soil Poll 5, 239-251.

Gatz, D.F., Bowersox, V.C. and Su, J., 1989. Lead and cadmium loadings to the Great
Lakes from precipitation. J. Gt. lakes Res. 15, 246-264.

Goodarzi, F., Sanei, H. and Duncan, W.F., 2001. Monitoring the distribution and
deposition of trace elements associated with a zinc-lead smelter in the Trail
area, British Columbia, Canada. J. Environ. Monit. 3, 515-525.

Graney, J .R., Halliday, A.N., Keeler, G.J., Nriagu, J 0., Robbins, J .A. and Norton,
S.A., 1995. Isotopic record of lead pollution in lake sediments from the
northeastern United States. Geochim. Cosmochim. Acta 59, 1715-1728.

Harrington, J.M., Laforce, M.J., Rember, W.C., Fendorf, SE. and Rosenzweig, RF,
1998. Phase associations and mobilization of iron and trace elements in Coeur
d'Alene Lake, Idaho. Environ. Sci. Technol., 650-656.

Heyvaert, A.C., Renter, J .E., Slotton, D.G. and Goldman, CR, 2000.
Paleolirnnological reconstruction of historical atmospheric lead and mercury
deposition at Lake Tahoe, California-Nevada. Environ. Sci. Technol. 34, 3588-
3597.

Johnson, L.B., Richards, C., Host, GE. and Arthur, J .W., 1997. Landscape inﬂuences
on water chemistry in Midwestern stream ecosystems. Freshwater Biology 37,
193-205.

Johnson, MG. and Nicholls, K.H., 1988. Temporal and spatial trends in metal loads to
sediments of Lake Sirncoe, Ontario. Water, Air and Soil Pollution 39, 337-354.

Kang, W., Trefry, J .H., Nelsen, TA. and Wanless, HR, 2000. Direct atmospheric

inputs versus runoff ﬂuxes of mercury to the lower Everglades and Florida
Bay. Environ. Sci. Technol. 34, 4058—4063.

13

Kerfoot, W.C., Lauster, G. and Robbins, J .A., 1994. Paleolimnological study of
copper mining around Lake Superior: Artiﬁcial varves from Protage Lake
provide a high resolution record. Limnology and Oceanography 39, 649-669.

Long, D.T., Eisenreich, SJ. and Swackhammer, D.L., 1995. Metal Concentrations in
Sediments of the Great Lakes, Great Lakes Protection Fund.

Mecray, E.L., King, J .W., Appleby, P.G. and Hunt, AS, 2001. Historical trace metal
accumulation in the sediments of an urbanized region of the Lake Champlain
watershed, Burlington, Vermont. Water Air Soil P01] 125, 201-230.

Muhlbaier, J. and Tisue, GT, 1981. Cadmium in the Southern Basin of Lake-
Michigan. Water Air Soil Poll 15, 45-59.

National Atmospheric Deposition Program (NRSP-3)/National Trends Network, 2003.
NADP Program Ofﬁce, Illinois State Water Survey, 2204 Grifﬁth Drive,
Champaign, IL 61820.

Nriagu, 10. (Editor), 1978. The Biogeochemistry of Lead in the Environment, 1A.
Elsevier, New York, 422 pp.

Nriagu, J .O., 1989. A Global Assessment of Natural Sources of Atmospheric Trace-
Metals. Nature 338, 47-49.

Paul, J .F., Comeleo, R.L. and Copeland, J., 2002. Landscape metrics and estuarine
sediment contamination in the mid-Atlantic and southern New England
regions. J. Environ. Qual. 31, 836-845.

Rarnanathan, A.L., Subramanian, V. and das, B.K., 1996. Sediment and heavy metal
accumulation in the Cauvery basin. Environ. Geol. 27, 155-163.

Rice, K.C., 1999. Trace-element concentrations in streambed sediment across the
conterminous United States. Environ. Sci. Technol. 33, 2499-2504.

Richards, C., Johnson, LB. and Host, GE, 1996. Landscape-scale inﬂuences on
stream habitats and biota. Canadian Journal of Fisheries and Aquatic Sciences
53, 295-31 1.

Sanudo-Wilhelmy, SA. and Gill, G.A., 1999. Impact of the Clean Water Act on the
levels of toxic metals in Urban Estuaries: the Hudson River estuary revisited.
Environ. Sci. Technol. 33, 3477-3481.

Shahin, U., Yi, S.M., Paode, RD. and Holsen, T.M., 2000. Long-term elemental dry

deposition ﬂuxes measured around Lake Michigan with an automated dry
deposition sampler. Environmental Science & Technology 34, 1887-1892.

14

 

Simonetti, A., Gariepy, C. and Carignan, J ., 2000. Pb and Sr isotopic compositions of
snowpack from Quebec, Canada: Inferences on the sources and deposition
budgets of atmospheric heavy metals. Geochim. Cosmochim. Acta 64, 5-20.

Simpson, S.J., Long, D.T., Geisey, J .P. and Fett, J .D., 2000. Inland lakes sediment
trends: sediment analysis results for ﬁve Michigan Lakes (MI/DEQ/SWQ-
01/030), Department of Environmental Quality, East Lansing, MI.
http://www.michigan.gov/deq/0,1607,7-1 35-33 1 3_3 686_3 728-32365—,00.html

Souch, C.J., F ilippelli, G.M., Dollar, N., Perkins, S. and Mastaler'z, M., 2003.
Accumulation rates of airborne heavy metals in wetlands. Phys. Geogr. 23, 21-
43.

Spiethoff, HM. and Hemond, HF, 1996. History of toxic metal discharge to surface
waters of the Aberjona Watershed. Environ. Sci. Technol. 30, 121-128.

Sweet, C.W., Weiss, A. and Vermette, S.J., 1998. Atmospheric deposition of trace
metals at three sites near the Great Lakes. Water Air Soil P01] 103, 423-439.

Winchester, J .W. and Nifong, GD, 1971. Water pollution in Lake Michigan by trace
elements from pollution aerosol fallout. Water, Air, Soil Pollut. 1, 50-64.

Yohn, SS. and Long, D.T., in preparation. The importance of the multi-element
approach for interpreting sediment histories.

Yohn, S.S., Long, D.T., Fett, J .D. and Patino, LC, 2004. Regional versus local
inﬂuences on lead and cadmium loading in the Great Lakes region. App].
Geochem. 19, 1157-1175.

Yohn, S.S., Long, D.T., Fett, J.D., Patino, L.C., Giesy, J .P. and Kannan, K., 2002a.
Assessing environmental change through chemical-sediment chronologies
from inland lakes. Lakes & Reservoirs: Research and Management 7, 217-230.

Yohn, S.S., Long, D.T., Giesy, J.P., Fett, J .D. and Kannan, K., 200]. Inland lakes
sediment trends: sediment analysis results for two Michigan Lakes.
MI/DEQ/WD-02/115, Department of Environmental Quality, East Lansing.

Yohn, S.S., Long, D.T., Giesy, J.P., Scholle, L.K., Patino, L.C., Fett, JD. and Kannan,
K., 2002b. Inland lakes sediment trends: sediment analysis results for ﬁve
Michigan Lakes. MI/DEQ/WD-03/052, Department of Environmental Quality,
East Lansing.

15

Yohn, S.S., Long, D.T. and Parsons, M.J., in preparation-a. Lead in the environment:
evaluating sources and pathways over time using lake sediments.
Environmental Science & Technology.

Yohn, S.S., Long, D.T. and Winkler, J.A., in preparation-b. The use of multivariate
statistics for the interpretation of sediment chronologies. Limnology and
Oceanography: Methods.

Yohn, S.S., Parsons, M.J., Long, D.T., Giesy, J.P., Scholle, L.K. and Patino, LC,

2003. Inland lakes sediment trends: sediment analysis results for six Michigan
Lakes, Michigan Department of Environmental Quality, East Lansing, MI.

16

CHAPTER 2
THE USE OF MULTIVARIATE STATISTICS FOR THE INTERPRETATION OF
SEDIMENT CHRONOLOGIES

Introduction

Sediment chronologies have long been used to understand both natural and
human inﬂuences on the enviromnent (Edgington and Robbins, 1976; Kerfooteta1.,
1994; Graney et al., 1995; Kolak et al., 1998; Khim et al., 1999; Lamoureux, 1999).
While earlier studies often focused on speciﬁc chemicals of environmental concem,
such as lead and cadmium (Iskander and Keeney, 1974; Dillion and Evans, 1982;
Evans and Dillon, 1982; Johnson and Nicholls, 1988), the increased ability for multi-
elemental analysis (e.g., inductively coupled plasma technology) has led to the
simultaneous examination of a large number of elements in lake sediments (Kerfoot
and Robbins, 1999; Sanei et al., 2001; Tuncer et al., 2001; Manta et al., 2002). The
analysis of a suite of elements may provide insight into the geochemical processes
occurring within the lake sediment and help clarify sources for different elements.

While the analysis of many elements may provide for a better understanding of
the lake system, it also leads to large quantities of data which may be difﬁcult to
interpret. For example, the analysis of 25 elements in a sediment core vertically
sectioned into 50 samples leads to over 1,200 individual data points. Because the data
set is large, many investigators are using multivariate statistics to examine structure in
the data that can aid in interpreting results (Kerfoot et al., 1994; Tuncer et al., 2001;

Manta et al., 2002).

17

 

Multivariate statistics are typically used for initial data analysis to determine
which elements follow similar trends over time in the sediment core, with the ultimate
objective of using elemental behavior as an indicator for source (Kerfoot et al., 1994;
Sanei et al., 2001; Tuncer et al., 2001). Similar vertical (e.g., temporal) trends are a
good indication that a particular suite of elements may originate from the same source
(e.g., aluminum ﬁorn clay), or are inﬂuenced by the same process (e.g., erosion).
Therefore, grouping elements with similar trends based on a statistical technique is
useful because it reduces the number of individual trends to interpret, and may lend
insight into the important sources or processes inﬂuencing that assemblage based on
the elements found in the group. A group of elements may contain element(s) with
well known sources (e.g., aluminum from natural terrestrial inputs such as clay); thus,
other less studied elements in that group (e.g., vanadium) may be considered to be
inﬂuenced by the same source and/or process (Manta et al., 2002). Additionally, a
group may contain a suite of elements that matches the loading history of a particular
point source, such as mining (Tuncer et al., 2001).

It is possible to determine which elements follow similar temporal trends by
visual comparison of each trend. However, as the number of variables increases, it
becomes more difﬁcult to group each element, and more likely that not all
relationships will be correctly identiﬁed. The potential beneﬁt of using multivariate
statistics to interpret sediment chronologies involving multi-element data sets is clear.
What is less apparent is which multivariate technique is the most appropriate to use,
the potential for incorrect interpretation, and what methodology should be used within

each technique (e.g., should ranked or “raw” data be used?) Both factor and cluster

18

analysis have been used for sediment chronologies (Kerfoot and Robbins, 1999;
Manta et al., 2002), with little discussion as to which technique is more appropriate to
use or what the relative merits of the two approaches are. Additionally, each statistical
technique has different options. For example, factor analysis could include principal
component analysis or common factor analysis, and the starting point for the analysis
could be a correlation matrix calculated using a parametric method (e.g., Pearson’s
product moment correlation coefﬁcient) or a nonparametric method (e.g., Spearman’s
correlation coefﬁcient). Different distance measures and clustering algorithms may be
used for cluster analysis. As the use of multivariate statistics to interpret sediment
chronologies increases, it is useful to describe and compare the statistical decision-
making processes used so that results may be compared among studies. The results of
earlier studies that have employed statistical techniques, such as factor and cluster
analysis, to reduce sediment chronology data were generally reported with little
description of the decision making processes involved (Kerfoot et al., 1994; Sanei et
al., 2001; Tuncer et al., 2001). Consequently, it is difﬁcult for other investigators to
reproduce the analyses that might aid in the development of a consistent set of
protocols so that comparable interpretations can be made among studies.

Therefore, we have chosen four lakes ﬁorn the Michigan Inland Lakes Trend
Monitoring Program (http://www.michigan.gov/deq/0,1607,7-135-3313_3686_3728-
32365-,00.html) that represent a variety of lake conditions to illustrate the impact of
the choice of statistical technique (either factor analysis or cluster analysis) along with
the associated decision-making processes on the interpretation of sediment

chronologies. The four lakes are a subset of 17 lakes for which sediment cores were

19

collected and analyzed for 18 elements. Statistical analyses were used for the purpose
of choosing a representative element for each source/process grouping, and identifying
elements that have been inﬂuenced by human inputs. For factor analysis, results from
principle component and common factor analysis were compared, and for cluster
analysis three clustering algorithms (Ward’s, average, and single linkage) were
examined. For both statistical methods, we explore the effect of outliers on the
resulting groups. For all methods, the resulting groupings were visually checked for
internal consistency.

The overall goal of this paper was to suggest a statistical technique and
decision making processes for grouping elements in sediment chronologies with
similar proﬁles, so that the proposed method may be tested in other regions and
modiﬁed if necessary. It is inevitable that the use of multivariate statistics for this
purpose will continue to increase, and it is essential for these statistics to be reported
consistently. This paper also continues to explore the inﬂuences on the geochemical

associations of chemicals in the environment.

Methods and Procedures
In this section the procedures for the lake coring and geochemical analyses are
described. Also, methods for identifying outlier points are outlined. In addition, a
relatively comprehensive review of the two primary statistical methods, factor analysis
and cluster analysis, is provided. We felt that a relatively lengthy description of these
methods was necessary as many readers may not be aware of some of the important

aspects and assumptions of these methodologies. Furthermore, an understanding of

20

 

 

the methods is required to provide context for our recommendations on the future use

of these methods in interpreting sediment chronologies.

Geochemical analyses
Elk, Gratiot, Gull, and Paw Paw lakes were chosen from the seventeen lakes

sampled as part of the Michigan Department of Environmental Quality’s (MDEQ)
Inland Lakes Sediment Trend Monitoring program. They were chosen to represent a
variety of lake sizes, sedimentation rates, sediment types (e.g., calcium carbonate rich
and carbonate poor), and extent of human development around the lake. Sediment
cores were collected from these four lakes during 1999 and 2001. Elk, Gull and Paw
Paw lakes are located in the Lower Peninsula of Michigan on Pleistocene glacial
deposits which are generally carbonate rich. Gratiot Lake is located in the copper rich
Keweenaw area (Upper Peninsula of Michigan), part of the Canadian Shield, with
little to no glacial deposits, and is calcium carbonate poor (Door and Eschman, 1970).
The four lakes have very different sediment chemistries and proﬁles. Cores collected
from Gull and Elk lakes comprise sediment deposited over several hundred years
(~500 and 700 years respectively), while Gratiot Lake includes only 200 years and
Paw Paw Lake 80 years of sediment history. Gull and Elk lakes are carbonate rich
(300-400 g/kg Ca), while Paw Paw Lake has moderate calcium concentrations (40
g/kg Ca) and Gratiot Lake has very low calcium (2 g/kg Ca). For many elements in
Gull and Elk lakes, the sediment proﬁle is characterized by a large increase in
concentration in the mid-18003, while Gratiot and Paw Paw lakes are less dominated

by one clear trend (Yohn et al., 2002b; Yohn et al., 2004).

21

 

Pour replicate sediment cores from 45 to 56 cm long were taken from the
deepest portion of each lake using a MC-400 Lake/Shelf Multi-corer (San Diego, CA)
deployed from either the EPA Research Vessel Mudpuppy or the MDEQ Monitoring
Vessel Nibi. A new set of cores was immediately taken if the cores showed any
evidence of disturbance. Cores were examined and described in terms of color,
texture, and signs of zoobenthos. Cores were immediately extruded and sectioned
open to the atmosphere at 0.5 cm intervals for the top 5-8 cm, and at 1 cm intervals for
the remainder of the core.

210Pb was measured on one core from each lake to determine sedimentation
rates, sediment ages, and focusing factors by the Freshwater Institute in Winnipeg,
Manitoba, Canada. Sediments from Gratiot and Paw Paw lakes were dated using the
constant ﬂux: constant sedimentation (CF:CS) model (Golden et al., 1993), and horn
Gull and Elk lakes using the segmented CF :CS mode] (SCFzCS) (Heyvaert et al.,
2000). The use omeb dating for these cores is discussed in detail elsewhere (Yohn
et al., 2004).

Sediments for meta] analysis were frozen for storage, freeze-dried and digested
by nitric acid in a CEM-MDS-81D microwave (EPA Method 3051) (Hewitt and
Reynolds, 1990). Standard reference material (N 1ST SRM 8704 Buffalo River
Sediment) and procedural blanks were processed as well as three replicate digestions
on two samples from each lake. The concentrated-acid digests were ﬁltered through
an acid-washed, distilled-deionized or E-pure (Bamstead International, Dubuque, IA)
water rinsed 0.40 um polycarbonate ﬁlter (Nucleporem). Sediments were analyzed

for a suite of metals and metalloids including magnesium, aluminum, potassium,

22

 

calcium, titanium, vanadium, chromium, manganese, iron, copper, zinc, arsenic,
strontium, molybdenum, cadmium, barium, lead, and uranium in a Micromass
Platform inductively coupled plasma mass spectrometer with hexapole technology
(ICP-MS-HEX). Strontium data were not collected for Elk and Gull lakes, and

potassium data were not collected for Paw Paw Lake.

Outlier analysis
All of the data analysis techniques used in this study are sensitive to the

presence of outliers (Hair, 1998). Outliers are data points signiﬁcantly different from
the rest of the data. Outliers may be the result of analytical/sampling errors or actually
reﬂect unusually high or low element concentrations in the sediment. Although
principal component, factor, and cluster analyses do not require data to be normally
distributed (Hair, 1998), the vertical concentration data were ﬁrst log-transformed
sincewhatmay appeartobeanoutlierinthe “raw”datamay notbesowhenthedata
are transformed. A log transformation was chosen as geochemical data are often log-
normally distributed. However, based on the Lillefors test (Wilkinson 2000), the log
transformation did not signiﬁcantly increase the number of elements that were
normally distributed. Therefore, the original, non-transformed data were used in the
statistical amlyses and further examined for the presence of outliers. Data points were
considered to be outliers if they were outside the outer “fence” as determined by
boxplot analysis. The outer fence is deﬁned as:

upper hinge + (3 x hspread) or

lower hinge - (3 x hspread)

23

 

where the “hinge” is the ﬁrst or third quartile, and “hspread” is the distance between
the hinges (SPSS 2000). This method ﬁnds data points that are far outside the range
of the majority of the data, and these data may be inﬂuential when statistical analyses
are performed.

Outliers were identiﬁed for all four of the lakes. Two or three samples near the
surface had outlying values for several metals in Elk Lake (As, Mo, Mn, Ba, Fe).
Gratiot Lake had one or two outliers for aluminum, magnesium, and chromium, Gull
Lake had ﬁve outliers for arsenic, and Paw Paw Lake had four outliers for manganese,
and one for barium and molybdenum.

Obviously, if outliers are due to analytical or sampling error, these data need to
be removed from the data set. However, outliers that represent data accurately may be
statistically inﬂuential, and one or two outlying data points could determine how an
element proﬁle groups. Consequently, it may not be appropriate to simply delete these
inﬂuential points, as these points may be representative of certain processes. Our
approach was to perform principal component, factor, and cluster analysis on all data
for each lake and on data with the outliers removed. In addition, the statistical
analyses were performed on ranked data, as the inﬂuence of outlier points is decreased
when the data are ranked. Removal of outliers in Gratiot, Gull, and Paw Paw lakes did
not signiﬁcantly change the statistical results, or the resultant element groupings;
therefore, only the results for the analyses for the “raw” and ranked data are shown
below. Removal of outliers for Elk Lake did have a signiﬁcant impact on the results,

and results from all data types are discussed below.

24

Factor analysis

Factor analysis (principal component or common factor analysis) is a
technique used to reduce the number of variables in large data sets to a smaller number
of components or factors, and is based on a correlation or covariance matrix (Klovan,
1975; Davis, 1986). Correlation, rather than covariance, should be used for datasets
with variables that have greatly different ranges of values (Stenson and Wilkinson,
2000), such as sediment concentrations of elements. The resulting components or
factors should represent the underlying structure of the data (Hair, 1998). The
components or factors can be thought of as composite variables, which characterize
more than one of the original variables and represent a common theme (Joreskog et
al., 1976). For example, in our dataset, since the variables are metals, an anticipated
factor is one that represents the terrestrial group of elements. I

The ﬁrst factor in factor analysis explains the largest amount of variance in the
data, and is a linear combination of the original variables (Davis, 1986). The second
factor then explains as much of the residual variance as possible, with each additional
factor explaining less variance (Klovan, 1975). Factors may also be rotated to provide
a more meaningful solution. In this case, the ﬁrst factor does not explain the
maximum possible variance; instead, a chosen number of factors are rotated to more
evenly distribute the variance among factors and improve the interpretation of the
factors. The loading of a variable on a factor is the correlation between that variable
and the factor, and is a value from -1 to 1 interpreted in the same manner as a
correlation coefﬁcient (Joreskog et al., 1976). There is no agreement on what values

should be considered high, moderate or low for loadings, which makes comparisons

25

among studies utilizing factor analysis difﬁcult. Consensus on appropriate values or at
least justiﬁcation of the values chosen would facilitate the comparison of data from
sediment chronology studies. Relatively high values were chosen for this study on the
assumption that elements inﬂuenced by the same sources and processes will have very
similar proﬁles and should be highly correlated. Elements with loadings greater than
0.8 (or < -0.8) were considered to be highly loaded on a factor, and elements with
loadings of 0.6-0.8 were considered to be moderately loaded. The squared loading
represents the percentage of variance for that variable explained by the factor. For
example, in our data we expect aluminum to load highly on a terrestrial factor, but
lead to load weakly on that factor. The eigenvalue of a factor represents the total
variance explained by that factor, and is the sum of squared loadings for a factor. If an
eigenvalue of a factor is greater than one, that factor explains more variance than one
variable (Hair, 1998).

For factor analysis to be an appropriate technique the variables must be
correlated. A general guideline is that the correlation matrix should have a substantial
number of correlations coefﬁcients greater than 0.3 (Hair, 1998). Both a typical
parametric (i.e., Pearson’s) correlation coefficient and a nonparametric (i.e.,
Spearman’s) correlation coefﬁcient that is based on ranked data were calculated.
Correlations among most of the elements within all of the four study lakes were
greater than 0.3. Gull Lake had the greatest correlation among elements (98% of the
possible variable pairs had values >0.3 using the parametric correlation measure and
96% using the nonparametric measure), and Elk Lake had the second highest (74%

and 82% for the parametric and nonparametric coefﬁcients, respectively). The

26

variables were slightly less correlated for Gratiot Lake (65% and 64%), and
considerably fewer variable pairs had correlations >0.3 at Paw Paw Lake (48% and
48%).

Multiple factors are calculated from factor analysis, and it is not always
straightforward to determine the number of factors that are meaningful. Additionally,
when factors are rotated, the number of factors to rotate must be determined
beforehand. There are several techniques to determine the appropriate number of
factors, but the three most common are 1) retaining all factors with an eigenvalue
greater than one, 2) retaining all factors that explain greater than 5% variance, and 3)
using a scree test to estimate the appropriate number of factors to retain. A scree test
simply involves plotting the eigenvalue for each factor versus the number of factors;
the point at which the curve begins to straighten out horizontally represents the
maximum number of variables to retain. The scree test typically results in more
factors being retained than the other two methods (Hair, 1998). Our approach was to
use factor analysis without rotation, and determine the appropriate number of factors
to retain with each of the criteria individually. The factors retained by each approach
were then rotated. Two additional models were also developed by retaining and
rotating one more and one less number of factors. Then each of these models was
examined to determine which was most appropriate.

There are several considerations when determining the most appropriate
model. The total amount of variance explained should be high, but increasing the
number of factors always increases the explained variance. Ifnetaining and rotating an

additional factor results in a factor with no high loadings, the solution with fewer

27

factors may be more appropriate, even though the total explained variance is slightly
lower. Another consideration is the attainment of simple structure. The concept of
simple structure is that each variable is well described by one, and only one factor, and
therefore the resultant factors completely describe the underlying structure of the data
set. In this case, all variables will load highly on one factor, and weakly on all other
factors, and each factor will have more than one variable loaded highly on it. While
complete simple structure is uncommon, the solution chosen should result in many
variables being highly loaded on only one factor, rather than moderately loaded on
two or more factors. The ﬁnal consideration is that it should be possible to assign a
meaning to each factor. For each factor, it should be considered which variables load
highly or moderately on that factor, and if this combination of variables represents an
underlying dimension in the data set. For example, aluminum, titanium, cobalt and
vanadium may load highly on one factor, which would represent terrestrial inputs. A1]
of these considerations must be taken into account when choosing the most
appropriate solution.

As indicated earlier, each lake was also analyzed using principal component
analysis. Principal component analysis takes into account all the variance among
variables, which includes the common variance, speciﬁc (unique) variance, and error,
while common factor analysis uses only the common variance and ignores variance
unique to the variable (Davis, 1986; Hair, 1998). The communality of an element is
the amount of total variance that is common with the other variables. When the
communalities of most of the variables exceed 0.6, the results of common factor

analysis and principal component analysis are essential identical (Hair 1998). For all

28

 

lakes, the only elements having a communality less than 0.6 were chromium in Elk
and Gratiot lakes, and titanium, arsenic, magnesium and manganese in Paw Paw Lake
(T able 1). Not surprisingly, the results of the principal component analysis and
common factor analysis differed little, and only the results from the common factor
analysis will be discussed below.

Table l. Communalities of each element for the

four study lakes and Elk Lake with outliers removed.

Paw Elk no
Gratiot Gull Paw Elk outliers

Mg 0.872 0.71 1 0.529 0.987 0.979

 

AI 0.798 0.993 0.861 1.000 0.998
K 0.790 0.946 0.966 0.967
Ca 0.884 0.921 0.888 0.828 0.822
Ti 0.795 0.985 0.374 0.944 0.941
V 0.788 0.996 1.000 0.978 0.975
Cr 0.263 0.945 0.814 0.532 0.673
Mn 0.996 0.877 0.235 0.819 0.765
Fe 0.868 0.982 0.907 0.865 0.997

Cu 0.909 0.975 0.852 0.943 0.942
Zn 0.968 0.997 0.770 0.976 0.967
As 0.976 0.944 0.407 0.773 0.824
Sr 0.952 0.907

Mo ' 0.394 0.847 0.854 0.638 0.232
Cd 0.870 0.942 0.910 0.773 0.835
Ba 0.930 0.946 0.544 0.971 0.908
Pb 0.933 0.930 0.796 0.944 0.930
U 0.871 0.788 0.765 0.789 0.805

Cluster Analysis

Hierarchical cluster analysis is a grouping technique designed to maximize
within group homogeneity and between group heterogeneity (Grifﬁth and Amrhein,
1997; Gotz and Lauer, 2003). Although cluster analysis is typically used to group
similar samples, it also can be used effectively to group variables (Hair, 1998). The
goal of cluster analysis is to create groups of samples or variables (in this study,

elements) that are similar in multi-dimensional space (Gotz and Lauer, 2003).

29

 

Similarity is based on some distance measure, such as Euclidean distance measure or
correlation (Grifﬁth and Amrhein, 1997). When clustering variables with very
different ranges of values, it is most appropriate to use a correlation measure rather
than Euclidean distance (Wilkinson et al., 2000). Therefore, Pearson’s product
moment correlation coefficient was used in the analyses described below, and the
distance measure for each pair of variables for a given lake is one minus the
correlation coefﬁcient. Hierarchical cluster analysis ﬁrst groups the two elements that
are the most similar (closely correlated, short distance). Then the next two elements
that are the most similar are clustered. This may mean that a third element is included
with the ﬁrst cluster, or that two new elements are clustered. This process is continued
until all new elements are included in one cluster. Non-hierarchical methods, often
called optimization methods, are also available. However, these methods require the
number of clusters be speciﬁed before the analysis, and typically use Euclidean
distance rather than correlation as the similarity metric.

While it is relatively straightforward to measure the distance between two
elements, there are different methods of determining the distance between two
clusters, and different clustering algorithms are used to determine which clusters are
most similar. Three clustering algorithms were used in cm analyses: single linkage,
average, and Ward’s method. All three algorithms look at all possible combinations of
clusters, and determine which two are the most similar, and then join those two
clusters. None of these methods allow regrouping of variables, that is, once a variable
is placed in a cluster it cannot be removed. The average algorithm uses distances

between the average values of clusters as a similarity measure, whereas single linkage

30

uses the distance from one element in one cluster to one element in another cluster,
and uses those elements that are closest together. The Ward’s method takes a slightly
different approach, and attempts to minimize the within cluster sum of squares. Each
of these methods has different biases. Ward’s method is biased towards producing
clusters of similar size. The average distance algorithm tends to produce clusters with
similar variance. Single linkage may form long chainlike clusters where individuals in
the groups may end up being dissimilar (Hair, 1998). It is common practice to apply
more than one clustering procedure as a check for the stability of the groups and the

existence of “natural” groupings.

Assessment

A review of previous literature revealed that multiple source and processes
should be inﬂuencing depth proﬁles of metals in the sediments of the four lakes under
study. Based on the literature, these sources and processes could be 1003er
categorized as terresuial, anthropogenic, carbonate, and diagenetic. These initial
deﬁnitions guided our interpretation of the statistical groupings from the factor
analysis and cluster analysis procedures. Below, we describe how well the resulting
groups ﬁt these initially deﬁned source/process groups, the potential relationships
between these a priori identiﬁed source/process groupings, and additional groupings
that may exist. In addition, these a priori deﬁned groupings help to 1) cast the results
of the statistical analysis in the context of widely-recognized underlying processes and
2) evaluate the physical plausibility of the resulting statistical groupings. The
terrestrial group refers to elements inﬂuenced by elastic mineral detritus entering the

lake, and, based on earlier work, is likely to include aluminum, titanium and iron,

31

potassium, cobalt, nickel, magnesium, sodium, scandium, and the rare earth elements
(Bruland et al., 1974; Kemp and Thomas, 1976; Johnson and Nicholls, 1988; Boyle et
al., 1999; Kerfoot and Robbins, 1999; On et al., 2001; Sanei et al., 2001; Yohn et al.,
2002a). The carbonate group refers to elements inﬂuenced by the precipitation and
dissolution of calcium carbonate in the water column and, in addition to calcium, may
include strontium, magnesium, iron, manganese, and barium (Auer et al., 1996; Sanei
et al., 2001). The diagenetic group is intended to include elements that are gained,
lost, or redistributed in the deposited sediment either due to a change in the oxidation
state of that metal or association with a redox inﬂuenced element, and may include
iron, manganese, arsenic, molybdenum, and uranium (Cooper and Morse, 1998;
Brown et al., 2000). Finally, the anthropogenic group refers to elements whose
concentrations in the sediment have been enriched by human actions, and may include
arsenic, cadmium, copper, chromium, mercury, lead, and zinc (Bruland et al., 1974;
Iskander and Keeney, 1974; Lantzy and Mackenzie, 1979; Evans and Dillon, 1982;

Spiethoff and Hemond, 1996).

Factor Analysis
A lake by lake assessment of the efficacy of factor analysis in delineating

element groups with similar proﬁles follows. Additionally, possible causes of the
differing success among lakes are discussed. The results from the common factor
analysis are discussed in the most detail, although at the end of this section these
results are compared to those from factor analysis using a nonparametric correlation

matrix.

32

I) Gratiot Lake
Factor analysis on sediment chronology data from Gratiot Lake resulted in

three factors with variance greater than 5%, and two factors with eigenvalues greater
than one (Table 2). The use of two rotated factors resulted in the greatest number of
high loadings, whereas the use of a third factor resulted in one factor that was
moderately loaded on only manganese and barium (Table 2). Therefore, the solution
with two rotated factors was chosen as the most appropriate solution. Additionally,
this solution makes some geochemical sense. Elements loading positively on Factor 1
represent terrestrial elements, and include titanium, aluminum, strontium, magnesimn,
calcium, uranium, and vanadium. Potassium, manganese and barium load negatively
on Factor 1, and their trends are characterized by an increase in concentration near the
sediment-water interface (Figure 1). This may represent diagenetic processes
(Kneebone and Hering, 2000) or a diluting phase (e.g., organic matter), since these
elements increase in concentration as the terrestrial elements decrease (Jaquet et al.,
1982)

Elements that are highly or moderately positively loaded on Factor 2 appear to
represent the anthropogenic elements (Bruland et al., 1974; Iskander and Keeney,
1974; Evans and Dillon, 1982; Spiethoff and Hemond, 1996), and include copper,
arsenic, cadmium, lead, zinc and iron (Table 2). All of these elements have their
highest concentrations in the 1970s and decrease until the present, and probably record
the history of copper mining and smelting in the area (Figure l). The multivariate
statistics were particularly useful in identifying that iron is being inﬂuenced by

anthropogenic inputs. Iron is frequently grouped with diagenetic or terrestrial

33

elements, but in Gratiot Lake, iron clearly follows a trend similar to the other
anthropogenic elements. In this case, factor analysis allowed the identiﬁcation of an
unexpected anthropogenic element that otherwise may have been overlooked or
misinterpreted. The identiﬁcation of unexpected relationships is an advantage of
using statistical methods rather than only visually evaluating the data.

Table 2. Factor analysis results for Gratiot Lake. “Big” indicates the eigenvalue for
the factor, “°/o” is the percent variance explained by each factor, and “1%” is the total

percentage variance explained by the solution. Bold indicates high loading (>0.8), and
italics moderate loading (0.6-0.79).

 

 

 

 

 

 

 

 

 

2 factors,
3 factors, no rotation 3 factors, Varimax Varimax
Factor Factor Factor Factor Factor Factor Factor Factor
1 2 3 1 2 3 1 2

Mg -0. 792 0.471 0.151 0.922 0.032 0.143 0.922 0.048
Al -0. 629 0.598 0.211 0.867 0.212 0.041 0.833 0.234
K 0.856 -0.181 0.158 -0. 741 0.221 -0.439 -0.842 0.238
Ca -0.879 0.277 0.190 0.914 -0. 184 0.123 0.906 -0.164
TI -0. 773 0.444 -0.006 0.845 0.033 0.283 0.896 0.034
V -0.859 0.072 0.213 0.809 -0.357 0.083 0.793 -0.335
Cr -0.348 0.365 0.093 0.488 0.150 0.046 0.479 0.162
Mn 0.888 -0.291 0.401 -O. 725 0.103 -O. 677 -0.882 0.147
Fe 0.681 0.570 0.283 -0.203 0.787 -0.455 -0.336 0.81 1
Cu 0.431 0.827 0.195 0.096 0.907 -0.276 0.003 0.929
Zn 0.551 0.814 -0.054 -0.087 0.977 -0.081 -0.109 0.980
As 0.517 0.810 -0.231 -0.1 17 0.978 0.096 -0.083 0.949
Sr -0.959 0.075 0.164 0.877 -0.396 0.162 0.885 -0.380
Mo 0.092 0.620 0.028 0.226 0.585 -0.020 0.208 0.594
Cd 0.435 0.799 -0.208 -0.047 0.926 0.101 -0.015 0.904
Ba 0.769 -0.175 0.555 -O.542 0.143 -0. 784 —O. 731 0.199
Pb 0.689 0.671 -0.088 -0.279 0.919 -0.102 -0.299 0.917
U -0.829 0.41 1 0.122 0.915 -0.035 0.179 0.928 -0.022
Eng 8.82 5.12 0.92 7.22 5.86 1.77 7.96 5.88
% 49% 28% 5% 40% 33% 10% 44% 33%
T% 82% 83% 77%

 

34

2000‘
19501
1900‘
1850‘
1800

 

Mg

2000
1950-
1900*
1850‘

Cr
1800*
10

2000‘
1950*
1900‘
1850‘

Date

 

 

Sr

30007501

”ma

 

 

1 pr

 

 

 

 

1800
5

 

 

25 0.3

4 Al

4 ————l
50 200 300 1.2E4 2E5 40

Mo C

 

 

 

K

 

 

 

--_,,l L“... .ELa

E4 1.7E 500 25001500 35005.00

‘Cu

Ti

 

 

 

_nl_

6004

Zn

J

l
v

 

 

L l

 

 

 

100

 

 

 

.__‘,____———

 

 

+ ~—
0.6 0.0

 

 

 

80 50

Ba Pb

 

 

 

 

1.550

 

800

Concentration (mg/kg)

     

 

120 0

4U

 

 

.__.____—.‘_4

300.7

 

1.4

Figure l. Sediment chemical concentration (mg/kg) proﬁles for Gratiot Lake.

Chromium and molybdenum are the only elements that do not load highly or

moderately on any factor (Table 2). This was anticipated, since these two variables

have a much lower communality than all the other elements (Table 1). The low

communality of chromium and molybdenum indicates that the proﬁles of these two

elements are dissimilar to proﬁles of other elements in the lake sediments, and

therefore were likely inﬂuenced by different sources and processes.

35

2) Gull Lake
The results of the factor analysis for Gull Lake portray a more complicated

grouping of elements compared to Gratiot Lake. Three factors have eigenvalues
greater than 1 and explain more than 5% of the variance (Table 3). Titanium,
vanadium, chromium, copper, arsenic, molybdenum, cadmium, lead, aluminum, zinc,
potassium, and uranium all load positively on the ﬁrst factor and all increase in
concentration in the 1900s (Figure 2) for Gull Lake. This group includes elements that
are typically classiﬁed as terrestrial (e.g., Al, Ti) (Kemp and Thomas, 1976; Heyvaert
et al., 2000), as well as elements that are typically anthropogenically inﬂuenced (e.g.
Pb, Zn) (Sweet et al., 1998; Callender and Rice, 2000). Mg, Ca, Fe and Ba all load
negatively on the ﬁrst factor, but do not clearly represent any one source/process
group. Only As and Ba are not well associated with Factor 1.

Unlike Gratiot Lake where an orthogonal rotation of the factors resulted in a
larger number of high loadings and a clearer interpretation of the element groupings,
rotation of either two or three factors for Gull Lake did not improve the interpretation.
Compared to the unrotated solution, the rotated factors were shifted toward the Ba and
As variables, when viewed in variable space, and the loadings for most of the other
variables decreased in magnitude. An oblique rotation would likely improve the
interpretation, however oblique methods require an a priori subjective decision on the

size of the angle between factors making them more difﬁcult to use.

36

Table 3. Factor analysis results for Gull Lake. “Big” indicates the eigenvalue for the

factor, “%” is the percent variance explained by each factor, and “T°/o” is the total

percentage variance explained by the solution. Bold indicates high loading (>0.8), and
italics moderate loading (0.6-0.79).

 

 

 

 

 

 

 

 

2 factors,
3 factors, no rotation 3 factors, Varimax Varimax
Factor Factor Factor Factor Factor Factor Factor Factor
1 2 3 1 2 3 1 2

Mg -0. 781 0.195 0.252 -0.323 -0. 680 -0.380 -0.419 -0. 720
Al 0.900 0.247 0.348 0.890 0.182 0.409 0.920 0.298
K 0.916 0.200 0.258 0.830 0.263 0.433 0.891 0.363
Ca -0.931 0.227 0.057 -0.563 -0. 707 -0.322 -0.568 -0.778
Ti 0.930 0.174 0.300 0.867 0.271 0.399 0.891 0.380
V 0.970 0.162 0.169 0.797 0.357 0.483 0.889 0.452
Cr 0.961 0.102 0.106 0.739 0.420 0.472 0.833 0.505
Mn -0.804 0.197 -0.438 -0.841 -0.412 —0.009 -0. 605 -0.509
Fe -0. 732 0.523 -0.416 -0. 740 -0.616 0.236 -0.390 -0.642
Cu 0.943 -0.245 -0.158 0.496 0. 769 0.371 0.545 0.825
Zn 0.979 0.142 -0.135 0.580 0.500 0.641 0.803 0.565
As 0.640 0.585 -0.438 0.180 0.115 0.948 0.653 0.181
Mo 0.816 -0.1 13 -0.410 0.241 0.705 0.540 0.470 0.702
Cd 0.898 0.366 0.036 0.674 0.224 0.661 0.917 0.304
Ba -0.647 0. 673 0.274 -0.163 0.959 -0.001 -0.014 -1.000
Pb 0.942 0.036 -0.202 0.494 0.584 0.587 0.695 0.637
U 0.825 -0.043 -0.326 0.316 0.626 0.545 0.534 0.642
Eig 12.75 1.60 1.38 6.59 5.05 4.09 8.13 6.06
% 75% 9% 8% 39% 30% 24% 48% 36%
T% 92% 93% 84%

 

37

Date

2000‘ ‘ ‘ 7 T * 1
1800 . i . .
1600‘ 4 4 1

Mg Al 1 x L Ti v

1400.1__ . _ “—4 ____1 F44 -___
6000 9000 4000100 6002E5 41350 800 15

2000 T ‘ _‘ ” T‘
J l 4 4
.1 i a
Mn Fe Cu Z ASi

n
1400 f—W . . l——’+ ~—~———~1 . ~
0 65 12000 400000 250 2000 35

 

 

 

 

 

 

 

 

 

 

 

 

1800

 

1600

 

 

 

 

 

 

 

 

 

 

      

 

 

 

 

 

 

 

2000 . ~— 4
1800
1600 I
M P
1400 LLB. H 1 1-1.139 __.P. - U
0 4 0 1120 180 0 175 0 2.5
Concentration (mg/kg)

Figure 2. Sediment chemical concentration (mg/kg) proﬁles of Gull Lake.

The impact of the time period encapsulated by the sediment core on the ability

of factor analysis to distinguish between important sources and processes in the post-
settlement period is also evident hour the results for Gull Lake. Whereas the sediment
age for Gratiot Lake extended back only approximately 200 years, the sediments for
the Gull Lake core were deposited over approximately 500 years. Consequently, there
were considerably more data points for the pre-settlement period compared to the
post-settlement period for Gull Lake. As a result, the pre-settlement period had a

greater inﬂuence on the factor analysis results, making it difﬁcult to distinguish

38

between different processes in the post-settlement period. To illustrate, the
standardized proﬁles of an element that is most oﬁen linked to terrestrial inputs to a

lake (Al) and an element that has dominant anthropogenic sources (Pb) were

 

 

   
 

 

 

compared (Figure 3).
2000 ‘* * ‘ '
Although these two
I950 1
elements loaded highly on
1900
the same factor, there are 0
‘e'u’ 1850 -
clear differences between a
1800 j
the two elements in the I
1750 -+— Al
post-settlement period, with l _._ Pb
. . 1700 -—~— .—-- --‘-— —-—--~--~ . -«
“me““at‘ons 0f 31mm“ -200 -100 0.00 1.00 2.00 3.00
increasing and peaking Standardized concentration
Figure 3. Standardized concentrations of
earlier than lead. The aluminum and lead in sediments of Gull Lake.

increases in aluminum concentration correspond with increases in the overall
sedimentation rate, likely due to high erosion rates related to the extensive clear
cutting in the area. Lead, on the other hand, is more related to atmospheric deposition
from sources such as coal burning and leaded gasoline (Eisenreich et al., 1986; Graney
et al., 1995; Yohn et al., 2002a). This comparison indicates that while factor analysis
can successfully group elements with similar overall proﬁles, identifying relatively
minor differences, especially in the post-settlement period when there are fewer data

points, is difﬁcult and may result in important source/process groups to be missed.

39

3) Paw Paw Lake

The unrotated solution for Paw Paw Lake had four factors with eigenvalues
greater than one, all of which explain more than 5% of the variance (Table 4). The
unrotated and rotated solutions (using either four or three retained factors) have many
moderate loadings (0.6-0.8), and several elements that only load weekly on factors.
The larger number of weak loadings compared to the solutions for Gratiot and Gull
Lakes reﬂect the lower communalities for several of the elements (Mg, Ti, As, and Ba)
at Paw Paw Lake (Table 1). This suggests that element proﬁles in Paw Paw Lake are
less similar to each other compared to the other lakes (Figure 4). Additionally, the
factors for Paw Paw Lake are difﬁcult to interpret in terms of the a priori deﬁned
source/process groups. When four factors are rotated, the ﬁrst factor appears to be a
terrestrial factor, but contains only two elements (V -Al). The second factor includes
chromium, lead and zinc, typically anthropogenic elements, as well as calcium and
strontium, which are generally inﬂuenced by calcium carbonate deposition. Factor 4
may represent diagenetic elements (Mo-U-As), but Factor 3 is difﬁcult to interpret

(Cd-Fe, and negatively loaded on Cu).

40

Table 4. Factor analysis results for Paw Paw Lake. “Big” indicates the eigenvalue for
the factor, % is the percent variance explained by each factor, and “T%” is the total
percentage variance explained by the solution. Bold indicates high loading (>0.8), and

italics moderate loading (0.6-0.79).

 

 

3 factors, Varimax
4 factors, no rotation 4 factors, Varimax rotation rotation
Factor Factor Factor Factor Factor Factor Factor Factor Factor Factor Factor
1 2 3 4 1 2 3 4 1 2 3

 

Mg 0.490 0.160 0.459 -0.228
Al -0.445 0.700 0.415 -0.014
Ca 0.863 0.164 -0.313 0.136
T1 -0.009 -0.002 0.582 -0.188
V -0.402 0.516 0.725 -0.217
Cr 0.509 0.703 0.226 -0.095
Mn -0.455 -0.151 0.068 0.029
Fe -0.685 0.374 0.083 0.539
Cu 0.693 -0.200 0.573 -0.060
Zn 0.353 0.800 0.063 -0.032
As -0.022 -0.100 0.390 0.494
Sr 0.903 0.145 -0.260 0.058
Mo 0.396 -0.638 0.376 0.385
Cd -0.058 0.760 -0.367 0.440
Ba -0.686 0.249 -0.081 0.066
Pb 0.544 0.679 —0.198 0.027

0.326 0.383 -0.502 0.155
0.798 0.154 0.444 —0.060
-0.495 0.749 -0.261 0.120
0.503 —0.085 -0.297 0.163
0.993 0.003 0.115 0.000
0.355 0.822 -0.109 0.017
0.127 -0.424 0.196 -0.022
0.316 -0.171 0.858 0.205
0.131 0.259 -0. 734 0.479
0.309 0.808 0.119 -0.082
0.121 -0.099 0.113 0.608
-0.453 0.752 -0.354 0.102
-0.250 -0.208 -0.443 0.743
-0.034 0.578 0.757 -0.040
0.267 -0.274 0.591 -0.221
-0.029 0.883 0.070 -0.100
0.017 0.192 -0.147 0.840

-0.057 0.385 0.577
0.889 0.307 0.030
-0.683 0.640 0.023
0.275 -0.036 0.526
0.895 0.161 0.374
0.110 0.858 0.248
0.278 -0.387 -0.088
0.658 -0.086 -0.291
-0.347 0.200 0.864
0.198 0.861 0.013
0.112 -0.108 0.312
-0.698 0.647 0.094
-0.444 -0.301 0.590
0.254 0.571 -0.507
0.591 -0.180 -0.413
-0.126 0.883 -0.121
-0.157 0.118 0.532

 

0 0.425 -0.111 0.489 0.577
Eig 4.81 3.70 2.52 1.39
% 28% 22% 15% 8%
T% 73%

2.94 4.19 3.24 2.05
17% 25% 19% 12%
73%

3.97 4.03 2.83
23% 25% 17%
65%

 

41

2000
1 980
1960
1940
l 920

4 1 1 -1
‘1 1 4
1 -<

V Cr

6E3 1E4 1E4 2. 5E4 2E4 9E4 10 60 20 40 10 50

2000 - 4 4 ~
1980 4 4 4 4
1960 4 4 4 4 i
1940 4 Mn 4 4 As 4

Fe Cu Zn Sr
1920 _. —~ #4 - 1

0 4E3 2E4 4E4 20 70 100 30015 30 30 90

 

 

 

 

 

 

 

 

 

 

 

 

 

Date

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

2000‘ 1
1980‘ i
1960‘
19404 4 4 4
Cd Ba Pb U
1920 T
l 6 0.5 1.3 200 500 0 200 0.8 1.5
Concentration (mg/kg)

Figure 4. Sediment chemical concentration (mg/kg) proﬁles of Paw Paw Lake.

4) Elk Lake

Three factors with eigenvalues greater than one and explaining greater than 5%
variance were found for Elk Lake (Table 5). The unrotated solution shows a high
degree of simple structure with all elements having high or moderate loadings on the
ﬁrst two factors. Rotating two factors changes the interpretation only slightly.
Interpretation of both the unrotated and rotated two-factor solutions indicates that the
factor analysis is separating elements that have high concentrations at the surface (Mn,
Fe, As, Mo, and Ba) from those that do not (Figure 5). When three factors are rotated,

the elements that had loaded highly of the ﬁrst factor of the two-rotation solution are

42

now distributed among two factors (Factor 1 and Factor 3), and the elements with high
surface concentrations remain grouped together on Factor 2. Even though the two
factor solution has a high degree of simple structure, the three factor solution is more
interpretable in terms of the a priori deﬁned physical processes and appears to be the
more appropriate solution. Factor 1 can be interpreted as representing the terrestrial
elements (Ti-V-Al-Mg-K-U), whereas the elements that load positively on Factor 3
(Cu-Cd-Pb-Zn) represent anthropogenic elements. Ca loads negatively on Factor 3
and represents the carbonate group. The elements associated with Factor 2 (As-Mo-
Mn-Ba-F e) may be associated with diagenesis.

Table 5. Factor analysis results for Elk Lake. “Big” indicates the eigenvalue for the
factor, “%” is the percent variance explained by each factor, and “T%” is the total

percentage variance explained by the solution. Bold indicates high loading (>0.8), and
italics moderate loadingAQ.6-0.79).

 

 

 

 

 

 

 

 

All data Outliers removed
2 factors, 3 factors, Varimax
3 factors, no rotation 3 factors, Varimax Varimax rotation

Factor Factor Factor Factor Factor Factor Factor Factor Factor Factor Factor

1 2 3 1 2 3 1 2 1 2 3
Mg 0.922 0.178 0.326 0.914 0.117 0.371 0.931 0.064 0.938 0.314 -0.023
At 0.949 0.186 0.256 0.891 0.105 0.443 0.966 0.062 0.930 0.358 0.076
K 0.914 0.156 0.325 0.902 0.135 0.366 0.918 0.083 0.943 0.271 0.071
Ca -0.881 -0.020 0.226 -0.485 -0.173 -0.750 -0.857 -0.203 -0.567 -0.692 -0.145
Ti 0.966 0.096 0.028 0.730 0.162 0.620 0.966 0.152 0.797 0.546 0.086
V 0.874 0.234 0.398 0.943 0.063 0.289 0.893 0.003 0.953 0.247 -0.072
Cr 0.677 0.211-0.172 0.431 -0.053 0.586 0.710 -0.032 0.488 0.646 -0.133
Mn 0.159 «0.891 0.024 —0.133 0.894 0.036 -0.075 0.911 -0.009 -0.257 -0.836
Fe 0.745 -0.491 0.261 0.551 0.700 0.266 0.596 0.653 0.975 0.136 -0.165
Cu 0.928 -0.128 -0.257 0.454 0.321 0.795 0.863 0.355 0.603 0. 757 0.075
Zn 0.865 0.106 -0.464 0.346 0.050 0.924 0.848 0.119 0.444 0.840 0.253
As 0.322 -0.808 -0.126 -0.090 0.835 0.261 0.106 0.863 0.487 0.763 0.066
Mo 0.099 -0.743 0.275 0.029 0.776 -0.184 -0.090 0.725 0.212 -0.410 -0.140
Cd 0.828 0.043 -0.292 0.411 0.127 0.766 0.808 0.168 0.521 0.663 0.352
Ba 0.695 -0.693 0.090 0.347 0.854 0.350 0.498 0.850 0.894 0.283 0.168
Pb 0.767 0.035 —0.595 0.171 0.073 0.954 0.725 0.157 0.270 0.906 0.193
U 0.652 0.529 0.289 0.803 -0.293 0.242 0.761 0335 0.840 0.246 0.200
Eig 10.00 3.23 1.49 5.91 3.64 5.18 9.46 3.64 8.38 5.07 1.11
% 59% 19% 9% 35% 21% 30% 56% 21% 49% 30% 7%

T% 87% 86% 77% 86%

 

43

  

 

 

 

 

 

 

 

 

 

 

 

2000 ~ —4 , —~ -

1800

1600 1 1 1 .

1400 . .

1200 ME. 1 911 awa—Kj 19”-- ___T.L l__.Yl
553 1.254 153 8B3 0 1500 355 555 20 804 16

     
 

 

 

 

Cu Zn

#_4.__

 

 

 

 

 

nﬂ_ ﬂ _' ﬂ,

--—4 4 4
30 0 3000 5E3 2E4 2 10 0 90 0

T

 

   

4m

      

 

 

 

 

 

 

 

Mo 1
0.4 0 0.8 50 100 0 70 0.2 0.6
Concentration (mg/kg)

Figure 5. Sediment chemical concentration (mg/kg) proﬁles of Elk Lake.

As shown earlier, the outlier analysis for Elk Lake identiﬁed 2-3 points near
the surface as outliers for several variables (As, Mo, Mn, Ba, Fe). Factor analysis on
the Elk Lake data with inﬂuential outlying data points removed resulted in a
considerably different interpretation. When three factors are rotated, Factor 2 no
longer simply represents those elements with high surface concentrations. Rather it
can now be interpreted as the anthropogenic factor, and includes the same elements as

were previously associated with Factor 3, as well as arsenic. Barium and iron now

load highly on the terrestrial factor (Factor 1), and Factor 3 contains only manganese.
Molybdenum does not load highly or moderately on any factor.

The choice between the solution from the factor analysis with all data points
and that with outlier points removed is not clear, and analysis both with and without
outliers may be necessary. Visual inspection of Figure 5 indicates that the overall
proﬁles of arsenic, molybdenum, manganese, barium and iron are dissimilar.
However, because all of these elements had high concentrations at the surface, they
were grouped together when all data points were included in the factor analysis.
While this may be appropriate because all of these elements have likely been
inﬂuenced by diagenetic processes, additional insight on the behavior of the elements

was gained by also analyzing the data with those inﬂuential points removed.

Factor Analysis of Ranked Observations

Factor analysis was also used on a Spearman’s correlation coefﬁcient matrix.
Spearman’s correlation coefﬁcient is similar to Pearson’s product moment correlation
coefﬁcient, except that it is performed on ranked data. Using ranked data reduces the
inﬂuence of outliers and may be more appropriate for sediment chronologies. When
data are ranked, the lowest concentration data point for each element is assigned the
number 1, the second lowest number 2, and so forth. If two data points have the same
concentration of an element, both data points are assigned the average of the two ranks
(e.g., if the 5th and 6‘” lowest concentrations have the same value, both will be
assigned the value of 5.5) (SPSS 2000).

For all lakes, using a Spearman’s correlation matrix (i.e., ranked data) for

factor analysis did not result in clearer element groupings compared to using the

45

standard Pearson’s product moment correlation matrix (i.e., “raw” data). Overall, the
results for the factor analysis using the ranked data were similar to those when applied
to the “raw” data, except that the loadings were generally lower. Ranked data reduces
the inﬂuence of large changes in concentration, and also increases in the importance of
small changes in concentration. For Elk and Gull lakes in particular, many elements

have multiple samples with similar concentrations (e.g., all lead concentrations older

 

than 1800 have similar

2000 ___ “—— - __ ~ " r 5
concentrations), and minor

1900
changes in these 1800
concentrations become 1700 i

8

magniﬁed when using a 1600

1500
ranked data (Figure 6).

1400 :
Because of this, the 1300 + Unmodiﬁed

. + Ranked
terrestrial and 1200 _ _. _. 1 . 4-4 .5 w. *T -___._4
anthropogenic elements 0.00 20.00 40.00 60.00 80.00
. . Concentration (mg/kg) or rank

were not differentrated Figure 6. Unmodiﬁed concentrations (mg/kg) and
using ranked data in Elk ranked values of lead 1n sediments of Elk Lake.

Lake, but were differentiated when the original observations were used to calculate the
correlation matrix. Therefore, with sediment chronology data and the types of
questions asked here, it may be more appropriate to apply factor analysis to parametric
rather than nonparametric correlation matrices. In other situations or questions, this

may not be true (e.g., Wayland et al., 2003). When using parametric correlation

46

coefﬁcients, it is essential to identify possible outliers using boxplot analysis and to

perform statistical analyses with and without these points.

Summary

Overall, factor analysis is effective for differentiating groups of elements with
broadly similar proﬁles, but not for identifying small differences among proﬁles,
especially if the differences are conﬁned to a time period (e.g., post-settlement) for
which there are few data points. Factor analysis was particularly successful for
Gratiot Lake, where there were a small number of groups with very distinct proﬁles.
Factor analysis was less successful in categorizing elements when correlation was
small from a number of the element proﬁles (e.g., Paw Paw Lake), when the
differences between proﬁles is conﬁned to a short segment of the proﬁle (e. g., Gull
Lake), or when outlier/inﬂuential points are present (e. g., Elk Lake).

One difﬁculty in using factor analysis for sediment chronologies is that the
processes inﬂuencing metal inputs to lakes are temporally related to each other. For
example, the majority of human disturbance in North America has occurred since the
1850s. This disturbance can inﬂuence both terrestrial elements (e.g., through land use
change, such as clear cutting) (Davis, 1976) and direct human inputs (e. g.,
atmospheric deposition of lead from leaded gasoline, cadmium from smelting)
(Iskander and Keeney, 1974; Johnson and Nicholls, 1988; Kada and Heit, 1992;
Callender and Rice, 2000; Heyvaert et al., 2000). While these are clearly diﬂ'erent
sources and processes inﬂuencing the two groups, the processes are temporally
related, and therefore are not always well differentiated by factor analysis (e.g., Gull

Lake). While it is possible to use a non-orthogonal rotation technique that allows

47

factors to correlate (e.g., Oblimin), it is then necessary to determine what extent off
correlation to allow among the factors. As an alternative to factor analysis, the use of

cluster analysis to group elements was explored on the same dataset.

Cluster analysis

Results ﬁom the three clustering algorithms were similar to each other for all
of the lakes, therefore, only results using the most common method, Ward’s, are
shown, with any differences in element clusters among the methods discussed.

The cluster tree for Gratiot Lake is shown in Figure 7. The cluster tree is read
left to right, where the horizontal lines represent similarity (reported as distance) and
the vertical lines show when two elements or clusters have been joined. A long
horizontal line before two clusters are joined indicates that there is a long distance
between the two clusters, or that they are relatively dissimilar. Hierarchical cluster
analysis does not create deﬁnite groups; instead, the number of groups is determined
by the user. A common approach is to consider where there is a large increase in
distance before two groups are combined (a long horizontal line to the leﬁ of the
vertical line joining the clusters). This indicates that the groups are relatively
dissimilar, and perhaps should be considered separate. For example, in Gratiot Lake,
three clusters might be chosen and interpreted as, K-Mn-Ba (diagenetic group), Fe-Cu-
Pb-Zn-As-Cd-Mo (anthropogenic group), and the last including the remainder of the
elements (terrestrial group). However, molybdenum and chromium are still fairly
dissimilar to their groups (anthropogenic and terrestrial). To make the ﬁnal decision

of which elements belong in which groups, standardized proﬁles (z-scores) for each

48

 

 

 

 

 

 

 

 

 

 

 

V
Mr? TI
V .—
U AS
MO —' ZN
CD PB
AS CU
ZN MO
PB U .—
CU MN
BF}? BF}? 4
M12] h——— Gratiot [$18 Gull
i r f 1 r r r r I 1 r i‘ r I
0 l 2 3 4 5 6 7 8 9 0 2 1 0 l 2
D lsmnces Distances
8115b 9
MG
ZN A115
8711—— n

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

__1 Paw Paw C A_______. Elk

I I I ﬂ I I I r I ﬂ

0 l 2 3 4 0 l 2 3 4 5
Distances Distances

Figure 7. Cluster trees for sediment chronologies for four Michigan lakes. Ward’s
clustering algorithm and Pearson’s correlation were used.

element in a group are plotted together, and examined for similarity. The investigator
then must judge if the proﬁles are similar enough for those elements to be included in
a group. This decision should be based on how similar the elements need to be based
on the purpose of the study. In the case of Gratiot Lake, molybdenum is dissimilar to
the rest of the anthropogenic group, while chromium is similar to the terrestrial group,
with the exception of one unusual point. Therefore, chromium was grouped as a

terrestrial element, while molybdenum was left unclassiﬁed (Table 6). This is slightly

49

different that the results of factor analysis, where neither chromium nor molybdenum
was placed in a group.

Table 6. Final groupings for each element in the
four study lakes, and for Elk Lake with outliers
removed (Elk OR). T = terrestrial, A =
anthropogenic, C = carbonate, D = diagenetic. D1
and D2 indicate separate diagenetic groups. A —
indicates that data were not collected for that
element.

 

Paw
Gratiot Elk Elk OR Gull Paw

Mg T T T C

Al T T T T T
K D T T T —
Ca T C C C C
T1 T T T T T
V T T T T T
Cr T T A

Mn D D D1

Fe A D T D1

Cu A A A A A
Zn A A A A A
As A D A 02
Sr T -— - - C

Mo D D2 D2
Cd A A . A A
Ba D D T C

Pb A A A A A
U T T T D2

 

The ﬁnal groupings of the elements based on cluster analysis and the
examination of standardized proﬁles for Gratiot Lake are very similar to the groups
determined by factor analysis, therefore either approach provides the same
information. However, the same is not true for Gull Lake (Figure 7). Cluster analysis
can diﬂ‘erentiate smaller differences in proﬁles, and therefore distinguish the
source/process groups in Gull Lake.

The ﬁrst group of elements in Gull Lake is the terrestrial elements, including

vanadium, titanium, aluminum, potassium, chromium and cadmium (Figure 7).

50

Cadmium and chromium are the last two elements to be included in this cluster, and
have been found to have anthropogenic inputs in other lakes (Kemp and Thomas,
1976; Johnson and Nicholls, 1988; Kada and Heit, 1992). Since part of the purpose of
this type of study is to identify elements that have been inﬂuenced by human inputs,
further investigation into the sources of cadmium and chromium may be warranted,
such as the examination of cadmium to aluminum ratios (Bruland et al., 1974;
Koelmans, 1998; Tuncer et al., 2001). However, the similarity of the trends of
cadmium and chromium to the terrestrial elements suggests that these elements are
more inﬂuenced by terrestrial inputs than direct anthropogenic inputs.

The anthropogenic elements are lead, copper and zinc (Figure 7), and a
diagenetic group includes molybdenum and uranium. Manganese and iron form a
second diagenetic group, while barium, calcium and magnesium form the carbonate
group. Although Mn-F e and Ba-Ca-Mg could be considered one cluster (Figure 7),
and have proﬁles that are similar overall (e.g., lower concentrations in the 1900s)
(Figure 2), the two groups are distinctly different when plotted together, and probably
represent different sources and/or process groups of diagenetic and carbonate.

The decision making processes used above was utilized for Paw Paw and Elk
lakes, and the resultant groups are shown in Table 6. Paw Paw Lake had six different
groups, including terrestrial (Al-Ti-V), anthropogenic (Cd-Pb-Cr-Zn), carbonate (Ca-
Sr), and two diagenetic groups (Mo-U-As and Fe-Ba-Mn) (Figure 7). The last
grouping included only copper and magnesium. The reason for the similar proﬁles
between these two elements is unclear. In Paw Paw Lake, factor analysis resulted in

moderate loadings and poor delineation of groups of elements. Cluster analysis, in

51

this case, provides more information about the similarities among element proﬁles
than factor analysis.

Elk Lake has only four clear groupings: terrestrial (U-V-Ti-Mg-Al-K),
anthropogenic (Cd-Zn-Pb-Cu), diagenetic (Mn-Mo-Fe-As-Ba) and carbonate (Ca)
(Figure 7). These groupings are similar to those found with factor analysis. Removal
of the inﬂuential outliers from the dataset results in regroupings of elements that are
also similar to factor analysis. Iron and barium group with the terrestrial elements,
and arsenic with the anthropogenic elements. Although molybdenum, manganese and
calcium cluster together, there is a relatively large distance before they are clustered,
indicating that the proﬁles are not very similar.

For all four lakes results ﬁ'om the three clustering algorithms were generally
very similar, which suggests that the groups were distinct ﬁom each other, and not an
artifact of the technique used. The use of more than one algorithm was useful to test if
the solution is robust. Elements that change groups with different algorithms
typically do not ﬁt well into any group, and should be considered individually, rather
than placed in a group.

Overall, cluster analysis was successful in grouping elements with similar
proﬁles. The use of hierarchical cluster analysis shows which elements are the most
similar within groups, and which elements ﬁt the group more poorly and perhaps
should be considered individually. For exploratory data analysis, this information is
particularly useful for interpreting possible sources and/or processes that inﬂuence

different elements.

52

Discussion

Factor analysis identiﬁed elements with broadly similar proﬁles, and the
resultant groups were consistent with visual examination in two of the four lakes. In
the remaining two lakes, subtle differences in the depth proﬁles were important for
differentiating source/process groups (e.g., terrestrial and anthropogenic in Gull Lake),
but were not identiﬁed by factor analysis. In the last 200 years, humans have
inﬂuenced the terrestrial, anthropogenic, and possibly carbonate groups. As a result,
the broad depth patterns in the sediment of these groups often correlate with each
other, and may be difﬁcult to differentiate with orthogonal factor analysis. Accepting
results from factor analysis without examination of the proﬁles may lead to missing
subtle but important differences among proﬁles, and therefore lead to incorrect
interpretations. Oblique rotation, which allows factors to be correlated to each other,
should be considered if factor analysis is used for grouping sediment proﬁles. It
appears to make little difference if principal component analysis or factor analysis is
used when analyzing sediment chronologies, as long as there is a high degree of
communality.

Cluster analysis, though typically used for grouping samples, was useful for
determining element groupings. One possible advantage of cluster analysis for the
grouping of variables is that, unlike factor analysis, this technique does not attempt to
create factors that are uncorrelated. Additionally, hierarchical cluster analysis not only
creates groups, but also shows which elements are the most similar, and may allow for

additional interpretation of data. Cluster analysis objectively determines the similarity

53

of element proﬁles, and provides the framework for determining which elements
belong in each source/process group.

Outliers generally appeared to be “true” data, and needed to be considered in
the interpretation of sediment chronologies. Often these outliers represented the
modiﬁcation of historical deposition records due to diagenetic processes, and it was
helpful to interpret these proﬁles both with and without the outliers present.

Regardless of the statistical approach used, a difﬁculty in differentiating
source/process groups is that humans have signiﬁcantly inﬂuenced sediment
chronologies in North America for only 150—200 years, but sediment cores may
represent a much longer time period (e.g., Gull and Elk lakes). It may be difﬁcult to
separate out both natural processes and human processes in sediment cores that
encompass hundreds of years, and have relatively few samples during the last 150
years. For example, in Gull Lake it was difﬁcult to differentiate terrestrial and
anthropogenic inputs. Possible approaches to avoid this issue include: I) truncating
the core, 2) “edit” the core by deleting some of the pre-settlement data points to make
the number of pre and post settlement points more equal in number, 3) run statistical
analyses on the entire core and then separately on the post settlement portion.
Regardless of the approach used, it is important to be careful when comparing groups
ﬁ'om lakes with large differences in the time periods captured by the cores.

While it would be ideal to have a statistical technique to objectively group
elements that were controlled by the same sources and/or processes, with no additional
judgment required to deﬁne these groups, it is unlikely that this will occur in a system

as complex as lake sediments. A close examination of each element will always be

54

necessary for the correct interpretation of sediment chronology data. However,
multivariate statistical analysis provides an important framework for data analysis, and
helps facilitate the task of interpreting proﬁles of a large number of elements and

greatly reduces the possibility of important relationships not being identiﬁed.

Comments and recommendations

We suggest the following approach to the analysis of multivariate data. First,
potentially inﬂuential points need to be identiﬁed using boxplot or some other
analysis. If these points can be attributed to analytical error, such as an unusual point
in only one element, then they should be removed ﬁom the data set. If inﬂuential
points represent the data, further analyses and interpretations should be done with and
without these points to determine their inﬂuence. Next, perform cluster analysis on
the dataset using correlation as the distance measure. The standardized proﬁles for
elements in each possible group should be plotted together to visually check their
similarity. The conﬁrmation of the groupings by actual examination of the proﬁles is
essential; without this step the potential for incorrect groupings is high, regardless of
the statistical technique used. Finally, if the core encompasses a very long span of
time, it may be truncated around 1800 (for cores collected in North America), and the
recent sediments evaluated again to identify subtle differences in cores during the time
period of human inﬂuence. The resultant element groupings from these steps can than
be interpreted to provide an understanding of the geochemistry of the sediments and

history of inputs to the lake.

55

References

Auer, M.T., Johnson, N.A., Penn, MR. and Bfﬂer, S.W., 1996. Pollutant sources,
depositional environment, and the surﬁcial sediments of Onondaga Lake, New
York J. Environ. Qual. 25, 46-55.

Boyle, J.F., Rose, N.L., Bennion, H., Yang, H. and Appleby, P.G., 1999.
Environmental impacts in the J ianghan Plain: evidence from lake sediments.
Water, Air, Soil Pollut. 112, 2140.

Brown, E.T., Callonnec, LL. and German, CR, 2000. Geochemical cycling of redox-
sensitive metals in sediments from Lake Malawi: A diagnostic paleotracer for
episodic changes in mixing depth. Geochim. Cosmochim. Acta 64, 3515-3523.

Bruland, K.W., Bertine, K., Koide, M. and Goldberg, ED, 1974. History of metal
pollution in Southern California costal zone. Environ. Sci. Technol. 8, 425-
432.

Callender, B. and Rice, K.C., 2000. The urban environmental gradient: Anthropogenic
inﬂuences on the spatial and temporal distributions of lead and zinc in the
environment. Environ. Sci. Technol. 24, 232-238.

Cooper, DC. and Morse, J.W., 1998. Biogeochemical controls on trace metal cycling
in anoxic marine sediments. Environ. Sci. Technol. 32, 327-330.

Davis, B.M., 1976. Erosional rates and land use history in southern Michigan.
Environmental Conservation 3, 139-148.

Davis, J .C., 1986. Statistics and Data Analysis in Geology. John Wiley & Sons, New
York.

Dillion, PJ. and Evans, RD, 1982. Whole-lake lead btn'dens in sediments of southern
Ontario, Canada. Hydrobiologia 91, 121-130.

Door, J.A. and Eschman, DR, 1970. Geology of Michigan. University of Michigan
Press, Ann Arbor, 476 pp.

Edgington, D.N. and Robbins, J .A., 1976. Records of lead deposition in Lake
Michigan sediments since 1800. Environ. Sci. Technol. 10, 266-274.

Eisenreich, S.J., Metzer, NA. and Urban, N.A., 1986. Response of atmospheric lead
to decreased use of lead in gasoline. Environ. Sci. Technol. 20, 171-174.

56

Evans, RD. and Dillon, P.J., 1982. Historical changes in anthropogenic lead fallout in
southern Ontario, Canada. Hydrobiologia 91, 131-137.

Golden, K.A., Wong, C.S., Jeremiason, J .D., Eisenreich, S.J., Sanders, M.G.,
Hallgren, J., Swackhammer, D.L., Engstrom, DR. and Long, D.T., 1993.
Accumulation and preliminary inventory of organochlorines in Great Lakes
sediments. Wat. Sci. Technol. 28, 19-31.

Gotz, R. and Lauer, R., 2003. Analysis of sources of dioxin contamination in
sediments and soils using multivariate statistical methods and neural networks.
Environmental Science & Technology 37, 5559-5565.

Graney, J .R., Halliday, A.N., Keeler, G.J., Nriagu, J .O., Robbins, J .A. and Norton,
S.A., 1995. Isotopic record of lead pollution in lake sediments ﬁom the
northeastern United States. Geochim. Cosmochim. Acta 59, 1715-1728.

Grifﬁth, DA. and Amrhein, CG, 1997. Multivariate statistical analysis for
geographers. Prentice Hall, Upper Saddle River, N.J., xix, 345 pp.

Hair, J .F ., 1998. Multivariate data analysis. Prentice Hall, Upper Saddle River, N.J.,
730, 12 pp.

Hewitt, AD. and Reynolds, C.M., 1990. Dissolution of metals ﬁom soils and
sediments with a microwave-nitric acid digestion technique. Atomic
Spectroscopy 11, 1425-1436.

Heyvaert, A.C., Renter, J .E., Slotton, D.G. and Goldman, GR, 2000.
Paleolimnological reconstruction of historical atmospheric lead and mercury
deposition at Lake Tahoe, California-Nevada. Environ. Sci. Technol. 34, 3588-
3597.

Iskander, I.K. and Keeney, DR, 1974. Concentration of heavy metals in sediment
cores from selected Wisconsin Lakes. Environ. Sci. Technol. 8, 165-170.

Jaquet, J .-M., Davaud, E., Rapin, F. and Vernet, J.-P., 1982. Basic concepts and
associated statistical methodology in the geochemical study of lake sediments.
Hydrobiologia 91.

Johnson, MG. and Nicholls, K.H., 1988. Temporal and spatial trends in metal loads to
sediments of Lake Simcoe, Ontario. Water, Air and Soil Pollution 39, 337-354.

Joreskog, K.G., Klovan, J .B. and Reyment, R.A., 1976. Geological factor analysis.

Methods in Geomathmatics. Elsevier Scientiﬁc Publishing Company, New
York, 178 pp.

57

Kada, J. and Heit, M., 1992. The inventories of anthropogenic Pb, Zn, As, Cd and the
radionuclides mCs and excess 2'on in lake sediments of the Adirondack
region, USA. Hydrobiologia 246, 231-241.

Kemp, A.L.W. and Thomas, R.L., 1976. Cultural impacts on the geochemistry of the
sediments of Lakes Ontario, Erie and Huron. Water, Air and Soil Pollution 5,
469-490.

Kerfoot, W.C., Lauster, G. and Robbins, J .A., 1994. Paleolimnological study of
copper mining around Lake Superior: Artiﬁcial varves from Protage Lake
provide a high resolution record. Limnology and Oceanography 39, 649-669.

Kerfoot, W.C. and Robbins, J .A., 1999. Nearshore regions of Lake Superior: Multi-
element signatures of mining discharges and a test of Pb-210 deposition under
conditions of variable sediment mass ﬂux. J. Gt. Lakes Res. 25, 697-720.

Khim, J., Villeneuve, D., Kannan, K., Lee, K., Snyder, S., Koh, C. and Giesy, J ., 1999.
Alkylphenols, polycyclic aromatic hydrocarbons (PAHs), and organochlorines
in sediment ﬁom Lake Shihwa, Korea: Instrumental and bioanalytical
characterization. Environ Toxicol Chem 8, 2424-2432.

Klovan, J .E., 1975. R- and Q-Mode factor analysis. In: R.B. McCammon (Editor),
Concepts in geostatistics. Springer-Verlag, New York.

Kneebone, RE. and Hering, J .G., 2000. Behavior of arsenic and other redox sensitive
elements in Crowley Lake, CA: A reservoir in the Los Angeles Aqueduct
system. Environmental Science & Technology 34, 4307-4312.

Koelmans, A.A., 1998. Geochemistry of suspended and settling solids in two
freshwater lakes. Hydrobiologia 364, 15-29.

Kolak, J.J., Long, D.T., Beals, T.M., Eisenreich, SJ. and Swackhamer, D.L., 1998.
Anthropogenic inventories and historical and present accumulation rates of
copper in Great Lakes sediments. Appl. Geochem. 13, 59-75.

Lamoureux, S., 1999. Spatial and interannual variations in sedimentation patterns
recorded in nonglacial varved sediments from the Canadian High Arctic.
Journal of Paleolimnology 21, 73-84.

Lantzy, R.J. and Mackenzie, F .T., 1979. Atmospheric trace metals, global cycles and
assessment of man's impact. Geochim. Cosmochim. Acta 43, 511-525.

Manta, D.S., Angelone, M., Bellanca, A., Neri, R. and Sprovieri, M., 2002. Heavy

metals in urban soils: a case study ﬁom the city of Palermo (Sicily), Italy. Sci.
Total Environ. 300, 229-243.

58

Qu, W.C., Dickman, M. and Wang, S.M., 2001. Multivariate analysis of heavy metal
and nutrient concentrations in sediments of Taihu Lake, China. Hydrobiologia
450, 83-89.

Sanei, H., Goodarzi, F. and Van der Flier-Keller, E., 2001. Historical variation of
elements with respect to different geochemical ﬁ'actions in recent sediments
ﬁom Pigeon Lake, Alberta, Canada? J. Environ. Monit. 3, 27-36.

Spiethoff, HM. and Hemond, HF, 1996. History of toxic metal discharge to surface
waters of the Aberjona Watershed. Environ. Sci. Technol. 30, 121-128.

Stenson, H. and Wilkinson, L., 2000. Factor Analysis. In: SPSS Inc. (Editor),
SYSTAT 10 manual, Statistics 1, Chicago, pp. 327-364.

Sweet, C.W., Weiss, A. and Vermette, S.J., 1998. Atmospheric deposition of trace
metals at three sites near the Great Lakes. Water Air Soil Poll 103, 423-439.

Tuncer, G., Tuncel, G. and Balkas, T.I., 2001. Evolution of metal pollution in the
Golden Horn (Turkey) sediments between 1912 and 1987. Marine Pollution
Bulletin 42, 350-360.

Wayland, K.G., Long, D.T., Hyndman, D.W., Pijanowski, B.C., Woodhams, SM. and
Haack, S.K., 2003. Identifying relationships between baseﬂow geochemistry
and land use with synoptic sampling and R-mode factor analysis. J. Environ.
Qual. 32, 180-190.

Wilkinson, L., Engehnan, L., Corter, J. and Coward, M., 2000. Cluster Analysis,
Systat 10, Statistics I. SPSS Inc., Chicago, pp. 53-85.

Yohn, S.S., Long, D.T., Fett, J .D. and Patino, LC, 2004. Regional versus local
inﬂuences on lead and cadmium loading in the Great Lakes region. Appl.
Geochem. 19, 1157-1175.

Yohn, S.S., Long, D.T., Fett, J.D., Patino, L.C., Giesy, J.P. and Kannan, K., 2002a.
Assessing environmental change through chemical-sediment chronologies
ﬁom inland lakes. Lakes & Reservoirs: Research and Management 7, 217-230.

Yohn, S.S., Long, D.T., Giesy, J .P., Scholle, L.K., Patino, L.C., Fett, J .D. and Kannan,
K., 2002b. Inland lakes sediment trends: sediment analysis results for ﬁve
Michigan Lakes. MI/DEQ/WD-03/052, Department of Environmental Quality,
East Lansing.

59

CHAPTER 3
THE IMPORTANCE OF THE MULTI-ELEMENT APPROACH FOR
INTERPRETING SEDIMENT HISTORIES

Abstract

Sediment chronologies are often used to provide a history of anthropogenic
inputs to lakes and their surrounding watersheds. However, analysis of only target
elements (e. g., lead, arsenic) may lead to incorrect interpretations and quantiﬁcation of
human inputs to lake sediments. This study uses sediment cores collected from eleven
lakes in Michigan, U.S.A., to determine a minimum suite of elements for
environmental monitoring of anthropogenic inputs using sediment chronologies.

Cluster analysis was used to group elements with similar proﬁles, and a
representative element was chosen from each group. Groupings were similar, but not
identical among lakes. A suite of six elements in addition to those elements directly
inﬂuenced by anthropogenic inputs was selected to represent terrestrial inputs
(aluminum and titanium), diagenesis (iron, manganese and molybdenum), and
carbonate deposition (calcitun). The use of these six elements allowed the
differentiation of anthropogenic and diagenetic inﬂuences on arsenic by comparison to
the diagenetic elements. Additionally, results from determination of anthropogenic
inputs using normalization to a terrestrial element were different ﬁom using a constant
background, indicating the importance of assessing terrestrial inputs. This suite of
elements should be considered the minimum analysis needed to provide sufﬁcient

information for an environmental monitoring program using sediment chronologies.

60

Introduction

Lake sediments act as recorders of events that occur both within the watershed
and the region, and can provide a history of events such as forest ﬁres, logging, and
urbanization. This is particularly useful because the water column of most lakes have
not been monitored over long periods of time, making the historical record found in
the sediments an important tool for environmental monitoring.

The application of sediment chronologies to environmental monitoring is most
useful if a quantiﬁable record of human inputs to a lake can be determined. To do
this, it is necessary to distinguish between natural inputs (inputs of metals from
naturally occurring erosion or atmospheric deposition), secondary human inputs
(increased inputs due to erosion from human activities, such as clear cutting), and
direct human inputs (selective enrichment of certain elements due to human activities,
such as lead enrichment due to the use of leaded gasoline). Increasing erosion within
a watershed will cause additional inputs of metals such as lead and cadmium to a lake,
but these metals are not selectively enriched compared to other terrestrial elements
such as aluminum (Heyvaert et al., 2000). Therefore, it is useﬁrl to differentiate direct
human inputs which enrich certain metals (e.g., lead, cadmium), from increased
erosion.

A further complication in the interpretation of sediment chronologies is the
potential mobilization of metals after deposition (early diagenesis), which may
obscure the depositional record (McKee et al., 1989). While some elements that have
signiﬁcant human sources (e.g., lead) are relatively unaffected by diagenesis, others,

such as arsenic, are more likely to be inﬂuenced by diagenetic processes (Brown et al.,

61

2000). It is critical to be able to identify when a sediment proﬁle of an element
represents a history of inputs to the lake, and when it represents post-depositional
processes.

As a result of these complications in the interpretation of sediment
chronologies, it may be necessary to analyze more than only elements with known
human sources to be able to quantify direct human inputs. When analyzing inorganic
contaminants in water, it is well known that there is a minimum suite of parameters
that must be quantiﬁed (e.g., major anions and cations, temperature, pH) to model the
water and understand the behavior of trace elements (Garrels and Thompson, 1962;
Long and Saleem, 1974). It is possible that there is a comparable set of elements to be
measured that are necessary to understand the history of trace element deposition in
the sediment. The following hypothesis will be explored: there is a relatively small set
of elements, in addition to elements of interest (e.g., lead, arsenic), which are
necessary to identify and quantify direct anthropogenic inputs to inland lakes and to
interpret depositional patterns. To explore this hypothesis, elements with similar
proﬁles will ﬁrst be grouped in each lake using cluster analysis to determine if there
are elements that are consistently representative of each group. Secondly, we will
investigate if it is necessary to analyze additional elements to interpret and quantify
anthropogenic inputs. Speciﬁcally, the importance of diagenetic and terrestrial

elements will be investigated.

62

Methods

Sediment cores were collected ﬁom eleven lakes throughout Michigan ﬁom
1999 to 2001: Elk, Gratiot, Higgins, Gull, Crystal M, Littleﬁeld, Cadillac, Crystal B,
Mullett, Paw Paw and
Whitrnore lakes
(Figure 1). Lakes in
the Lower Peninsula of
Michigan are located
on Pleistocene glacial
deposits, which are

generally carbonate

 

rich. Gratiot Lake is

located in the copper p" Paw 6"" . Whitman

 

rich Keweenaw area

(Figure 1) part of the Figure 1. Location of eleven study lakes in Michigan.

Canadian Shield, with sparse glacial deposits in this area (Door and Eschman, 1970).
The lakes vary greatly in surface area and depth and were chosen for sampling based
on depth, location, and accessibility. Lakes reﬂect a gradient of human disturbance,
from areas of high development to relatively remote lakes.

A set of four replicate sediment cores from 40 to 58 cm long were taken from
the deepest portion of each lake using a MC-400 Lake/Shelf Multi-corer (San Diego,
CA) deployed from either the US. Environmental Protection Agency Research Vessel

Mudpuppy or the Michigan Department of Environmental Quality Monitoring Vessel

63

Nibi. A new set of cores was immediately taken if the cores showed any evidence of
disturbance. Cores were examined and described in terms of color, texture, and signs
of zoobenthos. Cores were immediately extruded and sectioned open to the
atmosphere at 0.5 cm intervals for the top 5-8 cm, and at 1 cm intervals for the
remainder of the core.

2|”Pb was measured on one sub-core from each lake at the Freshwater Institute
in Winnipeg, Manitoba, Canada, under the direction of Paul Wilkinson, to determine
sedimentation rates, focusing factors, and sediment ages. Three different models were
used to extract dates from 2|”Pb data; the constant ﬂux, constant sedimentation model
(CF:CS) (Golden et al., 1993), the segmented CFzCS (SCFzCS) (Heyvaert et al.,
2000), and the constant rate of supply model (CRS) (Sanchez-Cabeza et al., 2000).

For all models, sediment deeper than the presence of excess 2|”Pb cannot be
dated. Dates older than this were determined by extrapolation, using the assumption
that sedimentation rates remain constant below this depth. For the CFzCS, and
SCFzCS model, the sedimentation rate in the lower portion of the core was used to
extrapolate dates. For the CRS model, the average sedimentation rate in the last ﬁve
samples with quantiﬁable excess 2|(’Pb was used. The sedimentation rate chosen to
use for extrapolation has a signiﬁcant effect on the resulting dates, and all dates older
than 1850 should be considered estimations.

Sediments for metals analysis were ﬁozen for storage, ﬁeeze-dried and
digested by nitric acid in a CEM-MDS-81D microwave (EPA Method 3051) (Hewitt
and Reynolds, 1990). Standard reference material (NIST SRM 8704 Buffalo River

Sediment) and procedural blanks were processed as well as three replicate digestions

on two samples ﬁom each lake. The concentrated-acid digests were ﬁltered through
an acid-washed, distilled-deionized or E-pure (Barnstead International, Dubuque, IA)
water rinsed 0.40 um polycarbonate ﬁlter (N uclepore). Sediments were analyzed for a
suite of metals and metalloids including Mg, Al, K, Ca, Ti, V, Cr, Mn, Fe, Cu, Zn, As,
Sr, Mo, Cd, Ba, Pb, and U using a Micromass Platform inductively coupled plasma
mass spectrometer with hexapole technology (ICP-MS-HEX), with a concentric
nebulizer for sample introduction.

All statistics were preformed using SYSTAT 10 (SPSS, 2000). Boxplots were
used to identify outliers for each element for each lake, and were considered to be
outliers if they were outside the outer fence (SPSS, 2000; Yohn et al., in preparation).
Elements were grouped by hierarchical cluster analysis on standardized data. Three
clustering algorithms were used, Ward’s, average and single linkage, to test the
robustness of the solution. Clusters were chosen based on the examination of

standardized proﬁles (Yohn et al., in preparation).

Results and discussion

me and dating

21°Pb proﬁles varied greatly among lakes, with some lakes having relatively
log linear decay, others showing abrupt changes in sedimentation rate, and others
appearing to have gradual shifts in sedimentation rate. Because of this, sedimentation
rates and ages in each lake were determined using the constant ﬂux, constant
sedimentation model (CF :CS) (Golden et al., 1993), the segmented CFzCS (SCF:CS)
(Heyvaert et al., 2000), and the constant rate of supply model (CRS) (Oldﬁeld and

Appleby, 1984; Sanchez-Cabeza et al., 2000), and then the models were evaluated to

65

ascertain which was the most appropriate for each lake based on the location of the
I37Cs peak and the stable lead peak (Table 1) (Yohn et al., 2004).

Table 1. Characteristics of 11 Michigan study lakes and selected data determined
from 210Pb analysis, including sampling year, lake surface area, lake depth (depth of
the water column above the coring location), model used for dating, the approximate
mixed depth, the average sedimentation rate, the focusing factor (FF), and the oldest
section in the core.

 

Lake
surface Lake Approx Sed Oldest
Lake Year area depth Dating mixed rate FF section

name sampled (km’) (m) model depth (glm’ly)

 

Cadillac 2001 4.7 8 CRS 14 cm 117 1.7 18298
Crystal B 2001 39 50 CR8 4 cm 624 2.9 1516"l
Crystal M 2000 2.9 17 CR8 6 cm 465 1.7 17321’
Elk 1999 31 59 SCF:CS 1 cm 337 2.1 1279 a
Gratiot 1999 5.8 24 crzcs 5 cm 255 2.5 18238
Gull 1999 8.2 34 scr:cs 3 cm 404 1.8 1496 5'
Higgins 1999 39 41 CF:CS 3 cm 232 2.0 17293
Littlefleld 2000 0.7 21 Pb 444 2.0b 1732 a
Mullett 2001 67 36 scr:cs 4 cm 801 3.6 1708 "

Paw Paw 2001 3.7 28 crzcs 3cm 828 27° 1923
Whitmore 2001 2.7 20 scrzcs 6cm 556 28° 1887

 

a. Estimated dates based on extrapolation.
b. A focusing factor could not be calculated for Littleﬁeld Lake, so the average
focusing factor of all lakes sampled previously (except Cass Lake) was used.
c. Estimated focusing factors based on extrapolation.

21on in the sediments of Littleﬁeld Lake has an unusual proﬁle, with the
lowest 21on activities in the middle of the core, and could not be dated using 21"Pb
(Y ohn et al., 2004). It is unlikely that an error in core collection would cause this
altered 21”Pb proﬁle. This proﬁle indicates that deposition of sediment at the sampling
site was disturbed either by natural or human processes. Dates for this lake were

estimated through the shape of the stable lead proﬁle. The sample with the highest

concentration of lead was assigned a date of 1972, and the sample where background

66

concentrations were reached was assigned a date of 1850. Sedimentation rates were
calculated from these dates and ages were estimated for each section (Y ohn et al.,

2004).

Outlier analysis

Although cluster analysis does not require data to be normally distributed, this
technique is sensitive to the presence of outliers (Hair, 1998; Yohn et al., in
preparation). Outliers are data with values in one or more variables that are
signiﬁcantly different from the rest of the data, and may represent analytical/sampling
errors, or accurately reﬂect concentrations in the sediment. If outliers are present due
to analysis or sampling error, these data should clearly be removed. However, even
outliers that represent data accurately will be statistically inﬂuential, and one or two
outlying data points may determine how an element factors or clusters. It is not
appropriate to simply delete these inﬂuential points, as these points may be
representative of certain processes. Therefore cluster analysis was performed on all
data for each lake and on data with outliers removed.

Outliers were present in nine of the eleven lakes. Two or three samples near
the surface were outliers for several metals in Elk Lake (As, Mo, Mn, Ba, Fe), and this
lake was analyzed both with and without these outliers present. Mullett (As, Mn, Ba)
and Higgins (Ba, Mn) lakes also had high concentrations near the surface of some
elements. Removal of these points did change results slightly. Results using all data
will be reported, but the effect of the outliers will be discussed. In Crystal M Lake,

titanium and molybdenum each had one high outlier. These data appeared to be

67

analysis error, and Crystal M Lake was analyzed without these samples. Outliers in
Gratiot Lake were present for aluminum (2 samples) and chromium (one sample), and
in Gull Lake for arsenic (ﬁve samples). However, removal of these data had little
impact on analysis, therefore all data were used. Crystal B, Cadillac, Paw Paw and
Whitrnore lakes also had one to three outliers, but removal of these data had no

signiﬁcant effect on the groupings, therefore all data were used.

Element grouping

Elements that have similar temporal trends in a sediment core are probably
inﬂuenced by similar sources and processes (Kemp and Thomas, 1976; Sanei et al.,
2001). Elements with similar proﬁles can be grouped using cluster analysis (Y ohn et
al., in preparation). If these groups of elements remain consistent among lakes, it
would be possible to choose one element to represent each group, and reduce the
number of elements analyzed without loss of information. Elements in the study lakes
were divided into groups with similar proﬁles and groups were categorized as
terrestrial, anthropogenic, diagenetic or carbonate. Elements that had proﬁles
dissimilar to other elements were not grouped. The category that each group was
assigned to was based on the expected sources or processes inﬂuencing the suite of
elements in that group.

The elements classiﬁed as terrestrial in the study lakes generally include
titanium, vanadium, aluminum, magnesium, chromium and potassium, and sometimes
uranium, barium and iron (Table 2). Changes in the inputs of terrestrial elements,

which are those that are inﬂuenced by the amount of allocthonous non-organic

68

material entering the lake, may be caused by increased erosion by natural (e.g., forest
ﬁres) or human processes (e.g., clear cutting) (Davis, 1976). Gratiot Lake, the only
lake sampled in the Upper Peninsula of Michigan, includes calcium in the terrestrial
elements. Gratiot Lake is the only non-carbonate rich lake where calcium may be
more inﬂuenced by terrestrial inputs than calcium carbonate dynamics.

Table 2. Classiﬁcation of elements into terrestrial (T), carbonate (C), diagenetic (D,
D1,DZ),and anthropogenic (A, A1, A2). Use of A2 indicates there was more than one
group of anthropogenic elements in the lake. Use of D1, D2 notation indicates that
there was more than one group of diagenetic elements in the lake. Unclassiﬁed
elements did not ﬁt clearly into a group, and elements classiﬁed twice appear to be
inﬂuenced by both classes. A (—) indicates that data were not collected for this
element. Lakes include Gratiot (Grat), Elk, Gull, Higgins (Hig), Littleﬁeld (Lit),
Crystal M (CrM), Cadillac (Cad), Crystal B (CrB), Mullett (Mul), Paw Paw (Paw),
and Whitrnore (Whit). OR indicates that outliers were removed.

 

 

Elk
Grat Elk 0R Gull Fﬁg Lit CrM Cad CrB Mul Paw Whit
Ti T T T T T T T T T
v T T T T T T T T T
Cr T T T T T A T
Cu A A A A A A A A A A2
As A D A A,02 01 01 01 D D D
Mo D 02 D1 01 02 D D
Cd A A A A A A A A A A A1
Pb A A A A A A A A A A A A1
Al T T T T T T T T T T T T
Zn A A A A A A A A A A A1
Sr T — - - C C C C C C C C
Mg T T T C T C, T T
K D T T T T T T T - — .-
Mn D D D1 02 02 D1 D D
Ba D D T C 02 02 C C D T
Ca T c c C C C C C C C C C
Fe A D T D1 D1 D T D
U T T T 02 01 01 02 T D

 

69

The terrestrial groupings are relatively similar among lakes (Table 2) and to

results from previous studies (Kemp and Thomas, 1976; Qu et al., 2001; Sanei et al.,

2001; Tuncer et al., 2001) despite a large variation of proﬁles among lakes (Figure 2).

Aluminum is
grouped with the
terrestrial elements
in all the lakes,
suggesting it may be
considered
representative of
this group.
Titanium,
vanadium, and
potassium are
clustered with the

terrestrial inputs in

Alumlnum concentrations (mg/kg)
10000

0 5000 15000 20000

1820 l

1800 1 " ' " ' ‘

0 2000 4000 6000 8000
Aluminum concentrations (mg/kg)

 

 

 

—o—Elk +Gull +Gratiot -o—Whitmore

Figure 2. Sediment aluminum concentration proﬁles
(mg/kg) of four lakes representing the diversity of
concentrations and proﬁles found among the study lakes.
Gratiot and Whitrnore lakes have signiﬁcantly higher
aluminum concentrations, and are plotted on the upper x-
axrs.

all but one or two lakes, and may also be considered representative.

The carbonate cluster includes calcium and strontium, and sometimes

magnesium and barium (Table 2). This portion of the sediment is most often thought

to be inﬂuenced by calcium carbonate deposition and therefore tends to have low

concentrations of most metals (Auer et al., 1996; Sanei et al., 2001). In many of the

lakes (Elk, Gull, Higgins, Crystal M, and Mullett Lake), calcium is negatively

correlated to aluminum (r = -0.746 to -0.801). These lakes are carbonate rich, and

70

increases in the amount of carbonate would dilute the concentration of the terrestrial
elements (Jaquet et al., 1982). Calcium is the most representative element of this
group (Table 2).

Diagenetic processes are apparent in all of the lakes, but the behaviors of
elements are not consistent among lakes. Molybdenum, uranium, iron, manganese,
barium, and arsenic all appear to be inﬂuenced by diagenetic processes in some of the
lakes (Table 2). In some lakes there are two diagenetic groups, the ﬁrst including
molybdenum and uranium, and the second iron, barium and manganese. However, in
other lakes, all of these elements have similar proﬁles. Arsenic in some lakes behaves
similarly to molybdenum, and in other lakes arsenic proﬁles are similar to iron or
manganese. However, despite the dissimilar behaviors, it appears that molybdenum,
iron and manganese represent all the diagenetic groups for the study lakes (Table 2).

Lead proﬁles are inﬂuenced dominantly by anthropogenic inputs in all of the
study lakes, and zinc, copper and cadmimn in the majority of lakes (Table 2). These
elements are enriched in lake sediments due to human actions and may enter lakes
ﬁ'om atmospheric deposition or ﬁom point source inputs within the watershed
(Bruland et al., 1974; Iskander and Keeney, 1974; Evans and Dillon, 1982; Spiethoff
and Hemond, 1996; Kerfoot and Robbins, 1999). Chromium concentrations in the
sediment are only inﬂuenced dominantly by anthropogenic inputs in Paw Paw Lake.
There are also anthropogenic inputs of arsenic in Gratiot, Elk and Gull lakes, and iron
in Gratiot Lake. Extensive copper mining and smelting occm'red near the Gratiot Lake

watershed, possibly providing a source for iron to the lake.

71

Anthropogenic elements ﬁ'om each lake did cluster together, despite the fact
that the sources for each metal may be different (e. g., copper from copper smelting
emissions, or lead from leaded gasoline). Most of these proﬁles increase from the
early 1900s to the 1970s, and decrease to the present. The overall pattern of industrial
growth combined with few environmental regulations in the early 1900s, the
implementation of environmental legislation in the 1970s (Eisenreich et al., 1986;
Callender and vanMetre, 1997), along with the decline of some major industries (e. g.,
steel production, copper mining) (Kerfoot et al., 1994; Perkins et al., 2000) may lead
to the overall similarity in the proﬁles of the anthropogenic metals, despite differences
in sources.

Outliers in Elk, Higgins and Mullet lakes are the result of a large increase in
concentration near the sediment surface, probably due to primary and secondary
diagenetic processes (Elk Lake: Mo, Ba, Mn, Fe, As; Higgins Lake: Mn, Ba; Mullett
Lake: As, Mn, Ba). The removal of outliers in Higgins and Mullet lakes shows that
some elements are grouped only because of high concentrations near the surface. In
Higgins Lake, manganese and barium are no longer grouped together when outliers
are removed. In Mullet Lake, manganese, barium and iron are still grouped after the
removal of outliers, but the proﬁle of arsenic is dissimilar. The removal of outliers in
Elk Lake allows some interpretation of the original sources and processes inﬂuencing
the inputs of elements that are affected by diagenesis. After the removal of outliers,
iron and barium cluster with the terrestrial elements. Examination of the proﬁles
indicates that iron and barium have proﬁles very similar to the terrestrial elements,

except for the increase in concentration at the surface due to diagenesis, indicating that

72

these two elements

 

 

 

2000 l ‘— T P —_ m? 'i are inﬂuenced by
1950 J l . .
both terrestrial mputs
19004 i
l and diagenetic
g 1850 |
1800 , processes (Figure 3).
1750 4 Similarly, the
1700 17‘ ‘ -.-—- —.— ‘_‘—“* concentration proﬁle
0 5000 10000 15000 20000 25000
Concentration (mglkg) of arsenic is very
—)(-—Al —I— Fe similar to the
Figure 3. Sediment aluminum and iron concentration .
proﬁles (mg/kg) in Elk Lake. anthropogenlc

elements, except for the high concentrations near the surface.

Despite the dissimilarity of element proﬁles among lakes, the placement of
elements into source/process groups is consistent enough to choose representative
elements for each of these groups. Aluminum is the most representative of the
terrestrial elements, calcium of the carbonate group, and iron, manganese, and
molybdenum represent the diagenetic group. This group of elements is the minimum
suite necessary, in addition to any anthropogenically inﬂuenced elements of interest,

to represent terrestrial inputs, carbonate deposition and diagenetic processes.

Identification and quantiﬁcation of human inputs

Element groupings are similar among lakes, and it is possible to select
representative elements to characterize source/process groups. When modeling

contaminants in water, the analysis of all the major constituents is necessary to

73

understand the behavior of the contaminant. Similarly, it may be necessary to analyze
elements representing each source/process group to understand histories of
contaminant inputs in the sediment. In particular, the analysis of diagenetic elements
may be necessary to understand arsenic proﬁles, and the analysis of terrestrial
elements to differential natural, secondary and direct human inputs.

Arsenic is an element of particular interest, as both anthropogenic inputs and
diagenetic processes inﬂuence arsenic (Harrington et al., 1998). In order to monitor
the anthropogenic inputs of arsenic using sediment chronologies, it is essential to be
able to differentiate early diagenesis from anthropogenic inputs. In this study, we
assume that if arsenic follows a trend similar to an element that is known to be
inﬂuenced by diagenetic processes and not anthropogenic inputs, it may be concluded
that arsenic is being inﬂuenced by diagenesis. If arsenic patterns are dissimilar to
diagenetic elements, it is possible that arsenic patterns are due to anthropogenic inputs.

It is clear from the results of the cluster analysis that arsenic concentration
proﬁles in the sediment may be altered by diagenetic processes, and often cannot be
interpreted simply as anthropogenic inputs. Diagenetic processes may cause proﬁles
that, for example, have a peak in arsenic concentrations near the sediment water
interface, such as Elk Lake (Figure 4). This pattern could easily be interpreted as an
increase in human inputs, however, the same pattern is seen in iron, manganese,
barium, and molybdenum, indicating that the high concentrations at the surface are
more likely a result of diagenetic processes (Figure 4). This demonstrates the need to
analyze for some diagenetic elements to interpret arsenic proﬁles. In all lakes where

arsenic is inﬂuenced by diagenesis, it has a similar trend to iron, manganese or

74

molybdenum (Table

 

 

 

 

2). There are clear 2000 1 x L:
l
anthropogenic inputs 195° ’ l
1900 l l
of arsenic in only i
g 1850 1
three lakes (Gratiot, 1300 l
Elk and Gull), and Elk 175° “
17004 ~-—*- «1- —- —
and GU" lakes are USU 0.00 10.00 20.00 30.00 40.00
inﬂuenced by 3090003390" ("'9’")
diagenesis. -o—As —x—-Mn

Figure 4. Arsenic and manganese sediment concentration
Quantiﬁcation proﬁles in Elk Lake. Both elements show a large
concentration increase near the surface, probably due to

of a terrestrial element diagenetic Processes.
may also be necessary to understand and measure direct human inputs. Unlike organic
contaminants (e. g., DDT), metals are present naturally in sediments, complicating the
determination of anthropogenic inputs. Metals such as lead and copper will enter the
lake from both direct anthropogenic sources and terrestrial inputs, and in order to
calculate the amount of human inputs, it is necessary to determine the terrestrial
inputs. While humans can inﬂuence terresnial inputs (e. g. increased erosion due to
land use change), these inputs are still of natural geological material, and should be
differentiated from direct anthropogenic inputs (e. g., atmospheric and point source
inputs).

There are two techniques used in determining natural or background
concentrations. The ﬁrst approach, the constant background technique, is to use the

average concentration in sediments that were deposited before signiﬁcant human

75

inﬂuence (Callender and Rice, 2000). This technique works on the assumption that
the concentration of an element due to natural inputs remains the same throughout the
depth of the core. This constant background concentration is subtracted from the total
to determine the anthropogenic concentration.

A second method, the watershed correction technique, uses normalization to a
terrestrial element to determine anthropogenic inputs. This technique assumes that the
concentration of a metal remains the same in the terrestrial inputs from the watershed,
but the amount of terrestrial input may vary over time. Since the concentration of
elements is assumed to remain the same in the terrestrial inputs, it follows that the
natural ratios of elements remain the same throughout the core. The ratio of an
anthropogenic element (e.g., lead), and an element not inﬂuenced by direct human
inputs (e.g., aluminum, iron or scandium) is determined in the pre-anthropogenic
sediments (Bruland et al., 1974; Koelmans, 1998; Tuncer et al., 2001), and the
concentration due to human inputs is calculated separately for each depth using the
equation:

MT
M; -(ﬁ x N1] = M’swm
86

MT Total concentration of the anthropogenic metal of interest in sample S

M66 Total concentration of anthropogenic metal of interest in background sample
MAW“o Anthropogenic concentration of metal of interest in sample S

All;5 Total aluminum concentration in background sample

Al:3 Total aluminum concentration in sample S

76

If the constant background technique is sufﬁcient to determine anthropogenic
inputs, then the analysis of terrestrial elements is not necessary in order to monitor
elements such as lead (Heyvaert et al., 2000). However, if there are large changes in
concentration of terrestrial elements over time, the use of the watershed correction
technique may be more appropriate (Kemp and Thomas, 1976; Johnson and Nicholls,
1988), and the analysis of a terrestrial element necessary. Differences in
anthropogenic inventories for lead and copper calculated with a constant background
and normalizing to aluminum were examined to determine if the analysis of terrestrial
elements is necessary. Anthropogenic inventories are a measure of the total amount of
contaminant input by humans per unit area (uglcmz) ﬁ'om the present until background
concentrations are reached (Golden et al., 1993).

The importance of the calculation method used depends both on the lake and
the element (Figure 5). Among lakes the difference between the two calculation
methods is greater for lakes like Elk and Gull, which have large shifts in the amount of
terrestrial inputs. For all lakes, the relative difference between the constant
background methods and the watershed correction technique are much higher for
copper then lead (Figure 5). This is because there is sufﬁcient anthropogenic lead
input to make the natural inputs relatively insigniﬁcant. As a result, the method used
to calculate natural inputs has little effect on the determination of human inputs.
Copper anthropogenic inputs are much lower than those of lead (Figure 5), with the
exception of Cadillac Lake, making the calculation of these inputs more sensitive to
the background value used. Copper sulfate was added to Cadillac Lake to control

swimmer’s itch, and caused the highly elevated copper concentrations. Overall, lakes

77

     
 

 

 

 

Copper inventories (HQ/cmz)
3 8 8 S 8

.a
O

 

 

 

 

 

 

 

 

O

 

 

V

 

 

Load inventories 119/
s 8 '8‘ 8

l l
I l

 

 

 

 

 

 

 

 

 

O
L

 

6%6‘9 (@466
Figure 5. Focusing corrected anthropogenic inventories
of (a) copper and (b) lead calculated using the constant
background (white) and watershed correction technique
(black). Inventories for copper in Cadillac Lake were
above the scale. ’The focusing factor of Littleﬁeld
Lake was estimated.

with large shifts in
terrestrial inputs (e. g.,
Gull and Elk lakes)
and anthropogenically
inﬂuenced elements
that are not highly
enriched are the most
sensitive the to the
background
calculation technique.
For many of
the lakes, the use of
aluminum
normalization
changes the results for
copper signiﬁcantly
(Figure 5a).
Therefore, in order to

quantify direct human

inputs to a lake it is essential to quantify a terrestrial element. Aluminum consistently

groups with the terrestrial elements in the study lakes, and is the most representative of

this group, but titanium and vanadium also ﬁequently represent the terrestrial elements

and may be considered representative.

78

Conclusions

When determining anthropogenic inputs from sediment chronologies, it is
necessary to evaluate terrestrial inputs and diagenetic processes as well as inputs of
anthropogenic metals. We suggest a suite of six elements in addition to elements
known or suspected to have human inputs that will represent the minimum analysis
needed to evaluate temporal trends and quantify direct anthropogenic inputs. These
are aluminum and titanium which represent the terrestrial elements; iron, manganese
and molybdenum to represent diagenetic processes; and calcium to represent
carbonate deposition. Although aluminum appears to be sufﬁcient to evaluate
terrestrial inputs, the analysis of a second terrestrial element would help validate the
results. Analysis of terrestrial elements is necessary when calculating anthropogenic
inputs, and also acts as an indicator of the amount of erosion and sediment input from
the watershed (Davis, 1976). Three elements are necessary to monitor diagenesis,
because elements respond differently in various lake environments. These three
elements appear to be sufﬁcient because, in the study lakes, arsenic always behaves
similarly to one of these elements when it is inﬂuenced by diagenesis. While the
analysis of calcium is not necessary to calculate anthropogenic inputs, analysis of this
element gives insight into the important processes of calcium carbonate deposition.
These six elements, in addition to elements with known anthropogenic sources, should
be considered the minimum number of elements needed to interpret patterns of human

inputs when using sediment chronologies.

79

References

Auer, M.T., Johnson, N.A., Penn, MR. and Effler, S.W., 1996. Pollutant sources,
depositional environment, and the surﬁcial sediments of Onondaga Lake, New
York. J. Environ. Qual. 25, 46-55.

Brown, E.T., Callonnec, LL and German, CR, 2000. Geochemical cycling of redox-
sensitive metals in sediments ﬁom Lake Malawi: A diagnostic paleotracer for
episodic changes in mixing depth. Geochim. Cosmochim. Acta 64, 3515-3523.

Bnrland, K.W., Bertine, K., Koide, M. and Goldberg, ED, 1974. History of metal
pollution in Southern California costal zone. Environ. Sci. Technol. 8, 425-
432.

Callender, E. and Rice, K.C., 2000. The urban environmental gradient: Anthropogenic
inﬂuences on the spatial and temporal distributions of lead and zinc in the
environment. Environ. Sci. Technol. 24, 232-238.

Callender, B. and vanMetre, RC, 1997. Reservoir sediment cores show US lead
declines. Environ. Sci. Technol. 31, A424-A428.

Davis, B.M., 1976. Erosional rates and land use history in southern Michigan.
Environmental Conservation 3, 139-148.

Door, J .A. and Eschman, DR, 1970. Geology of Michigan. University of Michigan
Press, Ann Arbor, 476 pp.

Eisenreich, S.J., Metzer, NA. and Urban, N.A., 1986. Response of atmospheric lead
to decreased use of lead in gasoline. Environ. Sci. Technol. 20, 171-174.

Evans, RD. and Dillon, P.J., 1982. Historical changes in anthropogenic lead fallout in
southern Ontario, Canada. Hydrobiologia 91, 131-137.

Garrels, RM. and Thompson, ME, 1962. A chemical model for sea water at 25C and
one atmosphere total pressure. American Journal of Science 260, 57-66.

Golden, K.A., Wong, C.S., Jeremiason, J .D., Eisenreich, S.J., Sanders, M.G.,
Hallgren, J ., Swackhammer, D.L., Engstrom, DR. and Long, D.T., 1993.
Accumulation and preliminary inventory of organochlorines in Great Lakes
sediments. Wat. Sci. Technol. 28, 19-31.

Hair, J .F ., 1998. Multivariate data analysis. Prentice Hall, Upper Saddle River, N.J.,
730, 12 pp.

80

Harrington, J .M., Laforce, M.J., Rember, W.C., Fendorf, SE. and Rosenzweig, RR,
1998. Phase associations and mobilization of iron and trace elements in Coeur
d'Alene Lake, Idaho. Environ. Sci. Technol., 650-656.

Hewitt, AD. and Reynolds, C.M., 1990. Dissolution of metals from soils and
sediments with a microwave-nitric acid digestion technique. Atomic
Spectroscopy 11, 1425-1436.

Heyvaert, A.C., Reuter, J .E., Slotton, D.G. and Goldman, CR, 2000.
Paleolimnological reconstruction of historical atmospheric lead and mercury
deposition at Lake Tahoe, California-Nevada. Environ. Sci. Technol. 34, 3588-
3597 .

Iskander, I.K. and Keeney, DR, 1974. Concentration of heavy metals in sediment
cores from selected Wisconsin Lakes. Environ. Sci. Technol. 8, 165-170.

Jaquet, J .-M., Davaud, E., Rapin, F. and Vemet, J .-P., 1982. Basic concepts and
associated statistical methodology in the geochemical study of lake sediments.
Hydrobiologia 91.

Johnson, MG. and Nicholls, K.H., 1988. Temporal and spatial trends in metal loads to
sediments of Lake Simcoe, Ontario. Water, Air and Soil Pollution 39, 337-354.

Kemp, A.L.W. and Thomas, R.L., 1976. Cultural impacts on the geochemistry of the
sediments of Lakes Ontario, Erie and Huron. Water, Air and Soil Pollution 5,
469-490.

Kerfoot, W.C., Lauster, G. and Robbins, J .A., 1994. Paleolimnological study of
copper mining around Lake Superior: Artiﬁcial varves from Protage Lake
provide a high resolution record. Limnology and Oceanography 39, 649-669.

Kerfoot, W.C. and Robbins, J .A., 1999. Nearshore regions of Lake Superior: Multi-
element signatures of mining discharges and a test of Pb-210 deposition under
conditions of variable sediment mass ﬂux. J. Gt. Lakes Res. 25, 697-720.

Koelmans, A.A., 1998. Geochemistry of suspended and settling solids in two
ﬁeshwater lakes. Hydrobiologia 364, 15-29.

Long, D.T. and Saleem, Z.A., 1974. Hydrogeochemistry of carbonate groundwaters of
an urban area. Water Resources Research 10, 1229-1238.

McKee, J .D., Wilson, T.P., Long, D.T. and Owen, R.M., 1989. Pore water proﬁles and

early diagenesis of Mn, Cu and Pb in sediments ﬁom large lakes. J. Gt. Lakes
Res. 15, 68-83.

81

Oldﬁeld, F. and Appleby, P.G., 1984. Empirical testing of 210Pb-dating models for
lake sediments. In: E.Y. Haworth and J .W.G. Lund (Editors), Lake Sediments
and Environmental History. University of Minnesota Press, Minneapolis, pp.
93-124.

Perkins, S.M., Filippelli, GM. and Souch, C.J., 2000. Airborne trace metal
contamination of wetland sediments at Indiana Dunes National Lakeshore.
Water Air Soil Poll 122, 231-260.

Qu, W.C., Dickman, M. and Wang, S.M., 2001. Multivariate analysis of heavy metal
and nutrient concentrations in sediments of Taihu Lake, China. Hydrobiologia
450, 83-89.

Sanchez-Cabeza, J .A., Ani-Ragolta, I. and Masque, P., 2000. Some considerations of
the Pb-210 constant rate of supply (CRS) dating model. Limnology and
Oceanography 45, 990-995.

Sanei, H., Goodarzi, F. and Van der F lier-Keller, E., 2001. Historical variation of
elements with respect to different geochemical ﬁactions in recent sediments
from Pigeon Lake, Alberta, Canada? J. Environ. Monit. 3, 27-36.

Spiethoff, HM. and Hemond, HR, 1996. History of toxic metal discharge to surface
waters of the Aberjona Watershed. Environ. Sci. Technol. 30, 121-128.

SPSS, 2000. SYSTAT 10, Chicago.

Tuncer, G., Tuncel, G. and Balkas, T.I., 2001. Evolution of metal pollution in the
Golden Horn (Turkey) sediments between 1912 and 1987. Marine Pollution
Bulletin 42, 350—360.

Yohn, S.S., Long, D.T., Fett, J .D. and Patino, LG, 2004. Regional versus local
inﬂuences on lead and cadmium loading in the Great Lakes region. Appl.
Geochem. 19, 1157-1175.

Yohn, S.S., Long, D.T. and Winkler, J.A., in preparation. The use of multivariate

statistics for the interpretation of sediment chronologies. Limnology and
Oceanography: Methods.

82

CHAPTER FOUR

Yohn, S.S., Long, D.T., Fett, J.D., Patino, L.C., Giesy, J .P. and Kannan, K., 2002.
Assessing environmental change through chemical-sediment chronologies ﬁom inland
lakes. Lakes & Reservoirs: Research and Management 7, 217-230.

83

CHAPTER 4
ASSESSING ENVIRONMENTAL CHANGE THROUGH CHEMICAL SEDIMENT
CHRONOLOGIES FROM INLAND LAKES

Abstract

This paper examines the hypothesis that the spatial pattern of metal
accumulation rates in inland lake sediments provides insight into the cause(s) for the
historical and continued loadings of contaminants to the environment. To address this,
copper and lead accumulation rates in the Great Lakes watershed are studied. A multi-
element approach was used, in which many non—toxic chemicals are measured along
with the toxic chemicals of interest, rather than a target speciﬁc approach. The multi-
element approach also allows for assessing the environmental state of the lake with
respect to its surrounding watershed.

Sediment cores were collected ﬁom the deepest portion of ﬁve Michigan
inland lakes, sectioned, metals extracted by nitric acid, microwave digestion and
leachates analyzed for 26 metals using an inductively coupled plasma mass
spectrometer with hexapole technology. Sedimentation rates, ages and focusing were
calculated via me. Dasymetric mapping was used to estimate population
distributions in lake watersheds. Watershed area was estimated from digital elevation
data using Arc/INFO.

The data show that as loadings ﬁom dominant (regional) sources decrease
(e.g., atmospheric deposition in the case of lead), new patterns provide insight into
causes for continued contaminant loadings (e. g., population density). The data also

show the universal response of the watersheds to the onset of the anthropogenic

84

loadings of lead but not that of copper. Individual watersheds had different onset
times for anthropogenic copper loadings. Recent lead loadings to the environment can
be related to population density and those of copper reﬂect in part a spatial pattern that
is similar to the regional pattern for lead during the mid 19708. As this pattern for lead
could be related to a regional gradient of lead concentrations in atmospheric
deposition, the same relation cannot be made for copper. The cause for this spatial

pattern for copper is not clear.

Introduction

The environment is being altered because of changes in population grth and
distribution (e.g., creation of mega cities), the use of chemicals (e.g., pesticides), and
land use (e.g., transition of agriculture to urban). Thus, questions are asked such as
what is the current state of the environment, what is the cause of the current state,
what are emerging and lingering issues, and what are future trends? Changes in the
accumulation rates or loadings of metals to lakes over time are recorded in the
sediments (Jaquet et al., 1982), making sediment chronologies an extremely useful
approach to answering these questions (Hermanson, 1998).

This paper examines changes in lead and copper accumulation rates (via
sediment chronologies) to the Great Lakes region, where regulatory measures have
worked to decrease loadings of these as well as several toxic chemicals. Because
loadings for many of these chemicals are not at background levels, inﬂuences on and
sources for these chemicals need to be determined. Major anthropogenic sources for
lead include coal combustion, ore smelting (lead and Zn), waste incineration, leaded

gasoline combustion and for copper; copper mining and smelting, other non-ferrous

85

smelters, plastic industry, steel works, agriculture, and sewage sludge (pig farming)
(Reimann and de Caritat, 1998). Important sources for lead and copper in the Great
Lakes basin are leaded gasoline and mining, respectively.

Sediment chronologies from large lakes (e.g., > 1,000 kmz) are often used to
address environmental issues, but these chronologies typically integrate the
contaminate accumulation signal from a large area, which may be at a scale too coarse
to address the problem stated above. On the other hand, studying the chemical
sediment chronology of inland lakes (e.g., < 100 kmz) provides important information
about a particular lake, however the sources inﬂuencing one lake may or may not have
the same effect on another lake. The results cannot necessarily be generalized to other
lakes in the region and thus, sediment chronologies from multiple inland lakes that are
spatially located over a region are an excellent compliment to those from the large
lakes. Inland lakes record information on a local scale that can be compared to other
inland lakes on a larger spatial scale. Most spatial studies of lakes have only
examined recent data (Johnson and Nicholls, 1988; Boyle et al., 1999; Callender and
Rice, 2000), rather than examining the change in patterns over time, as we do in this
study.

Interpreting and comparing sediment chronologies ﬁom inland lakes offer
certain challenges. In particular, terrestrial inputs from the watershed, anthropogenic
inputs ﬁom the watershed and atmosphere, and autochthonous materials are deposited
together in one mixed record, making it difﬁcult to differentiate the relative
importance of each different source (Edgington and Robbins, 1976; Balogh et al.,

1999). To address this problem we measure not only the toxic chemical(s) (the target

86

speciﬁc approach) of interest such as lead (Iskander and Keeney, 1974; Dillion and
Evans, 1982; Spiethoff and Hemond, 1996), but also many non-toxic chemicals such
as aluminum and barium (the multi-element approach). The multi-element approach
is integral to understanding patterns of metal accumulation within a lake (Bruland et
al., 1974; Johnson and Nicholls, 1988; Ramanathan et al., 1996; Heyvaert et al., 2000;
Kang et al., 2000). The multi-element approach allows: for determining current
contaminant levels and historical loading trends, predicting future changes in chemical
loadings, detecting new and emerging issues, understanding diagenesis and the
inﬂuences of terrestrial input ﬁom the watershed, differentiating terrestrial inputs ﬁom
anthropogenic inputs (e.g., Kemp and Thomas, 1976; Harrington et al., 1998; Brown
et al., 2000). Thus, this pilot study combines the spatial, temporal and multi-element
aspects of different studies performed in the past (e.g., Bruland et al., 1974; Dillion
and Evans, 1982; Birch et al., 1996; Heyvaert et al., 2000), to develop a more
complete understanding of what is inﬂuencing the chemical ﬂuxes and geochemistry
of lake sediments.

This study not only examines spatial patterns of metal accumulation rates, but
also the changes in spatial patterns over time. Our hypothesis is that the spatial pattern
of chemical accumulation rates provides insight into the cause(s) for the historical and
continued loadings of contaminants to the environment. It is developed around the
following:

0 environmental regulations are typically put in place to address dominant

contaminant sources whose inﬂuences are typically at regional scales,

87

0 spatial and temporal patterns of chemical loadings from dominant
inﬂuences/sources most likely obscure the patterns ﬁ'om less dominant sources
such as those that might operate at the watershed scale,

0 as loading from the dominant sources are reduced, the effect of less dominant
inﬂuences/sources on the spatial patterns of contaminant accumulation rates
may be revealed, and

o the newly detected patterns will provide insight into the underlying causes(s)

for the continued loading of the contaminants to the environment.

By understanding changes in the importance of different sources and processes
over time, it may be possible to determine the inﬂuences and efﬁcacy of
environmental legislation. In addition, this approach will help us to understand what
sources and processes are inﬂuencing contaminant accumulation rates of individual
lakes. This is important because it has been shown that contaminants can have an
adverse effect on the zoobenthos (Catallo et al., 1995), and that sediments may
provide a continuous source of toxic metals to the aquatic environment (Spiethoff and
Hemond, 1996). This information can help policy makers by indicating which sources
of contamination are the most signiﬁcant.

At the watershed scale, communities surrounding the lake potentially inﬂuence
chemical loading to lakes. These communities often do not contribute much in the
way of industrial contamination, but do act as sources of contamination to the lake via
household and yard wastes, storm drains, construction activities and materials, and

trafﬁc. For example, Callender and Rice (2000) found a relationship between ﬂuxes

88

of lead and zinc to lakes and population density. By comparing known contaminant
accumulation rates to population densities of each watershed, it may be possible to
evaluate impacts of the local community. The shores of inland lakes across the United
States are a preferred location for many homes and cottages, and many of these
residents are concerned about the health of their lake. By establishing a correlation
between population density and heavy metal contaminants, it is possible to
demonstrate to residents that they do inﬂuence the health of the lake. This may result
in actions such as allowing less development around the lakes, or updating sewer

systems, which would decrease metal loading to the lake.

Methods
Sediment cores were collected from Cass, Elk, Gratiot, Higgins and Gull Lakes

in the State of Michigan, USA during the summer of 1999 (Figure l). Sediment cores

were collected from the deepest ,1: __ Gratiot Lake
portion of each lake using a MC- 11\ \‘v’e’ﬂlﬁ
URL I (ﬁr.—-a_{.-__n/§,
. . N 31,/"J; {,- “L...
400 Lake/Shelf Multl-corer, whlch I 4 /b . % Higgins Lake
takes four replicate cores and is Elk Lake I (4’)»
designed to collect the ﬂocculent L a $7 Cass Lake
Gull Lake 0 f’

7’
g.—

__—- ‘

material at the sedirnent-water

Figure 1. Location of lakes sampled in
Michigan, U.S.A. For reference the Greater
Chicago metropolitan/industrial area is
southwest of Gull Lake.

interface. If the cores showed any
evidence of disturbance, a set of

new cores was immediately taken.
Cores were described and examined for color, texture, and signs of zoobenthos.

Sediment was extruded ﬁom the cores and sectioned at 0.5 cm intervals for the top 5

89

to 15 cm of the core, and at 1 cm intervals for the remainder of the core. An extruding
device was designed for this study to facilitate the extrusion in remote areas (i.e., no
electrical or hydraulic power is needed) and allows for precise sampling of the
ﬂocculent material at the sediment/water interface. Samples were stored in the ﬁeld
on ice in acid washed, plastic jars.

Upon return to the laboratory, sediments were ﬁozen until they could be
ﬁeeze-dried and digested by nitric acid in a CEM-MDS-8le microwave (Hewitt and
Reynolds, 1990). Standard reference material (Buffalo River SRM 2704) and
procedural blanks were processed to test for accuracy and contamination. The
concentrated-acid digests were ﬁltered through an acid washed, distilled-deionized
water (DDW) rinsed 0.40 pm polycarbonate ﬁlter (Nuleopore'm). Samples for Hg
analysis were preserved with a gold chloride solution (ﬁnal concentration 200 ppb
Au). Samples were then analyzed using a Micromass Platform“ inductively coupled
plasma mass spectrometer with hexapole technology (ICP-MS-HEX). Sediments
were analyzed for a suite of metals and metalloids including magnesilun, aluminum,
potassium, calcimn, sodium, titanium, vanadium, manganese, iron, cobalt, nickel,
copper, zinc, arsenic, cadmium, barium, lead, lithium, selenium, scandium,
phosphorus, boron, rubidium, and uranium. Solutions were generally analyzed by
direct injection with a 1:10 dilution, with the standards matrix matched to the samples;
typically adjustments were made for high calcium concentrations. In some cases,
magnesium, calcium, and potassium were analyzed via ﬂame atomic absorption

spectrometry (Perkin Elmer 5100PC'”).

90

2|“Pb was measured (Freshwater Institute in Winnipeg, Manitoba, Canada) on

one sub-core from each lake to determine sedimentation rates, sediment ages and

focusing factors. Results ﬁ'om 21on analyses were veriﬁed in selected lakes using

137C S

All spatial data were collected ﬁom secondary somces, and were in, or were

converted into the Michigan GEOREF coordinate system: oblique Mercator

projection, datum NAD83, spheroid GRS 1980. Sources and type of data included:

Digital Elevation Map (DEM) of Michigan: United States Geological
Survey (Aichele, pers. com.). DEM data was in 30 m2 grid size raster data,
l:24,000 scale.

Census Tract Data: Michigan Maps and Information, Topologically
Integrated Geographic Encoding and Referencing (TIGER) database
(http://www.crs.msu.edu/db/welcome.htrnl). Data were imported as
polygon coverages with census tracts. Population data from 1990 were
used, and were downloaded for each county.

State Ownership Data: Michigan Department of Natural Resources spatial
data library (www.dnr.state.mi.us/spatialdataliblary). Data were imported
as polygon coverages and included data for state owned mineral rights, and
state owned surface rights.

Lake Data: Michigan Department of Natural Resources spatial data library
(www.dnr.state.mi.us/spatialdatalibrary). Data were imported as shape

ﬁles with lake identiﬁcation, lake type and other data.

91

Watersheds were delineated around each of the ﬁve lakes of interest. All
watersheds were delineated with the WATERSHED command in Arc/INFO.

To determine population densities within the watershed, it was essential to
estimate as accurately as possible where people were living within the census tract.
Therefore, dasymetric mapping was used to estimate population distribution. Areas
that are state owned (surface or fee rights) or areas covered by lakes were considered
as non-livable areas. It was assumed that the population was evenly distributed
throughout the remaining livable area within each census tract.

Anthropogenic accumulation rates of heavy metals were plotted versus
population density, and correlation analysis was preformed using a best ﬁt, least
squares regression line. R2 and p values were calculated, with the slope being

considered signiﬁcantly different than zero if p <0.05.

Results and Discussion

Sedimentation rates and focusing

2|°Pb analysis provides information on sediment age, rates of sedimentation,
and the quality of the sediment core. Results ﬁom the ﬁve lakes showed no
signiﬁcant disturbance in the cores. However, results indicate that some bioturbation
and mixing are present near the surface of the cores. The mixing zone is determined
by plotting accumulated dry mass versus log activity of 210Pb, where the slope is
related to sedimentation rates and the non-linear portion near the surface is considered
the mixing zone (Figure 2).

Cass Lake has a signiﬁcantly higher sedimentation rate than the other lakes,

and the oldest sediments collected are ﬁom 1971 (Table 1). Because of this,

92

background concentrations and inventories could not be calculated for Cass Lake. The
oldest sediments collected from all other lakes were from the early 1800s or older

(Table 1).
Log Excess me (Bqlg)

0.0010 0.0100 0.1000 1.0000
A 0.0 L ‘ I‘—

1.61
"2.0‘
3

E 3.0‘

34.0

g 5.04
E 6.0‘
3
8 7.0 1 /
<

8.0 r

 

glcm2

 

 

 

 

Figure 2. Log 21on activity (Bq/g) versus accumulated dry mass (g/cmz) in
Elk Lake. Regression lines for the CF :CS (constant ﬂux-constant
sedimentation) model and the SCFzCS (segmented constant ﬂux-constant
sedimentation) model are shown.

Table 1. Date of the oldest section (date), focusing factors (FF) and average
sedimentation rates (rate) ﬁ'om sediment cores in ﬁve lakes in Michigan, USA.

 

 

Lake Date FF Rate m2
Cass 1971 6.00 3480
Elk 1279 2.05 420
Gratiot 1823 2.49 260
Gull 1496 l .78 500
Hi gg'ns l 729 2.02 240

 

93

Two models were used to date the sediments, the constant ﬂux, constant
sedimentation rate model (CF:CS) (and a modiﬁcation of this method, the segmented
CF :CS), and the constant rate of supply model (CRS). The CF:CS model assumes a
constant ﬂux of 210Pb to the sediment and a constant rate of sedimentation over some
period of time. Generally, the rate of sedimentation is assumed to stay the same
throughout the length of the core (Figure 2, CF:CS), but the segmented CF :CS
(SCF:CS) model may be used if there is a clear change in sedimentation rate (F igure 2,
SCF:CS) (Heyvaert et al., 2000). The CRS model also assumes constant ﬂux of 21"Pb
to the sediment, but does not assume that the sedimentation rate remains the same.
Instead, the CRS model determines a different sedimentation rate between each depth.
Some authors prefer the CRS model, ﬁnding that it gives more insight into changes in
sedimentation rates (Fritz et al., 1993; Gottgens et al., 1999). However, the CRS
method is particularly vulnerable to noisy data, as errors are propagated to date
assignments (Heyvaert et al., 2000). There is no consensus as to which model is more
appropriate in all cases (Oldﬁeld and Appleby, 1984). However, in this study, the
dates calculated with the CF:CS (Gratiot, Cass and Higgins Lakes) and SCF:CS (Gull
and Elk Lakes) models corresponded well with the dates from the CRS model, with
the dates generally no greater than 5 years apart in sediments younger than 1900, with
smaller differences in the most recently deposited sediments. All data were graphed
with CF:CS or SCF:CS calculated dates. All dates older than 1800 were determined
ﬁom data extrapolation assuming constant sedimentation rates, and should be

considered estimations.

94

Another factor that needed consideration in interpreting the data is sediment
focusing. This is the process by which ﬁne-grained sediment and associated
contaminants are redistributed laterally by currents and episodic storm events from
shallow nearshore (erosional) regions to more quiescent (depositional) areas (Davis
and Ford, 1982; Eisenreich et al., 1998). Sediment focusing occurs in different
intensities in lakes, making it difﬁcult to compare chemical concentrations, trends, and
rates of accumulation among lakes. However, a focusing factor can be used to
account for this sediment movement. For example, the total mass of an
anthropogenically-derived contaminant in a sediment core is known as the inventory
of the contaminant. Dividing by a focusing factor, the inventory can be corrected for
the amount of sediment focusing that has occurred at the collection site. The focusing
factor (FF) is calculated as the ratio of the integrated 21°Pb inventory measured in the
sediment to the expected inventory ﬁom 21(’Pb deposition from the atmosphere (15.5
pCi cm'z; ﬁom Golden et al., 1993):

FF _ 210 Pb inventory
15.5 pCi/cm2

 

Focusing factors for most lakes were near two (Table l), with the exception of

Cass Lake, where the estimated focusing factor is six.

95

Chemical sediment chronologies

There are two major sources of sediment in a lake: watershed (terrestrial) input
and within lake production (organic and carbonate) and these sources appeared to vary
greatly among the ﬁve different lakes, due to both the local geology/soils and the
trophic status of the lake. The four lakes located in the Lower Peninsula were rich in
calcium carbonate, but differed in the amount of organic matter. Cass Lake sediments
were enriched in organic matter and tending toward a eutrophic state, and Gull and Elk
Lakes appeared to be oligotrophic, and were poor in organic matter. Higgins Lake is
less enriched in calcium carbonate than the other Lower Peninsula lakes. Gratiot
Lake, located in the Upper Peninsula, has much lower calcium carbonate
concentrations, and much higher concentrations of terrestrial elements (e. g. aluminum,
iron).

Because of these differences, chemical sediment chronologies based on
absolute concentrations could not be compared among the lakes. For example, if over
time there is an overall increase in total sedimentation rate, but the rate of
anthropogenic input remains the same, the anthropogenic input will be diluted by the
increased amount of sediment, and concentration patterns in the sediment core will
appear to go down. This problem is accounted for by examining accumulation rates
(pg/mzly), rather than concentrations.

To determine anthropogenic accumulation rates for metals, background
concentrations (typically estimated ﬁom pre-1800s samples) were subtracted ﬁom
total concentrations for each sediment increment. It is generally considered that these

background concentrations/accumulation rates remain constant throughout time

96

(Figure 3). However, watershed inputs will tend to have higher natural concentrations
of metals than lake production, and changes in the proportions of these two sources
can cause changes in chemical concentration trends, even if there is no change in
anthropogenic input. Such changes in proportions may be due to natural (e. g., change
in erosion rates due to climate change) or anthropogenic causes (e.g., increased

erosion due to logging).

Copper Accumulation Rate (uglmzly)

0.00 0.20 0.40 0.60 0.80 1.00 1.20
2000- - ~~ -

 

    

 

19501

1900‘

   

E
>2 1850i Accumulation Rates
13004 ' + Pre-1800 background
4" Watershed background

1750i i + Total Flux

 

 

 

1700 . r a ﬂ . t
Figure 3. Copper accumulation rates; total, pre-1800 background values, and

watershed (aluminum corrected) background values for Gull Lake. The rate of
anthropogenic copper accumulation would be the difference between total and
background.

For example, Elk Lake shows the most signiﬁcant change in watershed input
(Figure 4) as seen by the increase in the accumulation rate of aluminum from the late
1800s until 1920, then the decrease to the present. Aluminum accumulation rates still
remain above backgrormd accumulation rates. This increase in watershed input is

likely due to the extensive logging and land use change in the watershed (Davis, 1976;

Bearing et al., 1987). Trends in barium, iron, potassium, magnesium, nickel, titanium,

97

Aluminum Accumulah'on Rates (pg/cmz/y)
1 00 200 300 400 500

0
2000—ﬁg; ' U w;
I ) ‘0 :‘J

 

 

 

 

 

 

 

 

 

1950: J5.;-§\\
§ 1900—
>' + Elk
1850- T—‘T Gratiot
- ‘ Gull
g * Higgins
1800 < i Cass
1750 4 4 4 i 4 r .
0 60 120 180 240 300

Figure 4. Aluminum accumulation rates over time in ﬁve inland lakes. For Cass
Lake accumulation rates use the upper scale, for all other used the lower scale.

and vanadium all appear to be inﬂuenced primarily by watershed input in Elk Lake.
Gull Lake also shows an increase in aluminum accumulation, increasing ﬁ'om 1800 to
1940, remaining high until 1968, and then decreasing until the present. This trend is
different than that seen in Elk Lake, but is still probably related to logging and other
land use changes, including extensive land development around the lake. In Gull
Lake, potassium, titanium, and vanadium appear to be related to the changes in
watershed input. There is no clear cause for the patterns of aluminum accumulation in
Cass, Higgins and Gratiot Lakes, but they are likely related to land use changes and
changes in lake productivity. In Cass Lake, cobalt, potassium, magnesium, nickel,
titanium, and vanadium appear to be related to changes in watershed input; in Higgins
Lake, barium and magnesium strongly correlate with aluminum, and cobalt, iron and

vanadium weakly correlate; in Gratiot Lake, cobalt, magnesium, nickel and titanium

98

strongly correlate, and vanadium and uranium correlate weakly. It is clear that a
variety of elements might be used to characterized inputs to a lake from the watershed,
but aluminum appears to be the most consistent indicator.

By comparing selected metals to aluminum, it may be possible to assess the
effects of the changing proportions of watershed inputs and within lake production on
the metal concentrations in the sediment. This method assumes that the
metal:aluminum ratio of material washed in ﬁom the watershed remains the same over
time (Bruland et al., 1974). A watershed background is determined for each sample
separately based on the aluminum accumulation rate. The watershed background
metal concentration is estimated by multiplying the aluminum concentration in the
sediment times the average metal/aluminum ratio for pre 18003 sediments. Using this
approach, anthropogenic input is considered to include direct human inputs and
atmospheric deposition, but does not include changing input of sediment ﬁom natural
or human processes. Anthropogenic accumulation rates have been calculated using

the following equation:

FCAccumjlpg/mzyr) = (C9 ij x104)! FF 1

FCAccum’j = focusing corrected sediment accumulation rate for ith metal at j'h lake,

Cf’. = watershed background corrected concentration of 1'11 metal in the sediment

(lug/g) ati'h lake.

W]. = mass sedimentation rate (g/cm2 yr) in jth laked based on 2|on dating,

FF]. = sediment focusing factor for j"I lake

99

Figures 5a and b are graphs of the changes in the focusing corrected
anthropogenic sediment accumulation rates over time for lead and copper respectively.
Although, watershed corrections could not be made for Cass Lake because the cores
only reached the 1970s, Cass Lake is included for relative comparisons.

Signiﬁcant anthropogenic accumulations for lead in all of the lakes begin
around 1875. Concentrations peak around the middle 1970s in all of the lakes.
Concentration peaks are highest for Cass Lake followed in order by Gull, Higgins,
Elk, and Gratiot lakes. Anthropogenic accumulation rates for lead in the lakes are
decreasing to the present.

Patterns for the temporal changes of copper accumulation in the lakes are not
as regular as they are for lead. Anthropogenic copper accumulation begins earliest in
Higgins Lake (1850) and most recently in Gull (1940) and Elk (1930) lakes.
Signiﬁcant accumulations begin in Gratiot Lake around 1900. Accumulation rates in
Gull, Higgins and possibly Elk peak around the mid 19705. Copper accumulation
rates in Gratiot Lake were highest around the late 1950s and also in recent times. Cass
Lake has the highest over all rates while those of Higgins, Gull, and Gratiot lakes are
similar. Elk has the lowest accumulation rates for copper.

The coincidence of the peak accumulation rates for lead suggests that a
regional source has inﬂuenced the anthropogenic loading of lead, but such a
conclusion is not clear for copper, which might be a combination of both a regional
source and local inﬂuences. The relative magnitudes of the peak concentrations for

lead in the mid 1970s is consistent with a regional gradient for lead deposition ﬁom

100

Lead Accumulation Rates (pg/mzly)
0 30000 60000 90000 120000

 

 

  

Year

 

‘ ‘ +Higgins L

1850 gv—ﬁﬂi ‘ *guuﬁo‘ ‘
O ra

1800 4 '5'“ i— 4

1 BCass l

 

 

 

 

 

 

 

 

 

 

 

 

 

 

a
1700
Copper Accumulation Rates (uglmzly)
6000 8000 10000 12000
2000 ' ' 4 ‘
1950
1900
a
5‘3 1850
+Higgins
W "-7. .- 77 .- *Gull . , ,, , ..
1800 i 0 Gratiot
- , ' Elk
1750 8 - ,_ - - P- ., - W fl.” ECass 7.7. H i. ..
l , ‘ b
1700? ' l T l ' l '
0 2000 4000 6000 8000

Figure 5. Graphs showing the metal accumulation rates for study lakes a. lead, b.
copper. The top axis in b. is for Cass Lake only.

101

the atnrosphere. Concentrations might be expected to be higher to the south (inﬂuence
from the great Chicago area, for example) and lower to the north (Eisenreich et al.,
1986). Such a regional gradient is not apparent for copper since Gull, Higgins, and
Gratiot lakes appear to have similar accumulation rates.

To gain further insight into the patterns of the anthropogenic accumulation
rates for lead and copper, the data from Figures 58 and b were replotted in Figures 6a
and b, respectively. In these plots the accumulation rates determined from each
increment were normalized to the highest accumulation rate in the particular core.
Such plots allow the exploration of how the temporal trends of the accumulation rates
compare among the lakes. For example, if the source for contaminant was the same
over a region, the trajectories for the change in accumulation rates among the lakes
should overlap.

As might be expected, the peak accumulation rates for lead reach the
normalized value of one in the mid 19705 (Figure 6a). The graph clearly shows the
time of onset of the anthropogenic additions of lead to be the same among the lakes.
Furthermore, except for Gratiot Lake, all of the lakes have a similar pattern of increase
to the peak rates in the mid 1970s and their trajectories overlap. The difference in
Gratiot Lake may be due to contribution of lead ﬁ'om smelting operations or the
inﬂuence of bioturbation, which can smear the sediment record (Robbins, 1982). The
decreasing trajectories in lead accumulation rates after the peak are not the same
among the lakes, however. It can be seen in recent sediments that Gull Lake has
returned to rates 15% of the peak with the other lakes following the order Cass (25%),

51k (45%), Higgins (70%), and Gratiot (70%).

102

Normalized Lead Accumulation Rates

 

 

 

 

 

 

 

‘ + Higgins

, _ _ . i Gull
185° ’ O Gratiot

' I Elk

B Cass a
1800 - _ — —*
Normalized Copper Accumulation Rates
0.0 0.2 0.4 0.6 0.8 1.0

2000 ~

 
    

 

* Higgins

'* Gull

, o Gratiot
I Elk

H Cass b

 

 

 

 

 

Figure 6. Graphs showing the metal normalized rates for study lakes. Accumulation
rates are normalized to the highest value in a core. a. lead, b. copper.

103

Figure 6b more clearly shows how the timing for the onset of copper
accumulations in the lakes differed. Additionally, once anthropogenic accumulations
of copper begin, the rates of increase among the lakes were not as similar as they are
for lead. Elk, Cass, Higgins, and Gull lakes have peaks in accumulation rates in the
mid 1970s. Elk Lake accumulation rates in recent sediments remain similar to peak
rates, unlike Cass, Higgins and Gull Lakes. Post mid 1970s there is very little
similarity in the rates of change for copper accumulation among the lakes. For
example, Gull Lake has returned to 40% of the peak with the other lakes following the
order Higgins (62%), Cass (78%), Elk (99%), Gratiot (100%).

For lead, the overlapping peak concentrations and trajectories of the temporal
change of accumulation prior to the mid 1970s further supports a regional source for
loadings in this portion of the Great Lakes. The consistent onset of the occurrence of
anthropogenic lead implies a rapid distribution ﬁom the source. These observations
are consistent with the atmosphere as being the dominant source for lead prior to the
mid 1970s. The decrease in accumulation rates after this time is related to
environmental legislation and the subsequent removal of lead from gasoline (Evans
and Dillon, 1982; Eisenreich et al., 1986; Graney et al., 1995). The lack of overlap of
the normalized accumulation trajectories after the mid 1970s is consistent with the
hypothesis that the source for lead in the environment is changing ﬁom regional
inﬂuences to more local or watershed scale inﬂuences.

Although atmospheric deposition has been suggested as a source for copper
(Eisenreich et al., 1986), the temporal relationships of the accumulation trajectories

are not as consistent with the hypothesis as they are for lead. The impact ﬁom humans

104

on copper loadings to the environment is clear, but a regional source does not appear

to be evident from the above analysis.

Anthropogenic inventories

To further examine the role of the atmospheric inﬂuence on lead and copper
accumulations in the sediment of the lakes, focusing corrected anthropogenic
inventories were compared. Anthropogenic inventories are a measure of the total
mass in a core of contaminant input by humans per unit area (uglcmz).

InVi (118/0312) = (21 lewd. x «14) x p x (DD/FF)

where:

Invi = focusing corrected inventory for itll metal in the sediment core (pg/cmz),

Csed“ = watershed background corrected concentration of metal in sediment (ug/ g dry
wt),

0 = Porosity.

p = sediment dry density (g/cm3),

d = thickness of sediment increment (cm),

i = number of depth increment, and

F Fj = focusing factor at jm core site.

If atmospheric deposition is the major source of a chemical to lakes, then
focusing and watershed corrected anthropogenic inventories should be the same in all
lakes or show a regional pattern that reﬂects the distribution of somoes of chemicals to

the atmosphere. If inventories for a contaminant are different among all lakes and no

105

regional trend is evident, then local sources dominant contaminant loadings in the
lakes. Regional patterns in inventories could indicate the roles of land use in
contaminant loadings.

The anthropogenic inventories of lead (Figure 78) among sampled inland lakes
are not equal, indicating that if the atmosphere was the source for lead, then the lead
concentrations in the atmosphere were not well mixed. The highest inventory is in
Gull Lake, which is located relatively close to the greater Chicago area, and the lowest
inventory is in Gratiot Lake, which is located relatively far away from major sources
of lead to the atmosphere. This regional north-south trend is consistent with results
from Lake Michigan, with the highest loading of lead occurring in its south basin
nearest Gull Lake and lowest in the northern basin near Elk Lake (Edgington and
Robbins, 1976). Figure 7a also shows the inﬂuence of correcting the inventory value
by using either pre 1800 background metal concentrations or watershed background
concentrations to determine anthropogenic concentrations in the sediments.

The pattern of inventories for copper (Figure 7b) is not similar to that of lead.
Gull and Higgins lakes have similar inventories, Elk Lake the lowest, and Gratiot the
highest. Gratiot Lake, however, is located on the Keweenaw Peninsula of Michigan, a
region of Michigan characterized by copper mining activities (Kolak et al., 1999).
Although copper was not mined in the watershed of Gratiot Lake, the lake’s proximity
to the mining activities made it susceptible to emissions (e.g., smelters), resulting in
the relatively high inventories. The lower copper inventory values for Elk Lake than

Higgins and Gull lakes may indicate some form of a regional gradient similar to lead.

106

The watershed corrected inventories for the inland lakes are around the average

inventory values in Lake Michigan for lead (Long et al., 1995) and copper (Kolak et

al., 1999) and shows that the environmental record from inland lakes can be compared

to the record from large lakes.

400

350
300
250
200
1 50
1
50
0

Pb Inventory (pg/cmz)

AUTO)
000

N
0

Cu Inventory (pg/cmz)
ES 8

O

 

 

 

Type of Background

Correction
I none
I: pre 1880s

7 I watershed

Gull Higgins Elk Gratiot
Lake
Type of Background Correction b

 

r.

 

    

 

I none in pre 188OSI watershed

 

 

 

 

 

Higgins
Lake

 

Elk Gratiot

Figure 7. Graphs showing the metal inventories as a function of background
correction for study lakes: 8. lead, b. copper. Lake Michigan (LM) values for lead are
from Long et al. (1995) and for copper from Kolak et a1. (1999).

107

Temporal and spatial patterns of accumulation

The previous analysis indicated that both dominant (regional) and local
(watershed) factors are involved in the accumulation of lead and copper in the
environment and that the relative contributions of these factors have changed over
time. Although spatial analysis of anthropogenic inventories is an important approach
that gives insight into possible regional and watershed inﬂuences, inventories cannot
be studied over time. Changes in the spatial pattern of accumulation rates over time
may however give insights into changes in regional and watershed inﬂuences on lead
and copper loadings to the environment.

Figures 8 a and b show the temporal and spatial changes in lead and copper
anthropogenic accumulation rates in the study lakes in Michigan. Metal accumulation
rates are shown for each decade starting with 1890. The rates are averages for a ten-
year period, ﬁve years before and ﬁve years after the decade. For example, the rates
for 1990 are calculated as the average ﬁom 1985 to 1995. The purpose of using
averages rather than the value for a particular year is that l) a particular year may not
be represented by a sample taken and would have to be estimated from the shape of
the sediment accumulation curve (e. g., Figure 6a), and 2) there is some degree of
inaccuracy in the estimates of the sediment ages. Considering the rates of change
observed in the sediment for metal accumulation, being off by only a few years can
signiﬁcantly affect accumulation values being compared. Using averages minimizes
both of these issues.

At the onset of signiﬁcant inputs of anthropogenic lead to the environment

(e.g., 1890), a regional pattern of higher accumulations rates to the south and lower to

108

 

 

 

 

     
  

  

.
I

 

 

11:7: '1? 1
: =1
4

1 890 G Gull Lake

I" T” “ T "F“ H Higgins Lake

l E Elk Lake

. ' Gr Gratiot Lake
a

MILE—-—

 

Pb (uglkmz/y)

 

 

 

  

 

 

 

 

_. 1920 _ g_

6 I
E l
i .
3’ G Gull Lake
0 H Higgins Lake

E Elk Lake

Gr Gratiot Lake

b

 

Figure 8. Graphs showing decadal metal accumulation rates for study lakes; 8. lead,
b. copper.

109

the north is evident. By the 1930s, the pattern is well developed and is maintained
through to the 1980s. By 1990, however, the pattern appears to break down. Gull
Lake no longer has the highest rates of lead accumulation and Gratiot and Elk are
similar. Higgins Lake in 1990 has the highest rate of lead accumulation.

In contrast, no clear pattern emerges for copper accumulation until 1960s
(Figure 8b). Prior to that time Higgins Lake and eventually Gratiot Lake had the
highest rates of accumulation. In 1960, copper accumulation in Gratiot remains high,
however, an accumulation pattern similar to lead is seen for Gull, Higgins, and Elk
lakes. Gull has the highest rates of copper accumulation and Elk the lowest. Although
the rates are slightly lower, this pattern remains in 1990.

We interpret the changes in the spatial and temporal patterns of lead
accumulation to be due to change from more regional inﬂuences on lead loadings
(e.g., atmosphere and leaded gasoline emissions) to more local inﬂuences (e.g.,
watershed and population density). For copper, local inﬂuences (e.g., copper mining)
play a role. If the pattern of copper accumulation in Gull, Higgins, and Elk lake
represent a regional inﬂuence, the cause of this inﬂuence is unclear and most likely it

is not related to the atmosphere.

Population effects on lead and copper accumulation

To explore the role of watershed scale inﬂuences, such as population density
(e.g., Callender and Rice, 2000), population densities for the watershed surrounding a
lake were compared to accumulation rates within the lake. Areas in the watershed

ranged ﬁom 11.6 to 120.8 km2 (Table 2), and populations ranged from 7 to 5,830 with

110

population densities of 0.4 to 946 people/m2. Because of the assumptions made in
determining population within the watershed (e. g., that population is spread evenly
throughout all areas that are not state owned or water), there is clearly error present in
the watershed populations. The most signiﬁcant source of error is likely from not
accounting for the tendency of people to live near water, which ‘would result in an

underestimation of population density within the watershed.

Table 2. Areas, populations, and population densities for watersheds of the ﬁve

 

 

study lakes in Michigan.
Watershed Watershed area Watershed Population density
(kmz) population (per kmz)
Cass Lake 11.6 5830 946
Elk Lake 120.8 1280 20
Gratiot Lake 26.3 7 0.4
Gull Lake 14.7 326 52
Mus Lake 116.5 2561 75

 

Additional error was incurred when delineating the watersheds, particularly for
lakes with little topography (e. g., Cass Lake), but this source of error was probably not
signiﬁcant compared with the error inherent is assuming an even distribution of people
throughout each census tract. Finally, no attempt was made to account for seasonal
increases in population, which may be signiﬁcant around some of the lakes (e.g.,
Gratiot Lake population may increase almost tenfold, B. Hay, pers. com., 2000).
However, the results still do reﬂect the observed relative density of population around
lakes, with Cass having signiﬁcantly higher development around the lake, and Gratiot

Lake being the least developed.

111

Population densities for each watershed in 1990 are plotted against 1990
accumulation rates for lead and copper in Figures 9 a, b, c, d). In Figure 9a, all ﬁve
lakes are included and the trend of the data clearly show that there is a relationship
between population density and lead accumulation rate, but this relationship does not
appear to be linear. Cass Lake, near Detroit, has the highest population density and
highest rate of lead accumulation. Recent lake sediments have been highly disturbed
by waste additions from a treatment plant, perhaps causing the unusually high
sedimentation rates found for the lake. The non-linearity between Cass Lake and the
other lakes may be related to underestimated population densities in the other four
lakes, to a non-linear relationship between population density and accumulation rate,
or to additional sources of lead and copper not related to population density present in
some of the lakes. Replotting the data without Cass Lake (Figure 9b) shows that there
is a linear relationship between population density and lead accumulation rate. The
intercept (e.g., zero population density) of about 5,000 ug/mZ/yr for lead accumulation
might be considered the current regional lead accumulation rate.

Accumulation rates of copper exhibited a different pattern, with Gratiot Lake
appearing to have a higher than expected accumulation rate (Figure 9 c, d). These
high rates again show the importance of the local copper mining activities and,
therefore, the higher input rate not related to population density would be expected.
Overall, there is no clear relationship between 1990 population density and copper
1990 accumulation rates.

Because of the error present in the population density estimations, and because

of the small sample size of ﬁve lakes, great care must be taken in interpreting the data.

112

Pb (pg/WM)

Pb (HQ/mzlyr)

CU (lug/WM)

CU (HQ/WM)

 

60000
50000 ‘
40000 ‘
30000 “
20000 ‘
10000 (

 

 

 

0
0.0

200.0 400.0

f

800.0

T

600.0 1000.0

 

30000
25000 4
20000 ‘

 

0.0 20.0 40.0 60

y = 177.63x + 4935.1

I I

T

.0

   
  
  

R2 = 0.9841 b

 

f

80.0 100.0 120.0 140.0

 

12000
10000“
80001
60001
4000*
20001

 

 

 

0
0.0

T

200.0 400.0

600.0 800.0 1000.0

 

2500

1500 ‘
1000 ‘
500 ‘

 

2000 ‘0

y = 6.531x + 1333.4

X

A
4‘-

R2 = 0.1221
(1

 

 

0

f

0.0 10.0 20.0 30.0 40.0 50.0 60.0 70.0 80.0

Population density

Figure 9. Graphs showing the metal accumulation rates versus watershed population
densities for study lakes; a. lead with Cass Lake, b. lead without Cass Lake, c. copper,
d. copper without Cass Lake. Lake symbols: (6) Cass, (I) Higgins, (A) Gull, (X)

Elk, (O) Gratiot.

113

It is crucial to realize that because Cass Lake has a signiﬁcantly higher
population density, and signiﬁcantly higher rates of heavy metal input, it is possible to
have statistically signiﬁcant results when there appears to be little correlation between
population density and accumulation rates for the other four lakes. Therefore,
although preliminary conclusions may be drawn from these data, further samples are
needed to verify the results.

The rates of input of lead, but not copper to inland lakes in Michigan are
related to population density within the watershed. The relationship may be due to a
direct relationship between the two factors (e. g., lead in sewage where the amount of
sewage would be directly related to the number of people), or due to an indirect
relationship (e. g., population related to amount of industry or road density).

Whether the relationship is direct or indirect, the results indicate that local
processes are important in the current chemical loading to lakes. This supports the
earlier conclusion that the dominant some for lead has changed in the last decade
from regional atmospheric sources to more localized sources (Dillion and Evans,
1982; Evans and Dillon, 1982; Eisenreich et al., 1986). These results are similar to the
results of Callender and Rice (2000), who found a relationship between population
density and concentrations of lead and zinc in ﬁve reservoirs located along a rural-

urban gradient.

Summary
Lead and copper concentrations in sediment cores from ﬁve inland lakes in
Michigan were studied to test the hypothesis that the spatial pattern of chemical

accumulation rates provides insight into the cause(s) for the historical and continued

114

loadings of contaminants to the environment. The data show that the loadings of lead
and copper to the environment are generally decreasing ﬁom historic high loadings in
the mid 19708, but have not decreased to background values. However, the data do
show that as loadings ﬁom dominant (regional) sources for chemical loadings
decrease (e. g., atmospheric deposition in the case of lead), new patterns provide
insight into causes for continual contaminant loadings (e.g., population density).

The data also show the universal response of the watersheds to the onset of the
anthropogenic loadings of lead, but not that of copper. Individual watersheds had
different onset times for anthropogenic copper loadings. The magnitude of recent
loadings of lead to the environment can be related to population density. The
magnitude of recent COpper loadings reﬂect, in part, a spatial pattern that is similar to
the regional pattern for lead during the mid 1970s. This pattern for lead could be
related to a regional gradient of lead concentrations due to atmospheric deposition, but
the same relation cannot be made for copper. The cause for this spatial pattern for
copper is not clear.

Finally, it is clear that sediment chronologies provide important insights into
contaminant loadings to the environment over various temporal and spatial scales. We
also suggest that using an approach that combines both spatial and temporal analysis
gives the most robust information about the history and current state of chemical

loadings to the environment.

115

Acknowledgements
This study was made possible by a grant from the Michigan Department of
Environmental Quality. The lakes discussed in this report were sampled on the
Environmental Protection Agency's Research Vessel Mudpuppy, with the assistance of
its captain and crew. We would like to thank the MSU Physics Machine Shop in

fabricating the extruding system used for this research.

116

References

Balogh, S.J., Engstrom, D.R., Alemendinger, J .E., Meyer, ML. and Johnson, K.D.,
1999. History of mercury loading in the Upper Mississippi River reconstructed
from sediments of Lake Pepin. Environ. Sci. Technol. 33, 3297-3302.

Birch, L., Hanselmann, K.W. and Bachofen, R., 1996. Heavy metal conservation in
Lake Cadagno sediments: historical records of anthropogenic emissions in a
meromictic alpine lake. Water Resources 30, 679-687.

Boyle, J .F ., Rose, N.L., Bennion, H., Yang, H. and Appleby, P.G., 1999.
Environmental impacts in the J ianghan Plain: evidence from lake sediments.
Water, Air, Soil Pollut. 112, 21-40.

Brown, E.T., Callonnec, LL. and German, CR, 2000. Geochemical cycling of redox-
sensitive metals in sediments from Lake Malawi: A diagnostic paleotracer for
episodic changes in mixing depth. Geochim. Cosmochim. Acta 64, 3515-3523.

Bruland, K.W., Bertine, K., Koide, M. and Goldberg, ED, 1974. History of metal
pollution in Southern California costal zone. Environ. Sci. Technol. 8, 425-
432.

Callender, E. and Rice, K.C., 2000. The urban environmental gradient: Anthropogenic
inﬂuences on the spatial and temporal distributions of lead and zinc in the
environment. Environ. Sci. Technol. 24, 232-238.

Catallo, W.J., Schlenker, M., Gambrell, RP. and Shane, B.S., 1995. Toxic chemicals
and trace metals from urban and rural Louisiana lakes: recent historical proﬁles
and toxicological signiﬁcance. Environ. Sci. Technol. 29, 1439-1446.

Davis, B.M., 1976. Erosional rates and land use history in southern Michigan.
Environmental Conservation 3, 139-148.

Davis, ME. and Ford, MS, 1982. Sediment focusing in Mirror Lake, New
Hampshire. Limnology and Oceanography 27, 137-150.

Dealing, J.A., Hakansson, H., Leidberg-Jonsson, B., Persson, A., Skansjo, S.,
Widholm, D. and El-Daoushy, F., 1987. Lake sediments used to quantify the
erosional response to land use change in southern Sweden. Oikos 50, 60-78.

Dillion, PJ. and Evans, RD, 1982. Whole-lake lead burdens in sediments of southern
Ontario, Canada. Hydrobiologia 91, 121-130.

117

Edgington, D.N. and Robbins, J .A., 1976. Records of lead deposition in Lake
Michigan sediments since 1800. Environ. Sci. Technol. 10, 266-274.

Eisenreich, S.J., Capel, P.D., Robbins, J .A. and Bourbonniere, R., 1998. Accumulation
and diagenesis of chlorinated hydrocarbons in lacustrine sediments. Environ.
Sci. Technol. 23, 1116-1126.

Eisenreich, S.J., Metzer, NA. and Urban, N.A., 1986. Response of atmospheric lead
to decreased use of lead in gasoline. Environ. Sci. Technol. 20, 171-174.

Evans, RD. and Dillon, P.J., 1982. Historical changes in anthropogenic lead fallout in
southern Ontario, Canada. Hydrobiologia 91, 131-137.

Fritz, S.C., Kingston, J .C. and Engstrom, DR, 1993. Quantitative trophic
reconstruction from sedimentary diatom assemblages: a cautionary tale.
Freshwater Biology, 1-23.

Gottgens, J.F., Rood, B.E., Delﬁno, J.J. and Simmers, B., 1999. Uncertainty in
paleoecological studies of mercury in sediment cores. Water, Air, Soil Pollut.
110, 313-333.

Graney, J .R., Halliday, A.N., Keeler, G.J., Nriagu, J .O., Robbins, J .A. and Norton,
S.A., 1995. Isotopic record of lead pollution in lake sediments from the
northeastern United States. Geochim. Cosmochim. Acta 59, 1715-1728.

Harrington, J.M., Laforce, M.J., Rember, W.C., Fendorf, SE. and Rosenzweig, RF,
1998. Phase associations and mobilization of iron and trace elements in Coeur
d'Alene Lake, Idaho. Environ. Sci. Technol., 650-656.

Hermanson, M.H., 1998. Anthropogenic mercury deposition to arctic lake sediments.
Water, Air, Soil Pollut. 101.

Hewitt, AD. and Reynolds, C.M., 1990. Dissolution of metals ﬁ'om soils and
sediments with a microwave-nitric acid digestion technique. Atomic
Spectroscopy 11, 1425-1436.

Heyvaert, A.C., Reuter, J.E., Slotton, D.G. and Goldman, CR, 2000.
Paleolimnological reconstruction of historical atmospheric lead and mercury
deposition at Lake Tahoe, California-Nevada. Environ. Sci. Technol. 34, 3588-
3597.

Iskander, I.K. and Keeney, DR, 1974. Concentration of heavy metals in sediment
cores from selected Wisconsin Lakes. Environ. Sci. Technol. 8, 165-170.

118

Jaquet, J .-M., Davaud, B., Rapin, F. and Vemet, J.-P., 1982. Basic concepts and
associated statistical methodology in the geochemical study of lake sediments.
Hydrobiologia 91.

Johnson, MC. and Nicholls, K.H., 1988. Temporal and spatial trends in metal loads to
sediments of Lake Simcoe, Ontario. Water, Air and Soil Pollution 39, 337-354.

Kang, W., Trefry, J .H., Nelsen, TA. and Wanless, HR, 2000. Direct atmospheric
inputs versus runoff ﬂuxes of mercury to the lower Everglades and Florida
Bay. Environ. Sci. Technol. 34, 4058-4063.

Kemp, A.L.W. and Thomas, R.L., 1976. Cultural impacts on the geochemistry of the
sediments of Lakes Ontario, Erie and Huron. Water, Air and Soil Pollution 5,
469-490.

Kolak, J.J., Long, D.T., Kerfoot, W.C., Beals, TM. and Eisenreich, S.J., 1999.
Nearshore versus offshore copper loading in Lake Superior sediments:
Implications for transport and cycling. J. Gt. Lakes Res. 25, 611-624.

Long, D.T., Eisenreich, SJ. and Swackhammer, D.L., 1995. Metal Concentrations in
Sediments of the Great Lakes, Great Lakes Protection Fund.

Oldﬁeld, F. and Appleby, P.G., 1984. Empirical testing of 210Pb-dating models for
lake sediments. In: E.Y. Haworth and J .W.G. Lund (Editors), Lake Sediments
and Environmental History. University of Minnesota Press, Minneapolis, pp.
93-124.

Ramanathan, A.L., Subramanian, V. and das, B.K., 1996. Sediment and heavy metal
accumulation in the Cauvery basin. Environ. Geol. 27, 155-163.

Reimann, C. and de Caritat, P., 1998. Chemical Elements in the Environment:
F actsheets for the Geochemist and Environmental Scientist. Springer-Verlag,
Berlin.

Robbins, J .A., 1982. Stratigraphic and dynamic effects of sediment reworking by
Great Lakes zoobenthos. Hydrobiologia 92, 611-622.

Spiethoff, HM. and Hemond, HR, 1996. History of toxic metal discharge to surface
waters of the Aberjona Watershed. Environ. Sci. Technol. 30, 121-128.

119

CHAPTER 5

Yohn, S.S., Long, D.T., Fett, J .D. and Patino, LC, 2004. Regional versus local
inﬂuences on lead and cadmium loading in the Great Lakes region. Appl. Geochem.
19,1157-1175.

120

CHAPTER 5
REGIONAL VERSUS LOCAL INFLUENCES ON LEAD AND CADMIUM
LOADING TO THE GREAT LAKES REGION

Abstract

Environmental legislation has reduced the anthropogenic loadings of lead and
cadmium to the Great Lakes region over the last three decades. However, the
accumulation rates of these metals still remain above background values. Because
environmental legislation was targeted at major somces (e. g., lead in gasoline) whose
inﬂuence on the environment was on a regional scale, local sources (e.g., watershed
scale) for the metals may now play a more signiﬁcant role. The relative importance of
regional versus local scale inﬂuences on metal inputs to the environment is poorly
understood. In this study, sediment chronologies of lead and cadmium were examined
from 12 inland lakes that cover the broad geographic area of the State of Michigan.
These chronologies were compared temporally and spatially and to watershed
population densities and metal production records to gain an understanding of local
and regional inﬂuences on metal inputs to the Great Lakes region. Results show that
anthropogenic lead loading during the 1930s and 1970s was dominated by regional
sources, such as coal burning and use of leaded gasoline. Current loadings are now
more related to local inﬂuences such as watershed population densities, rather than
atmospheric deposition. Anthropogenic cadmium loadings to the Great Lakes region
have been dominated by both regional and local sources over time. Lead may also
have shown the inﬂuence of local sources over time, if the inﬂuence of emissions from

gasoline had not been present. This work shows that lead and cadmium loadings in

121

the Great Lakes region are strongly related to watershed population densities,

however, the speciﬁc sources and pathways for the metal cycling are unclear.

Introduction

Over the last three decades signiﬁcant progress has been made in reducing the
levels of metals released to the environment. This reduction of metal inputs has been
recorded in lake sediments. For example, the removal of lead ﬁom gasoline has
signiﬁcantly reduced lead concentrations in lake sediments (Graney et al., 1995;
Callender and vanMetre, 1997). However, despite the reduction of major contaminant
sources, concentrations of contaminant metals in sediments, air and water still remain
above natural levels (Sanudo-Wilhelmy and Gill, 1999; Simonetti et al., 2000a;
Alfaro-De la Torre and Tessier, 2002).

Even as point sources for metals are regulated and emissions ﬁom these
sources decline, humans continue to alter the environment in ever more signiﬁcant
ways (e.g., urban sprawl) which will have unknown impacts on metal inputs to the
environment. Therefore it is essential to understand metal loadings to both assess the
effects of environmental legislation, and to determine the emerging sources for metal
pollution. This study compares patterns of metal accumulation in inland lakes over
space and time to try to determine the current and historical inﬂuences on lead and
cadmium loadings to the Great Lakes region. Michigan has a surface area of
approximately 151,586 kmz, extends seven degrees of longitude and six degrees of
latitude (Encyclopadia Britannica Premium Service, 2003), and has a large number

of lakes, making it an ideal location for this study.

122

Humans have signiﬁcantly altered the global cycles of both lead and cadmium
(Graney et al., 1995; Simonetti et al., 20008; Alfaro-De la Torre and Tessier, 2002).
Atmospheric levels of lead have been enriched signiﬁcantly in North America ﬁom
the mid-1800s to the present (Shahin et al., 2000). Lead isotope ratios in the Great
Lakes area have been used to show that the major source of lead to the atmosphere in
the early 1900s was coal burning, from 1930-1980 leaded gasoline was dominant, and
current somces are mixed and less clear (Graney et al., 1995). In the United States,
automobile emissions to the atmosphere in 1993 were only 1% of those in 1970
(Bollhofer and Rosman, 2001). Lead also has local sources, and may be found in
wastewater effluents and stormwater runoff (Callender and Rice, 2000). Som'ces for
cadmium to lakes have been less extensively studied. Known sources to the
environment include both regional sources (e. g., coal combustion, smelters, steel
mills) and local sources (e.g., tire wear, sewage sludge and wastewater) (Reimann and
de Caritat, 1998). Lead and cadmium are of particular interest because these elements
are known health hazards for both humans and wildlife (Catallo et al., 1995; Callender
and vanMetre, 1997), and are present in concentrations above the threshold
concentration level (below this concentration toxic effects are unlikely to occur:
MacDonald et al., 2000) in several of the study lakes (Y ohn et al., 2002).

A common approach to understanding factors that inﬂuence contamination
loading to the environment is to study one lake closely (Bruland et al., 1974; Evans
and Dillon, 1982; Johnson and Nicholls, 1988; Balogh et al., 1999). However, while
the extensive study of one lake over time provides important information about that

particular lake, the results cannot necessarily be generalized to other lakes in the

123

region. To create a more general model of factors inﬂuencing metal input to lakes,
multiple lakes can be sampled, and the inﬂuence of a particular source determined
over space (Y ohn et al., 2002). These studies usually examine only contemporary
effects (Johnson and Nicholls, 1988; Boyle et al., 1999; Callender and Rice, 2000),
rather than examining the change in patterns over time. Our study examined spatial
patterns, as well as the changes in those spatial patterns over time.

The dominant sources for contaminants to the environment are often difﬁcult
to determine, but spatial and temporal deposition patterns may show if the input of a
metal was locally or regionally inﬂuenced. If a contaminant is regionally inﬂuenced,
it must have an atmospheric pathway which allows the metal to move beyond the local
watershed. Deposition will be controlled by the distribution of sources in the region,
rather than by the presence of the source in the local watershed. A metal with locally
inﬂuenced inputs is affected dominantly by sources within the watershed.

Knowing if inputs of a contaminant are locally or regionally inﬂuenced can be
useful for policy makers. Reducing inputs of a metal that is regionally inﬂuenced
requires legislation on the state or federal level (e.g., reduction of emissions ﬁom
automobiles), while a metal that is locally inﬂuenced may be effected by county and
municipal governments or lakes associations (e.g., reduction of pesticides used on
lawns bordering the lake). However, once it is known if a metal is locally or
regionally inﬂuenced, it is also desirable to determine sources for that metal to the
environment. Ideally, the actual ﬂux from a particular source is known (e.g.,
atmospheric ﬂux or sewer outfall) (Bruland et al., 1974; Evans and Dillon, 1982;

Johnson and Nicholls, 1988; Balogh et al., 1999). Unfortunately, this information is

124

rare, and therefore a quantiﬁable factor that acts as an indicator of that source(s) may
be used instead (e.g., copper mining production records as an indicator of input from
mining) (Kerfoot et al., 1994). Although this factor does not provide the exact
proportion of the input that comes from that source, use of the factor can indicate if a
particular source inﬂuences chemical input into the lake at different time periods.

Om approach was to ﬁrst determine if inputs of lead and cadmium were locally
or regionally inﬂuenced by examining sediment chronologies ﬁom multiple lakes over
time. If sediment chronologies ﬁom multiple lakes have very similar patterns over
time (e.g., concentrations increase from background at similar times and peaks occur
at the same time) then inputs of that metal are regionally inﬂuenced. Secondly, spatial
patterns can be examined for different time intervals. If deposition of a metal is
regionally inﬂuenced then all lakes should have similar inputs, or a spatial gradient
should exist. If lakes in close proximity have very different inputs, this implies that
the metal is inﬂuenced by sources within each watershed (locally inﬂuenced).

Finally, we compared two factors which represent possible sources to the
environment: watershed population densities and industrial production records.
Population densities are representative of multiple sources, including sewage and
storm water discharge, yard waste and other local inputs (Callender and Rice, 2000).
Population densities are also positively correlated to trafﬁc density (Callender and
Rice, 2000), and may be representative of sources related to automobiles and
highways (e. g., tire wear, auto emissions, road maintenance). National production
records (e. g., production of leaded gasoline, or smelter production of cadmium)

provide an estimation of the patterns of output from these industries. They do not

125

report how much of a metal is released into the environment, but if patterns of
production are similar to depositional patterns, this may be a good indication of
source.

An important assumption in using sediment cores as records of environmental
change is that the chemicals of interest are not remobilized and redistributed in the
sediment after deposition. For example, because of its oxidation-reduction properties,
iron can be highly redistributed in the sediment and changes in vertical concentration
often do not reﬂect changes in input (Kneebone and Hering, 2000). On the other
hand, elements such as lead and cadmium are much less inﬂuenced by such processes,
and therefore we will assume that concentrations of cadmium and lead reﬂect changes

in inputs.

Methodoloy

Sediment cores were collected from twelve lakes throughout Michigan ﬁom
1999 — 2001 (Figlue 1, Table l). Lakes in the Lower Peninsula ofMichigan are
located on Pleistocene glacial deposits, which are generally carbonate rich, whereas
Gratiot Lake is located in the Upper Peninsula on part of the Canadian Shield, with
little to no glacial deposits in this region (Door and Eschman, 1970). The lakes vary
greatly in surface and watershed area (Table 1), and were chosen based on depth,
location, and accessibility. Lakes were also chosen to represent a broad range of
human disturbance, from areas of high development (e.g., Cass Lake) to relatively

remote lakes (e.g., Gratiot Lake).

126

  

Mullett

  
    
   

  

Elk

0 Higgins
. Cadillac

ittleﬁeld O

Gull O
Paw Paw

 

Figure l. Lakes sampled in Michigan, USA. For

reference, Gary, Indiana and Chicago, Illinois are located

to the southwest of Michigan.

Table 1. Characteristics of 12 Michigan study lakes. Lake depth is the depth of the
water column above the coring location.

 

 

Lake
surface Lake Watershed Watershed
Location Year area depth area to lake area
Lake (couny) sampled (kmz) (m) (ka) ratio
Cadillac Wexford 2001 4.7 8 48 10.2
Cass Oakland 1999 5.2 37 9.1 1.8
Crystal 8 Benzie 2001 39 50 106 2.7
Crystal M Montcalm 2000 2.9 17 12 4.1
Elk Grand Traverse 1999 31 59 217 6.9
Gratiot Keweenaw 1999 5.8 24 31 5.3
Gull Kalamazoo 1999 8.2 34 61 7.4
Higgins Roscommon 1999 39 41 108 2.8
Littleﬁeld Isabella 2000 0.7 21 17 22.9
Mullett Cheboygan 2001 67 36 718 10.7
Paw Paw Berrien 2001 3.7 28 30 8.0
Washtenaw/
Whitmore Livingston 2001 2.7 20 5.6 2.0

 

127

Four replicate sediment cores from 40 to 58 cm long were taken from the
deepest portion of each lake using a MC-4OO Lake/Shelf Multi-corer (Ocean
Instruments, San Diego, CA) deployed from either the EPA Research Vessel
Mudpuppy or the Michigan Department of Environmental Quality Monitoring Vessel
Nibi. If the cores showed any evidence of disturbance, a new set of cores was
immediately taken. The cores were described and examined for color, texture, and
signs of zoobenthos. Cores were immediately extruded and sectioned at 0.5 cm
intervals for the top 5-8 cm, and at 1 cm intervals for the remainder of the core.

2”Pb was measured on one core from each lake to determine sedimentation
rates, sediment ages and focusing factors (Freshwater Institute in Winnipeg, Manitoba,
Canada). Sediments dates and accumulation rates were calculated using three models:
the constant ﬂux: constant sedimentation model (CF:CS) (Golden et al., 1993), the
segmented CF:CS (SCF:CS) (Heyvaert et al., 2000), and the constant rate of supply
model (CRS) (Appleby and Oldﬁeld, 1978). The evaluation of the models is
described in the results section.

Sediments for metals analysis were frozen for storage, ﬁeeze-dried and
digested with nitric acid in a CEM-MDS-81D microwave (CEM, Matthews, NC; EPA
Method 3051). Standard reference material (NIST RM 8704 Buﬂ‘alo River Sediment)
and procedural blanks were processed. Three replicate digestions were performed on
two samples from each lake. The concentrated-acid digests were ﬁltered through an
acid-washed, distilled-deionized water (DDW) rinsed 0.40 pm polycarbonate ﬁlter

(Nuclepore). Samples were then analyzed using a Micromass Platform inductively

128

coupled plasma mass spectrometer with hexapole technology (ICP—MS—HEX), using a
concentric nebulizer for sample introduction. Due to high concentrations of calcium
in the extracts from some lakes (2,000 mg/kg in solution), all standards were matrix
matched with calcium concentrations similar to those in the extract. Indium and
bismuth were used as internal standards. Sediments were analyzed for a suite of
metals and metalloids, including Mg, Al, K, Ca, Ti, V, Cr, Mn, Fe, Cu, Zn, As, Sr,
Mo, Cd, Ba, Pb, and U. Reproducibility of the reference material for all elements was
better than 15% relative standard deviation, and recovery of cadmium and lead were
greater than 90%. Recoveries of other elements (e. g. aluminum) were much lower
(20%), due to the use of partial digestion.

Spatial data were collected from secondary sources, and were in, or were
projected into the Michigan GEOREF coordinate system: oblique Mercator projection,
datum NAD83, spheroid GRS 1980. Sources for data included the United States
Geological Survey, Michigan Maps and Information (TIGER base data), and
Michigan Department of Natural Resource’s spatial data library. The digital elevation
models (DEMs) used to delineate watersheds were obtained ﬁom the Shuttle Radar
Topography Mission (SRTM) dataset. Census data at the township scale were used
for population density calculations. Population data were collected from 1870-1990.
Watersheds were delineated in the same manner as described in Yohn et al., 2002.
Production data were collected from the Minerals Yearbook, United States

Department of the Interior, and included data available from 1859-1999.

129

Results and discussion

me results

The most recently deposited sediment in a lake may be disturbed, or mixed, by
resuspension of sediment or bioturbation by zoobenthos. The depth of the mixed zone
in lake sediments can be inferred from the depth in which the mph activity remains
relatively constant. While mixing rates are sufficient to homogenize 210Pb, other
chemicals are often not homogenized (Golden etal., 1993; Simpson et al., 2000).
Mixed zones for the study lakes were less than 6 cm deep, with the exception of

Cadillac Lake. Cadillac Lake had a very deep mixed zone, with the top 14 cm having

 

 

 

 

 

 

 

 

 

0.0 p—----——--——~—~-A a .— -— 0.0 —-~—- —- —— .- _. a -——-——»-3,—~a
0.5 — a } l 1.0 L b x
1.0 — .o 2.0 ~ .'
O. O
1.5 ~ 3.0 » .'
2.0 »- ,l 4.0 — 3'
a? 2.5 - 3" 5.0 - . .°
5 3.0 . .o' 6.0 ~ . °
3, 3.5 ,_ .3. 7.0 — '.
E 4.0 . ° J 8.0 l l
2‘ 4.5 ——~~——~~L — .._ ~—~ - —~ — 9.0 h.______.__ ——- g..- .____i
1: 1.503 1.5-02 1.5-01 1.E+00 1.E+01 1.5-03 1.5-02 1.5-o1 1.E+00
g 0.0 __-______.-,_ m— 7 0.0 —————-— —— ___, ——+ -
_ 1.0 r . 1_o e 1'.
g 2.0 h C ..." 2.0 _ d .3
8 3.0 L .0. 3.0 - a.
< 4.0 ~ g' 4.0 L g'
5.0 ~ -‘ 5.0 ’1’
6.0 - .3! 6.0 c
7.0 ~ . s’ . 7.0 ~ g
3.0 ~ ’1. J 8.0 ~
9.0 ~-—-——~—~L—————4—~———-~- 9.0 .
1.503 1.5-02 1.5-01 1.E+00 1.5-03 1.5-02 1.501 1.E+00
Excess 21"Pb (Bq 9“)

Figure 2. 2"’Pb activity (Bq/g) proﬁles for Gratiot (a), Elk (b), Crystal M (c) and
Littleﬁeld (d) lakes.

130

relatively constant 21”Pb activities (Figure 2). Cadmium and lead were not
homogenized in the mixed zones of the study lakes.

Three different models were used in this study to date sediment cores using
unsupported 21°Pb activities, the constant ﬂux constant sedimentation model (CF :CS)
(Golden et al., 1993), the segmented CF :CS (SCF:CS) (Heyvaert et al., 2000) and the
constant rate of supply model (CRS) (Appleby and Oldﬁeld, 1978; Sanchez-Cabeza et
al., 2000). Sedimentation rates in each lake were determined using all models possible
for that lake, and then the models were evaluated to ascertain which was the most
appropriate to use in determining sediment ages. There is no consensus as to which
model is more appropriate in all cases (Oldﬁeld and Appleby, 1984), and several
factors were considered when choosing a model. Visual examination of the 2|”Pb
proﬁle gave some insight into the most appropriate model to be used. The CRS model
is more appropriate for lakes with large mixing zones, and the SCF:CS or CRS models
are more appropriate for lakes with clear changes in sedimentation (Heyvaert et al.,
2000). Additionally, this study uses two other indicators to determine the most
appropriate model to use: proﬁles of cesium-137 ( 137Cs) activity and stable lead
concentration proﬁles. 137C5 is an artiﬁcial radionuclide that was produced by
atmospheric testing of nuclear weapons in the late 19503 and early 1960s, with peak
fallout occurring in the early 1960s (Walling and Qingping, 1992). The second
indicator is the stable lead peak. Stable lead has an historical pattern of deposition that
is very consistent among lakes, with lead concentrations increasing ﬁ'om the mid-
18005 to the early to mid-19705, and decreasing to the present. The peak in lead

concentrations in the mid-19708 due to the removal of lead from gasoline is consistent

131

enough to use for dating veriﬁcation (Callender and vanMetre, 1997 ; Alfaro-De la
Torre and Tessier, 2002).

Gratiot, Higgins, Cass and Paw Paw Lakes all showed a log-linear decay of
21°Pb over depth (Figure 2a), suggesting that sedimentation rates remained constant
over time. These lakes were dated using the CF:CS model. Cass Lake had a very high
sedimentation rate, and the oldest section in the collected core was only 1971 (Table
2). Mullett, Whitrnore, Gull and Elk Lakes also had a log-linear decay of me, but
one or more breaks in the slope were present (Figure 2b). This implies that
sedimentation rates changed over time in these lakes, and therefore they were dated
using the SCF:CS model. The CRS model was most appropriate for Crystal M,
Crystal B and Cadillac Lakes (Figure 20).

2'on in the sediments of Littleﬁeld Lake did not show a log-linear decay with
depth, and could not be dated using me (Figure 2d). It is unlikely that an error in
core collection would cause this 21“Pb proﬁle, but rather this proﬁle indicates that the
sediment or process of deposition was somehow disturbed. Although dating with
21”Pb is not feasible for this lake, we attempted to estimate dates through the
examination of the stable lead proﬁle. The sediment sample with the highest
concentration of lead was assigned the date of 1972 (Eisenreich et al., 1986). The
amount of sediment deposited from the sample with the highest lead concentration to
the surface (year 2000) was determined (g/mz), and divided by 28 years (1972-2000)
to obtain a sedimentation rate of 692 g/mz/y. With this sedimentation rate applied to
the entire core, the peak in lead occurred at the appropriate time, but lead

concentrations decreased towards background much earlier than in other cores.

132

Therefore a separate sedimentation rate was determined from 1972-1850, where the
date 1850 was assigned to a sample with low lead concentrations. Lead
concentrations begin to increase from background near 1850 in Michigan lakes
(Simpson et al., 2000), which is similar to values found in Ontario lakes (Blais and
Kalff, 1995). This sedimentation rate (252 g/mz/y) was used to date the lower section
of the core (from 1972 to 1733).

Table 2. Selected data determined from 210Pb analysis, including the model

used for dating, the approximate mixed depth, the average sedimentation
rate, the focusirg factor (FF), and the oldest section in the core.

 

 

M.,... 3333332.: 33233233 a .3333:
Cadillac CRS 14 cm 117 1.7 1329‘
Cass CF:CS 3 cm 3460 so“ 1971
Crystal B CR8 4 cm 624 2.9 1516‘
Crystal M CRS 6 cm 465 1.7 17328
Elk SCF:CS 1 cm 337 2.1 1279a
Gratiot CF:CS 5 cm 255 2.5 1623‘
Gull SCF:CS 3 cm 404 1.8 1496‘
Higgins CF:CS 3 cm 232 2.0 1729’
Littleﬁeld Pb NA 444 2.0” 1732‘
Mullett SCF:CS ‘ 4 cm 801 3.6 1706’
Paw Paw CF:CS 3 cm 828 2_7° 1923
Whitmore SCF:CS 6 cm 556 26° 1887

 

a. Estimated dates based on extrapolation.
b. A focusing factor could not be calculated for Littleﬁeld Lake, so the

average focusing factor of all lakes sampled previously (except Cass Lake)
was used.
0. Estimated focusing factors based on extrapolation.
Comparison among lakes of sediment chronology data is complicated by the
process of sediment focusing, which is the tendency of ﬁne-grained sediment to move
towards the deepest portion of the lake. This process can be accounted for through the

use of a focusing factor. The focusing factor (FF) was calculated by dividing the

133

actual 210% inventory in each lake (Kada and Heit, 1992) by the expected 21on

inventory (Golden et al., 1993).

actual 210Pb inventory

 

FF =
theoretical 210Pb inventory

Focusing factors were calculated for each of the study lakes in Michigan, and
ranged ﬁom 1.7 to 6.0 (Table 2). The focusing factors for Cass, Whitmore and Paw
Paw Lakes were estimated using extrapolation to calculate the inventory of excess
21"Pb, since supported 2|“Pb levels were not reached. A focusing factor could not be
calculated for Littleﬁeld Lake, so the focusing factor was estimated by taking the

average of the focusing factor of all lakes sampled previously, excluding Cass Lake.

Temporal patterns

Lead

Lead concentrations vary considerably among the twelve study lakes (Figure
'3), but the proﬁles are very similar in shape (Figure 4). All lakes increase from
background concentrations in the mid-1800s and peak between 1973 and 1981. The
similarity of trends in the study lakes indicates that lead was regionally inﬂuenced
until the 19808. This is ﬁrrther corroborated by the similarity of the proﬁles among
lakes in this study to other lakes in the Great Lakes region (Graney et al., 1995;
Alfaro—De la Torre and Tessier, 2002; Outridge et al., 2002). However, rates of
decrease in concentration ﬁom the peak to the present differ among study lakes.
There has been little documentation of this phenomenon for multiple lakes at this

spatial scale.

134

2005 Wj‘iA‘ 1

 

1965f '
1925:
1885:
1845: 1
1805 ... . 1+J
MR

 

 

 

 

 

 

 

 

01965
g 1925
1885

1845
1805 Me
2005 (fr '

 

 

 

-k
r
, I
"z

i

Littleﬂeld

1965
1925
1885
1845
1805

 

 

 

0 100200300 0 100200300 0 100200300

 

 

Whither:

100 200300

Concentrations of lead (mglkg)
Figure 3. Lead concentrations (mg/kg) over time in 12 Michigan lakes.

 

 

 

 

 

 

 

1995
1975 ~
1955 L
1935 —
'8; 1915 .
O
1895 ~
1875 .
. +Lillleiield‘ —-—Crystal M
1355 . —x—Elk +Glatiot
' . —l—l-iggins —-—Cadillac
1835 . +Mlllett +Paw Paw
1815 ’ __l l l __,____l__________
0.00 020 0.40 0.60 0.80

 

+ Cass
+ Gull

—0— Crystal 8
—e— Whitmore

 

 

1.00 1.20

Lead sediment concentrations normalized to peak .
Figure 4. Lead concentrations normalized to peak concentration in 12

Michigan lakes.

135

' .1972

' 71972

 

There are several possibilities for the differing rates. Mixing of the sediment and
focusing can cause broadening of contaminant peaks (Crusius and Anderson, 1995),
and different amounts of focusing and/or mixing could cause the differences in the
proﬁles among lakes. However, the increasing trend is very similar among lakes,
despite differences in within lake processes, indicating that these explanations in
themselves are not sufficient. Another possibility is that the input of lead is decreasing
at similar rates among lakes, but input rates of a diluting component of the sediment
(e.g., production of organic matter, calcium carbonate) are different among some
lakes. Proﬁles of lead accumulations rates, which should accormt for the effects of
dilution, were examined among lakes, but rates of decrease still differed among lakes.
The difference in decreasing rates of lead may also be caused by a shift from regional
to local inﬂuence. Local sources for lead may be point sources within the watershed,
or the input of soils that have enriched lead concentrations due to current and historic

atrrrospheric deposition.

Cadmium

Patterns of deposition and possible sources for cadmium to lake sediments
have not been as extensively studied as lead. Similar to lead, cadmium proﬁles also
increase ﬁ'om background levels, peak in the subsurface and decrease to the present
(Figure 5). The pattern of cadmium deposition in the study lakes is similar to the
cadmium proﬁle found in another lake in the region (Lake Tantare in Quebec, Canada:

Alfaro-De la Torre and Tessier, 2002). However, despite the overall similar trend,

136

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

2°05 7_r""rhf"' ‘1 ‘ CT' . ":_1 .fmf"?‘iﬂi WV 9 [—8"i_ "Tff'«
4

1965: ' 2 2 i 2 ‘1- . _‘ /- 21972
1925: E 5‘ i : 5 {I ;
1885’ . 1 V i ‘f' : :5 2

’ g . 7. n l.
1845 : Cass ‘ Cadlllac i 1‘, Crystal 3‘ ( " Crystal M:

__... __.___r__._.___ l”. ._._.,,._ 3___ i __h_ i an .
1605
““553?_fFTT_W iifﬁ‘iwﬁﬁﬂTili‘ A l “,"—1

“965:3, f l g; f! 2 f; 11972
31925 i“) . ,2 _.‘- ‘ g." . a

1665; i +-{ i > r’
1645: Elki ~ Gratiot, 3' Gull, Li Higginsi
1805 r . i u 1 . . r t r i i l l t l '1}. t . . .
20057 ﬁ—i , - 1 .W .- 1 .. ; T—
1985 ; 71-?) i : ‘2 1L ti xi: _‘ 3371972
1925: f’ j j r- 2 . “ f {"” .
16653,} ;{I _ i q
1645 j Littleﬁeld 2f. Mullett 2 Paw Paw] Whitmore:
1805 W_L___l_ Lit t l l r l l r 1 t l l r r r r r

0 1 2 3 0 1 2 3 0 1 2 3 0 1 2 3

Date

Concentrations of cadmium (mg/kg)

Figure 5. Cadmium concentrations (mg/kg) over time in 12 Michigan
lakes.

 

 

1995 r '7

1975 .

1955 ~

1935 -

 

1915 —
1695 h

1875 t

it ' . A " +Gratiot +Gull
1355 ” g ‘ +Higgins —-Cadillac
‘\

 

 

 

1835 i: , +Cl‘Y8tBIB +Mlll8u
i‘ +PawPaw +Whitmore
1615 ’ . 1 a . . . .

 

0.00 0.20 0.40 0.60 0.80 1 .00 1 20
Cadmium sediment concentrations normalized to peak

Figure 6. Cadmium concentrations normalized to peak
concentration in 12 Michigan lakes.

137

proﬁles are not as similar among lakes as they are for lead proﬁles (Figure 6). Unlike
lead, cadmium concentrations in the sediments increase ﬁom background at different
times among lakes, and peak at different times. The overall shape of the cadmium
proﬁles in study lakes and Lake Tantare implies regional human perturbation of the
cadmium cycle, however, the diﬁ‘erences among the proﬁles indicate that cadmium

has been locally inﬂuenced much more than lead.

Spatial patterns

Further insight into inﬂuences on lead and cadmium loadings in the region
were obtained by examining the spatial patterns of focusing-corrected anthropogenic
accumulation rates. If lead or cadmium inputs are regionally inﬂuenced, then
accumulation rates should be the same in all lakes or show a regional pattern that
reﬂects the distribution of sources for these elements to the atmosphere. If
accumulation rates differ greatly in nearby lakes, or no regional gradient exists, then
the deposition of that metal may be inferred to be locally inﬂuenced. A metal with
locally inﬂuenced inputs will probably also be atmospherically deposited, but if local
inﬂuences are dominant, they could obscure the regional pattern.

The portion of the anthropogenic elemental abundances of lead and cadmium
in the sediment was determined by normalizing to a relatively inert element
(aluminum). Normalization corrects for changing proportions of terrestrial inputs and
diluting phases (e.g., calcium carbonate) (Bruland et al., 1974; Kemp and Thomas,
1976; Callender and vanMetre, 1997; Koelmans, 1998; Tuncer et al., 2001; Alfaro-De

la Torre and Tessier, 2002). For most lakes, the metal to aluminum background ratio

138

was calculated as the average ratio of all samples older then 1800. However, sediment
cores taken from Whitmore and Paw Paw Lakes were not deep enough to reach
background sediments (oldest sections were 1887 and 1923, respectively). Therefore,
background metal to aluminum ratios in these lakes were estimated from Higgins
Lake, which had the similar concentrations of selected major sediment components
(e.g., aluminum and calcium) to both Paw Paw and Whitmore Lakes. Background
concentrations were not calculated for Cass Lake, because the sediment core ﬁom
Cass Lake only reached 1971. Because such a small time interval was sampled, and
because Cass Lake had a sedimentation rate much higher than any other lake, we did
not feel that it was reasonable to estimate anthropogenic inputs.

Focusing corrected anthropogenic accumulation rates were calculated using the

equation (Wong et al., 1995):

i 2 i 4
FCAccumj(pg/m yr) = (Cj x W] x10 )IFFj

FCAccum'} = focusing corrected sediment accumulation rate for im metal at jﬂ'
lake,
'. = watershed background corrected concentration of i‘h metal in the

I
sediment (pg/g) at jd' lake,
W I- = mass sedimentation rate (g/cm2 yr) in jth laked based on 2")Pb dating,

w,- = sediment focusing factor for j‘“ lake

Spatial patterns were examined over three time intervals: 1925-1935 (1930),
1965-1975 (1970) and 1985-1995 (1990). The early industrial period is represented
by 1930, the period of park anthropogenic inputs by 1970 (e.g., Figures 3,5), and 1990

represents recent deposition. Focusing-corrected anthropogenic accumulation rates

139

were averaged over a decade to correspond to population density data (decadal census
data). Using decadal averages also accounts for samples (0.5 or 1 cm sections of
sediment) including deposition ﬁ'om varying numbers of years among lakes. For
example, the sample with the highest concentration of lead in Elk Lake includes
sediment deposited from 1972-1976, and in Gull Lake this sample includes 1975-
1979. Averaging over a longer time period rather than attempting to choose one

comparable sample from each lake reduces the inﬂuence of the varying time intervals.

Lead

In 1930, the two lakes with the highest lead accumulation rates, Paw Paw and
Gull lakes, are in the southwest corner of Michigan (Figure 7a). The other study lakes
have relatively low anthropogenic accumulation rates (<10,000 ug/mzly). Rates have
increased in all lakes in 1970, and lakes in southern Michigan continue to have the
highest values (Figure 7b). Lead accumulation rates in Whitmore Lake, near Detroit,
increased more than other lakes. In both the 1930s and the 1970s there was a general
trend of higher accumulation rates in the south and decreasing northwards. This
regional trend has also been seen in the Great Lakes, with higher atmospheric
deposition and sediment concentrations in southern than northern Lake Michigan
(Edgington and Robbins, 1976; Eisenreich, 1980). In 1990 the regional gradient is no
longer apparent G’igure 7c). Accumulation rates of lead in Gull and Paw Paw Lakes
decreased substantially, and no longer had rates higher than lakes ﬁlrther north.
Furthermore, some lakes that are in relatively close proximity had very different

accumulation rates (e.g., Littleﬁeld and Cadillac). Lead accumulation rates for

140

Gratiot Lake remained below 10,000 ug/mZ/y for all three time intervals, and are not

shown in Figure 7.

Focusing-corrected
lead anthropogenic
accumulation rates
(us/m2!»

. 0 - 10000
10001 - 20000
20001 - 30000
30001 - 40000
40001 - 50000
50001 - 60000

 

Figure 7. Average lead focusing-corrected anthropogenic accumulation rates
(pg/mZ/y) for: (3) 1925-1935, (b) 1965-1975, (c) 1985-1995. Cass Lake is not
shown because anthropogenic accumulation rates could not be calculated. Gratiot
Lake and the Upper Peninsula are not shown; Gratiot Lake has an accumulation
rate between 0 — 10000 ug/mZ/y for all time periods.

The regional gradient present in 1930 and 1970 suggests that lead is regionally
inﬂuenced. Many studies have shown that the atmosphere is an important pathway for
lead (e.g., Edgington and Robbins, 1976; Evans and Dillon, 1982; Eisenreich et al.,
1986; Outridge et al., 2002), and Graney, et al., (1995) suggest from isotope data that
lead emissions to the atmosphere were inﬂuenced by both coal burning and leaded
gasoline in the 1930s and dominantly by leaded gasoline in the 1970s. The general
population distribution in Michigan decreases towards the north, therefore, sources for
lead to the atmosphere such as coal burning stoves or automobiles would be expected
to produce the regional gradient seen in 1930 and 1970. Furthermore, prevailing
winds are predominantly from the southwest in this region (Landis et al., 2002), and

may have carried atmospheric lead into Michigan from the extensive steel mills in and

141

around Gary, Indiana (Souch et al., 2003), and the large population center of Chicago,
Illinois. This may account for the higher lead accumulation rates in Paw Paw and Gull
Lakes in 1930 and 1970, but is unlikely to be the only contributor of lead to Michigan
lakes. Sources impacting lead inputs in 1990 are less apparent, but the lack of spatial
gradient indicates that lead has become more locally inﬂuenced as regional sources
(e.g., leaded gasoline) have declined.

Atmospheric deposition has been estimated ﬁ'om measurements and modeling
in the Great Lakes region (Eisenreich and Strachan, 1992; Hoff et al., 1996; Sweet et
al., 1998; Simonetti et al., 2000b; Alfaro-De la Tone and Tessier, 2002), making it
possible to compare atmospheric deposition rates of lead to anthropogenic
accumulation rates in lake sediments. Accumulation rates in the study lakes are all
signiﬁcantly higher than measures of total atmospheric deposition within the region
(Table 3), indicating another source for lead in addition to atmospheric deposition.
There appears to still be a regional gradient of atmospheric deposition in 1990, with
the lowest rates of deposition measured in the north, over Lake Superior, and higher
rates measured over Lake Michigan (Eisenreich and Strachan, 1992; Sweet et al.,
1998). If this gradient is ignored, we can estimate from the literature the average rate
of atmospheric deposition to this region (~1700 pg/mzly), and use this to evaluate the
approximate percentage of the lead input to each lake from atmospheric deposition
(estimated atmospheric deposition from the literature / anthropogenic accumulation in
lake sediments; Table 3), with the remaining portion being due to local sources. All
lakes have less than 50% of lead inputs contributed from atmospheric deposition, with

Gratiot Lake having the highest atmospheric component (~30%), and Cass, Cadillac

142

Table 3. Summary of lead and cadmium accumulation rates in this study (average
1985-1995) and accumulation rates measured in the sediment, atmospheric deposition,
and snowpack from previous studies. Data from this study are focusing-corrected
anthropogenic accumulation rates, data from other studies are total accumulation rates.
Percent atmospheric is the estimated percent of the sediment accumulation rate due to
atmospheric deposition, with the parenthetical values in the ﬁrst line representing the
estimated rate of atmospheric deposition (pg/mZ/y) used in the calculation.

 

 

 

 

Accum rate pg/mzly % atmospheric
Location Medium Pb Cd Pb Cd
This study 1985-1995 (1700) (100)
Cass‘ sediment 55000 220 3% 45%
Elk sediment 8900 60 18% 167%
Gratiot sediment 4900 50 33% 200%
Gull sediment 18000 70 9% 143%
Higgins sediment 15000 170 11% 59%
Crystal M sediment 26000 210 6% 48%
Littleﬁeld sediment 17000 100 9% 100%
Cadillac sediment 34000 240 5% 42%
Crystal B sediment 12000 210 13% 48%
Paw Paw sediment 25000 130 6% 77%
Whitmore sediment 45000 340 4% 29%
Mullett sediment 9200 80 17% 125%
Lake Tantare, Canadaa sediment 61
L. Superior, 1993-1994b atmospheric 1500 460
L. Superiorc atmospheric 817 148
L. Michigan. 1993-1994b atmospheric 1600 450
L. Michiganc atmospheric 1747 153
L. Erie, 1993-1994b atmospheric 1800 490
L. Hum"d atmospheric 1700 110
L. Ontariod atmospheric 2200 140

Eastern Ontario, 1996" snowpack 720-1800 60—140

Northeast USA, 1996’ snowpack 140-340 25-60

* Accumulation rates for Cass Lake are total (not anthropogenic) focusing corrected
accumulation rates.

a. Headwater lake in an ecological reserve in Quebec, Canada (Alfaro-de la Torre and
Tessier, 2002).

b. Sweet et al., 1998

c. Eisenreich and Strachan, 1992

d. Hoff et al., 1996

e. Simonetti et al., 20000

143

and Whitmore Lakes having the lowest (55%). These three lakes are all located in
relatively highly developed areas, while Gratiot Lake is in a more remote location.
Some of the difference in the estimation of the percentage of total inputs due to
atmospheric deposition among lakes is probably due to a regional atmospheric
gradient. Atmospheric deposition rates are probably lower than estimated in the north
(e.g., near Gratiot Lake), causing an overestimation of the contribution from the
atmosphere. However, even if the lowest deposition rate (817 pg/mz/y, Eisenreich and
Strachan, 1992) for the northern region of the state is used, atmospheric deposition
still accounts for a portion in Gratiot Lake (17%) which is greater than that of the
urban lakes. From these data it can be inferred not only that lead is locally inﬂuenced
in 1990, but also that the relative contribution of atmospheric deposition varies among
lakes. This may account for varying rates of decrease in lead sediment concentrations

among lakes that was described above.

Cadmium

As might be anticipated from the results shown in Figure 8, spatial patterns of
cadmium deposition over time are less evident than lead. The regional trend of higher
accumulation rates in the south that was present for lead is absent for cadmium over
all time intervals. Similar to lead, in 1930 Paw Paw Lake has the highest
accumulation rates, but unlike lead, Crystal B Lake also has relatively high
accumulation rates. In 1970 the southern lakes have high accumulation rates of
cadmium, with Whitmore Lake having the highest accumulation rate. However, lakes

further north (e.g., Crystal B) also have high accumulation rates. In 1990, cadmium

144

accumulation rates in Gull and Paw Paw have decreased signiﬁcantly, and Whitmore,
Crystal M, Cadillac and Crystal B Lakes have the highest rates, resulting in a spatial
pattern that is similar to lead in 1990. A regional gradient does not exist in 1990, but
those lakes with relatively high cadmium accumulation rates also have high values for
lead. Accumulation rates of cadmium remain low (< 100 ug/mZ/y) in Gratiot Lake

throughout all time intervals.

Focusing-corrected
cadmium
anthropogenic
accumulation rates
(HQ/"I’M
- 0 - 100

s 101 - 200

e 201 - 300

e 301 - 400

O 401 - 500
Figure. 8. Average cadmium focusing-corrected anthropogenic accumulation rates
(pg/mZ/y) for: (6) 1925-1935, (b) 1965-1975, (c) 1985-1995. Cass Lake is not shown
because anthropogenic accumulation rates could not be calculated. Gratiot Lake and
the Upper Peninsula are not shown; Gratiot Lake has an accumulation rate between 0

— 100 ug/mZ/y for all time periods.

 

Relatively high accumulation rates of cadmium in Gull and Paw Paw Lakes in
1930 and 1970 may have been inﬂuenced by the Gary / Chicago area (Souch et al.,
2003), but except for the direct inﬂuence of this area there appears to be no regional
trend in cadmium accumulation rates. This suggests that the primary inﬂuence on
cadmium is local contributions. There was probably also atmospheric deposition of
cadmium in these time periods, but the overall spatial pattern of deposition is locally

inﬂuenced. The decrease in accumulation rates in Gull and Paw Paw Lakes in 1990

145

suggests a decrease in the inﬂuence of atmospheric contaminants from this urban
region. It is unclear what is locally inﬂuencing cadmium inputs to lakes in this region.

From atmospheric deposition rates in the region (Eisenreich and Strachan,
1992; Hoff et al., 1996; Sweet et al., 1998; Simonetti et al., 2000b; Alfaro-De la Torre
and Tessier, 2002), it may be inferred that cadmium loadings were greatly inﬂuenced
by atmospheric deposition in 1990 (Table 3). Rates of atmospheric deposition of
cadmium range ﬁom approximately 100 - 150 ug/mZ/y, if data from Sweet et al.,
(1998) are excluded. Atmospheric deposition rates reported by Sweet et al. (1998) in
the Great Lakes region are higher than both anthropogenic and total accumulation
rates in our study lakes, and are signiﬁcantly higher than other estimations of
atmospheric deposition (Eisenreich and Strachan, 1992; Hoff et al., 1996; Simonetti et
al., 2000a), and therefore are not included in this analysis.

Five of the study lakes have cadmium accumulation rates equal to or lower
than the rate of atmospheric deposition estimated from the literature (Table 3, ~100
pg/mzly), indicating that accumulation rates calculated from lake sediments
underestimate cadmium deposition, or that rates of atmospheric deposition are
overestimated. Despite this discrepancy, these data suggest that cadmium inputs to
ﬁve of the study lakes were inﬂuenced primarily by atmospheric deposition (Elk,
Gratiot, Gull, Littleﬁeld and Mullett Lakes, accumulation rates _<.100 pg/mzly). Five
other study lakes are probably locally inﬂuenced, all with accumulation rates >200
pg/mzly (Cass, Crystal M, Cadillac, Crystal B, and Whitmore Lakes). Higgins and
Paw Paw Lakes have intermediate accumulation rates (170 and 130 ug/mz/y), and

with the discrepancies in the data, it is difﬁcult to determine the inﬂuence of local

146

inputs on these two lakes. Overall, atmospheric deposition is a more signiﬁcant
source for cadmium than lead.

The similarity of the spatial pattern of lead and cadmium in 1990, with
Whitmore, Crystal M, Cadillac and Crystal B Lakes having higher accumulation rates
(Figure 7c, 8c), suggests that sources for cadmium and lead may have been the same
for some lakes, or may be related to similar processes (e.g., population growth within
the watershed). Those lakes that are most inﬂuenced by local sources for lead
(Cadillac, Whitmore, Cass and Crystal M) are also inﬂuenced for cadmium (Table 3).
Spatial patterns may have been dissimilar in 1930 and 1970 because major
atmospheric sources were present for lead and not cadmium (e.g., leaded gasoline).

These regional sources for lead may have masked any local inﬂuences.

Population density
The examination of spatial and temporal patterns of cadmium and lead provide

insight into the relative importance of local and regional inﬂuences on cadmium and
lead loading to the environment, as well as determine if the patterns are consistent
with speciﬁc sources suggested in the literature. These interpretations may be further
tested by the comparison of accumulation rates of metals in the sediment to watershed
population densities. Although humans themselves are not sources for metals, many
human activities do release metals to the environment, and population density values
can serve as an approximate indicator for these sources. Population density has been

related to sewage and storm water discharge, yard waste and other local inputs

147

(Callender and Rice, 2000), as well as sources related to automobiles and highways
(e.g., tire wear, auto emissions, road maintenance) (Rice, 1999).
Watershed population densities for most lakes have increased over time, with

greater rates of increase after the 19403 (Figure 9). The highest rates of watershed

 

 

 

 

 

 

 

 

100 ——~ ---~ 350
90 ~ . +Elk
é +Gull
5 70 1 #8 250 +Higgins
2 a. 60 ~ +CryslalM
o u 200 .
a 5 50 A +thtleﬁe|d
g 3 —--Cadillac
11:. Q ~- 150
8 g 40 ~ —l—CrystalB
'3 v 30 _2 j 100 +PawPaw
"E 20 a +Whitmore
g _. 50 —-—Mullett
; 10 1 —e—Cass*
o - o
1650 2000

 

Figure 9. Watershed population densities (people/kmz) over time. *Cass Lake is
plotted on the right y-axis scale.

population growth are for lakes located in southern Michigan (Paw Paw, Gull,
Whitmore, Cass and Crystal M Lakes). Cadillac Lake also ins a high rate of
population growth, but the growth occurs at a much earlier time period than the other
lakes. The city of Cadillac, located within the Cadillac Lake watershed, was
incorporated in 1877, and grew quickly until the 1920s, when growth slowed (Figure

9). This settlement pattern is unique among the study lakes. The settlement pattern of

148

Gratiot Lake watershed is also unique; population densities have decreased over time,
with the highest population density in 1870. Settlement in this area was highest in the
late 18003 due to the extensive copper mining, which began in the 18403 and peaked

between 1890 and 1930 (Kerfoot et al., 1994).

Lead

Population density data corroborate the interpretations of the spatial and
temporal data. In 1930 there is little relationship between population density and lead
accumulation, supporting the interpretation that lead was regionally inﬂuenced (Figure
10a). There appears to be a positive correlation between population density and lead
accumulation rates, with the exception of Cadillac Lake (r2 = 0.77 with Cadillac
removed), however, this relationship is an artifact of Paw Paw Lake having a high
population density and high lead accumulation rate, and does not hold for the
remainder of the lakes (r2 = 0.22 with Paw Paw and Cadillac removed). All but four
lakes (Cadillac, Paw Paw, Gull and Crystal M) have both low lead accumulation rates
and low population densities (Figure 10a), with sources for lead probably including
coal burning and the use of leaded gasoline throughout the state. The high
accumulation rates in Crystal M, Gull, and Paw Paw Lakes, located in southwestern
Michigan, are probably being inﬂuenced by the Gary / Chicago urban and industrial
area rather than the population density arormd the lake.

There is a positive relationship between watershed population densities and
lead accumulation rates in 1970 (Figure 10b), with Cadillac and Paw Paw Lakes, and

to a lesser extent Crystal B and Elk Lakes, have lower lead accumulation rates then

149

25000

20000

1 5000

.3
o
O
O
0

0'1
0
O
O

0

60000
50000

.3
O
O
O
CO

50000

Focusing-corrected lead anthropogenic accumulation rates (pg/mzly)
8 ‘5’ 8
O O O
O O O
O O O

1 0000

0

40000 .
30000 ,
20000 ~

 

 

 

 

l
s
A
U I
0 20 40
:1
A A
_ >i
0 20 40
A
L s
A+
I
'- x
)k
0 20 40

60 80
Watershed population density (people/km?)

 

l
_._______ 1,, __. J

100

 

I
u___._-...

100

O

r>+0xx>

Cadillac
Crystal 8
Crystal M
Elk
Gratiot
Gull
Higgins
Littletield
Mullett

° Paw Paw

Whitmore

Figure 10. Average focusing-corrected lead anthropogenic accumulation rates
(pg/mzly) versus watershed population densities for (a) 1925-1935, (b) 1965-1975, (c)

1985-1995.

150

expected ﬁ'om the trend of the other lakes. However, interpretation of this relationship
is complicated by the similarity between the regional population gradient and
watershed population densities. Iflead, as suggested above, was regionally inﬂuenced
in 1970 by sources such as leaded gasoline, it should not be closely related to local
watershed population densities, instead, it should be related to the regional population
distribution (i.e., a watershed with low population density in a highly populated region
would still have high lead accumulation rates). In the 19703, the overall population
density in Michigan was highest in the south and decreased to the north, and
watershed population densities of the lakes in this study were generally consistent with
that gradient, with the exception of Cadillac Lake. In other words, watershed
population densities follow the same south - north gradient seen in lead and seen in
the overall population distribution. Cadillac Lake, however, has a very high
population density within the watershed, but is in an area of low regional population
density. The lead accumulation rate in Cadillac Lake is much lower than anticipated
for the watershed population density (Figure 10b), but ﬁts into the regional gradient
(Figure 7b). This implies that lead inputs to the study lakes are mostly inﬂuenced by
the regional population gradient, which is consistent with lead being regionally
inﬂuenced, with the probable source of atmospheric deposition of lead ﬁom the use of
leaded gasoline.

In 1990 there was a positive relationship between population density and lead
accumulation rates (Figure 10c). Unlike 1970, the lead accumulation rates do not
seem to follow the statewide population distribution, but seem more closely linked to

the watershed population density. This supports the conclusion that lead was locally

151

inﬂuenced in 1990, and that regional atmospheric deposition was no longer the major
source for lead to the study lakes.

The presence of two separate linear trends in 1990 suggests that watershed
population density is not sufficient to explain lead accumulation rates, and other
factor(s) must also be important. Other sources may have been present for the lakes
on the upper trend line (e.g., Whitmore, Crystal M, Higgins Lakes) that are not related
to population density, or, the sources may have been similar to all lakes, but the
pathway of lead to the lake is being inﬂuenced by the watershed. The characteristics
(morphology, lakezwatershed area ratio) of a watershed are known to inﬂuence
nutrient and trace metal cycling (Dillion and Evans, 1982; Hurley et al., 1995). Elk,
Gull, Cadillac and Paw Paw Lakes all have high watershed to lake ratios (Table 1),
and follow a lower trajectory than the other lakes. Additionally, there are many
additional factors beyond those related to population density (e.g., land use within the
watershed) that were not evaluated in this study may also be inﬂuencing the loading of
lead to the study lakes.

Despite the presence of two separate trends, both lines appear to cross the y-
intercept (population density of zero) at an anthropogenic accumulation rate of
approximately 5000 ug/mzly (Figure 10c). This suggest that there is a regional lead
input at that rate, and if all local inﬂuences were removed, lakes in Michigan would

still have anthropogenic lead accumulation rates of at least 5000 ug/mzly.

152

Cadmium

Cadmium accumulation rates are positively correlated with population density
over all three time intervals (Figure 11 a,b,c), with the exception of Cadillac Lake
(1930, 1970) and Paw Paw Lake (1970). In 1990, there seems to be two separate
relationships between population density and accumulation rate, with Paw Paw,
Cadillac, Gull, Elk Lakes following a lower trajectory than the rest of the lakes.

The spatial and temporal patterns for cadmium deposition were more locally
inﬂuenced than lead, an interpretation which is supported by the relationship between
cadmium and population densities for all time intervals. In 1970, Cadillac and Paw
Paw Lakes had lower accumulation rates than expected from their population densities
for both cadmium and lead (Figure 11b). However, cadmium accumulation rates did
not have the same spatial pattern as lead or the regional population gradient (Figure
8b). From this, it can be inferred that sources for cadmium were different than lead,
and inﬂuenced locally rather than regionally. In 1990, the apparent presence of two
separate relationships between population densities and cadmium accumulation rates
is similar to lead. The lakes that followed the lower trajectory for lead are the same
for cadmium (Figure 100, 11c). This suggests that the input of lead and cadmium was
being inﬂuenced by some characteristic of the watershed, perhaps the lake to
watershed area ratio, or some factor that has not been examined.

The y-intercept of the relationship between cadmium anthropogenic
accumulation rates and watershed population densities in 1990 is approximately 50
ug/mzly, suggesting that at zero population density (no local inﬂuences) a lake in

Michigan would still have an anthropogenic accumulation rate of cadmium of at least

153

250 ,

200

150

100

50

400
350
300
250
200
1 50
1 00

50

Focusing-corrected cadmium anthropogenic accumulation rates (pg/mzly)

Ll

 

‘1

o 20
Watershed population density (people/kmz)

 

 

 

40

 

 

- 2 L31
1
i
1
1
1
. ;
____,_____r l
60 80 100
. l
e i
!
i
so so 100
_ H “Twit
D C
oi
l
O
60 80 100

O

>+.XX>I

0

Cadillac
Crystal B
Crystal M
Elk
Gratiot
Gull
Higgins
Littleﬁeld
Mullett
Paw Paw

Whitmore

Figure 11. Average focusing-corrected cadmium anthropogenic accumulation rates
(pg/mzly) versus watershed population densities for (a) 1925-1935, (b) 1965-1975, (c)

1985-1995.

154

50 ug/mzly. This is similar to the estimated rate of atmospheric deposition to the

region (Table 3).

Production records

Spatial and temporal patterns can provide insight into whether inputs of a
metal are locally or regionally inﬂuenced, but cannot provide information about actual
sources, except through comparison to previous work (e.g., Graney et al., 1995).
Additionally, from the presence of a relationship between population densities and
anthropogenic accumulation rates it may be inferred that inputs of the metal are locally
inﬂuenced, and that the source is related to population density. To gain more insight
into speciﬁc sources for a metal with regionally inﬂuenced inputs, we can compare
industrial production records for that metal (e.g., production of cadmium from
smelters) to temporal deposition patterns in lakes. National production records
provide an estimation of the output over time ﬁom these industries. They do not
report how much of a metal is released into the environment, but if temporal records
of production are similar to depositional histories in the sediment, this may be a good
indication of source. For example, the production of leaded gasoline probably releases
relatively little lead to the atmosphere, but is probably highly correlated to the amount
of leaded gasoline combustion. To determine if metal deposition histories were
similar to production records, production records and sediment concentration proﬁles
were normalized to their peak values and compared to each other.

Spatial and temporal patterns suggested that lead inputs were regionally

inﬂuenced in 1930 and 1970, but locally inﬂuenced in 1990. The recorded use of lead

155

in gasoline matches the lead proﬁles recorded in the lake sediments very well ﬁ'om the
19503 to the 19703, but decreases at a rate faster than lead in the sediment (Figure 12).
This further supports the conclusion that in 1970 the lead was regionally inﬂuenced,
with the use of leaded gasoline as an important source for lead, but more locally
inﬂuenced in 1990. While the importance of leaded gasoline as a source of lead to the
environment is well documented (e.g., Edgington and Robbins, 1976; Evans and
Dillon, 1982; Eisenreich et al., 1986; Graney et al., 1995), the good correlation
between the suggested source and concentration proﬁles in the sediment suggests that

this technique may be used on cadmium, where sources are more poorly deﬁned.

 

2005
1985
1 965
1945
1925
1905
1885
1865 - ' + Lead in gasoline
1345 -— Elk Lake

1825 p + Gull Lake

1805 _ .l. l, l _ l l J
0.00 0.20 0.40 0.60 0.80 1.00 1.20

Normalized values

Figure 12. Production of leaded petroleum, and concentration proﬁles of Elk and
Gull Lakes normalized to the peak values. Elk and Gull Lakes are plotted to represent
the typical proﬁles seen in the Michigan inland lakes sampled in this study.

Date

 

 

 

 

 

 

Spatial and temporal patterns indicate that cadmium is locally inﬂuenced;
therefore no consistent relationship between production records and cadmium

concentration proﬁles would be anticipated. However, it is possible that lakes in the

156

southwest of Michigan are inﬂuenced by industrial activities in and around Gary,
Indiana and Chicago, Illinois (Souch et al., 2003). This cannot be determined from the
spatial pattern of cadmium deposition, because the extent of the study area is large
enough to preclude the possibility of this one source being dominant for the entire
study area. Therefore, this hypothesis was investigated using production records.

Concentration proﬁles in Gull and Paw Paw Lakes have a similar proﬁle to the
total United States production of cadmium and smelter production of cadmium (Figure
13). No other lakes have a similar proﬁle to any production records for cadmium.
These results support the hypothesis that these two lakes, located in the southwest
comer of Michigan, may have been inﬂuenced by atmospheric deposition from the
industrial center of Gary, Indiana to the west, while other lakes in the state were more
locally inﬂuenced. Unlike lead, the importance of the industrial production of

cadmium to the region has not been previously studied.

2005 - — a
1965 3 ~‘ ‘
1965
1945
1925
g 1905
1885 +Cd production
1865 +Cd smelter

1845 + Gull Lake
1825 + Paw Paw Lake

1805 _ l 1 L l L
0.00 0.20 0.40 0.60 0.80 1.00 1.20
Normalized values

Figure 13. Total production of cadmium and smelter production of cadmium in the
United States, and concentration proﬁles of Gull and Paw Paw Lakes normalized to

peak values.

 

 

 

  

 

 

157

Conclusions

Sediment chronologies for lead and cadmium from twelve lakes throughout
Michigan were compared spatially and temporally and to watershed population
densities and metal production records. The broad geographic extent of this study
allowed for an understanding of the relative importance of local versus regional
inﬂuences on the loading of lead and cadmium to this region, as well as providing
insight into possible sources for these metals.

Overall, inputs of lead to Michigan Lakes appear to be regionally inﬂuenced in
1930 and 1970. The data in this study are consistent with previous studies that suggest
atmospheric deposition of lead emitted from the burning of coal (1930) and leaded
gasoline (1970) controls lead loading to the environment. It is likely that local sources
were also present, but were insigniﬁcant compared to atmospheric inputs. In the
19903, inputs of lead are locally inﬂuenced, though there still appears to be a
contribution ﬁ'om regional atmospheric deposition. Although watershed population
densities are positively correlated to anthropogenic lead accumulation rates in the
19903, they are not a good predictor of lead accumulation and other factors must also
be important (e.g., watershed: lake area ratios).

Unlike lead, cadmium appears to be regionally and locally inﬂuenced over all
time intervals. Again, there is a positive relationship between watershed population
densities and anthropogenic cadmium accumulation rates, but there are clearly other
factors inﬂuencing cadmium inputs. It is unclear what the speciﬁc sources and
pathways for the cycling of cadmium are, but it appears that atmospheric deposition

contributes a signiﬁcant amount of total cadmium deposition in 1990.

158

The approach of using sediment chronologies ﬁom multiple lakes over a broad
geographic extent provided information on both if a metal was locally or regionally
inﬂuenced, as well as giving insight into some speciﬁc sources. Results for lead
corroborate well with many other studies (Edgington and Robbins, 1976; Dillion and
Evans, 1982; Evans and Dillon, 1982; Graney et al., 1995; Callender and Rice, 2000),
indicating that this technique is a useful approach, and may be used on other elements.
Additionally, this technique can be used to evaluate the contribution of other factors
(e. g., land use) to metal loadings in lake sediments. Expanded assessment of
Michigan lakes with this multi-elemental technique may eventually produce predictive
models for metal loading, greatly facilitating environmental monitoring and

protection.

Acknowledgements

We would like to thank the Michigan Department of Environmental Quality
for frmding this project, as well assistance ﬁ'om Bill Taﬁ, Neal Godby, Rick Lundgren,
and Sarah Walsh. Many thanks as well to the captain and crew of the R/V Mudpuppy,
the Michigan State University physics shop for fabricating our extruding system, Paul
Wilkinson for assistance with 2"’Pb dating, and Lydia Scholle for extensive help in the
laboratory and ﬁeld. Merideth Lindeman and Nellie Niagolova provided valuable
assistance in the ﬁeld and thanks to Chuck Yohn, Gabriel Filippelli, and an
anonymous reviewer for helpful reviews of the manuscript. Funding for S. Yohn was

provided by an EPA Science To Achieve Results (STAR) Fellowship.

159

References

Alfaro-De la Torre, MC. and Tessier, A., 2002. Cadmium deposition and mobility in
the sediments of an acidic oligotrophic lake. Geochim. Cosmochim. Acta 66,
3549-3562.

Appleby, P.G. and Oldﬁeld, F., 1978. The calculation of lead-210 dates assuming a
constant rate of supply of unsupported 2|on to the sediment. Catena 5, 1-8.

Balogh, S.J., Engstrom, D.R., Alemendinger, J .E., Meyer, ML. and Johnson, K.D.,
1999. History of mercury loading in the Upper Mississippi River reconstructed
from sediments of Lake Pepin. Environ. Sci. Technol. 33, 3297-3302.

Blais, J .M. and Kalff, J ., 1995. The inﬂuence of lake morphometry on sediment
focusing. Limnology and Oceanography 40, 582-5 88.

Bollhofer, A. and Rosman, K.J.R., 2001. Isotopic source signatures for atmospheric
lead: The Northern Hemisphere. Geochim. Cosmochim. Acta 65, 1727-1740.

Boyle, J .F., Rose, N.L., Bennion, H., Yang, H. and Appleby, P.G., 1999.
Environmental impacts in the J ianghan Plain: evidence from lake sediments.
Water, Air, Soil Pollut. 112, 21-40.

Bruland, K.W., Bertine, K., Koide, M. and Goldberg, ED, 1974. History of metal
pollution in Southern California costal zone. Environ. Sci. Technol. 8, 425-
432.

Callender, E. and Rice, K.C., 2000. The urban environmental gradient: Anthropogenic
inﬂuences on the spatial and temporal distributions of lead and zinc in the
environment. Environ. Sci. Technol. 24, 232-238.

Callender, E. and vanMetre, P.C., 1997. Reservoir sediment cores show US lead
declines. Environ. Sci. Technol. 31, A424-A428.

Catallo, W.J., Schlenker, M., Gambrell, RP. and Shane, B.S., 1995. Toxic chemicals
and trace metals from urban and rural Louisiana lakes: recent historical proﬁles
and toxicological signiﬁcance. Environ. Sci. Technol. 29, 1439-1446.

Crusius, J. and Anderson, RR, 1995. Sediment focusing in six small lakes inferred
from radionuclide proﬁles. Journal of Paleolimnology 13, 143-155.

Dillion, PJ. and Evans, RD, 1982. Whole-lake lead burdens in sediments of southern
Ontario, Canada. Hydrobiologia 91, 121-130.

160

Door, J.A. and Eschman, DR, 1970. Geology of Michigan. University of Michigan
Press, Ann Arbor, 476 pp.

Edgington, D.N. and Robbins, J .A., 1976. Records of lead deposition in Lake
Michigan sediments since 1800. Environ. Sci. Technol. 10, 266-274.

Eisenreich, S.J., 1980. Atmospheric input of trace metals to Lake Michigan. Water,
Air, Soil Pollut. 13, 287-301.

Eisenreich, S.J., Metzer, NA. and Urban, N.A., 1986. Response of atmospheric lead
to decreased use of lead in gasoline. Environ. Sci. Technol. 20, 171-174.

Eisenreich, SJ. and Strachan, W.M.J., 1992. Estimating atmospheric deposition of
toxic substances in the Great Lakes - an update, Canada Centre for Inland
Waters, Burlington, Ontario, January 31-Febuary 2, 1992.

Encyclopedia_Britannica_Premirun_Service, Michigan, 2003.
<http://www.britannica.com/needmore>

Evans, RD. and Dillon, P.J., 1982. Historical changes in anthropogenic lead fallout in
southern Ontario, Canada. Hydrobiologia 91, 131-137.

Golden, K.A., Wong, C.S., Jeremiason, J.D., Eisenreich, S.J., Sanders, M.G.,
Hallgren, J ., Swackhammer, D.L., Engstrom, DR. and Long, D.T., 1993.
Accumulation and preliminary inventory of organochlorines in Great Lakes
sediments. Wat. Sci. Technol. 28, 19-31.

Graney, J.R., Halliday, A.N., Keeler, G.J., Nriagu, J .O., Robbins, J .A. and Norton,
S.A., 1995. Isotopic record of lead pollution in lake sediments from the
northeastern United States. Geochim. Cosmochim. Acta 59, 1715-1728.

Heyvaert, A.C., Reuter, J.E., Slotton, D.G. and Goldman, CR, 2000.
Paleolimnological reconstruction of historical atmospheric lead and mercury
deposition at Lake Tahoe, California-Nevada. Environ. Sci. Technol. 34, 3588-
3597 .

Hoff, R.M., Strachan, W.M.J., Sweet, C.W., Chan, C.H., Shackleton, M., Bidleman,
T.F., Brice, K.A., Burniston, D.A., Cussion, S., Gatz, D.F., Harlin, K. and
Schroeder, W.H., 1996. Atmospheric deposition of toxic chemicals to the
Great Lakes: A review of data through 1994. Atrnos. Environ. 30, 3505-3527.

Hurley, P.J., Benoit, J .M., Babiarz, C.L., Shafer, M.M., Andren, A.W., Sullivan, J .R.,

Hammond, R. and Webb, D.A., 1995. Inﬂuences of watershed characteristics
on mercury levels in Wisconsin rivers. Environ. Sci. Technol. 29, 1867-1875.

161

Johnson, MG. and Nicholls, K.H., 1988. Temporal and spatial trends in metal loads to
sediments of Lake Simcoe, Ontario. Water, Air and Soil Pollution 39, 337-354.

Kada, J. and Heit, M., 1992. The inventories of anthropogenic Pb, Zn, As, Cd and the
radionuclides ”70s and excess 2’on in lake sediments of the Adirondack
region, USA. Hydrobiologia 246, 231-241.

Kemp, A.L.W. and Thomas, R.L., 1976. Cultural impacts on the geochemistry of the
sediments of Lakes Ontario, Erie and Huron. Water, Air and Soil Pollution 5,
469-490.

Kerfoot, W.C., Lauster, G. and Robbins, J .A., 1994. Paleolimnological study of
copper mining arotmd Lake Superior: Artiﬁcial varves ﬁ‘om Protage Lake
provide a high resolution record. Limnology and Oceanography 39, 649-669.

Kneebone, PE. and Hering, J .G., 2000. Behavior of arsenic and other redox sensitive
elements in Crowley Lake, CA: A reservoir in the Los Angeles Aqueduct
system. Environmental Science & Technology 34, 4307-4312.

Koelmans, A.A., 1998. Geochemistry of suspended and settling solids in two
freshwater lakes. Hydrobiologia 364, 15-29.

Landis, M.S., Vette, AF. and Keeler, G.J., 2002. Atmospheric mercury in the Lake
Michigan basin: Inﬂuence of the Chicago/Gary urban area. Environ. Sci.
Technol. 36, 4508-4517.

MacDonald, D.D., Ingersoll, CG. and Berger, T.A., 2000. Development and
evaluation of consensus-based sediment quality guidelines for freshwater
ecosystems. Archives of Environmental Contamination and Toxicology 39, 20-
3 l .

Oldﬁeld, F. and Appleby, P.G., 1984. Empirical testing of 210Pb-dating models for
lake sediments. In: E.Y. Haworth and J.W.G. Lund (Editors), Lake Sediments
and Environmental History. University of Minnesota Press, Minneapolis, pp.
93-124.

Outridge, P.M., Hermanson, M.H. and Lockhart, W.L., 2002. Regional variations in
atmospheric deposition and sources of anthropogenic lead in lake sediments
across the Canadian Arctic. Geochim. Cosmochim. Acta 66, 3521-3531.

Reimann, C. and dc Caritat, P., 1998. Chemical Elements in the Environment:

F actsheets for the Geochemist and Environmental Scientist. Springer-Verlag,
Berlin.

162

Rice, K.C., 1999. Trace-element concentrations in streambed sediment across the
conterminous United States. Environ. Sci. Technol. 33, 2499-2504.

Sanchez-Cabeza, J .A., Ani-Ragolta, I. and Masque, P., 2000. Some considerations of
the Pb-210 constant rate of supply (CRS) dating model. Limnology and
Oceanography 45, 990-995.

Sanudo-Wilhelmy, SA. and Gill, G.A., 1999. Impact of the Clean Water Act on the
levels of toxic metals in Urban Estuaries: the Hudson River estuary revisited.
Environ. Sci. Technol. 33, 3477-3481.

Shahin, U., Yi, S.M., Paode, RD. and Holsen, T.M., 2000. Long-term elemental dry
deposition ﬂuxes measured around Lake Michigan with an automated dry
deposition sampler. Environmental Science & Technology 34, 1887-1892.

Simonetti, A., Gariepy, C. and Carignan, J ., 2000a. Pb and Sr isotopic compositions of
snowpack from Quebec, Canada: Inferences on the sources and deposition
budgets of atmospheric heavy metals. Geochim. Cosmochim. Acta 64, 5-20.

Simonetti, A., Gariepy, C. and Carignan, J., 2000b. Pb and Sr isotopic evidence for
sources of atmospheric heavy metals and their deposition budgets in
northeastern North America. Geochim. Cosmochim. Acta 64, 3439-3452.

Simpson, S.J., Long, D.T., Geisey, J .P. and Fett, J .D., 2000. Inland lakes sediment
trends: sediment analysis results for ﬁve Michigan Lakes (MI/DEQ/SWQ-
01/030), Department of Environmental Quality, East Lansing, MI.
http://www.michigan.gov/deq/0, l 607,7-1 35-33 13_3686_3728-32365-,00.html

Souch, C.J., F ilippelli, G.M., Dollar, N., Perkins, S. and Mastalerz, M., 2003.
Accumulation rates of airborne heavy metals in wetlands. Phys. Geogr. 23, 21-
43.

Sweet, C.W., Weiss, A. and Vermette, S.J., 1998. Atmospheric deposition of trace
metals at three sites near the Great Lakes. Water Air Soil Poll 103, 423-439.

Tuncer, G., Tuncel, G. and Balkas, T.I., 2001. Evolution of metal pollution in the
Golden Horn (Turkey) sediments between 1912 and 1987. Marine Pollution
Bulletin 42, 350-360.

Walling, DE. and Qingping, H., 1992. Interpretation of caesium-137 proﬁles in

lacustrine and other sediments: the role of catchment-derived inputs.
Hydrobiologia 235, 219-230.

163

Wong, C.S., Sanders, G., Engstrom, D.R., Long, D.T., Swackhammer, BL. and
Eisenreich, S.J., 1995. Accumulation, inventory, and diagenesis of Chlorinated
hydrocarbons in Lake Ontario sediments. Environ. Sci. Technol. 29, 2661-
2672.

Yohn, S.S., Long, D.T., Fett, J .D., Patino, L.C., Giesy, J .P. and Kannan, K., 2002.

Assessing environmental change through Chemical-sediment chronologies
from inland lakes. Lakes & Reservoirs: Research and Management 7, 217-230.

164

CHAPTER 6
LEAD IN THE ENVIRONMENT: EVALUATING SOURCES AND PATHWAYS
OVER TIME USING LAKE SEDIMENTS

Introduction

Humans have signiﬁcantly modiﬁed the natural cycle of contaminants such as
lead, resulting in enrichment of these metals in the environment and toxicological
concerns for both wildlife and humans (Nriagu, 1978; Nriagu, 1989; Eisenreich and
Strachan, 1992; Catallo et al., 1995; Sanudo-Wilhelrny and Gill, 1999).
Environmental legislation, such as the Clean Air and Water Acts, reduced inputs of
lead to the environment (Graney et al., 1995; Long et al., 1995; Callender and
vanMetre, 1997; Callender and Rice, 2000; Simonetti et al., 2000; Mecray et al.,
2001), but concentrations in the water, air and sediment remain elevated (Graney et
al., 1995; Long et al., 1995; Callender and vanMetre, 1997; Mecray et al., 2001).
Humans continue to alter the environment in ever increasing ways, including
population growth and urban sprawl, which may result in emerging contaminant
somces. In order to continue developing strategies for the reduction of anthropogenic
lead loadings to the environment, an understanding of the current sources of lead and
pathways through the environment are necessary. Additionally, by understanding the
history of lead inputs, it may be possible to determine if sources have remained the
same over time or if new sources are emerging. One technique for understanding the
history of lead inputs to a region is through the examination of sediment chronologies

from multiple lakes.

165

Lake sediments act as collectors of contaminants both from the local watershed
and regional atmospheric deposition and can provide information on current and
historic rates of contaminant inputs. As a result, sediments are a valuable tool in
monitoring chemical loading to the environment. While sediments can provide a
history of the rate of contaminant inputs, it is possible to gain further insight by
comparing loadings ﬁom multiple lakes to characteristics of the watersheds that act as
a proxy for potential sources and pathways through the environment (e.g., urban land
cover, soil erodibility). Ideally, anthropogenic accumulation rates in the sediment
could be compared to actual rates of inputs of different sources to determine the
relative importance of each source (Edgington and Robbins, 1976), but these data are
not available for most lakes. Therefore, quantiﬁable characteristics of the watershed
(e.g., population density) have been chosen to represent source(s) (Callender and Rice,
2000). In this study, the relationship between lead accumulation rates in 16 Michigan
lakes to watershed characteristics, such as population density and percentage urban
land cover, are explored. This comparison is done for two time periods (1970-1980
and 1990-2000) to identify potential lead sources and pathways through the
environment, and if and how these relationships have changed over time.

Although there have been many studies on the relationship between watershed
characteristics and nutrient concentrations in aquatic systems (Detenbeck et al., 1993;
Richards et al., 1996; Soranno et al., 1996), few studies have examined the
relationship between metal deposition in sediments and landscape variables (Rice,
1999; Callender and Rice, 2000). Watershed variables in this study were selected to

represent physical, transport, and source characteristics for both time periods.

166

The physical characteristics (watershed area, lake area, watershed: lake area
ratio) may give insight into the importance of atmospheric deposition. For example, a
lake with a small watershed area to lake area ratio may be dominated by atmospheric
deposition, whereas a lake with a large ratio may be more inﬂuenced by inputs from
the watershed (Dillion and Evans, 1982).

Three characteristics (K factor, slope, pattern of sulfate deposition) were
chosen to represent the pathway (transport) of metals to a lake. Metals deposited
within the lake watershed ﬁ'orn processes such as pesticides and fertilizers applications
or ﬁom atnrospheric deposition (Muhlbaier and Tisue, 1981; Graney et al., 1995; Rice,
1999) must be mobilized from the landscape to be deposited in the lake. Erosion
might be one of those processes and the amount of metal reaching the lake via this
pathway may be dependant on soil erodibility. The average slope of the watershed
and K factor (a measure of soil erodibility) would represent variables that would
inﬂuence the amount of metals reaching a lake by erosion within the watershed
(Detenbeck et al., 1993). Steep slope and high K factors would represent high erosion
potential.

Another important pathway is atmospheric transport and deposition
(Winchester and Nifong, 1971; Gatz, 1975; Gatz et al., 1989; Nriagu, 1989; Callender
and vanMetre, 1997; Sweet et al., 1998; Rice, 1999; Callender and Rice, 2000;
Simonetti et al., 2000; Goodarzi et al., 2001; Souch et al., 2003), however there are
few direct measures of the importance of this pathway for a speciﬁc lake watershed. It
may be possible to use a surrogate to estimate the importance of atmospheric

deposition. One might assume that a major control on the chemistry of metals in the

167

atmosphere is the burning of fossil fuels such as coal and oil. Considering this,
sulfate, which is also released to the atmosphere ﬁ'om fossil fuel combustion (N riagu,
1989; Berner and Berner, 1996) and has been monitored through the region (National
Atmospheric Deposition Program (N RSP-3 )/N ational Trends Network, 2003), may
serve as a proxy for trace metals. Therefore, atmospheric deposition rates of sulfate
will be used to estimate the relative distribution of atmospheric deposition of lead
ﬁ'om the burning of coal and oil. Sulfate deposition rates will not, however, represent
atmospheric deposition from the burning of leaded gasoline, which is known to be an
important source of lead to the environment in the 19703 (Iskander and Keeney, 1974;
Edgington and Robbins, 1976; Eisenreich et al., 1986; Graney et al., 1995; Callender
and vanMetre, 1997; Benoy and Kalaff, 1999; Heyvaert et al., 2000).

The watershed characteristics that represent possible sources include Toxic
Release Inventory data, population density, and land cover. Releases of toxic
chemicals are required to be reported by the Emergency Planning and Community
Right-to-Know Act of 1986 (EPCRA), including the location and quantity of release.
These data represents point sources that may be unique to a particular lake. Watershed
population density and urban land cover both represent sources related to human
development, including: releases from the wear of automobile parts (Rice, 1999),
automobile emissions (Callender and vanMetre, 1997; Rice, 1999), road dust (Shahin
et al., 2000), wastewater, and household and yard waste (Callender and Rice, 2000).
Agricultural land use represents pesticide and fertilizer usage (Muhlbaier and Tisue,
1981; Rice, 1999), and forest and wetland land cover correspond to areas that should

have less anthropogenic sources.

168

Each of the landscape characteristics representing sources and pathways was
quantiﬁed for the entire watershed, but it is conceivable that sources closer to the lake
will be more important than those ﬁrrther away (Comeleo et al., 1996; Richards et al.,
1996; Basnyat et al., 1999). Previous work on relating water quality parameters and
aquatic sediments provides mixed results on the success of quantifying watershed
characteristics for the total watershed versus a portion of the watershed Close to the
waterbody. In some cases the best predictions of water quality variables occurred
using whole watershed data, and other cases the predictions were better when
proximity to the waterbody was taken into account (Comeleo et al., 1996; Richards et
al., 1996; Johnson et al., 1997; Basnyat et al., 1999). To study if landscape closer to
the lake is more inﬂuential on metal accumulation patterns in lakes than those further
away, the study variables (land cover, K factor, and slope) were characterized in a 100
m buffer around the lake (Richards et al., 1996; Johnson et al., 1997). A second
approach to evaluating the importance of proximity to the lake was also used for land
cover, and each land cover type was weighted by the inverse of the distance to the lake
along the ﬂow path, so that land cover close to the lake was weighted highly (Comeleo
et al., 1996). Landscape characteristics within the 100 m buffer, and measures of land
cover that were ﬂow path weighted were then correlated to metal accumulation rates
in the same manner as the whole watershed characteristics.

Historical sources of lead have been relatively well deﬁned, with the dominant
sources including the burning of coal and leaded gasoline (Edgington and Robbins,
1976; Eisenreich et al., 1986; Graney et al., 1995; Callender and vanMetre, 1997;

Callender and Rice, 2000), but current sources are not well deﬁned (Graney et al.,

169

1995) and need further investigation. The analysis of data from the 19703 and the
19903 will allow the comparison of a time period with known sources to a period
when further investigation is warranted. Additionally, the impact of environmental
legislation in the 19703 can be evaluated both by examining changes in the
anthropogenic accumulation rates and the relative inﬂuence of the physical, nansport

and source characteristics between the two time periods.

Methods

Four replicate sediment cores were collected from each of 16 lakes in
Michigan, U.S.A. (Figure 1) from 1999-2002 using a MC-400 Lake/Shelf Multi-corer
(Ocean Instruments, San Diego, CA) deployed from either the EPA Research Vessel
Mudpuppy or the Michigan Department of Environmental Quality Monitoring Vessel
Nibi. Cores were immediately extruded and sectioned at 0.5 cm intervals for the top
5-8 cm, and at 1 cm intervals for the remainder of the core. 2|”Pb was measured on
one sub-core from each lake at the Freshwater Institute in Winnipeg, Manitoba,
Canada, under the direction of Paul Wilkinson, to determine sedimentation rates,
focusing factors, and sediment ages. Three different models were used to extract dates
from 21“Pb data: the constant ﬂux, constant sedimentation model (CF:CS) (Golden et
al., 1993), the segmented CF :CS model (SCF:CS) (Heyvaert et al., 2000), and the
constant rate of supply model (CRS) (Oldﬁeld and Appleby, 1984; Sanchez-Cabeza et
al., 2000). The choice of model was determined based on the location of the mCs
peak and the stable lead peak (Y ohn et al., 2004). The focusing factor for each lake

was calculated by dividing the actual 2'on inventory by the expected (Golden et al.,

170

1993). This factor is used to account for the tendency of ﬁne-grained materials to be

transported to the deepest part of the lake (Hermanson, 1998).

 

 

4 Gratiot
Round
' A
Mullett
, Torch
Crystal B ' Higgins
Cadillac 0 ' H°U9ht00
Littleﬁeld .
.r Crystal M
Gull . O 9353
Paw Paw — . Whitmore

 

Figure 1. Location of lakes sampled in Michigan, U.S.A.

Sediments for metals analysis were ﬁ'ozen for storage, freeze-dried and
digested by nitric acid in a CEM-MDS-81D microwave (CEM, Matthews, NC; EPA
Method 3051). Standard reference material (N IST SRM 8704 Buffalo River
Sediment) and procedural blanks were processed. Three replicate digestions were
performed on two samples ﬁ'om each lake. The concentrated-acid digests were
ﬁltered through an acid-washed, distilled-deionized water (DDW) rinsed 0.40 pm
polycarbonate ﬁlter (Nuclepore). Samples were then analyzed using a Micromass
Platform inductively coupled plasma mass spectrometer with hexapole technology

(ICP-MS-HEX). Sediments were analyzed for a suite of metals and metalloids

171

including Mg, Al, K, Ca, Ti, V, Cr, Mn, Fe, Cu, Zn, As, Sr, Mo, Cd, Ba, Pb, and U.
Lead accumulation rates were determined from mass sedimentation rates and lead
concentrations. Anthropogenic lead concentrations were determined from total lead
concentrations with the assumption that natural lead : aluminum ratios remain constant
over time, and additional lead is due to human inﬂuences (Bruland et al., 1974;
Koelmans, 1998; Tuncer et al., 2001). Anthropogenic accumulation rates were
calculated from these concentrations and mass sedimentation rates determined by the
21(’Pb. These accumulation rates were then divided by the focusing factor, resulting in
focusing corrected anthropogenic accmnulation rates. These rates were used in further
comparisons among lakes and correlation calculations.

Watershed characteristic data were acquired using ArcView v. 3.2 and ArcInfo
v. 8.2, and are described below.

Lake area, watershed area, watershed: lake area ratio (Ianz): Lake areas were
taken from Michigan Department of Conservation bathymetric maps. Watersheds
were delineated around each of the sixteen lakes of interest (Y ohn et al., 2004) from
30 m digital elevation models (National Aeronautics and Space Administration, 2002),
with the study lake not considered as part of the watershed. Cass and Whitmore Lake
watersheds were deﬁned using 90 m SRTM data because 30 m data were not available
for this region of the state. If the watershed of a lake contained another lake (e.g.,
Mullett Lake watershed contains Burt Lake), the watershed of the second lake was
removed from the watershed of the study lake.

K factor: The K factor is a unitless variable that is a relative measure of the

susceptibility of bare cultivated soil to erosion. K factors were calculated from the

172

State Soil Geographic (STATSGO) database (U .S. Department of Agriculture, 1994).
Area weighted averages were used to determine the average slope and K factor for
each watershed.

Slope (degrees): Average slope of the land within the watershed was
calculated in ArcInfo using the same digital elevation models that were used for
determination of the watershed.

Sulfate deposition (kg ha'1 y"): Sulfate deposition data were collected for 10
sites from Michigan, Wisconsin and Illinois from the National Acid Deposition
Program database (National Atmospheric Deposition Program (NRSP-3)/National
Trends Network, 2003). Data were collected from all available years ﬁ'om each site,
but ﬁnal data are from 1990—2000. Very few stations collected data ﬁ'om 1970-1980,
making it difﬁcult to estimate sulfate deposition values for this time period. However,
the patterns of sulfate deposition appear to remain the same within the state over time,
therefore the 1990-2000 data were also used to represent the spatial gradient of
atmospheric deposition ﬁ'om 1970-1980. Inverse distance weighting was used to
estimate sulfate deposition values for each watershed. _

Population @eople) and population density (people km'z): Census data at the
census tract scale were used for population calculations. Watersheds generally only
encompassed a portion of each census tract; therefore, dasymetric mapping was used
to estimate population distribution. Areas that are state owned (Michigan Department
of Natural Resources, 2000) or areas covered by lakes (Michigan Department of
Natural Resources, 1992) were considered as non-livable areas. It was assumed that

the population was evenly distributed throughout the remaining livable area within

173

each census tract. This assumption is does not reﬂect the tendency of people to live

near a lake, and therefore population densities are probably underestimated. Census
data ﬁ'om 1970 and 1980 were averaged to represent the 19703 time period, and data
from 1990 and 2000 were averaged to represent the 19903 time period.

Land cover (kmz, %): Land use/cover data were collected for 1978 (Michigan
Department of Natural Resources, 1999) and 1997-2001 (Michigan Department of
Natural Resources, 2003). Land cover groupings were different for the two time
periods, therefore the Level 1, or coarsest, land cover groupings were used so that the
two time periods could be compared. Total area in the watershed and percentage land
covers were calculated for each watershed, and for a 100 m buffer around the lake for
each of the two time periods. 19703 data were taken from Michigan Resource
Information System land cover maps; coverages derived from 1:24,000 color-inﬁared
and black and white aerial photographs. 19903 data were taken from 1997-2001
IFMAP/GAP land cover maps; raster dataset with 30 m cell size was derived from
Landsat Thematic Mapper imagery.

Flow weighted [and cover: The ﬂow path (m2), or distance from the lake along
the path that water would ﬂow, for each grid cell in each watershed was determined
using the FLOWLENGTH command in ArcInfo. Inverted ﬂow path was then calculated
(l/ﬂowlength). Separate grid coverages were created for each land cover, with a value
of 1 for each cell that contains that land cover, and a value of 0 for all other land
covers. Each of these coverages was multiplied by the inverse ﬂow path coverage and
summed. This was completed for each land cover for each time period. Additionally,

the same calculations were performed using the inverse ﬂow path squared (l/dz).

174

Percentages were calculated dividing the result for one land cover type by the sum for
all land covers.

Toxic Release Inventory (T R1): Toxic release inventory sites that released
copper, cadmium, lead, or zinc for all of Michigan were identiﬁed from
http://www.epa.gov/triexplorer/. These sites were plotted spatially, and watersheds of
each of the study lakes were examined to determine if any TRI sites were present
within the watershed. No TRI sites were present in any of the watersheds; therefore
these data were not used.

Variables were log transformed to improve the distribution, with the exception
of sulfate deposition, where log transformation resulted in a skewed distribution. Data
that were outliers were removed to prevent these points ﬁom becoming overly
inﬂuential during correlation analysis (Y ohn et al., in preparation). No data were
removed from variables with high correlations with lead except slope in the 100 m
buffer for Crystal B Lake.

Correlation coefﬁcients were calculated between lead accumulation rates and
watershed characteristics using SYSTAT 10. Multiple regression was used to
determine the two watershed characteristics that best explained the variance in lead
accumulation rates for each time period. The number of predictor variables was
limited to two due to the small number of lakes sampled (Hair, 1998). All possible
pairs of the ten variables with the highest correlation coefﬁcients were modeled using
multiple regression to determine which pair had the highest 1'2 value. Only models

where both variables had a p < 0.05 were considered.

175

Results and discussion

Patterns of lead deposition

Anthropogenic accumulation rates of lead in most of the study lakes begin to
increase between 1850 and 1900, continue to increase until the 19703 and decrease to
the present (Figure 2). The similarity among lakes in terms of when the ﬁrst
signiﬁcant anthropogenic inputs are detected (late 18003) and peak accumulation rates
(mid 19703) suggests that a common source for lead existed throughout this time
period in Michigan, but with varying intensity in different lakes. This is consistent
with results from many studies, that suggest atmospheric deposition of lead from coal
and leaded gasoline in particular were dominant sources tmtil the 19703 (Eisenreich et
al., 1986; Graney et al., 1995; Long et al., 1995; Callender and vanMetre, 1997;
Callender and Rice, 2000). The overall decrease in lead accumulation rates from the
19703 to the 19903 is mainly attributed to a reduction in lead inputs to the environment
from the enactment of environmental legislation of the 19703 (Clean Air Act)
(Eisenreich et al., 1986; Callender and vanMetre, 1997). However, the relative
decrease in accumulation rates from the 19703 until the present varies greatly among
lakes (Figures 2, 3), with some lakes decreasing signiﬁcantly (e.g., Gull Lake) and
other lakes decreasing very little or increasing (e.g., Crystal M, Cadillac). The
differing responses among lakes may indicate that while sources that were dominant in
the 19703 have been reduced, other sources have emerged. These may have been
present in the 19703 but masked by the dominant sources, or may represent new

sources.

176

 

 

[ T-a-Cadilac
j |+Cm
i—e—CrystalB
' §+Crysinl~l
E +Elt
+Gl'atiot
+th
* E—-—l~iggins
I 2+?!»
g—e—Litllelleld
. ;--Mu-eu
l : s—PawPaw
. . 2 1 . l 1 ea...
0 20000 40000 60000 60000100000120000 i+roreh

 

 

Lead anthro accum rates (pa in2 y")

Figure 2. Anthropogenic accumulation rates of lead (pg 111’2 y") in 16 Michigan
lakes.

 

 

 

 

5.0 i r
is. Higher in 19909
'E
3 . -
c 4.5 *- e .9
3
3 o
3 , ~ .
§ 40 - ° —
'8 o
2 ﬂ
3 0
§ - Higher in 1970s
92 3.5 1 l

3.5 4.0 4.5 5.0

19703 109 lead accumulab’on (pg m“2 y")

Figure 3. Log of average anthropogenic lead accumulation rates from the period 1990
- 2000 plotted versus average rates from 1970 - 1980.

177

Lead inputs - watershed characteristics relationships

There were no signiﬁcant relationships (p < 0.05) between any of the physical
characteristics (lake size, watershed size, lakezwatershed ratio) and anthropogenic
accumulation rates of lead in the 19703 or the 19903 (Table 1). Additionally, there
were no Toxic Release Inventory (TRI) sites located in any of the study watersheds,
suggesting that point source inputs from major industry should not be a source for any
of the lakes.

There is a strong positive correlation between anthropogenic lead inputs and
population density during both time periods, with the higher correlation in the 19903.
This is similar to other studies that have found relationships between total lead
concentrations in stream sediments and population density (Callender and Rice, 2000),
and between the sum of copper, mercury, lead and zinc concentrations in stream
sediments and population density (Rice, 1999).

Atmospheric deposition correlates positively with sulfate deposition rates
during both time periods, with a higher correlation in the 19703. Sulfate deposition
was chosen to act as a proxy atmospheric transport from sources such as the burning
of coal (Graney et al., 1995; Bemer and Berner, 1996), and the positive correlation
may represent the importance of extensive coal burning that occurred in the Gary,
Indiana, and Chicago, Illinois industrial region, as well as in the industrial region near
Detroit. Coal burning also probably occurred at smaller quantities throughout the state
from residential use, power production and smaller industrial sites. The lower
correlation between lead accumulation rates and sulfate deposition rates in the 19903

is not surprising because the 19703 the Clean Air Act has served to reduce

178

contaminants released to the air, and there has been a general decline in steel
production since the 19803 (Perkins et al., 2000).
Table 1. Correlation coefﬁcients (r) between anthropogenic lead accumulation rates

from 1970-1980 (19703) and 1990-2000 (19903) and watershed characteristics. Only
correlations that are statistically signiﬁcant (p < 0.05)are shown.

 

 

 

19703 19903

watershed 100 m buffer watershed 100 m buffer
Watershed area NA NA
Lake area NA NA
Watershedzlake area NA NA
ratio
Population density 0.825 NA 0.851 NA
Population (total #) 0.523 NA NA
Sulfate deposition 0.798 NA 0.722 NA
K factor
Slope -0.881 -0.919 -0.755 -0.828

 

The K factor measure of erodibility also showed no correlation with
anthropogenic lead inputs for either time period. This suggests that the pathway of
erosion, at least as evaluated by the erosion tendency of the soil type, is not controlling
lead deposition patterns through the state.

However, both average slope of the watershed and within a 100 m buffer of the
lake have a high and negative correlation with anthropogenic lead accumulation rates.
Because of the negative correlation, slope of the land clearly does not represent the
transport of lead to the lakes through erosion (anticipated positive correlation). What
source or process this parameter is acting as a proxy for is unclear. One possibility is
that slope of the land may act as a measure of human development, because areas with

steep slopes may be less likely to be urbanized. However, the watersheds of most of

179

the study lakes have relatively shallow slopes (< 5 degrees), which would not impede
development of the land.

Because sources of lead to the environment in the 19703 have been extensively
studied, we may attempt to use the results of previous work to interpret the very high
correlation between lead accumulation rates and slope in the 100 m buffer. The
burning of leaded gasoline is well documented to be the dominant source for lead to
the environment during this time period (for example: Eisenreich et al., 1986; Graney
et al., 1995; Callender and vanMetre, 1997), and temporal patterns of lead deposition
in many of the study lakes match production records of leaded gasoline (Y ohn et al.,
2004). Additionally, measurements of atmospheric deposition in the 19703 in the
region are similar to the anthropogenic accumulation rates of lead in the study lakes
(Gatz, 1975; Edgington and Robbins, 1976; Cole et al., 1990; Callender and vanMetre,
1997). It is therefore conceivable that slope within the 100 m buffer is somehow
acting as a proxy for the burning of gasoline, or at least the regional pattern of
atmospheric deposition of lead.

There is no obvious causal relationship between the two, but slope within the
100 m buffer and the atmospheric deposition of lead from the burning of gasoline may
coincidentally have a similar spatial pattern across the state. Southern Michigan has
relatively little topography, and slope of the land within the watersheds are shallow.
In the northern portion of the state, slopes tend to be steeper, with the steepest in the
Upper Peninsula. The overall population distribution in Michigan follows a similar
pattern, with the greatest population density in the south and the least in the north.

This may lead to a similar distribution of automobiles releasing lead as slope within

180

the 100 m buffer. One might anticipate that population density should be a better
predictor than slope if automobile exhaust were the source of lead, but lead emitted to
the atmosphere from automobile exhaust may be deposited locally within the
watershed or may move outside of the watershed. As a result, even a lake with little
traffic within the watershed may have high atmospheric deposition of lead if there are
many roads near the watershed. Therefore population density within a particular
watershed may be less important than the overall distribution of automobiles
throughout the state. Overall, it remains unclear why slope within the 100 m buffer
has such a high correlation with lead accumulation rates in the 19703, but may be due
to a coincidental similar spatial distribution to population throughout the state. There
is no evidence that there is a cause and effect relationship between slope and lead
accumulation rates.

There is a positive correlation between percentage of urban land use and lead
accumulation rates during both time periods (Table 2), but the correlation is much
stronger in the 19903. The higher correlation in the 19903 suggests that local sources,
which occur within the watershed, are more important than regional atmospheric
sources in the 19903. This is supported by the higher correlation between lead and
population density in the 19903, and the lower correlation between sulfate and lead
accumulation rates.

There is a negative correlation between lead accumulation rates and forest
during both time periods, and rangeland in the 19703 (Table 2). Forest was the
dominant land cover type for many of the watersheds, and should not act as a source

of lead. There is also a negative relationship between agricultural land cover and lead

181

accumulation rates for both time periods, suggesting that the use of pesticides did not
contribute signiﬁcant amounts of lead to the study lakes.

While urban land cover consistently had the highest correlation with lead when
the percentage of land cover in the entire watershed was used, other land covers
correlated higher with different proximity measures. Agricultural and forest land
cover had the highest correlations when ﬂow inverted squared values were used, and
rangeland and wetlands also had the highest correlations when using measures that
included ﬂow distance to the lake (Table 2). Land cover composition either in the 100
m buffer or when weighted by distance to the lake along the ﬂow path differed greatly
from the composition of the entire watershed. In general, for both time periods, urban
land cover was higher in the 100 m buffer and ﬂow weighted measures than in the
entire watershed, and in many cases signiﬁcantly higher. For example, in the
watershed of Gull Lake in the 19703 there was 9% urban in the total watershed and
84% urban in 100 m buffer. For most lakes, urban land cover was dominant within
the 100 m buffer, suggesting concentrated developments near the lake compared to the
rest of the watershed. Agriculture tended to be much lower in the 100 m buffer and

ﬂow weighted measures, and forest and rangeland were generally lower.

182

Table 2. Correlation coefficients (r) between anthropogenic lead accumulation rates
from 1970-1980 (19703) and 1990-2000 (19903) and land cover variables. Only
correlations that are statistically signiﬁcant (p < 0.05) are shown.

 

 

 

 

 

 

urban agriculture openland/range
Land cover measure 19703 19903 19703 19903 19703 19903
Total watershed % 0.627 0.899
Total watershed area
100 m buffer % 0.535 0.737
100 m buffer area -O.588
Flow inverted % 0.771
Flow inverted value -0.599 -0.642 -0.749
Flow inv. sq. % 0.675
Flow inv. sq. value -0.659 -0.714 -0.668

forest wetlands

Land cover measure 19703 19903 19703 19903
Total watershed % -O.782 -0.642 0.683
Total watershed area
100 m buffer % -O.743 -0.618
100 m buffer area 0681
Flow inverted % -0.765 -0.657 0.686
Flow inverted value -0.762 -0.732
Flow inv. sq. % -0.666 -0.575
Flow inv. sq. value -0.829 -0.782

 

Multivariate regression

Interpretation of correlations between watershed characteristics and lead

accumulation rates is difﬁcult, because many of the watershed Characteristics are

correlated to each other (e.g., population density and % urban within the watershed).

Therefore, to determine the best combination of predictors of lead to the environment,

it is not appropriate to choose the variables that are most highly correlated with lead,

because the variables may be correlated to each other and actually act as a proxy for

similar sources or processes. To avoid this problem, we used multivariate regression

to select the two watershed characteristics that together best described the variation in

lead accumulation rates among lakes. The number of watershed characteristics

183

included in the model was limited to two because of the small number of lakes
sampled.

In the 19703, the two watershed characteristics resulting in the highest multiple
1'2 value were the average slope within a 100 m buffer, and atmospheric deposition
rates of sulfate (r2 = 0.922, p = < 0.000) (Figure 4). This model is represented by the
equation:

Log (LEAD70rw) = 3.969 + 0.43 x $04., - 0.823 x 109 (SLOPEthm)

Where:

LEAD70fw represents the average focusing corrected anthropogenic
accumulation rates in the 19703 (pg m'2 y"),

SO4d represents sulfate deposition rates in the watershed (kg ha'1 y'l), and

SLOPEIOOm represents the average slope of the land within a 100 m buffer of
the lake (degrees).

The standardized coeﬂicient, which is not dependant on the scale of the
variable, can be used to determine the relative importance of the two variables in the
model. In the 19703, slope within the 100 m buffer is the more important variable
(standardized coeﬂicient of SLOPElogm = -0.641, SO4d = 0.408).

These two variables could be interpreted as representing atmospheric
deposition ﬁom leaded gasoline (the dominant source, represented by slope within 100
m buﬁ’er) and coal burning (represented by S04 deposition). While this interpretation

is consistent with previous studies on the sources of lead to the environment during

184

this time period, the cause of the relationship between slope within the 100 m buffer

and lead deposition rates is unclear and needs further investigation.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

8 5.0 T l 8 5.0 I T
Us 0 [s e
s . 92 -
'- 4.5 . - ‘- 4.5— - -
E [— o O E Q
3 3
3 . . 8 a
CU . a O
a ’ 2 °
5 40 '- . e - 5 4O '— . s . _
C C
W s (U s
.8 O .3 O
2 52
3.5 L J 3.5 l l
:8; -0.5 0.0 . 0.5 1.0 g 5 10 15 20
Log slope in 100 m buffer Sulfate depostion
a:
‘- 4.6 P . 0 - 2 4.6 - s . -
2 . 0 .
E 44 E 44
E . ’- 0 -‘ E ' -- . —t
3i ' 3 °
8 4.2 - . - 8 4.2 - . -
e a . ’ '
5 4.0 - - a 5 4.0 - - -
g s .e s s e .
g 3.6- , — g 3.6- , a
2 2
3’ 3.8 '1 1 3:35 i 1 °
.1 -1 0 1 2 _r -0.5 0.0 0.5 1.0

Log percentage urban in watershed

Log slope in 100 m buffer

Figure 4. Log lead anthropogenic accumulation rates (pg m‘2 y") for the 19703 and
19903 versus the two variables for each time period that together explains the greatest
variance in lead accumulation rates.

185

In the 19903, percent urban land cover in the watershed and average slope in a
100 m buffer best predicted lead accumulation rates (r2 = 0.879, p < 0.000) (Figure 4):
Log (LEADQOM) = 3.951 + 0.464 x log (URBAN90) - 0.465 x log (SLOPEmOm)
Where:

LEAD90fmil is the average focusing corrected anthropogenic accumulation rate
in the 1990s (pg 111'2 y’l),

URBAN9O is the percentage urban land cover in the watershed in the 19903,
and

SLOPEloom is the average slope within a 100 m buffer of the lake (degrees).

In the 1990s, urban land cover has a higher standardized coefﬁcient, and this

variable contributed more to the model (URBAN 90 = 0.624, SLOPEIOOm = -0.405).

The high correlation between percentage urban land use within the watershed
and lead deposition rates indicate that currently there is a source of lead that is present
throughout the state that is related to urban land cover and acts at the watershed scale,
such as municipal wastewater, or road dust. Slope within a 100 m buﬂ'er may
continue to represent regional atmospheric deposition, similar to the 19703, but this
source is much less important in the 19903, and has a lower standardized coefﬁcient.
The importance of slope in the 100 m buffer in the model indicates that there may be
continued sources of lead to the atmosphere, including coal ﬂy ash and smelting
(Callender and Rice, 2000).

The change in the variables included in the model from the 19708 and the

19903, and the decrease in importance of the slope in the 100 m buffer may reﬂect a

186

decrease in the importance of atmospheric deposition, and an increase in the
importance of watershed scale sources, which is consistent ﬁom the conclusions made
from the correlation coefﬁcients. This is also supported by measurements of
atmospheric deposition rates of lead in the region, which are similar to accumulation
rates of lead in lake sediments in the 19703 (Gan, 1975; Edgington and Robbins,
1976; Cole et al., 1990; Callender and vanMetre, 1997), but cannot account for all the
lead deposited in lake sediments in the 19903 (Hoff et al., 1996; Sweet et al., 1998;
Simonetti et al., 2000). The reduction of atmospheric sources is consistent with
previous work showing the impacts of the Clean Air Act (Callender and vanMetre,

1 997).

For both time periods, two watershed characteristics (slope within a 100 m
buffer and sulfate deposition in the 1970s, and percentage urban land cover and slope
within a 100 m buffer in the 1990s) were able to explain greater than 85% of the
variance in anthropogenic lead deposition rates in the study lakes. These correlations
are very high given the variety of lakes examined and error inherent in estimating both
anthropogenic lead accumulation rates and watershed characteristics. Although this
technique does not identify the exact sources or processes that are resulting in lead
deposition in the study lakes, it successfully distinguished changes in sources over
time, and identiﬁed urban areas for ﬁirther study to determine the actual current

somces of lead.

187

References

Basnyat, P., Teeter, L.D., Flynn, K.M. and Lockaby, B.G., 1999. Relationships
between landscape characteristics and nonpoint source pollution inputs to
coastal estuaries. Environmental Management 23, 539-549.

Benoy, GA. and Kalaﬂ‘, J ., 1999. Sediment accumulation and Pb burdens in
submerged macrophyte beds. Limnology and Oceanography 44, 1081-1090.

Berner, E.K. and Berner, R.A., 1996. Global environment : water, air, and
geochemical cycles. Prentice Hall, Upper Saddle River, N.J., xiv, 376 pp.

Bruland, K.W., Bertine, K., Koide, M. and Goldberg, ED, 1974. History of metal
pollution in Southern California costal zone. Environ. Sci. Technol. 8, 425-
432.

Callender, E. and Rice, K.C., 2000. The urban environmental gradient: Anthropogenic
inﬂuences on the spatial and temporal distributions of lead and zinc in the
environment. Environ. Sci. Technol. 24, 232-238.

Callender, E. and vanMetre, P.C., 1997. Reservoir sediment cores show US lead
declines. Environ. Sci. Technol. 31, A424-A428.

Catallo, W.J., Schlenker, M., Gambrell, RP. and Shane, B.S., 1995. Toxic chemicals
and trace metals from urban and rural Louisiana lakes: recent historical proﬁles
and toxicological signiﬁcance. Environ. Sci. Technol. 29, 1439-1446.

Cole, K.L., Engstrom, D.R., Futyma, RP. and Stottlemyer, R., 1990. Past
Atmospheric Deposition of Metals in Northern Indiana Measured in a Peat
Core from Cowles Bog. Environmental Science & Technology 24, 543-549.

Comeleo, R.L., Paul, J .F., August, P.V., Copeland, J ., Baker, C., Hale, SS. and
Latimer, R.W., 1996. Relationships between watershed stressors and sediment
contamination in Chesapeake Bay estuaries. Landscape Ecology 11, 307-319.

Detenbeck, N.E., Johnston, CA. and Niemi, G.J., 1993. Wetland Effects on Lake
Water-Quality in the Minneapolis St-Paul Metropolitan-Area. Landscape
Ecology 8, 39—61 .

Dillion, PJ. and Evans, R.D., 1982. Whole-lake lead burdens in sediments of southern
Ontario, Canada. Hydrobiologia 9], 121-130.

Edgington, D.N. and Robbins, J .A., 1976. Records of lead deposition in Lake
Michigan sediments since 1800. Environ. Sci. Technol. 10, 266-274.

188

Eisenreich, S.J., Metzer, N.A. and Urban, N.A., 1986. Response of atmospheric lead
to decreased use of lead in gasoline. Environ. Sci. Technol. 20, 171-174.

Eisenreich, SJ. and Strachan, W.M.J., 1992. Estimating atmospheric deposition of
toxic substances in the Great Lakes - an update, Canada Centre for Inland
Waters, Burlington, Ontario, January 31-Febuary 2, 1992.

Gatz, D.F., 1975. Pollutant aerosol deposition into southern Lake Michigan. Water Air
Soil Poll 5, 239-251.

Gatz, D.F., Bowersox, V.C. and Su, J ., 1989. Lead and cadmium loadings to the Great
Lakes from precipitation. J. Gt. Lakes Res. 15, 246-264.

Golden, K.A., Wong, C.S., Jeremiason, J .D., Eisenreich, S.J., Sanders, M.G.,
Hallgren, J ., Swackhammer, D.L., Engstrom, DR. and Long, D.T., 1993.
Accumulation and preliminary inventory of organochlorines in Great Lakes
sediments. Wat. Sci. Technol. 28, 19-31.

Goodarzi, F., Sanei, H. and Duncan, W.F., 2001. Monitoring the distribution and
deposition of trace elements associated with a zinc-lead smelter in the Trail
area, British Columbia, Canada. J. Environ. Monit. 3, 515-525.

Graney, J .R., Halliday, A.N., Keeler, G.J., Nriagu, J .O., Robbins, J .A. and Norton,
S.A., 1995. Isotopic record of lead pollution in lake sediments from the
northeastern United States. Geochim. Cosmochim. Acta 59, 1715-1728.

Hair, J .F ., 1998. Multivariate data analysis. Prentice Hall, Upper Saddle River, N.J.,
730, 12 pp.

Hermanson, M.H., 1998. Anthropogenic mercury deposition to arctic lake sediments.
Water, Air, Soil Pollut. 101.

Heyvaert, A.C., Reuter, J .E., Slotton, D.G. and Goldman, CR, 2000.
Paleolimnological reconstruction of historical atmospheric lead and mercury
deposition at Lake Tahoe, California-Nevada. Environ. Sci. Technol. 34, 3588-
3597.

Hoﬁ‘, R.M., Strachan, W.M.J., Sweet, C.W., Chan, C.H., Shackleton, M., Bidleman,
T.F., Brice, K.A., Burniston, D.A., Cussion, S., Gatz, D.F., Harlin, K. and
Schroeder, W.H., 1996. Atmospheric deposition of toxic chemicals to the
Great Lakes: A review of data through 1994. Atmos. Environ. 30, 3505-3527.

Iskander, I.K. and Keeney, D.R., 1974. Concentration of heavy metals in sediment
cores from selected Wisconsin Lakes. Environ. Sci. Technol. 8, 165-170.

189

Johnson, L.B., Richards, C., Host, GE. and Arthur, J.W., 1997. Landscape inﬂuences
on water chemistry in Midwestern stream ecosystems. Freshwater Biology 37,
193-205.

Koelmans, A.A., 1998. Geochemistry of suspended and settling solids in two
freshwater lakes. Hydrobiologia 364, 15-29.

Long, D.T., Eisenreich, SJ. and Swackhammer, D.L., 1995. Metal Concentrations in
Sediments of the Great Lakes, Great Lakes Protection Fund.

Mecray, E.L., King, J .W., Appleby, P.G. and Hunt, A.S., 2001. Historical trace metal
accumulation in the sediments of an urbanized region of the Lake Champlain
watershed, Burlington, Vermont. Water Air Soil Poll 125, 201-230.

Michigan Department of Natural Resources, 1992. MIRIS base data. MDNR, Lansing,
MI.

Michigan Department of Natural Resources, 1999. 1978 Land cover / use. MDNR,
Lansing, MI.

Michigan Department of Natural Resources, 2000. State ownership. MDNR, Lansing,
MI.

Michigan Department of Natural Resources, 2003. IFMAP/GAP land cover. MDNR
Forest, Mineral, and Fire Management Division, Lansing, MI.

Muhlbaier, J. and Tisue, G.T., 1981. Cadmium in the Southern Basin of Lake-
Michigan. Water Air Soil Poll 15, 45-59.

National Aeronautics and Space Administration, 2002. Shuttle Radar Topography
Mission (SRTM) Spatial Dataset. US. Geological Survey, Sioux Falls, SD.

National Atmospheric Deposition Program (NRSP-3)/National Trends Network, 2003.
NADP Program Ofﬁce, Illinois State Water Survey, 2204 Griffith Drive,
Champaign, IL 61820.

Nriagu, J.O. (Editor), 1978. The Biogeochemistry of Lead in the Environment, 1A.
Elsevier, New York, 422 pp.

Nriagu, J .O., 1989. A Global Assessment of Natural Sources of Atmospheric Trace-
Metals. Nature 338, 47-49.

190

 

01c

Ri.

Se

Oldﬁeld, F. and Appleby, P.G., 1984. Empirical testing of 210Pb—dating models for
lake sediments. In: E.Y. Haworth and J .W.G. Lund (Editors), Lake Sediments
and Environmental History. University of Minnesota Press, Minneapolis, pp.
93-124.

Perkins, S.M., Filippelli, GM. and Souch, C.J., 2000. Airborne trace metal
contamination of wetland sediments at Indiana Dunes National Lakeshore.
Water Air Soil Poll 122, 231-260.

Rice, K.C., 1999. Trace-element concentrations in streambed sediment across the
conterminous United States. Environ. Sci. Technol. 33, 2499-2504.

Richards, C., Johnson, LB. and Host, GE, 1996. Landscape-scale inﬂuences on
stream habitats and biota. Canadian Journal of Fisheries and Aquatic Sciences
53, 295-311.

Sanchez-Cabeza, J .A., Ani-Ragolta, I. and Masque, P., 2000. Some considerations of
the Pb-210 constant rate of supply (CRS) dating model. Limnology and
Oceanography 45, 990-995.

Sanudo-Wilhelmy, SA. and Gill, G.A., 1999. Impact of the Clean Water Act on the
levels of toxic metals in Urban Estuaries: the Hudson River estuary revisited.
Environ. Sci. Technol. 33, 3477-3481.

Shahin, U., Yi, S.M., Paode, RD. and Holsen, T.M., 2000. Long-term elemental dry
deposition ﬂuxes measured around Lake Michigan with an automated dry
deposition sampler. Environmental Science & Technology 34, 1887-1892.

Simonetti, A., Gariepy, C. and Carignan, J ., 2000. Pb and Sr isotopic compositions of
snowpack from Quebec, Canada: Inferences on the sources and deposition
budgets of atmospheric heavy metals. Geochim. Cosmochim. Acta 64, 5-20.

Soranno, P.A., Hubler, S.L., Carpenter, SR. and Lathrop, R.C., 1996. Phosphorus
loads to surface waters: A simple model to account for spatial pattern of land
use. Ecological Applications 6, 865-878.

Souch, C.J., Filippelli, G.M., Dollar, N., Perkins, S. and Mastalerz, M., 2003.
Accumulation rates of airborne heavy metals in wetlands. Phys. Geogr. 23, 21-
43.

Sweet, C.W., Weiss, A. and Vermette, S.J., 1998. Atmospheric deposition of trace
metals at three sites near the Great Lakes. Water Air Soil Poll 103, 423-439.

191

Tuncer, G., Tuncel, G. and Balkas, T.I., 2001. Evolution of metal pollution in the
Golden Horn (Turkey) sediments between 1912 and 1987. Marine Pollution
Bulletin 42, 350-360.

US. Department of Agricultm'e, 1994. State Soil Geographic (STATSGO) database
for Michigan. US. Department of Agriculture, Natural Resources
Conservation Service, Fort Worth, Texas.

Winchester, J .W. and Nifong, GD, 1971. Water pollution in Lake Michigan by trace
elements from pollution aerosol fallout. Water, Air, Soil Pollut. 1, 50-64.

Yohn, S.S., Long, D.T., Fett, J .D. and Patino, L.C., 2004. Regional versus local
inﬂuences on lead and cadmium loading in the Great Lakes region. Appl.
Geochem. 19, 1157-1175.

Yohn, S.S., Long, D.T. and Winkler, J .A., in preparation. The use of multivariate

statistics for the interpretation of sediment chronologies. Limnology and
Oceanography: Methods.

192

CHAPTER 7
IDENTIFYING CHANGES IN SOURCES FOR CADMIUM, COPPER AND ZINC
TO INLAND LAKES OVER THREE DECADES

Introduction

Sediment chronologies from lakes throughout the United States have recorded
a decrease in the inputs of many metals to the environment since the 1970s and
passing of the Clean Air and Water Acts (Graney et al., 1995; Long et al., 1995;
Callender and vanMetre, 1997; Callender and Rice, 2000; Simonetti et al., 2000a;
Mecray et al., 2001). However, despite these decreases, sediment accumulation rates
and concentrations of some toxic metals remain elevated in lake sediments (Graney et
al., 1995; Long et al., 1995; Callender and vanMetre, 1997; Mecray et al., 2001),
suggesting that historical sources of metals are still present, or that new sources have
emerged. An understanding of historical and current sources for metals to the
environment through the examination of lake sediments would both lend insight into
the effects of environmental legislation and potentially help further reduction of
contaminant inputs to the environment.

Lake sediments act as collectors of contaminants both ﬁom the local watershed
and regional atmospheric deposition and can provide information on current and
historic rates of contaminant inputs. As a result, sediments are a valuable tool in
evaluating chemical loading to the environment. There are many approaches to
understanding sources for metals to lakes, including the thorough study of one lake
(Edgington and Robbins, 1976; Kerfoot et al., 1994; Mecray et al., 2001), or the

sampling of many lakes (Dillion and Evans, 1982; Catallo et al., 1995). The

193

advantage of sampling multiple lakes is the potential identiﬁcation and reduction of
sources that inﬂuence a large area rather than just one lake. Additionally, surface
sediments may be used to identify current sources (Catallo et al., 1995; Callender and
Rice, 2000), or sediment chronologies may be used to evaluate current and historical
sources (Kerfoot et al., 1994; Long et al., 1995; Kolak et al., 1998). This study uses
sediment chronologies from multiple lakes to identify changes in the sediment
accumulation of cadmium, copper, and zinc in seventeenteen Michigan lakes over
both space and time, and to evaluate the changing sources for these metals.

A three step approach will be used to help identify sources for metals to lake
sediments and evaluate pathways through the environment for two time periods, 1970-
1980 (19703) and 1990 — 2000 (19905): 1) comparison of sediment chronologies
(anthropogenic accumulation rates) from multiple lakes, 2) estimation of the
importance of atmospheric deposition from spatial patterns and comparison to
measures of atmospheric deposition during both time periods, and 3) evaluation of
potential sources by correlations with watershed characteristics.

The ﬁrst step of this approach, the comparison of sediment chronologies,
provides information about the increase or decrease of metal inputs to lake sediments
from the 19703 to the 19903, and therefore the effectiveness of environmental
legislation. Additionally, the comparison of patterns of anthropogenic accumulation
over time among lakes may indicate if sources are inﬂuenced on a regional or
watershed scale. Similarities in proﬁles among lakes suggest that there are common
sources for that metal to many lakes, and that the increase and decrease of these

sources is similar throughout the state (regionally controlled). This may be due to

194

regional distribution of a metal (e.g., ﬁom atmospheric deposition), or from regional
controls on a source (e.g., environmental legislation). If proﬁles among lakes are
dissimilar, this suggests that either sources for this metal are different among lakes, or
that sources are similar, but the strength of the source is dependant on local factors
(e.g., traffic density or urban development in the watershed).

The second step, estimation of the importance of atmospheric deposition,
includes both the examination of spatial patterns and comparison to measured rates of
atmospheric deposition. When atmospheric deposition is the major pathway of metal
inputs, the accumulation rate of metals should be dependant more on the regional
distribution of sources than the sources present within a lake’s watershed, and should
have a regional gradient across the state. If atmospheric deposition is not signiﬁcant,
than accumulation rates should be inﬂuenced more by characteristics of the watershed
than regional location. Rates of contaminant deposition in the lake sediments can also
be compared to measured rates of atmospheric deposition in the region to further
assess the importance of this pathway. The size of the lake should not inﬂuence the
focusing corrected anthropogenic accumulation rates in the sediment, allowing
atmospheric deposition rates to be compared to sediment accumulation rates.

Finally, potential sources for metals may be identiﬁed by correlations between
anthropogenic accumulation rates and watershed characteristics. Ideally,
anthropogenic accumulation rates in the sediment could be compared to actual rates of
inputs of different sources to determine the relative importance of each source
(Edgington and Robbins, 1976), but these data are not available for most lakes.

Therefore, quantiﬁable characteristics of the watershed are chosen to represent

195

source(s) (Callender and Rice, 2000). Although these characteristics do not
necessarily identify a speciﬁc source, or the exact proportion of the metal that is from
that source, correlation with watershed characteristics can provide an indication of the
major sources for metals to lakes in a region. This method will help identify those
sources that are common to multiple lakes, and that can be represented by a watershed
characteristic.

The watershed characteristics included in this study represent physical
(watershed area, lake area, watershed: lake area ratio), transport (K factor of
erodibility, slope, pattern of sulfate deposition), and source characteristics (total
population, population density, land cover, point source data) for both time periods.
The physical characteristics may give insight into the importance of atmospheric
deposition; a lake with a small watershed to lake area ratio may be dominated by
atmospheric deposition, whereas a lake with a large ratio may be more inﬂuenced by
inputs from the watershed (Dillion and Evans, 1982).

The transport characteristics represent the pathway of metals to a lake. The
presence of a source of metals in a watershed will not be recorded in the lake
sediments unless those metals are transported to the lake. The average slope of the
watershed and K factor (measure of erodibility) were included to represent the
erodibility of soils within the watershed (Detenbeck et al., 1993). Any metal
deposited on the land from sources such as pesticides or fertilizers (Muhlbaier and
Tisue, 1981; Rice, 1999) must be eroded before reaching the lake, and may be
dependant on soil erodibility. Sulfate deposition rates were estimated to evaluate the

importance of atmospheric deposition. The atmosphere can be an important pathway

196

of metals to lakes (Winchester and Nifong, 1971; Gatz, 1975; Gatz et al., 1989;
Nriagu, 1989; Callender and vanMetre, 1997; Sweet et al., 1998; Rice, 1999;
Callender and Rice, 2000; Simonetti et al., 2000a; Goodarzi et al., 2001; Souch et al.,
2003), but accurate estimations of deposition rates of metals for each watershed are
difficult to determine. Sulfate deposition, however, has been monitored throughout
Michigan (National Atmospheric Deposition Program (NRSP-3)/National Trends
Network, 2003), and can be used to estimate the relative distribution of atmospheric
deposition due to coal and oil bmning. Both sulfate and trace metals are released with
the burning of fossil fuels (Nriagu, 1989; Bemer and Berner, 1996). Sulfate
deposition rates will not, however, represent atmospheric deposition ﬁom other
sources (e.g., copper ﬁom copper smelting, lead from the burning of leaded gasoline).

The source characteristics include Toxic Release Inventory data, population
density, and land cover. Releases of toxic chemicals are required to be reported by the
Emergency Planning and Community Right-to-Know Act of 1986 (EPCRA),
including the location and quantity of release. These data represent point source
inputs that may be unique to that lake. Watershed population density and urban land
cover both represent sources related to human development, including: releases from
the wear of automobile parts (Rice, 1999), automobile emissions (Callender and
vanMetre, 1997; Rice, 1999), road dust (Shahin et al., 2000), wastewater, and
household and yard waste (Callender and Rice, 2000). Agricultural land use
represents pesticide and fertilizer use (Muhlbaier and Tisue, 1981; Rice, 1999). Forest
and wetland land cover correspond to areas that should have few anthropogenic

80111088.

197

All of the characteristics are quantiﬁed on the watershed scale, but it is
conceivable that sources that are closer to the lake will be more important than those
further away (Comeleo et al., 1996; Richards et al., 1996; Basnyat et al., 1999).
Previous work on water quality parameters and estuary sediments provides mixed
results, with some variables predicted best with whole watershed data, and others
when proximity to the waterbody was taken into account (Comeleo et al., 1996;
Richards et al., 1996; Johnson et al., 1997; Basnyat et al., 1999), and conclusions are
mixed among studies. To address the issue of proximity, land cover, K factor, and
SIOpe were quantiﬁed in a 100 m buffer around the lake (Richards et al., 1996;
Johnson et al., 1997), and, in an alternate approach, land cover was weighted by the
inverse of the distance to the lake along the ﬂow path (Comeleo et al., 1996).

Each of these watershed characteristics was correlated separately with
anthropogenic inputs of each metal during both time periods, and the individual
correlation coefficients were examined. However, watershed characteristics may be
correlated to each other (e.g., urban land cover and population density), and may not
represent separate sources. Therefore, multiple linear regression was used to identify
the two variables that best explained the variance in each metal for each time period.
Each model was limited to two variables because of the relatively small sample size.

This overall approach should indicate whether watershed or regional scale
processes inﬂuence metal inputs, and suggest possible sources for cadmium, copper,
and zinc to Michigan lakes by correlation with watershed characteristics.

Additionally, the impact of environmental legislation in the 19705 will be shown both

198

by changes in the anthropogenic accumulation rates and by different relationships with

watershed characteristics in the two time periods.

Methods

Methods are described both in the previous chapters and in the Appendix,
however, a brief description of sampling and analysis techniques, and description of
watershed characteristics is given below.

Sediment cores were collected from 17 lakes in Michigan (Figure 1). Lakes
were selected to represent both urban and remote areas and to be distributed
throughout the state. Only lakes that had a maximum depth greater than 8 m were
considered (Yohn et al., 2003), and lakes with one basin were preferred to those with
complicated bathymetry. Sediment cores were collected ﬁom the deepest portion of
each lake using a MC-400 Lake/ Shelf Multi-corer deployed ﬁ'om the Monitoring
Vessel Nibi. Cores were then extruded and sectioned at 0.5 cm intervals for the top 8

1
cm, and at 1 cm intervals for the remainder of the core.

21”Pb was measured on one sub-core from each lake to determine
sedimentation rates, sediment ages and focusing factors (Freshwater Institute in
Winnipeg, Manitoba, Canada). Results ﬁ'om all lakes were veriﬁed using 137Cs.
Porosity and accumulated dry mass were also determined from this sub-core.

Sediments were frozen, freeze-dried and digested by nitric acid in a CEM-
MDS-81D microwave (Hewitt and Reynolds, 1990). Standard reference material
(NIST RM 8704 Buffalo River Sediment) and procedural blanks were processed to
test for accuracy and contamination. The concentrated-acid digests were ﬁltered

through an acid-washed, e-pure (Barnstead) rinsed 0.40 pm polycarbonate ﬁlter.

199

Samples were then analyzed using a Micromass Platform inductively coupled plasma
mass spectrometer with hexapole technology (ICP-MS-HEX). Sediments were
analyzed for a suite of metals and metalloids including Mg, Al, K, Ca, Ti, V, Cr, Mn,

Fe, Co, Ni, Cu, Zn, As, Sr, Mo, Cd, Ba, Pb, and U.

a} Gratiot

    
   

 

 

Imp / g ‘ , ‘K
W'tCh ' »‘ ~ﬁ'- “ Mullett
Elk . 3 Torch
r
Crystal B ' Higgins
Cadillac 0 ‘ HOUthOH
Littleﬁeld Q
. Crystal M
Gull .0 0353
Paw Paw - . Whitmore

 

Figure 1. Location of study lakes in Michigan, USA.

Calculation of anthropogenic input rates from total sediment concentrations are
described in detail in Chapter 2, and are summarized below. Anthropogenic
concentrations are determined from total sediment concentrations by normalizing to a

terrestrial element (e. g., Al) for each depth using the equation:

200

MISKnthro = Mgotal _ E x TS
Tee

Where MEG and T30 are the concentrations of the anthropogenic metal of
interest and the terrestrial element respectively in pre-anthropogenic sediments, Ts is
the concentration of the terrestrial element in sample S, MST“ is the total

concentration of the metal of interest in sample S, and MsAmmo is the anthropogenic

concentration of the metal of interest in sample S.

Accumulation rates are calculated ﬁom the anthropogenic concentrations
using:

Anthropogenic accumulation rate (pg m“2 y'1) = anthro oonc (mg/kg) x mass
sedimentation rate (9 m2 y")

The mass sedimentation rate is taken from the dating model used for that particular
lake.

A focusing factor was calculated for each lake to account for the tendency of
ﬁne-grained particles to move towards the deepest portion of the lake (Golden et al.,
1993; Hermanson, 1998). The focusing factor was calculated using the equation

where the theoretical 2'on inventory used was 0.574 liq/cm2 (Golden et al., 1993):

actual 21oPb inventory

FF = 210
theoretical Pb inventory

 

To account for diﬂ‘erent amounts of focusing that occurs among lakes, the

anthropogenic accumulation rate is divided by the focusing factor.

201

In some cases constant metal: aluminum ratios were not present in the oldest
portions of the core, and anthropogenic accumulation rates could not be calculated.
Anthropogenic accumulation rates that could not be calculated include: cadmium:
Witch Lake; copper: Imp, Littleﬁeld, Round and Torch lakes; zinc: Imp, Littleﬁeld,
Mullett, Torch and Witch lakes.

Characteristics of the watersheds of each of the study lakes were determined
using ArcView and ArcInfo, and are described below. Complete methods are
described in the Appendix. Spatial data were collected from secondary sources, and
were in, or were projected into the Michigan GEOREF coordinate system: oblique
Mercator projection, datum NAD83, spheroid GRS 1980. Slope, K factor, and land
cover were calculated both for the entire watershed and for a 100 m buffer around the
study lake.

Lake area, watershed area, watershed: lake area ratio (kmz): Lake areas were
taken from Michigan Department of Conservation bathymetric maps. Watersheds
were delineated around each of the seventeen lakes of interest (Y ohn et al., 2004),
with the study lake not considered as part of the watershed. Cass and Whitmore Lake
watersheds were deﬁned using 90 m SRTM data because 30 m data were not available
for this region of the state. If the watershed of a lake contained another lake (e.g.,
Mullett Lake watershed contains Burt Lake), the watershed of the second lake was
removed from the watershed of the study lake. Lakes are effective sediment traps
(Wetzel, 2001), and outﬂow from a lake is likely to have very low concentrations of
metals. Therefore, it is likely that the watershed of a lake within the watershed of a

study lake is unlikely to signiﬁcantly contribute metals to the study lake. The

202

watershed: lake area ratio was determined by dividing the area of the watershed by the
area of the lake.

K factor, slope corrected K factor: K factors to estimate soil erodibility were
calculated from the State Soil Geographic (STATSGO) database (U .S. Department of
Agriculture, 1994). Area weighted averages were used to determine the average slope
and K factor for each watershed. The K factor is standardized to a 9% slope. To
create a variable that included both slope and the K factor, the K factor was multiplied
by the slope divided by 9%. For slopes greater than 9%, the K factor increases when
slope is included, indicating higher erosion potential. Slopes less than 9% would
result in a lower K factor, and a lower erosion potential. Unfortunately, the
STATSGO database deﬁnes very broad groupings of soils for large spatial areas, and
may not be detailed enough for this analysis.

Slope (degrees): Average slope for the watershed was calculated in ArcInfo
using the same digital elevation models that were used for determination of the
watershed.

Sulfate deposition (kg ha" y"): Sulfate deposition data were collected for 10
sites from Michigan, Wisconsin and Illinois from the National Acid Deposition
Program database (National Atmospheric Deposition Program (NRSP-3)/National
Trends Network, 2003). Data were collected ﬁ'om all available years ﬁ'om each site,
but ﬁnal data are from 1990-2000. Very few stations collected data from 1970-1980,
making it difﬁcult to estimate sulfate deposition values for this time period. However,
the patterns of sulfate deposition appear to remain the same within the state over time,

therefore the 1990-2000 data were also used to represent the spatial gradient of

203

atmospheric deposition ﬁ'om 1970-1980. Inverse distance weighting was used to
estimate sulfate deposition values for each watershed.

POpulatr'on meople) and population density (people km’z): Census data at the
census tract scale were used for population calculations. Population data from one
data set was used for population density calculations from 1870-1990, and a separate
data set was used for 2000 (Michigan Center for Geographic Information, 2002).
Because of this, census tracts are different for 2000, and some population densities
changed signiﬁcantly from 1990 to 2000 due to this. Watersheds generally only
encompassed a portion of each census tract; therefore, to determine population
densities within the watershed, it was essential to estimate as accurately as possible
where people were living within the township. Thus, dasymetric mapping was used to
estimate population distribution. Areas that are state owned (Michigan Department of
Natural Resources, 2000) or areas covered by lakes (Michigan Department of Natural
Resources, 1992) were considered as non-livable areas. It was assumed that the
population was evenly distributed throughout the remaining livable area within each
census tract. The preference of people to live near a lake (Stewart, 1994) was not
accounted for, causing a possible underestimation of population densities within
watersheds. Additionally, no attempt was made to account for seasonal increases in
population, which may be signiﬁcant around some lakes. Census data from 1970 and
1980 were averaged to represent the 19708 time period, and data from 1990 and 2000
were averaged to represent the 19908 time period. The number of people within the
Watershed was divided by the area of the watershed (not including the lake area) to

determine the population density of the watershed (people km'z)

204

Land cover (kmz, %): Land use/cover data were collected for 197 8 (Michigan
Department of Natural Resources, 1999) and 1997-2001 (Michigan Department of
Natural Resom'ces, 2003). Land cover groupings were different for the two time
periods, therefore the Level 1, or coarsest, land cover groupings were used so that the
two time periods could be compared (Table l). The Level 1 land cover groupings
were generally comparable between time periods, with the exception of the forested
and wetland land covers. The 1978 dataset includes lowland forest in the forested land
cover, whereas the 1997-2001 dataset includes lowland forest in the wetlands land
cover, and categorizes upland forest separately. The I997-2001 dataset was modiﬁed
to include lowland forest into a new forest category (instead of upland forest), thus
making the two time periods more comparable. Total area in the watershed and
percentage land covers were calculated for each watershed (not including the lake),
and for a 100 m buffer around the lake for each of the two time periods.

Table 1. Level 1 land cover groupings from two time periods.
The 1997-2001 dataset was modiﬁed to correspond better to

the 1978 dataset by moving the lowland forest land cover
from the wetlands category to the forest category.

 

 

1978 1997-2001
Urban Urban
Agriculture Agriculture
Rangeland Upland openland
Forested Forest = upland forest + lowland forest
Water Water
Wetland Wetland (not including lowland forest)
Banen Bare/sparsely vegetated

 

Flow weighted land cover: The ﬂowpath (m2), or distance ﬁ'om the lake along
the path that water would ﬂow, for each grid cell in each watershed was determined

Using the FLOWLENGTH command in ArcInfo. Inverted ﬂowpath was then calculated

205

(ll

of

(l/ﬂowlength). Separate grid coverages were created for each land cover, with a value
of l for each cell that contains that land cover, and a value of 0 for all other land
covers. Each of these coverages was multiplied by the inverse ﬂowpath coverage and
summed. This was completed for each land cover for each time period. Additionally,
the same calculations were performed using the inverse ﬂowpath squared (l/dz).
Percentages were calculated dividing the result for one land cover type by the sum for
all land covers.

Toxic Release Inventory (T RI): Toxic release inventory sites that released
copper, cadmium, lead, or zinc for all of Michigan were identiﬁed from
http://www.epa.gov/triexplorer/. These sites were plotted spatially, and watersheds of
each of the study lakes were examined to determine if any TRI sites were present
within the watershed. No TRI sites were present in any of the watersheds; therefore
these data were not used.

All watershed characteristics were examined for outliers using boxplots. All
outliers were removed before correlation analysis to prevent these points from having
signiﬁcant inﬂuence on the correlation coefﬁcients (Y ohn et al., in preparation).

Correlation coefﬁcients were calculated between lead accumulation rates and
watershed characteristics using SYSTAT 10. Multiple regression was used to
determine the two watershed characteristics that best explained the variance in lead
accumulation rates for each time period. The number of predictor variables was
limited to two due to the small number of lakes sampled (Hair, 1998). All possible

pairs of the ten variables with the highest correlation coefﬁcients were modeled using

206

multiple regression to determine which pair had the highest 1'2 value. Only models

where both variables had a p < 0.05 were considered.

Results
Sediment chronologies
All three metals show a general trend of anthropogenic accumulation rates
beginning to increase between 1850 and 1900, peaking in the 19705 and decreasing to
the present. However, for each of the metals there are considerable differences
among lakes, both in the actual values and in the patterns over time.

Cadmium accumulation rates generally increase from background values

—-— Cadillac

-t— Cass

-o- Crystal B

-— Crystal M

+ Elk

+ Gratiot

+ GU"

_._. Higgins

+ Houghton

— o Imp

+ Littleﬁeld

+ Mullett

—a— Paw Paw

, + Round

i + Torch

1850 L e 7* ,. W , an , 4 +Whitmore
0.00 0.10 0.20 0.30 0.40 0.50

Cadmium anthropogenlc accumulation rate (mg m‘2 y")

 

Figure 2. Sediment anthropogenic accumulation rates (mg m'2 y") of
cadmium in sixteen Michigan lakes.

207

within the time period of 1850-1900, but accumulation rates peak at different time

periods in the study lakes (Figure 2). Some lakes (e.g., Crystal B, Paw Paw) have

their highest cadmium accumulation rates in the early to mid-1900s, while other lakes

(e.g., Gull, Higgins) have peak accumulation rates near the 19705. Additionally, the

overall shape of the proﬁles varies greatly among lakes, from broad peaks with long

periods of similar accumulation rates (e.g., Gratiot, Round), to proﬁles with one

distinct peak (e. g., Gull, Crystal B). This suggests that, over time, sources or the

intensity of the sources has varied among lakes.

Cadmium
anthropogenic
accumulation rates have
decreased in most lakes
from the 19705 to the
1990s (Figure 3), with two
lakes having higher
cadmium accumulation
rates in the 1990s
(Cadillac, Crystal B), and
Gull Lake having much

lower accumulation rates

row
83

8

Log anthro Cd accum rates 1990s
.-» n
8 8

.3
N
O

 

 

 

 

 

 

1.504 ---.~e---—----—.—- A
1.50 1.70 1.90 2.10 2.30 2.50 2.70

Log anthro Cd accum rates 1970:

Figure 3. Average cadmium sediment log
anthropogenic accumulation rates (pg 111’2 y") in the
19903 plotted versus average values for the 19705.
The line represents a 1:1 relationship, or no change
between the two time periods.

 

 

z.r_

T—__——

in the 19905. The dissimilarity among lakes supports the observations made from the

sediment chronologies, and suggests that either the sources for cadmium or the

208

intensity of the sources was dependant on local watershed characteristics for at least
one of the time periods.

Copper proﬁles are the most dissimilar among lakes of the three metals (Figure
4). Three lakes clearly have much higher copper accumulation rates than the other
lakes. Both Cadillac and Houghton lakes have high accumulation rates due to the use
of copper sulfate (Yohn et al., 2002; Yohn et al., 2003). This is a local source that is
not represented in any of the watershed characteristics, and therefore these lakes will
be disregarded in further analysis of copper. Additionally, Witch Lake has a very
unusual copper proﬁle with very high copper accumulation rates. The cause of this
proﬁle is unclear, but Witch Lake is located in an area with extensive mining, and the
proﬁle is probably due to local mining activity. Gratiot Lake may also have been
inﬂuenced by mining, and is located in a region with extensive copper mining. While
copper accumulation rates are not exceptionally high, Gratiot Lake was consistently an
outlier in preliminary correlation analysis. Because both of these lakes have a known
source which is not accounted for by any of the watershed characteristics, they will
also not be included in further analyses.

The remainder of the lakes also have copper proﬁles that are somewhat
dissimilar, suggesting that copper is more inﬂuenced by local factors than regional
over time. Copper accumulation rates have not decreased signiﬁcantly since the 19705
(Figure 5), with copper accumulation rates in most lakes remaining constant or

decreasing only slightly over the last three decades.

209

 

0.0.0

Date

 

2000

1990 g
1980 7
1970 ‘

1960
1950
1940
1930
1920
1910
1900
1890
1880
1870
1860
1850

 

 

‘ + Mullett
‘ ‘ —e— Paw Paw

‘ + Witch

EadirhcA
‘+ Cass
, —0— Crystal B
f + Crystal M
l... Elk
—r—(3ﬁﬁkﬂ
+ Gull
—.— Higgins
1+ Houghton
. Imp

‘ + Round
+ Whitmore

 

20 40 60 80 1 00

l

2.5 5.0 7.5 10.0 12.5 15.0

Copper anthropogenic accumulation rate (mg m" y")

M+Cass

+ Crystal B
-— Crystal M
+ Elk

‘ar-Gﬁﬂkn
+ Gull

l + Higgins

r o — Imp

‘ + Mullett
+ Paw Paw
—~—Round
+ Whm

 

Figure 4. Sediment copper anthropogenic accumulation rates (mg m"2 y").
Cadillac, Houghton and Witch lakes are not present on the bottom graph.

210

:5
.5
0

9°
to
o

9’
~l
o

'8

3.30

3.10 ,

Log anthro Cu accum rates 1990.

2.90 -

 

 

 

 

 

 

 

2.70 — —w . a . . r
2.70 2.90 3.10 3.30 3.50 3.70 3.90 4.10

Log anthro Cu accum rates 1970:

Figure 5. Average copper sediment log anthropogenic
accumulation rates (ug rn’2 y'l) in the 19905 plotted versus
average values for the 19705. The line represents a 1:1
relationship, or no change between the two time periods.
Cadillac, Gratiot, Houghton and Witch lakes are not
included.

Zinc anthropogenic accumulation rates begin increasing around 1900 for most
lakes and many of the lakes have the highest anthropogenic accumulation rates in the
19705 (Figure 6). Additionally, most lakes have lower or the same accumulation rates
in the 19905 than the 19705 (Figure 7), with the exception of Cadillac Lake. The
overall similarity in the proﬁles of many of the lakes suggests that there may be

regional soruees, but local inﬂuences are also clearly important.

211

‘—: Cgiilaci ‘

 

2000 l
13:: + Cass
1970 ‘ -*- Crystal M
1960 ‘ + Elk
1950 + Gratiot
1940

,8, 1930 -°- GU“

Q 1920 w —o— Higglns
1900 l.
1890 ‘3 ' 'mp

\

1870 ‘ + ROUNd
1860

 

+ Whitmorel
0 20 40 60 80
Zinc anthropogenic accumulation rate (mg may")

Figure 6. Sediment zinc anthropogenic accumulation rates (mg m‘2 y'l).

 

 

 

 

 

5.007» - e 7*.
é
‘-
! 4.50 .
E
3
0
0
I
5
2 -
5 4.00
C
U
8’
—l
3.50 — if. — a
3.50 4.00 4.50 5.00

Log anthro Zn accum rates 1970:

Figure 7. Average zinc sediment log anthropogenic accumulation rates (ug m'2 y")
in the 1990s plotted versus average values for the 19705. The line represents a 1:]
relationship, or no change between the two time periods.

212

Sp:

Spatial patterns and atmospheric deposition

Sediment chronologies provide a history of metal inputs to the lake, and may
suggest if inputs of a metal are locally or regionally inﬂuenced over time. However,
looking at the spatial distribution of metals in Michigan lakes and at estimates of
atmospheric deposition rates can provide some insight into the importance of
atmospheric deposition during a speciﬁc time period.

There are major sources for metals to the atmosphere to the southwest of
Michigan in the extensive urban and industrial areas of Chicago and Gary (Harrison
and Winchest, 1971; Shahin et al., 2000; Landis et al., 2002; Souch et al., 2003), and
in the southeast of Michigan where Detroit is located. Additionally, the southern part
of Michigan is more developed than the northern Lower Peninsula or Upper Peninsula.
As a result, if metals are released to the atmosphere from either industrial sources, or
ﬁom local sources (e. g., coal burning for heat, gasoline burning) there may be a
regional gradient of metal deposition across Michigan, with the highest accumulation
rates in the south. If atmospheric deposition is not the pathway, then metal
accumulation rates in each lake should by related to characteristics of the watershed,
rather than a statewide gradient.

Cadmium does show a regional gradient in the 19708 (Figure 8), with the
highest accumulation rate in southeast Michigan, and decreasing northwards. The
lakes in southwest Michigan (Paw Paw and Gull) do not have signiﬁcantly higher
cadmium accumulation rates than those in mid-Michigan. This suggests that cadmium
released to the atmosphere from the Chicago / Gary area (Cole et al., 1990) is either

not reaching Michigan in signiﬁcant quantities, or is well distributed throughout much

213

of the state. Atmospheric deposition rates in the 19708 estimated from a bog
downwind of the Chicago / Gary region (Cole et al., 1990) are much higher than those
measured in and around Michigan (Shahin et al., 2000; Landis et al., 2002) (Table 2),
indicating that while some cadmium from this industrial area may enter Michigan,
much of it is deposited closer to the original source.

Atmospheric deposition to Lake Michigan in the 19705 was measured at a
deposition rate of 0.2 mg m"2 y'1 (Eisenreich, 1980), a rate similar to or greater than
many of the study lakes. This further supports the theory that atmospheric deposition
was an important pathway in the 19708. Only Cass and Whitmore lakes have
accumulation rates much higher than the measured atmospheric deposition rates.
These lakes may have higher rates because of additional atmospheric deposition of

cadmium ﬁom the urban area of Detroit, or ﬁ'om sources within the watershed.

214

    

1.6 - 1.82
1.82 - 2.04
2.04 - 2.26
2.26 - 2.48
2.48 - 2.7

O O 0 :13

Figure 8. Average log cadmium anthropogenic accumulation
(ug m'2 y") rates in sediments of Michigan lakes.

215

Table 2. Accumulation rates of cadmium, copper, and lead (mg m'2 y'l) during the

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

19708. Rates for this study are focusing corrected anthro genie accumulation rates.
. . Cd Cu Zn
Slte Medium (m 1111.2 y.1) m4 y.1) (m m.2 -11
This study, 19708
Cadillac Sediment 0.17 17
Cass Sediment 0.35 9.9 74
Crystal B Sediment 0.17 1.2
Crystal M Sediment 0.22 1.0 18
Elk Sediment 0.1 1 1.0 16
Gratiot Sediment 0.06 5.0
Gull Sediment 0.22 3.8 37
Higgins Sediment 0.19 1.7 17
Houghton Sediment 0.22 18
Imp Sediment 0.07
Littleﬂeld Sediment 0.19
Mullett Sediment 0.10 1.1
PawPaw Sediment 0.20 8.2 43
Round Sediment 0.1 1 8.0
Torch Sediment 0.13
Whitmore Sediment 0.41 3.6 46
N. Indiana (1973-1978)‘ 809 1.9 17 189
Lake Michigan (1975-76)b Atmospheric 0.2 2.1 19

 

 

 

 

 

 

a. (Cole et al., 1990)
b. (Eisenreich, 1980)

In the 1990s the regional gradient of cadmium accmnulation rates is no longer
apparent, suggesting that source(s) that are inﬂuenced on the watershed scale (Figure
8). This is supported by the measurements of atmospheric deposition (Table 3), which
show atmospheric deposition rates ranging ﬁ'om 0.06 to 0.14 mg m'2 y‘1 in the region.
Several of the study lakes (Cadillac, Cass, Crystal B, Crystal M, and Whitmore) have
accumulation rates much higher than that anticipated from atmospheric deposition,
and these lakes are probably also being inﬂuenced by local sources.

Copper accumulation rates in the 19703 are higher in the Southern portion of the state

and lower in the northern (Figure 9), but do not show a consistent regional gradient

216

because of high accumulation rates in Higgins Lake. From the atmospheric deposition
rates (Table 2), it appears that this pathway may be important for many lakes, but
lakes in the southern Michigan have much higher accumulation rates than those
estimated ﬁom atmospheric deposition to Lake Michigan (Table 2). Additionally,
Higgins Lake has a higher copper accumulation rate than Crystal M Lake, but is
located much farther north, suggesting that watershed scale inputs are probably also

important.

217

Table 3. Accumulation rates (mg m'2 y") of cadmium, copper, and zinc during the
19908. Rates from this stud are focusing corrected anthropogenic accumulation rates.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

. . Cd Cu Zn
Site Medium (m9 m-2 y.1) ( m g m.2 y.1) (m m.2 -11
This study, 19908
Cadillac Sediment 0.26 34
Cass Sediment 0.18 9.3 54
Crystal 8 Sediment 0.24 2.1
Crystal M Sediment 0.21 2.6 18
Elk Sediment 0.06 0.92 7.8
Gratiot Sediment 0.04 4.1
Gull Sediment 0.04 2.5 13
Higgins Sediment 0.14 1.3 12
Houghton Sediment 0.16 13
Imp Sediment 0.06
Littlefield Sediment 0.10 0.60
Mullett Sediment 0.06 1 .0
PawPaw Sediment 0.12 8.5 34
Round Sediment 0.10 7.9
Torch Sediment 0.08
Whitmore Sediment 0.28 4.0 35
L. Tantare, Canadaal sediment 0.06
|__ Superior, 1993-1 9946 atmospheric 3.1 8.8
L. Michigan, 19931994? atmospheric 1.9 6.0
L. Erie, 199349945 atmospheric 4.2 17
L. Huronc atmospheric 0.1 1
L. Ontarioc atmospheric 0.14
Eastern Ontario, 1993? snowpack 0.06-0.14 0.52-1.3 47-117
Northeast USA, 19936 snowpack 0025-006 025-062 80—199

 

 

 

 

 

 

a. (Alfaro-De la Torre and Tessier, 2002)

b. (Sweet et al., 1998)
c. (Hoff at al., 1996)

d. (Simonetti et al., 2000b)

218

 

Figure 9. Average log copper anthropogenic accumulation (ug m'2 y")
rates in sediments of Michigan lakes.

219

The importance of atmospheric deposition of copper in the 19908 is difﬁcult to
assess. There is an overall regional pattern of higher accumulation rates in the south
(Figure 9), which may indicate that atmospheric deposition is important. However,
estimates of atmospheric deposition from the Great Lakes region are lower than the
accumulation rates in the sediments of some of the study lakes (Table 3), suggesting
that local sources may also play a role.

Zinc accumulation rates in the study lakes have a consistent regional gradient
in the 19708 (Figure 10) that suggests the importance of atmospheric deposition. This
is corroborated by the measurement of atmospheric deposition (Table 2), which is
similar to or greater than many of the study lakes. The southern lakes have much
higher accumulation rates than that anticipated from atmospheric deposition in Lake
Michigan, but it is unclear of this is due to local sources, or greater atmospheric
deposition in these urban areas.

Similar to cadmium, the regional gradient in zinc accumulation rates that was
present in the 19708 is no longer apparent in the 19908 (Figure 10), suggesting that
local sources have become more signiﬁcant. Measurements of atmospheric deposition
from Lakes Michigan and Superior range from 8.8 - 6.0 mg m"2 y'l, and are lower than

many of the study lakes, supporting this conclusion (Table 3).

220

 

Figure 10. Average log zinc anthropogenic accumulation (pg m'2 y'l)
rates in sediments of Michigan lakes.

221

Watershed characteristics

The size of both the study lakes and their watersheds varied considerably, from
0.3 to 81 kmz, and 1.7 to 1286 km2 respectively. Watershed to lake area ratios vary
from 1.5 (Torch Lake) to 22.1 (Littleﬁeld Lake).

Both the K factor, a measure of erodibility, and average slope, were calculated
for both the entire watershed and a 100 m buffer around the lake. Average slopes of
the entire watersheds are generally shallow (<5 degrees) as well as slopes within the
100 m buffers. Slopes within the 100 m buffers were not consistently steeper or
shallower than the entire watershed. The watershed for Crystal B Lake had a much
higher slope in the 100 m buffer (8.7 degrees) than in the entire watershed or than any
other watershed. This datum was an outlier, and slope of the 100 m buffer around
Crystal Lake was not included in further analyses. The K factor for the 100 m buffer
was not consistently greater or less than the average for the entire watershed.

Sulfate deposition rates varied among lakes, with the lowest values in the
Upper Peninsula, and highest values near the southwest comer of Michigan (Figure
11). The prevailing winds in the southern Lower Peninsula are westerly, and carry
acid deposition from the Chicago, IL, and Gary, IN, industrial and urban centers
(Winchester and Nifong, 1971; Perkins et al., 2000; Souch et al., 2003).

Population densities were much higher for the Cass Lake watershed than any
other watershed (Appendix A). Population densities for Witch, Paw Paw, and
Cadillac lake watersheds were also high. All watersheds had a higher population

density in the 19908 than the 19708, with the exception of Round and Witch lake

222

watersheds. The Upper
Peninsula lakes (Gratiot,
Imp, Round and Witch)

have the lowest

     
 

watershed population Sulfate
densities. deposition
The dominant 3.5 - 10.7
land cover in most ' 10-7 ‘ 12.9
watersheds was forest, 3 12-9 ' 15-1

15.1 - 17.3
. 17.3- 19.5

Figure 11. Estimated rates of sulfate deposition rates
watersheds, and urban (kg ha" y") to Michigan study lakes.

with agriculture being

dominant in a few

dominant in Cass Lake watershed. Comparison of land cover between the two time
periods is difﬁcult because of the differences in the original data. 19708 land cover
maps were delineated from 124,000 color-infrared and black and white photos
(Michigan Department of Natural Resources, 1999), whereas the 19908 land cover
data are provided as a 30 m grid derived from satellite imagery (Landsat Thematic
Mapper) (Michigan Department of Natural Resources, 2003). As a result, where large
areas may be grouped as one land cover in the 19708 (e.g., urban in Cass Lake
watershed), these areas are divided into several land covers in the 19908 (Figure 12).
The percentage urban within most watersheds decreases from the 19708 to the 19908,

and this decrease in urban land cover is likely due to the increased resolution in the

223

 

i i '
,. .rs - .
"h: " ‘

Figure 12. Urban land cover in the Cass Lake watershed in 1975 and 1995.
Urbm land cover is in red, and water is in blue. Land covers of the other colors
vary between years. The watershed is outlined in green.

19908 land cover data than an actual decrease in urban land use. Because of this,
comparisons between the two time periods must be evaluated carefully.

The composition of land cover within the 100 m buﬂ‘er was signiﬁcantly
different than percentage land cover for the entire watershed. In general, for both time
periods, percentage urban was higher in the 100 m buffer, in many cases signiﬁcantly
(e.g., 9% urban in total watershed, 84% urban in 100 m buffer for Gull Lake in the
19708). Conversely, agriculture was much lower in the 100 m buffer. Forest and
rangeland was generally somewhat lower in the 100 m buffer, and wetlands varied by

watershed. For most lakes, urban land cover was dominant within the 100 m buffer,

224

suggesting concentrated developments near the lake compared to the rest of the
watershed. Percentage land cover that was weighted by the inverse of the distance
along the ﬂow path to the lake showed a similar trend of higher urban land cover, and
lower agricultm'e and forest. The use of inverse distance squared increased the

magnitude of these differences.

Correlations

Correlations between log anthropogenic accumulation rates of cadmium,
copper, and zinc, and watershed characteristics were determined (Tables 4-9). These
correlations do not necessarily represent a cause — effect relationship, but may give
insight into possible sources for these metals, and their pathways through the
environment.

There was no relationship between anthropogenic accumulation rates of
cadmium, copper or zinc in the 19708 or 19908 and lake or watershed size, or
watershed to lake area ratio (Tables 4,6,8). Additionally, there were no Toxic Release
Inventory (TRI) sites located in any of the study watersheds, suggesting that point
source inputs from major industry should not be a source for any of the lakes.

Accumulation rates of all three metals for both time periods correlated
positively with population density during both time periods. This is similar to other
studies that have found relationships between metal concentrations in stream
sediments and population density (Rice, 1999; Callender and Rice, 2000).

Atmospheric deposition correlates positively with sulfate deposition rates
during both time periods and for all metals, with generally higher correlations in the

19708 (Tables 4, 6, 8). Sulfate deposition was chosen to act as a proxy atmospheric

225

transport from sources such as the burning of coal (Graney et al., 1995; Berner and
Berner, 1996), and the positive correlation may represent the importance of extensive
coal burning that occurred in the Gary, Indiana, and Chicago, Illinois industrial region,
as well as in the industrial region near Detroit. Coal burning also probably occurred at
smaller quantities throughout the state from residential use, power production and
smaller industrial sites. The lower correlations between cadmium and zinc
accumulation rates and sulfate deposition rate in the 19908 is not surprising because
the 19708 the Clean Air Act has served to reduce contaminants released to the air, and
there has been a general decline in steel production since the 19808 (Perkins et al.,
2000).

There was also no relationship to the K factor measure of erodibility. This
suggests that the pathway of erosion, at least as evaluated by the erosion tendency of
the soil type, is not controlling copper, cadmium, or zinc deposition patterns through
the state.

An unanticipated relationship was the negative correlation between metal
accumulation rates and average slope of the watershed or within a 100 m buffer.

Slope was chosen as a variable to represent erosion potential of the soil, with an
anticipated positive relationship between metal accumulation rates and slope. Clearly
slope is not representing the extent of erosion, but must instead be representing a
different source or pathway. Slope may represent a measure of development in the
watershed, with steeper slopes being less developed, but in general, slopes are shallow
in the watersheds of the study lakes (< 5 degrees), and development should not be

limited by these slopes. A similar negative relationship between slope and lead

226

accumulation rates exists in the 19708, when the major source of lead to the
environment was the burning of leaded gasoline (Graney et al., 1995; Callender and
vanMetre, 1997). In Michigan, average slope in the watershed, or in a 100 m

Table 4. Pearson product moment correlations (r) between log anthropogenic
cadmium accumulation rates and watershed characteristics. Only correlations where p

< 0.05 are shown. Data for within a 100 m buffer were only collected for K factors
and

708 908
watershed 100 m watershed 1
area NA
Lake area NA
.lake area ratio NA

NA

NA
Sulfate NA
K

-0.604

 

Table 5. Pearson product moment correlations (r) between log anthropogenic
cadmium accumulation rates and land cover variables. Only correlations where p <
0.05 are shown.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

urban n’culture openlan

708 908 708 908 708 908
Total watershed % 0.720 0.725
Total watershed area
100 m buffer % 0.724 0.695
100 m buffer area -0.558
Flow inverted % 0.714
Flow inverted value 0733 -0.818
ﬂow inv. sq. % 0.655 J
Flow inv. sq. value -0.691 -0.754

forest wetlands

708 908 708 908
Total watershed % -0.688 0.673
Total watershed area
100 m buffer % -0.764 -0.607
100 m buffer area 0613
Flow inverted % ~0.718 -0.502
Flow inverted value -0.672 -0.518
Flow inv. sq. % -0.705 -0.527
Flow inv. sq. value -0.784 -0.570

 

 

 

 

 

 

227

Table 6. Pearson product moment correlations (r) between log anthropogenic copper
accumulation rates and watershed characteristics. Only correlations where p < 0.05
are shown. Data for within a 100 m buffer were only collected for K factors and

708 908
watershed 100 m watershed 100 m

area NA NA

Lake area NA NA
lake area ratio NA

.855 NA 0.856 NA

NA NA

Sulfate 0.668 0.644 NA

K

-0.766 -O.789 -0.732

 

Table 7. Pearson product moment correlations (r) between log anthropogenic copper
accumulation rates and land cover variables. Only correlations where p < 0.05 are
shown.

 

 

 

 

 

 

 

 

 

 

 

 

 

urban aﬂ'culture openland/range
708 908 708 908 708 908
Total watershed % 0.740 0.899
Total watershed area
100 m buffer %
100 m buffer area -0.713
Flow inverted % 0.731
Flow inverted value
flow inv. sq. % _
Flow inv. sq. value -0.667
forest wetlands

 

708 908 708 908

 

Total watershed 96
Total watershed area
100 m buffer %

100 m buffer area
Flow inverted %
Flow inverted value
Flow inv. sq. %

Flow inv. sq. value

 

 

 

 

 

 

 

 

 

 

 

 

 

 

228

Table 8. Pearson product moment correlations (r) between log anthropogenic zinc
accumulation rates and watershed characteristics. Only correlations where p < 0.05
are shown. Data for within a 100 m buffer were only collected for K factors and

908
watershed 100 m

708
watershed 100 m
area NA
Lake area NA
:lake area ratio NA
0 NA 0. 9
0.662

0.749

0753 NA
0.827 NA

 

-0.894 -0.847 -0.829 -0.885
Table 9. Pearson product moment correlations (r) between log anthropogenic zinc
accumulation rates and land cover variables. Only correlations where p < 0.05 are

shown.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

urban aﬂculture openland/range_
708 908 708 908 708 908
Total watershed % 0.874 0.918
Total watershed area 0.621
100 m buffer % 0.702 0.701 -0.701
100 m buffer area -0.871
Flow inverted % 0.663 0.759
Flow inverted value
ﬂow inv. sq. % 0.682 _
Flow inv. sq. value 0685 -0.677
forest wetlands
708 908 708 908
Total watershed % -0.778 0.734
Total watershed area
100 m buffer % —0.700
100 in buffer area
Flow inverted % -0.781 0.739
Flow inverted value -0.773 -0.810 -0.731
Flow inv. sq. % -0.636
Flow inv. sq. value 0807 -0.833 -0.765

 

 

229

buffer, generally follow a regional pattern of shallowest slopes in the south, and

steepest slopes in the north. This corresponds to a pattern of greater development and

higher population densities in the south and less development in the north (Figure 13).

As a result, it is possible that slope is representing the regional development gradient

across the state, rather than a speciﬁc source within each watershed. However, the

exact meaning of this variable remains unclear.

  
  
 

Log watershed slope
(degrees)

Population densitg 19708
(people km )

IIIIEU

-0.018 - 0.106
0.106 - 0.23
0.23 - 0.354
0.354 - 0.478
0.478 - 0.602
0. 602- 0.726

0.183 - 1.944
1.944 - 4.651
4.651 - 10.384
10.384 - 18.355
18.355 - 34.369
34.369 -4641.216

Figure 13. Average slope (log 2degrees) of study lake watersheds and

population density (people km 2) distribution in Michigan. Population
density 18 categorized so that there are equal areas of each grouping,
resulting in categories of unequal interval.

Percentage urban land cover within the watershed has a consistently positive

correlation with accumulation rates of all three metals in both time periods (Tables 5,

230

7, 9). In almost all cases, the total percentage urban within the watershed has a higher
correlation with metal accumulation rates than any of the landcover measures that are
based on proximity to the lake (100 m buffer, ﬂow path weighted). This suggests that
the overall urbanization of the watershed is more critical than where the development
is located. Additionally, correlations with urban land cover are generally higher in the
19908, further suggesting the importance of local sources during this time period.
Correlations between agricultural or openland / range and metal accumulation
rates were negative if signiﬁcant, and generally lower than correlations with urban
land cover. These negative relationships suggest that the use of pesticides or
fertilizers did not contribute signiﬁcant cadmium, copper or zinc to Michigan lakes.
Correlations with forest land cover were also negative. Copper accumulation rates
correlated poorly with land cover in general, with the exception of urban in the 19908

(Table 7).

Multivariate regression

Interpretation of correlations between watershed characteristics and lead
accumulation rates is difﬁcult, because many of the watershed characteristics are
correlated to each other (e.g., population density and % urban within the watershed).
Therefore, to determine the best combination of predictors of cadmium, copper, and
zinc to the environment, it is not appropriate to choose the variables that are most
highly correlated with these metals, because the variables may be correlated to each
other and actually act as a proxy for similar sources or processes. To avoid this

problem, we used multivariate regression to select the two watershed characteristics

231

that together best described the variation in each of these metals accumulation rates
among lakes. The number of watershed characteristics included in the model was
limited to two because of the small number of lakes sampled. Only models where
both variables contributed signiﬁcantly (p < 0.05) were considered, and in some cases
no two watershed characteristics explained signiﬁcantly greater variance than one

variable.

Cadmium
The two best predictors of anthropogenic cadmium accumulation rates in the
19708 are average slope of the watershed and sulfate deposition (1’2 = 0.85, n = 16,
Figure 14):
Log (CD703...) = -0.607 x 109 (SLOPE) + 0.039 x (804(3) + 1.873

Where:

CD701},at represents focusing corrected cadmium anthropogenic accumulation

rates in the 19708 (pg 111'2 y'l),

SLOPE represents average slope in the watershed (degrees), and

S04d is the estimated rate of sulfate deposition in the watershed (kg ha" y").
The standardized coefﬁcients, which account for the differences in scale among
variables, indicate the relative importance of each variable to the model. For cadmium

in the 19708, the two variables are of similar importance (standardized coefﬁcient of

SLOPE = -0.545, 804d = 0.505).

232

 

 

 

 

 

 

 

 

2 2-7 i l 2.7 l r
E 2.6b .. '1 26L. . .1
g 2.5— ~ 25 ° -
“g 2.4— .. ~ 2.4— -
'5 2.3~ ., . a 2.3— ,‘,' '.—
E 2-2- ° ‘ 1 2.2— ° ° ..
g 2.1 - ' ‘ 2.1 h ° ~
2.0— ° 3 ~ 20} '. . -
8 1.9- - 1gp _
§ 1.8- ° . ‘ 1.8- .° ‘
1:05 0.0 0.5 1.0 1'75 10 15 20
Log slope Sulfate deposition

Figure 14. Average log anthropogenic accumulation rates (mg
m' y") of cadmium in sediments deposited in the 19708 plotted
versus log average slope of the watersheds (degrees) and sulfate
deposition rates (kg ha'1 y'l).

The importance of sulfate deposition and slope suggest that atmospheric
deposition was an important source of cadmium during the 19708. This is in
agreement with the conclusions made from the spatial pattern of cadmium
accumulation rates and estimates of atmospheric deposition of cadmium. Cadmium is
released to the atmosphere from the combustion of fossil ﬁiels (e.g., coal, oil)
(Harrison and Winchest, 1971; Mecray et al., 2001), and atmospheric deposition was
found to be a signiﬁcant component of cadmium inputs to southern Lake Michigan
(Muhlbaier and Tisue, 1981). All of these results suggest that the atmosphere was a
dominant pathway of cadmium to Michigan lakes during the 19708.

In the 19908, the best predictor of cadmium accumulation rates in the
sediments of the study lakes was percentage urban land cover within the watershed (r2

= 0.53, n = 16, Figure 15):

233

Log (CDQOraar) = 0.523 x 109 (URBAN90) + 1.665

Where:

CD90fw represents focusing corrected anthropogenic cadmium accumulation

rates in the 19908 (ug m’2 y'l), and

URBAN 90 represents percentage urban land cover within the watershed.

Only 53% of the variance of cadmium can be explained by watershed

characteristics in the 19908, as opposed to the
85% explained in the 19708. This may be
explained by the increasing importance of local

inputs that are not well described by the

watershed characteristics examined in this study.

This conclusion is supported by the lack of a
regional gradient of accumulation rates, and
from comparison to measures of atmospheric
deposition. The exact sources for cadmium are

unclear, but appear to be related to urban land

 

 

 

 

Q 2.5 l I
ii 2.4- . ’ ~
8 2.3- ° . ..
E 2.2— - . —
321- . .
§2.0~ . ' 4
cu 1.9~ - e
8 13- . . . J
8’1.7~ . —
_l r ' 4
1'6-1 0 1 2
Log%urban

Figure 15. Average log
anthropogenic accumulation
rates (mg m'2 y") of cadmium
in sediments deposited in the
19908 plotted versus log percent
urban landcover of the
watersheds.

cover, and may include sources such as releases from wear of automobile parts,

wastewater, or household waste such as batteries (Rice, 1999; Callender and Rice,

2000). It is probable that there is also some continued atmospheric deposition of

cadmium, but local sources are more signiﬁcant for some Michigan lakes.

234

Copper
Watershed population density and slope corrected K factor are the two
variables that best predict the anthropogenic accumulation rates of copper in the 19708
(12 = 0.84, n = 9, Figure 16):
Log (CU70M) = 0.489 x 109 (POPDENS70) — 0.947 x log (KFACT3) + 1.776

Where:

CU70fam- is focusing corrected anthropogenic copper accumulation rates in the
1970s (ug m" y").

POPDENS70 is population density in the 1970s (people km'z), and

KFACTS is the average K factor in the watershed corrected for slope.

POPDENS70 has a higher standardized correlation coefﬁcient than KFACTs (0.732, -
0.409).

These two variables suggest that copper inputs were controlled at the
watershed scale rather than at the regional scale, which is supported by the lack of
regional pattern in anthropogenic accumulation rates. The relationship with the K
factor suggests that erosion may be an important pathway of copper to Michigan lakes;
however, this relationship is not strong (Figure 16), and this variable is less important
to the predictive model than population density. Due to regional inﬂuence of mining
in northern Michigan, and the usage of copper sulfate in two southern lakes, only nine

lakes were included in the analyses, and results must be interpreted with care.

235

 

 

 

 

 

 

 

 

2 4.0 l l V 2 4.0 i l 1' r i l i
8 3.9V 0 4 E 3.9_ . _
c 3'8“ - c: 3.8- _
% 3.7r _ .‘g‘, 3.7? _
5 3.6?- o . ' '5 3.6- . . _‘
E 3.5- - g 3.5- _
g 3'45 _ § 3.4e _
3.3e - 3.3+ ,
3 3.2— ' — 5 32. ° -
8’ 3"” . o i 8, 3.1T . . 4
L09 population density .\- I" A- A- .o- ,e- 5;. 9. 9.

Log K factor. slope ccrr

Figure 16. Average log anthropogenic accumulation
rates (mg 111'2 y") of copper in sediments deposited in the
19708 plotted versus log population density

(people km'z), and log slope corrected k-factor.

Similar to cadmium, the best predictor of copper accumulation rates in the
19903 is urban land cover (r2 = 0.81, n = 10, Figure 17):
Log (CU90fm) = 1.094 x 109 (URBAN90) + 2.410
Where:

CU90M is focusing corrected anthropogenic accumulation rates of copper in

the 19908 (ug m'2 y"), and

URBAN90 is percentage urban land cover in the watershed.

These results suggest that watershed scale sources remain inﬂuential during the
19908 with sources related to urban land cover. Possible sources may include wear
ﬁom household plumbing and automobiles, and household waste (Rice, 1999;
Callender and Rice, 2000; Mecray et al., 2001).

In addition to identifying possible source of copper, examination of the
relationships between copper accumulation rates and watershed characteristics
provided evidence that Gratiot Lake was inﬂuenced by copper mining in the region.

Although this lake is located in a region with extensive copper mining, no mining

236

occurred within the watershed of Gratiot

Lake, and copper accumulation rates were

not higher than most other lakes (Figure 4).

Therefore, although it was likely that
smelting and other sources in the region
inﬂuenced the lake, there was little direct
evidence until copper accumulation rates
were correlated with watershed
characteristics. Gratiot Lake consistently
had higher copper accumulation rates than

anticipated from the watershed

 

 

 

 

9 4.0 r T. o
t!
C
.9 .
E 3.5— . —
a 0
5 <9 '
01 30* ° . a
3
O .
31’ 25 . -
'-1 0 1 2
Log % urban

Figure 17. Average log
anthropogenic accumulation
rates (mg m'2 y") of copper in
sediments deposited in the
19908 plotted versus log
percent urban landcover of the
watersheds. Gratiot Lake is
circled, and was not included
in regression analysis.

characteristic (e.g., Figure 17). This lends strong support to the conclusion that there

was a source for copper that inﬂuenced Gratiot Lake and not other lakes in the state.

This also suggests that this approach will be useful in identifying lakes with unique

sources within their watersheds.

Zinc

The best predictor of zinc accumulation rates in Michigan lakes in the 19708 is

population density (r2 = 0.92, n = 10, Figure 18) if Cadillac Lake is removed:

237

Log (ZN70faar) = 0.466 x 109 (POPDENS70) + 3.678

Where:

ZN 70f”, is focusing corrected anthropogenic accumulation rates of zinc in the

1970s (pg m'2 y"), and

POPDENS70 is population density in the watershed in the 19708

(people km’z)

Cadillac Lake is an outlier in this relationship, and the inclusion of this one

lake reduces the r2 to 0.82: The cause of this is unclear, but Cadillac Lake is one of

two lakes (Cadillac and Cass) with a city located in the watershed. Only some of the

City of Cadillac is located within the

watershed, and it is possible that sewage and

stormwater is directed outside of the

watershed. This could result in the lower

than anticipated zinc accumulation rates for

the population density of the watershed.

Zinc is known to have both regional
atmospheric sources [fossil fuel combustion
(Callender and Rice, 2000), iron and steel

industry (Winchester and Nifong, 1971)] and

:5 :5 .0'
o 01 0

Log Zn accumulation rate

.01
or

 

l

 

 

 

0 , 2 , 3
Log population density

Figure 18. Average log

anthropogenic accmnulation rates

(mg 111‘2 y") of zinc in sediments

deposited in the 19708 plotted
versus log population density
(people km'z). The circled data

point represents Cadillac Lake.

watershed scale sources [corrosion of metal products containing zinc (Callender and

Rice, 2000), wear from automobile tires (Callender and Rice, 2000)]. The strong

positive correlation between population density and zinc anthropogenic accumulation

rates in the 19708 suggests that watershed scale inﬂuences are signiﬁcant during this

238

time period. Atmospheric sources are probably still present, as suggested by the

measurement of atmospheric deposition of zinc, but local sources related to population

density seem to be more signiﬁcant.

Percentage urban land cover in the watershed and slope in a 100 m buffer are

the best predictors of zinc accumulation rates in the 19908 (r2 = 0.909, n = 1 1, Figure

19):

Log (ZN90raar) = 0.506 x 109 (URBAN90) - 0.649 x(SLOPE100m) + 3.980

Where:

ZN 90f”, is focusing corrected anthropogenic accumulation rates of zinc in the

19908 (pg m’2 7").

URBAN 90 is percentage urban land cover in the 19908, and

SLOPEmom is average slope (degrees) in a 100 m buffer around the lake.

URBAN90 is weighted more highly in the model (standardized coefﬁcient URBAN90

= 0.581, SLOPEIOOm = -0.445).

 

9'
o

P
01
r

:5
o
l

 

Log Zn accumulation rate

 

 

9°
.01
.8

0

1
Log % urban

5.0

4.5

4.0

' .i 0.0
09 slope.
Figure 19. Average log anthropogenic accumulation rates
(mg m'2 y") of zinc in sediments deposited in the 19908
plotted versus log percentage urban in the watershed, and

 

I

 

l

 

 

0.5

log slope within a 100 m buffer of the lake (degrees).

239

1.0

100 in buffer

This model may reﬂect the continued importance of watershed scale sources of
zinc, including wear of tires and corrosion of metal products, as well as some

contribution from atmospheric deposition.

Summary and Conclusions

Anthropogenic accumulation rates of both zinc and cadmium have decreased
from the 19708 to the 19908 in the majority of Michigan lakes sampled. This suggests
that the combination of reduced industrial activity (e.g., reduction of steel production,
Perkins et al., 2000) and increased environmental legislation has effectively reduced
inputs of cadmium and zinc to Michigan lakes.

Sources of cadmimn appear to have shifted ﬁom dominantly atmospheric in
the 19708 to watershed scale sources dominating in the 19908, though atmospheric
sources are probably still present. Atmospheric sources may include burning of fossil
fuels, while watershed scale sources may include tire wear and other sources related to
urban land cover.

Zinc has both watershed scale and regional sources as well, with the burning of
fossil fuels releasing zinc to the atmosphere, and tire wear and metal corrosion
potentially releasing zinc within the watershed. Zinc anthropogenic accumulation
rates correlate with watershed scale characteristics during both time periods,
population density in the 19708, and percentage urban land cover in the 19908, but it is
unclear whether sources have remained the same over time or if new sources have
appeared. Zinc also correlates with slope in a 100 m buffer in the 19908, which may
suggest that atmospheric deposition is also an important pathway during this time

period.

240

Copper anthropogenic accumulation rates have not decreased in many lakes
from the 19708 to the 19908, suggesting that sources of copper to the environment
have not been reduced. Copper may be released to the atmosphere ﬁom the burning
of coal and fossil fuels, or from wear of disk pads on automobiles, wear of copper
pipes, and other urban related sources. Anthropogenic accumulation rates of copper
correlate with watershed scale characteristics during both time periods, suggesting
that, even if atmospheric deposition contributes to copper accumulation in lake
sediments, local sources are important as well.

For all metals, percentage urban land cover within the entire watershed is one
of the best predictor of rates metal accumulation rates in the 19908. This indicates that
urban land cover is one of the dominant sources of these metals to Michigan lakes, and
that the location of the urban land cover within the watershed does not matter. The
exact sources of metals ﬁ'om the urban land cover cannot be determined by this study,
and future work should focus on sources and pathways of metals to the lake from
urban land cover. These sources could potentially then be reduced, resulting in
sediment quality improvement and decreased loading of these trace metals to the
environment.

The approach of examining anthropogenic accumulation rates of metals in lake
sediments over space and time, and correlating these patterns to watershed
characteristics has successfully provided insight into effects of environmental
legislation, and the changing sources of metals over time. It is also useful in
identifying lakes with unique sources, such as the inﬂuence of mining on Gratiot Lake.

Additionally, this approach has identiﬁed the area to examine for current sources of

241

these metals to urban land cover. This knowledge could help to further reduce

anthropogenic inputs of cadmium, copper and zinc to the environment.

242

References

Alfaro-De la Torre, MC. and Tessier, A., 2002. Cadmium deposition and mobility in
the sediments of an acidic oligotrophic lake. Geochim. Cosmochim. Acta 66,

3549-3562.

Basnyat, P., Teeter, L.D., Flynn, K.M. and Lockaby, B.G., 1999. Relationships
between landscape characteristics and nonpoint source pollution inputs to
coastal estuaries. Environmental Management 23, 539-549.

Berner, E.K. and Berner, R.A., 1996. Global environment : water, air, and
geochemical cycles. Prentice Hall, Upper Saddle River, N.J., xiv, 376 pp.

Callender, E. and Rice, K.C., 2000. The urban environmental gradient: Anthropogenic
inﬂuences on the spatial and temporal distributions of lead and zinc in the
environment. Environ. Sci. Technol. 24, 232-238.

Callender, E. and vanMetre, P.C., 1997. Reservoir sediment cores show US lead
declines. Environ. Sci. Technol. 31, A424—A428.

Catallo, WJ., Schlenker, M., Gambrell, RP. and Shane, B.S., 1995. Toxic chemicals
and trace metals from urban and rural Louisiana lakes: recent historical proﬁles
and toxicological signiﬁcance. Environ. Sci. Technol. 29, 1439-1446.

Cole, K.L., Engstrom, D.R., Futyma, RP. and Stottlemyer, R., 1990. Past
Atmospheric Deposition of Metals in Northern Indiana Measured in a Peat
Core from Cowles Bog. Environmental Science & Technology 24, 543-549.

Comeleo, R.L., Paul, J.F., August, P.V., Copeland, J., Baker, C., Hale, SS. and
Latimer, R.W., 1996. Relationships between watershed stressors and sediment
contamination in Chesapeake Bay estuaries. Landscape Ecology 11, 307-319.

Detenbeck, N.E., Johnston, CA. and Niemi, G.J., 1993. Wetland Effects on Lake
Water-Quality in the Minneapolis St-Paul Metropolitan-Area. Landscape
Ecology 8, 39-61.

Dillion, PJ. and Evans, R.D., 1982. Whole-lake lead burdens in sediments of southern
Ontario, Canada. Hydrobiologia 91, 121-130.

Edgington, D.N. and Robbins, J .A., 1976. Records of lead deposition in Lake
Michigan sediments since 1800. Environ. Sci. Technol. 10, 266-274.

Eisenreich, S.J., 1980. Atmospheric input of trace metals to Lake Michigan. Water,
Air, Soil Pollut. 13, 287-301.

243

Gatz, DR, 1975. Pollutant aerosol deposition into southern Lake Michigan. Water Air
Soil Poll 5, 239-251.

Gatz, D.F., Bowersox, V.C. and Su, J ., 1989. Lead and cadmium loadings to the Great
Lakes from precipitation. J. Gt. Lakes Res. 15, 246-264.

Golden, K.A., Wong, C.S., Jeremiason, J.D., Eisenreich, S.J., Sanders, M.G.,
Hallgren, J., Swackhammer, D.L., Engstrom, DR. and Long, D.T., 1993.
Accumulation and preliminary inventory of organochlorines in Great Lakes
sediments. Wat. Sci. Technol. 28, 19-31.

Goodarzi, F., Sanei, H. and Duncan, W.F., 2001. Monitoring the distribution and
deposition of trace elements associated with a zinc-lead smelter in the Trail
area, British Columbia, Canada J. Environ. Monit. 3, 515-525.

Graney, J .R., Halliday, A.N., Keeler, G.J., Nriagu, J .0, Robbins, J .A. and Norton,
S.A., 1995. Isotopic record of lead pollution in lake sediments from the
northeastern United States. Geochim. Cosmochim. Acta 59, 1715-1728.

Hair, J .F ., 1998. Multivariate data analysis. Prentice Hall, Upper Saddle River, N.J.,
730, 12 pp.

Harrison, PR. and Winchest, J .W., 1971. Area-Wide Distribution of Lead, Copper,
and Cadmium in Air Particulates ﬁ'om Chicago and Northwest Indiana. Atmos.
Environ. 5, 863-&.

Hermanson, M.H., 1998. Anthropogenic mercury deposition to arctic lake sediments.
Water, Air, Soil Pollut. 101.

Hewitt, AD. and Reynolds, C.M., 1990. Dissolution of metals from soils and
sediments with a microwave-nitric acid digestion technique. Atomic
Spectroscopy 11, 1425-1436.

Hoff, R.M., Strachan, W.M.J., Sweet, C.W., Chan, C.H., Shackleton, M., Bidleman,
T.F., Brice, K.A., Burniston, D.A., Cussion, S., Gatz, D.F., Harlin, K. and
Schroeder, W.H., 1996. Atmospheric deposition of toxic chemicals to the
Great Lakes: A review of data through 1994. Atrnos. Environ. 30, 3505-3527.

Johnson, L.B., Richards, C., Host, GE. and Arthur, J.W., 1997. Landscape inﬂuences

on water chemistry in Midwestern stream ecosystems. Freshwater Biology 37,
193-205.

244

Kerfoot, W.C., Lauster, G. and Robbins, J .A., 1994. Paleolimnological study of
copper mining around Lake Superior: Artiﬁcial varves ﬁ'om Protage Lake
provide a high resolution record. Limnology and Oceanography 39, 649-669.

Kolak, JJ., Long, D.T., Beals, T.M., Eisenreich, SJ. and Swackhamer, D.L., 1998.
Anthropogenic inventories and historical and present accumulation rates of
copper in Great Lakes sediments. Appl. Geochem. 13, 59-75.

Landis, M.S., Vette, AF. and Keeler, G.J., 2002. Atmospheric mercury in the Lake
Michigan basin: Inﬂuence of the Chicago/Gary urban area. Environ. Sci.
Technol. 36, 4508—4517.

Long, D.T., Eisenreich, SJ. and Swackhammer, D.L., 1995. Metal Concentrations in
Sediments of the Great Lakes, Great Lakes Protection Fund.

Mecray, E.L., King, J.W., Appleby, P.G. and Hunt, A.S., 2001. Historical trace metal
accumulation in the sediments of an urbanized region of the Lake Champlain
watershed, Burlington, Vermont. Water Air Soil Poll 125, 201-230.

Michigan Center for Geographic Information, 2002. Statewide Layer: Michigan
geographic ﬁamework. Michigan Center for Geographic Information, Lansing,
MI.

Michigan Department of Natural Resources, 1992. MIRIS base data. MDNR, Lansing,
MI.

Michigan Department of Natural Resources, 1999. 1978 Land cover / use. MDNR,
Lansing, MI.

Michigan Department of Natural Resources, 2000. State ownership. MDNR, Lansing,
MI.

Michigan Department of Natural Resources, 2003. IFMAP/GAP land cover. MDNR
Forest, Mineral, and Fire Management Division, Lansing, MI.

Muhlbaier, J. and Tisue, G.T., 1981. Cadmium in the Southern Basin of Lake-
Michigan. Water Air Soil Poll 15, 45-59.

National Atmospheric Deposition Program (NRSP-3)/National Trends Network, 2003.
NADP Program Ofﬁce, Illinois State Water Survey, 2204 Griffrth Drive,
Champaign, IL 61820.

Nriagu, J .O., 1989. A Global Assessment of Natural Sources of Atmospheric Trace-
Metals. Nature 338, 47-49.

245

Perkins, S.M., Filippelli, GM. and Souch, C.J., 2000. Airborne trace metal
contamination of wetland sediments at Indiana Dunes National Lakeshore.
Water Air Soil Poll 122, 231-260.

Rice, K.C., 1999. Trace-element concentrations in streambed sediment across the
conterminous United States. Environ. Sci. Technol. 33, 2499-2504.

Richards, C., Johnson, LB. and Host, GE, 1996. Landscape-scale inﬂuences on
stream habitats and biota. Canadian Journal of Fisheries and Aquatic Sciences
53, 295-311.

Shahin, U., Yi, S.M., Paode, RD. and Holsen, T.M., 2000. Long-term elemental dry
deposition ﬂuxes measured around Lake Michigan with an automated dry
deposition sampler. Environmental Science & Technology 34, 1887-1892.

Simonetti, A., Gariepy, C. and Carignan, J ., 2000a. Pb and Sr isotopic compositions of
snowpack from Quebec, Canada: Inferences on the sources and deposition
budgets of atmospheric heavy metals. Geochim. Cosmochim. Acta 64, 5-20.

Simonetti, A., Gariepy, C. and Carignan, J., 2000b. Pb and Sr isotopic evidence for
sources of atmospheric heavy metals and their deposition budgets in
northeastern North America. Geochim. Cosmochim. Acta 64, 3439—3452.

Souch, C.J., F ilippelli, G.M., Dollar, N., Perkins, S. and Mastalerz, M., 2003.
Accumulation rates of airborne heavy metals in wetlands. Phys. Geogr. 23, 21 -
43.

Stewart, 8.1., 1994. The seasonal home location decision process: toward a dynamic
model. Ph.D. Thesis, Michigan State University, East Lansing, MI, 156 pp.

Sweet, C.W., Weiss, A. and Vermette, S.J., 1998. Atmospheric deposition of trace
metals at three sites near the Great Lakes. Water Air Soil Poll 103, 423-439.

US. Department of Agriculture, 1994. State Soil Geographic (STATSGO) database
for Michigan. US. Department of Agriculture, Natural Resources
Conservation Service, Fort Worth, Texas.

Wetzel, R.G., 2001. Limnology : lake and river ecosystems. Academic, San Diego,
Calif. London, xvi, 1006 pp.

Winchester, J .W. and Nifong, GD, 1971. Water pollution in Lake Michigan by trace
elements from pollution aerosol fallout. Water, Air, Soil Pollut. 1, 50-64.

246

Yohn, S.S., Long, D.T., Fett, J .D. and Patino, L.C., 2004. Regional versus local
inﬂuences on lead and cadmium loading in the Great Lakes region. Appl.
Geochem. 19, 1157-1175.

Yohn, S.S., Long, D.T., Giesy, J .P., Scholle, L.K., Patino, L.C., Fett, J .D. and Kannan,
K., 2002. Inland lakes sediment trends: sediment analysis results for ﬁve
Michigan Lakes. MI/DEQ/WD-O3/052, Department of Environmental Quality,
East Lansing.

Yohn, S.S., Long, D.T. and Winkler, J .A., in preparation. The use of multivariate
statistics for the interpretation of sediment chronologies. Limnology and
Oceanography: Methods.

Yohn, S.S., Parsons, M.J., Long, D.T., Giesy, J .P., Scholle, L.K. and Patino, L.C.,

2003. Inland lakes sediment trends: sediment analysis results for six Michigan
Lakes, Michigan Department of Environmental Quality, East Lansing, MI.

247

CHAPTER 8
SUMMARY AND RECOMMENDATIONS

Summary

The working hypothesis of this research is: sources and pathways controlling
anthropogenic accumulation rates of metals sediments have changed over time, and
these changes should be reﬂected both in the spatial patterns of metal accumulation
rates in lake sediments and the changing correlations between metal accumulation
rates and characteristics of the lake watersheds. This was investigated through a three
step approach: 1) identiﬁcation of groups of elements that are inﬂuenced by the same
sources and processes, and selection of elements that are representative of each group
to assist in calculating anthropogenic accumulation rates, 2) determination of
correlations between different watershed characteristics (e.g., percentage urban land
cover) and calculated anthropogenic accumulation rates of each of the elements
determined to have direct anthropogenic inputs for each of the two time periods (1970-
1980, 1990-2000), 3) examination of temporal changes in correlations with different
watershed characteristics.

Four groupings of elements, terrestrial, diagenetic, carbonate and
anthropogenic, were present in the majority of lakes and were differentiated using
cluster analysis. The two groupings that were the most useful in quantifying
anthropogenic accumulation rates were the diagenetic and terrestrial groups. The
elements representative of the diagenetic group (iron, manganese, and molybdenum),
were used to help determine if elements with known anthropogenic sources (e.g.,

arsenic) were also inﬂuenced by diagenesis. The elements representing terrestrial

248

inputs (aluminum and titanium), were used to differentiate natural inputs, inputs due to
increased erosion, and direct human inputs for cadmium, copper, lead and zinc. The
effects of logging and other land cover changes were clearly seen in the temporal
patterns of the terrestrial elements in some lakes (e.g., Elk and Gull lakes).

To evaluate the relationships between watershed characteristics and
anthropogenic accumulation rates of lead, copper, cadmium, and zinc, multiple
variables representing different sources and pathways to the lake with a two variable
linear regression model were used. The six most effective variables for predicting
anthropogenic accumulation rates of the four metals include sulfate deposition, slope
corrected K factor, percentage urban land cover, population density, average slope
within the watershed, and average slope in a 100 m buffer around the lake.

Sulfate deposition was chosen as a watershed characteristic to represent the
pathway of atmospheric deposition, and may represent metals released from the
burning of coal and other fossil fuels ﬁ‘om both the urban and industrial area of
Chicago, Illinois and Gary, Indiana, and smaller sources throughout Michigan. The
slope corrected K factor should represent the pathway of erosion, and transport of
metals deposited on land to the lake. Percentage urban land cover in the watershed
and watershed population densities may represent similar sources, including: releases
from wear of automobile parts (Rice, 1999), car emissions (Callender and vanMetre,
1997; Rice, 1999; Shahin et al., 2000), road dust (Shahin et al., 2000), wastewater and
household and yard waste (Callender and Rice, 2000). The signiﬁcance of average
slope of the watershed and slope within a 100 m buffer are less clear, because the

negative relationship between metal accumulation rates and slope clearly shows that

249

slope is not representing the pathway of erosion. There is no apparent causal
relationship between slope and metal anthropogenic accumulation rates, but watershed
slopes do tend to follow a similar pattern to the overall population density and urban
development in the state, with shallower slopes and greater development in the south-
east. As a result, slope may be representing local atmospheric deposition that is not
conﬁned to the watershed, but inﬂuenced by the overall population distribution in the
area. The meaning of this variable, however, needs to be ﬁrrther investigated for
clariﬁcation.

Correlations between anthropogenic accumulation rates of lead, copper,
cadmium, zinc and watershed characteristics were surprisingly high; multiple 13 values
using one or two watershed characteristics as predictor variables were greater than 0.8
for all cases except cadmium in the 19908 (1’2 = 0.53). The watershed characteristics
that were included in the multiple regression models for each of the elements include:

Cadmium 19708: average slope of the watershed, sulfate deposition rates

(:2 = 0.85)
19908: percentage lu'ban land cover in watershed (1'2 = 0.53)

Copper 19708: watershed population density, slope corrected K factor

(r2 = 0.84)
19908: percentage urban land cover in watershed, slope in 100 m
buffer around lake (r2 = 0.81)
Lead 19708: sulfate deposition rates, slope in 100 or buffer around lake

(r2 = 0.92)

250

 

19908: percentage urban land cover, slope in 100 or buffer
(:1 = 0.92)
Zinc 19708: watershed population density (r2 = 0.92)
19908: percentage urban land cover, slope in 100 m buffer

(1'2 = 0.91)

There are changes over time in anthropogenic accumulation rates, spatial
patterns and correlations with watershed characteristics. Anthropogenic accumulation
rates of cadmium, lead, and zinc decrease in the majority of lakes from the 19708 to
the 19908, indicating that inputs of these metals to the environment have been reduced,
probably due to environmental legislation. Copper anthropogenic accumulation rates
have not consistently decreased since the 19708, suggesting that historic sources have
not decreased or that new sources have emerged.

Both spatial patterns and multivariate regression models for lead and cadmium
suggest that there has been a shift from atmospheric deposition in the 19708 to more
local sources in the 19908 that are related to urban land cover. The major atmospheric
sources that were present in the 19708, such as the burning of leaded gasoline in
automobiles or burning of coal for major industries, were reduced by environmental
legislation such as the Clean Air Act. As a result, anthropogenic accumulation rates
are lower in the 19908 and the dominant somees are different than the 19708. Copper
and zinc do not show this same change over time; with population density the best
predictor in the 19708 and urban land cover in the 19908, suggesting that local somees

were important during both time periods. Atmospheric transport and deposition is

251

 

probably present during both time periods, but appears to be less important than local
sources. Percentage urban land cover within the watershed is the best predictor of
anthropogenic accumulation rates of all four metals in the 19908, suggesting that
sources within the watershed contribute more signiﬁcantly than atmospheric
deposition. Future work may identify speciﬁc sources of these metals and reduce their
inputs to the environment.

This research both corroborates previous work and provides new insight into
the changing sources and pathways of trace metals in the environment. Lead in the
environment has been extensively studied, and this work is in agreement with previous
studies concluding that atmospheric transport of lead from the binning of leaded
gasoline and coal burning were important sources in the 19708, and that the Clean Air
Act signiﬁcantly reduce lead inputs to the environment (Callender and vanMetre,
1997; Edgington and Robbins, 1976; Eisenreich et al., 1986; Evans and Dillon, 1982;
Graney et al., 1995; Nriagu, 1978). However, current sources of lead are not well
deﬁned. This work has identiﬁed a deﬁnitive correlation between urban land cover
and anthropogenic accumulation rates of lead in the 19908, indicating that current
sources are present that need to be identiﬁed and reduced.

Cadmium, copper, and zinc in the environment have not been as intensively
studied as lead, and trends over time have not been thoroughly explored. The regional
decrease of cadmium and zinc inputs to the environment after the 19708 corroborates
previous work (Alfaro-De la Torre and Tessier, 2002; Mecray et al., 2001), but shows
that these decreases are present over a broader region than their studies. The relatively

constant accumulation rates of copper over the last three decades were unexpected,

252

and are in contrast with previous work on the Great Lakes (Kolak et al., 1998). This
study shows that environmental legislation has been less effective for copper, and
methods for reducing copper loading to the environment must be reevaluated.

The impact of urban land cover has been determined previously, but this is the
ﬁrst research to show strong correlations between several metals and urban land cover
over a broad geographic scale. Additionally, little work has been done on historic
regional sources of cadmium, copper, and zinc. Much of the previous work done on
these elements included the sampling of only one lake, making it difﬁcult to determine
if the results represented only that speciﬁc watershed or what was occurring in the
regional environment. The relatively large number of lakes sampled in this work
allows generalized conclusions to be made about the change in sources ﬁom regional
to atmospheric for cadmium, and the continuing importance of local sources in both
the 19708 and 19908 for copper and zinc.

The approach of sampling spatially and correlating a current environmental
parameter with watershed characteristics is becoming common for stream, rivers,
estuaries, and lakes (Detenbeck et al., 1993; Gergel et al., 2002; Grifﬁth, 2002;
Hunsaker and Levine, 1995; Jones et al., 2001; Paul et al., 2002; Rice, 1999; Soranno
et al., 1996), but this study has taken this approach one step further by combining
historical and regional aspects, and is the ﬁrst such study to use sediment chronologies
for such a large number of lakes. The high correlations found between metal
accumulation rates and watershed characteristics show that this approach can
successfully provide new insight into the history of metal loading to a region, and the

sources and pathways of those metals through the environment.

253

 

Recommendations

This work has many practical applications, both with the current results and as

guidance for future work. Outlined below are suggestions for future work, and a

discussion of the beneﬁts and possible applications of this study.

1.

Correlate level 3 urban land cover with metal accumulation rates to obtain
more information about possible sources that are associated with urban
land cover. The 19908 land cover data can be sub-divided into low
intensity urban (210% cover with manmade structures) and high intensity
urban (225% covered with manmade structures), which can be further
divided into “ ° rt”, “road/parking lot”, and “other high intensity urban”.
It is possible that one of these subcategories, such as road/parking lot, may
correlate as well, or better than total percentage urban in the watershed.
This would result in a clearer idea of possible sources of cadmium, copper,
lead and zinc to the environment.

Investigate actual urban sources. This work has identiﬁed urban land cover
as representing signiﬁcant sources of trace metals to the environment in the
19908, but to identify the actual sources further work is needed. A more
detailed study with measurements of releases of metals from potential
sources could provide the information necessary to reduce trace metals
inputs to the environment.

Determine the signiﬁcance of the slope and slope within a 100 m buffer
variables. The meaning of these variables remains unclear, and yet these

two watershed characteristics correlate very highly with metal

254

 

anthropogenic accumulation rates. Other variables that might represent
regional population distribution (e.g., population density in a 10 km buffer
around the lake) could be quantiﬁed, and correlated with both metal
accumulation rates and slope. Additionally, lakes that have relatively steep
slopes and are located in southern Michigan could be identiﬁed and
sampled to help determine if a causal relationship exists. Ifthere is no
causal relationship, and slope represents a regional gradient that is present
only in Michigan, another variable should be identiﬁed before this model
can be applied to other regions.

. Install atmospheric deposition traps near lake shorelines to compare actual
rates of atmospheric deposition to deposition rates in the lake sediment.

. Identify a quantiﬁable factor to represent the intensity of recreation use of
the lake. Many Michigan lakes have extensive recreational usage which
may act as a source of trace metals, but is not accounted for by any of the
watershed characteristics in this study.

. Test the multivariate regression models with additional Michigan lakes for
model validation. As part of the Michigan Department of Environmental
Quality’s Sediment Trend Monitoring Program, Michigan lakes will
continue to be sampled and these lakes can be used for model validation.

. Test the multivariate regression models in other regions. The approach for
developing these statistical models should have widespread applicability in
any region with sufficient lakes or reservoirs. However, it is unclear if the

variables that correlate highly will remain the same in the different region,

255

 

 

particularly the slope variable. Using this approach in a different region
will allow a comparison of the importance of different watershed
characteristics over a broader spatial area.

Overall, this work has beneﬁts at many different scales. Locally, contaminant
sediment concentrations and accumulation rates of trace metals are of great interest to
those living around a lake. The sediment chronologies collected also provide detailed
histories of contaminant inputs, and if metals accumulation rates are increasing or
decreasing. Regionally, information on metal accumulation rates and concentrations
across the state help identify those lakes with high levels of contamination.
Additionally, the multivariate regression models that were determined can be used to
help identify sources that are present in most Michigan lakes, as well as identify lakes
with unique sources (e.g., copper in Gratiot Lake). The importance of urban land
cover in the 19908 was identiﬁed which will allow for further investigation and
possible reduction of sources of trace metals. It is now also possible to assess whether
a lake is likely to have high accumulation rates of trace metals based on the
characteristics of the watershed. This will help identify those lakes that are most
appropriate to sample in the future.

Scientiﬁcally, this approach to identifying potential sources and evaluating
changes over time was successful, and may be applied to other regions. This provides
an important method for evaluating not only current sources and pathways, but also
comparing these results to historical data. Although the approach of correlating
watershed characteristics to current water quality variables has been extensively

investigated, the success of using lake sediment chronologies may result in a better

256

 

understanding in of the changing importance of different sources and pathways of

contaminants in the environment.

257

 

References

Alfaro-De la Torre, MC. and Tessier, A., 2002. Cadmium deposition and mobility in
the sediments of an acidic oligotrophic lake. Geochim. Cosmochim. Acta 66,
3549-3562.

Callender, E. and Rice, K.C., 2000. The urban environmental gradient: Anthropogenic
inﬂuences on the spatial and temporal distributions of lead and zinc in the
environment. Environ. Sci. Technol. 24, 232-238.

Callender, E. and vanMetre, P.C., 1997. Reservoir sediment cores show US lead
declines. Environ. Sci. Technol. 31, A424-A428.

Detenbeck, N.E., Johnston, CA. and Niemi, G.J., 1993. Wetland Effects on Lake
Water-Quality in the Minneapolis St-Paul Metropolitan-Area. Landscape
Ecology 8, 39-61.

Edgington, D.N. and Robbins, J .A., 1976. Records of lead deposition in Lake
Michigan sediments since 1800. Environ. Sci. Technol. 10, 266-274.

Eisenreich, S.J., Metzer, N.A. and Urban, N.A., 1986. Response of atmospheric lead
to decreased use of lead in gasoline. Environ. Sci. Technol. 20, 171-174.

Evans, RD. and Dillon, P.J., 1982. Historical changes in anthropogenic lead fallout in
southern Ontario, Canada. Hydrobiologia 91, 131-137.

Gergel, S.E., Turner, M.G., Miller, J .R., Melack, J .M. and Stanley, EH, 2002.
Landscape indicators of human impacts to riverine systems. Aquatic Sciences
64, 1 18-128.

Graney, J .R., Halliday, A.N., Keeler, G.J., Nriagu, J .O., Robbins, J .A. and Norton,
S.A., 1995. Isotopic record of lead pollution in lake sediments from the
northeastern United States. Geochim. Cosmochim. Acta 59, 1715-1728.

Grifﬁth, J .A., 2002. Geographic techniques and recent applications of remote sensing
to landscape-water quality studies. Water Air Soil Poll 138, 181-197.

Hunsaker, CT. and Levine, D.A., 1995. Hierarchical Approaches to the Study of
Water-Quality in Rivers. Bioscience 45, 193-203.

Jones, K.B., Neale, A.C., Nash, M.S., Van Remortel, R.D., Wickham, J .D., Riitters,
K.H. and O'Neill, R.V., 2001. Predicting nutrient and sediment loadings to
streams from landscape metrics: A multiple watershed study from the United
States Mid-Atlantic Region. Landscape Ecology 16, 301-312.

258

 

Kolak, J.J., Long, D.T., Beals, T.M., Eisenreich, SJ. and Swackhamer, D.L., 1998.
Anthropogenic inventories and historical and present accumulation rates of
copper in Great Lakes sediments. Appl. Geochem. 13, 59-75.

Mecray, E.L., King, J .W., Appleby, P.G. and Hunt, A.S., 2001. Historical trace metal
accumulation in the sediments of an urbanized region of the Lake Champlain
watershed, Burlington, Vermont. Water Air Soil Poll 125, 201-230.

Nriagu, J .0. (Editor), 1978. The Biogeochemistry of Lead in the Environment, 1A.
Elsevier, New York, 422 pp.

Paul, J.F., Comeleo, R.L. and Copeland, J ., 2002. Landscape metrics and estuarine
sediment contamination in the mid-Atlantic and southern New England
regions. J. Environ. Qual. 31, 836-845.

Rice, K.C., 1999. Trace-element concentrations in streambed sediment across the
conterminous United States. Environ. Sci. Technol. 33, 2499-2504.

Shahin, U., Yi, S.M., Paode, RD. and Holsen, T.M., 2000. Long-term elemental dry
deposition ﬂuxes measured around Lake Michigan with an automated dry
deposition sampler. Environmental Science & Technology 34, 1887-1892.

Soranno, P.A., Hubler, S.L., Carpenter, SR. and Lathrop, RC, 1996. Phosphorus

loads to surface waters: A simple model to account for spatial pattern of land
use. Ecological Applications 6, 865-878.

259

 

APPENDICIES

260

 

APPENDIX A

SEDIMENT CHEMICAL AND WATERSHED CHARACTERISTICS DATA

261

APPENDIX A. SEDIMENT CHEMICAL AND WATERSHED
CHARACTERISTICS DATA

Sediment concentration, 2|"Pb activities and dates, and core descriptions can be found
in Michigan Department of Environmental Quality reports including:

Cass, Gratiot. Gull. Elk and Higgig lakes:

Simpson, S.J., Long, D.T., Geisey, J.P. and Fett, J .D., 2000. Inland lakes sediment
trends: sediment analysis results for ﬁve Michigan Lakes (MI/DEQ/SWQ-01/030),
Department of Environmental Quality, East Lansing, MI.
http://www.michigan.gov/deq/0, 1 607,7-1 35-33 13_3686_3 728-32365-—,00.html

Cgstal M and Littleﬁeld lakes:
Yohn, S.S., Long, D.T., Giesy, J .P., Fett, JD. and Kannan, K., 2001. Inland lakes

sediment trends: sediment analysis results for two Michigan Lakes. MI/DEQ/WD-
02/115, Department of Environmental Quality, East Lansing.

Cadillac. Crvsg B. Millet. Paw Paw and Whitmore lakes:

Yohn, S.S., Long, D.T., Giesy, J .P., Scholle, L.K., Patino, L.C., Fett, J .D. and Kannan,
K., 2002. Inland lakes sediment trends: sediment analysis results for ﬁve Michigan
Lakes. MI/DEQ/WD-03/052, Department of Environmental Quality, East Lansing.

Houghton. Hubbard, Imp, Round. Torch. Witch lakes:

Yohn, S.S., Parsons, M.J., Long, D.T., Giesy, J.P., Scholle, L.K. and Patino, L.C.,
2003. Inland lakes sediment trends: sediment analysis results for six Michigan Lakes,
Michigan Department of Environmental Quality, East Lansing, MI.

 

262

Table l. Focusing corrected anthropogenic inventories (ug/cmz) of cadmium, copper,
lead and zinc in ﬁfteen Michigan lakes. Inventories could not be calculated for Cass
and Witch lakes.

 

 

Lake Cadmium Copper Lead Zinc
Cadillac 1 .0 246 140 121
Crystal B 14 98

Crystal M 2.0 6 169 1 1 1
Elk 0.9 7 74 103
Gratiot 0.6 15 51 48
Gull 1.5 14 208 193
Higgins 1 8 14 112 132
Houghton 1 .4 173 92 1 16
Imp 0.7 51

Littleﬁeld“ 1.2 108

Mullett 1 .0 8.5 81

Paw Paw“ 227

Round 1 .2 78 69
Torch 1 .2 61

 

‘ Focusing factor (2.0) was estimated ﬁ'om lakes sampled in 1999-2000.
*“ Background concentrations were estimated ﬁem Higgins Lake.

Table 2. Focusing corrected anthropogenic accumulation rates (pg/mzly) of cadmium,
copper, lead and zinc averaged from 1970 - 1980.

 

 

 

Lake Copper Cadmium Lead Zinc
Cadillac 41546 172 25447 17481
Cass 9876 345 75462 74481
Crystal B 1165 165 9921

Crystal M 1043 219 26113 17908
Elk 1044 105 14604 15531
Gratiot 1471 55 6299 4913
Gull 3810 217 42479 36941
Higgins 1702 190 17068 16653
Houghton 41448 219 18172 17788
Imp 285 67 7098 2196
Littleﬁeld 191 31246

Mullett 1136 99 13686

Paw Paw 8198 197 40381 42589
Round 108 10402 8041
Torch 130 9196

Whitmore 3647 405 56650 46397

 

 

263

 

Table 3. Focusing corrected anthropogenic accumulation rates (ug/mz/y) of cadmium,
copper, lead and zinc averaged from 1990 — 2000.

 

 

 

 

 

 

 

Lake Copper Cadmium Lead Zinc
Cadillac 82245 264 37193 34077
Cass 9299 182 40067 54103
Crystal B 2058 238 1 3837
Crystal M 2637 208 24814 18275
Elk 922 63 7685 7824
Gratiot 1678 44 4484 4102
Gull 2537 43 10969 12875
Higgins 1296 141 13049 1 1547
Houghton 34166 161 12787 1 3025
Imp 651 58 5862 2210
Littleﬁeld 598 98 14692
Mullett 1038 63 7670
Paw Paw 8491 123 18881 34324
Round 103 9961 7945
Torch 79 8535
Whitmore 4023 278 31858 34897
Table 4. Watershed characteristics that do not vary over time.
Watershed Watershed Lake Watershed: SO, K-mctor K-factor Slope
parameter area area lake ratio slope

corr.
Lake/Units km’ km’ kglhaly degrees
Cadillac 42.8 4.7 9.2 16.7 0.15 0.14 2.62
Cass 41.9 5.2 8.1 16.2 0.22 0.11 0.96
Crystal 8 66.2 39.3 1.7 15.7 0.16 0.18 4.82
Crystal M 9.2 2.9 3.2 16.4 0.25 0.15 2.11
Elk 98.1 31.3 3.1 14.7 0.19 0.27 3.38
Gratiot 21.1 5.8 3.6 8.5 0.16 0.12 5.32
Gull 53.4 8.2 6.5 18.5 0.26 0.12 1.96
Higgins 67.2 38.9 1.7 14.9 0.12 0.08 2.46
Houghton 380.3 81.2 4.7 15.1 0.13 0.06 1.87
Imp 1.7 0.3 5.0 8.8 0.31 0.20 3.77
Littleﬁeld 16.4 0.7 22.1 15.8 0.21 0.09 3.20
Mullett 1285.8 70.3 18.3 12.6 0.16 0.14 3.36
Paw Paw 25.9 3.7 6.9 19.4 0.28 0.08 1.99
Round 15.5 7.0 2.2 12.2 0.18 0.12 2.74
Torch 115.7 76.0 1.5 14.1 0.19 0.29 4.31
Whitmore 9.1 2.7 3.3 16.6 0.23 0.14 1.10
Witch 11.1 0.9 13.0 9.3 0.29 0.16 2.90

 

264

Table 4 (cont’d)
Watershed k-factor, k-factor. Slope,
parameter 100 in slope, 100 m

 

 

 

100 m

Lake/Units degrees
Cadillac 0.14 0.07 1 .398
Cass 0.28 0.09 0.785
Crystal 6 0.17 0.17 8.741
Crystal M 0.32 0.16 2.636
Elk 0.18 0.28 3.419
Gratiot 0.16 0.09 4.213
Gull 0.26 0.12 1.681
Higgins 0.13 0.10 3.29

Houghton 0.13 0.06 1.991
Imp 0.31 0.20 5.661
Littleﬁeld 0.26 0.01 1.28

Mullett 0.15 0.07 2.727
Paw Paw 0.33 0.09 1.896
Round 0.19 0.10 3.04

Torch 0.19 0.29 4.643
Whitmore 0.20 0.18 0.924
Witch 0.29 0.16 3.136

 

Table 5. Watershed characteristics for the 19708 (1970-1980). “Ag” indicates
agriculture, “100 m” indicates within 100 m buffer of the lake, “ﬁv” indicates ﬂow
inverted value, “ﬁp” indicates ﬂow inverted percent, “ﬁsv” indicates ﬂow inverted
squared value, “ﬁsp” indicates ﬂow inverted squared percent.

 

 

 

Watershed Population Population % urban % ag % % %
parameter density range forest wetland
Lake/Units people/km people % % % % %
Cadillac 102.6 4394.7 19.7 40.7 10.2 24.4 4.1
Cass 736.6 30580.5 60.5 2.4 13.9 8.6 12.2
Crystal 8 23.5 1572.0 12.4 12.9 21.2 51.9 1.2
Crystal M 21.9 201.9 11.1 58.2 5.0 16.9 8.2
Elk 20.9 2084.7 6.3 28.0 19.8 41.0 3.9
Gratiot 1.1 24.1 0.8 0.0 0.0 98.1 0.6
Gull 37.9 1998.9 9.7 59.0 7.9 14.1 5.8
Higgins 14.2 981.8 15.0 0.5 3.5 77.4 3.4
Houghton 12.9 4776.6 7.1 1.6 5.3 74.5 9.8
Imp 1.4 2.4 7.3 0.0 0.0 86.7 4.9
Littleﬁeld 8.0 131.5 1.8 9.3 13.8 69.7 4.2
Mullett 6.6 8474.7 3.0 7.4 15.2 70.8 2.9
Paw Paw 82.5 2127.0 17.3 30.1 13.2 35.0 2.1
Round 5.5 84.9 4.8 13.8 11.4 68.5 1.0
Torch 16.1 1961.6 7.3 23.5 21.9 44.8 1.9
Whitmore 64.3 558.1 22.1 23.5 20.5 23.7 8.7
Witch 4.8 53.3 1.8 2.9 6.7 80.6 6.5

 

265

Table 5 (cont’d).

 

 

 

 

 

 

Watershed Urban Ag Range Forest Wetland %Forest %Ag
parameter 100 m 100 m
Lake/Units km2 ltrn2 km2 km2 km2 % %
Cadillac 8.45 17.41 4.35 10.43 1.76 3.06 0.00
Cass 25.34 1.02 5.83 3.59 5.11 2.68 0. 00
Crystal 8 8.20 8.55 14.06 34.34 0.80 26.34 0.47
Crystal M 1.02 5.38 0.46 1. 56 0.76 0.56 12.84
Elk 6.20 27.43 19.46 40. 22 3.85 20.92 7.88
Gratiot 0.18 0.00 0.00 20. 67 0.13 79.07 0.00
Gull 5.18 31.51 4.22 7.55 3.09 4.86 0.07
Higgins 10.07 0.33 2.33 52.01 2.31 12.24 0.00
Houghton 27.04 6.16 20.09 283. 39 37.41 4.19 0.38
Imp 0.12 0.00 0.00 1.48 0.08 60.99 0.00
Littleﬁeld 0.30 1.53 2.26 11.42 0.69 43.16 0.00
Mullett 38.45 94.59 195.22 910. 66 36.78 38.83 0.43
Paw Paw 4.50 7.81 3.42 9.08 0.54 1.90 0.64
Round 0.74 2.14 1.77 10. 63 0.15 43.66 0.00
Torch 8.48 27.20 25.39 51.88 2.21 24.02 0.15
Whitmore 2.00 2.13 1.86 2.15 0.79 3.00 0.27
Witch 0.20 0.33 0.74 8.94 0.72 75.67 0. 00
Watershed %range %urban %wetland Forest Ag Range Urban
parameter 100m 100m 100m 100m 100m 100m 100m
Lake/Units % % 91. km’ km’ km’ km’
Cadillac 1.36 86.67 1.95 0.03 0.00 0.01 0.92
Cass 1.58 63.80 8.65 0.04 0.00 0.02 0.94
Crystal 8 0.64 68.18 1.43 0.96 0.02 0.02 2.49
Crystal M 0.00 76.74 1.37 0.00 0.09 0.00 0.53
Elk 9.40 49.83 5.83 1.01 0.38 0.46 2.41
Gratiot 0. 00 10.52 2.25 0.87 0.00 0. 00 0.12
Gull 3. 42 84.25 0.00 0.10 0.00 0. 07 1. 72
Higgins 0. 40 82.48 0.00 0.42 0.00 0.01 2. 81
Houghton 1.24 86.39 4.14 0.25 0.02 0.07 5.09
Imp 0.00 35.62 0.00 0.17 0.00 0.00 0.10
Littleﬁeld 1.60 25.52 9.10 0.34 0.00 0.01 0.20
Mullett 5.22 38.65 12.30 2. 83 0.03 0.38 2. 81
Paw Paw 2. 70 82.42 4.67 0.03 0.01 0.04 1.24
Round 10. 28 37.67 0.60 0.49 0.00 0.12 0. 42
Torch 10. 99 62.34 0.83 1.61 0.01 0.74 4.18
Whitmore 7. 88 83.19 2.91 0.02 0.00 0.06 0. 64
Witch 8. 72 10.53 0.00 0.38 0.00 0.04 0. 05

 

266

Table 5 (cont’d).

 

 

 

 

 

 

 

Watershed Wetland Urban Ag Range Forest Wetland Urban
parameter 100 m ﬁv ﬁv ﬁv flv frv ﬁp
Lake/Units km1r %
Cadillac 0.02 22.26 5.10 1.96 4.43 1.87 62.5
Cass 0.13 2. 58 0 05 0.47 0.32 0.56 64.9
Crystal 8 0.05 49. 57 5.17 7.75 48.86 0. 76 44.2
Crystal M 0.01 10.16 5.43 0. 31 1. 36 0.52 57. 2
Elk 0.28 44.84 21.34 24. 68 33. 02 7.83 34.0
Gratiot 0.02 2. 35 0. 00 0. 00 154.75 2.18 1.5
Gull 0.00 29.11 12.56 4. 23 8. 58 2.18 51.4
Higgins 0.00 61.95 0.32 1.78 43. 67 1.35 56.8
Houghton 0.24 88.42 2.25 5. 87 65.66 16.04 49.6
Imp 0.00 1. 93 0. 00 0. 00 5. 95 0.24 23.8
Littleﬁeld 0.07 1 .66 0. 59 1.41 14.65 2. 00 8. 2
Mullett 0.90 56.92 12.69 29.00 127. 08 13. 34 23. 8
Paw Paw 0.07 20.69 4.23 2.32 5. 31 1.66 60.5
Round 0.01 6.61 2.59 4. 64 18. 39 0.18 20.4
Torch 0.06 73.13 16.11 39. 43 76. 63 3.07 35.1
Whitmore 0.02 1.04 0. 18 0.36 0.34 0.16 50.3
Witch 0.00 1.18 0. 21 1. 34 14.09 0. 42 6.9
Watershed Ag Range Forest Wetland Urban Ag Range
parameter ﬁg ﬁp ﬁp lip fisv ﬁsv fisv
Lake/Units % % % %
Cadillac 14.3 5.5 12. 4 5.3 0.316 0. 003 0. 009
Cass 1.1 11.8 8.1 14.1 0.014 0.000 0.001
Crystal 8 4.6 6.9 43.6 0.7 1. 096 0.011 0.012
Crystal M 30.6 1.7 7. 7 2.9 0.217 0.019 0.000
Elk 16.2 18.7 25.1 5.9 0.996 0.057 0. 192
Gratiot 0.0 0.0 97.2 1.4 0.065 0.000 O. 000
Gull 22.2 7.5 15.1 3.8 0.542 0.008 0. 017
Higgins 0.3 1.6 40.0 1.2 1.199 0. 000 0. 003
Houghton 1.3 3.3 36.8 9.0 1. 588 0.010 0. 023
Imp 0.0 0.0 73.3 2.9 0.049 0.000 0. 000
Littleﬁeld 2.9 7.0 72.1 9.8 0. 027 0.000 0. 001
Mullett 5.3 12.1 53.2 5.6 1.169 0.022 0.124
Paw Paw 12.4 6.8 15.5 4.9 0.341 0. 005 0.005
Round 8.0 14.3 56.7 0.6 0.159 0. 004 0.047
Torch 7.7 18.9 36.8 1.5 1.583 0. 023 0.303
Whitmore 8. 5 17. 2 16.3 7.6 0.009 0.000 0.002
Witch 1.2 7. 8 81. 7 2.4 0.023 0.000 0.013

 

267

Table 5 (cont’d).

 

 

 

Watershed Forest Wetland Urban Ag Range Forest Wetland
parameter fisv ﬁsv ﬁsp ﬁsp ﬁsp lisp ﬁsp
Lake/Units % % % % %
Cadillac 0.015 0.007 90.5 0.7 2.6 4.2 1.9
Cass 0.001 0.002 74.7 0.1 7.6 6.4 11.2
Crystal 8 0.308 0.002 76.7 0.8 0.8 21.6 0.2
Crystal M 0.003 0.005 89.2 7.6 0.1 1.1 1.9
Elk 0.317 0.120 59.2 3.4 11.4 18.8 7.2
Gratiot 2.786 0.053 2.2 0.0 0.0 96.0 1.8
Gull 0.038 0.003 89.2 1.4 2.7 6.2 0.5
Higgins 0.192 0.002 85.9 0.0 0.2 13.7 0.1
Houghton 0.124 0.123 85.0 0.5 1.2 6.6 6.6
Imp 0.071 0.001 40.8 0.0 0.0 58.6 0.6
Littleﬁeld 0.103 0.040 16.0 0.1 0.7 59.8 23.4
Mullett 0.553 0.134 58.4 1.1 6.2 27.6 6.7
Paw Paw 0.014 0.028 86.5 1.4 1.4 3.6 7.2
Round 0.145 0.000 44.6 1.1 13.3 40.9 0.1
Torch 0.688 0.024 60.4 0.9 1 1.6 26.2 0.9
Whitmore 0.001 0.000 73.4 1.9 15.1 7.0 2.7
Witch 0.175 0.000 11.0 0.1 6.2 82.5 0.1

 

Table 6. Watershed characteristics for the 19908 (1990-2000). “Ag” indicates
agriculture, “up” indicates upland/openland, “100 m” indicates within 100 or buffer of
the lake, “ﬁv” indicates ﬂow inverted value, “ﬁp” indicates ﬂow inverted percent,
“ﬁsv” indicates ﬂow inverted squared value, “ﬁsp” indicates ﬂow inverted squared

 

 

 

 

percent.
Watershed Population Population % urban % ag % up % %
parameter density forest wetland
Lake/Units people/km people % % % % %
Cadillac 120.3 5154.5 12.0 33.5 18.7 30.9 2.0
Cass 832.2 34550.5 31.0 0.5 19.1 38.1 10.2
Crystal 8 29.8 1989.0 6.0 10.8 22.3 58.9 0.9
Crystal M 29.5 271.3 11.1 49.4 2.8 21.1 13.5
Elk 36.0 3588.9 4.7 24.7 17.7 49.6 1.8
Gratiot 1.6 34.8 0.9 0.0 6.2 89.0 3.5
Gull 45.8 2416.5 5.3 57.7 6.3 22.2 5.8
Higgins 35.3 2434.6 7.0 0.1 15.3 72.9 3.7
Houghton 23.5 8660.2 4.1 0.8 15.3 62.2 15.7
Imp 1.7 2.9 1.7 0.0 0.5 94.9 0.7
Littleﬁeld 11.6 189.0 2.6 5.1 13.1 73.6 4.2
Mullett 9.8 12600.1 2.8 7.5 17.6 68.0 3.1
Paw Paw 101.7 2621.0 12.4 33.6 14.1 31.9 6.6
Round 4.8 74.1 5.3 8.3 9.3 67.2 9.1
Torch 22.1 2682.8 4.1 18.1 19.3 55.3 1.6
Whitmore 144.8 1256.4 12.3 17.0 16.0 41.8 11.7
_Witch 3.1 34.6 4.0 0.0 4.9 86.7 2.1
268

Table 6 (cont’d).

 

 

 

 

 

 

 

Watershed Urban Ag Up Forest Wetland %Forest %Ag
parameter 100 m 100 m
Lake/Units kmz km’ km’ km2 km2 % %
Cadillac 5.2 14.3 8.0 13.2 0.9 25.4 0.8
Cass 13.0 0.2 8.0 15.9 4.3 48.7 0.0
Crystal B 3.9 7.2 14.8 39.0 0.6 45.4 1.9
Crystal M 1.0 4.6 0.3 1.9 1.2 10.7 7.9
Elk 4.6 24.3 17.3 48.7 1.8 48.8 7.5
Gratiot 0.2 0.0 1.3 18.8 0.7 68.9 0.0
Gull 2.8 30.8 3.4 11.8 3.1 46.1 0.7
Higgins 4.7 0.1 10.3 49.0 2.5 47.8 0.2
Houghton 15.7 3.2 58.1 236.6 59.6 21.1 0.0
Imp 0.0 0.0 0.0 1.6 0.0 79.7 0.0
Littleﬁeld 0.4 0.8 2.1 12.1 0.7 29.7 1.3
Mullett 35.5 95.9 226.8 874.1 40.4 44.3 0.6
Paw Paw 3.2 8.7 3.7 8.3 1.7 25.2 3.9
Round 0.8 1.3 1.4 10.5 1.4 45.7 2.7
Torch 4.8 21.0 22.4 64.0 1.8 50.1 3.5
Whitmore 1.1 1.5 1.5 3.8 1.1 29.1 0.2
Witch 0.4 0.0 0.5 9.6 0.2 74.7 0.0
Watershed %up %urban %wetland Forest Ag Up Urban
parameter 100m 100m 100m 100m 100m 100m 100m
Lake/Units % % % km’ km2 km? km?
Cadillac 6.1 47.3 5.6 0.27 0.01 0.06 0.51
Cass 13.3 16.0 19.3 0.73 0.00 0.20 0.24
Crystal 3 11.3 24.5 4.3 1.53 0.06 0.38 0.82
Crystal M 4.9 63.1 2.4 0.07 0.05 0.03 0.43
Elk 11.1 9.7 8.5 1.98 0.30 0.45 0.39
Gratiot 7.3 3.3 14.5 0.71 0.00 0.08 0.03
Gull 13.0 21.6 9.9 0.94 0.01 0.27 0.44
Higgins 11.7 17.4 6.3 1.59 0.01 0.39 0.58
Houghton 9.1 46.2 9.8 1.03 0.00 0.44 2.24
Imp 1.0 7.3 1.3 0.23 0.00 0.00 0.02
Littlefield 9.1 15.4 20.1 0.23 0.01 0.07 0.12
Mullett 9.0 16.8 14.0 2.32 0.03 0.47 0.88
Paw Paw 8.2 42.3 20.0 0.38 0.06 0.12 0.64
Round 12.1 4.6 24.4 0.47 0.03 0.12 0.05
Torch 14.2 9.0 6.2 3.29 0.23 0.93 0.59
Whitmore 13.2 41.7 8.0 0.22 0.00 0.10 0.31
Witch 5.5 8.5 2.7 0.38 0.00 0.03 0.04

 

269

Table 6 (cont’d).

 

 

 

 

 

 

Watershed Wetland Urban A9 Up Forest Wetland Urban
parameter 100 m ﬁv fiv fiv fiv ﬁv ﬁp
Lake/Units km2 %
Cadillac 0.06 12.97 4.31 4.76 10.31 1.61 38.18
Cass 0.29 0.77 0.01 0.88 1.95 0.46 18.89
Crystal 8 0.14 19.92 4. 02 13.56 68.18 3.08 18.31
Crystal M 0.02 8. 06 4. 86 0.69 2. 63 1.05 46.61
Elk 0.34 11. 57 23.79 20.74 66.27 6.97 8. 94
Gratiot 0.15 1.17 0.00 5.49 139.68 11.42 0.74
Gull 0.20 8.14 12.71 7.68 22.64 4.47 14.63
Higgins 0.21 16.75 0.26 12.95 69. 27 4.91 16.09
Houghton 0.48 50.77 1.63 21.93 72.27 27.87 29.10
Imp 0.00 0.29 0.00 0.03 7. 46 0.03 3.68
Littleﬁeld 0.16 1.15 0.52 2.51 12.19 3.09 5. 92
Mullett 0.74 23.80 13.25 33.77 145.51 16.51 10. 22
Paw Paw 0.30 11.24 5.84 3.60 10.57 3.84 32.04
Round 0.25 1. 34 1.88 2.94 19.95 5.46 4. 26
Torch 0.41 17.45 17.63 34.87 117.49 7.49 8. 95
Whitmore 0.06 0.45 0.10 0.37 0.97 0.19 21.43
Witch 0.01 1.33 0.00 1.01 14.39 0.41 7.78
Watershed Ag Up Forest Wetland Urban Ag Up
parameter ﬁp ﬁp ﬁp ﬁp fisv ﬁsv ﬁsv
Lake/Units % % % %
Cadillac 12.69 14.02 30.35 4.75 0.163 0.004 0. 028
Cass 0.14 21.55 48.06 11.37 0.003 0.000 0. 004
Crystal 8 3.70 12.47 62.69 2. 83 0.432 0.020 0.149
Crystal M 28.13 3.98 15.21 6. 06 0.168 0.016 0.012
Elk 18.39 16.03 51.24 5. 39 0.164 0.150 0. 207
Gratiot 0.00 3.48 88.54 7.24 0.020 0.000 0. 066
Cu" 22.85 13.81 40.69 8.03 0.109 0. 015 0.125
Higgins 0.25 12.44 66.52 4. 71 0.237 0. 004 0.147
Houghton 0.93 12.57 41.42 15. 98 0.841 0. 001 0.191
Imp 0. 00 0.41 95.55 0. 36 0. 005 0.000 0.000
Littlefield 2. 68 12.92 62.62 15.87 0.015 0.002 0.017
Mullett 5.69 14.50 62.49 7. 09 0.348 0.019 0.217
Paw Paw 16.64 10. 25 30.13 10.94 0.170 0. 021 0.036
Round 5.94 9. 30 63.19 17.31 0.008 0. 009 0. 034
Torch 9.04 17.89 60.27 3.84 0. 246 0. 092 0.365
Whitmore 4. 93 17.86 46.69 9. 09 0. 003 0. 000 0.002
Witch 0. 00 5.91 83.90 2. 41 0.021 0. 000 0.013

 

270

 

Table 6 (cont’d).

 

 

 

Watershed Forest Wetland Urban Ag Up Forest Wetland
parameter tisv ﬁsv ﬁsp ﬁsp lisp ﬁsp tisp
Lake/Units % % % % %
Cadillac 0.100 0.018 52.119 1.223 8.829 31.943 5.885
Cass 0.011 0.002 13.226 0.006 22.061 55.093 9.614
Crystal 8 0.627 0.082 32.977 1.539 11.375 47.824 6.285
Crystal M 0.027 0.007 72.863 6.965 5.319 11.647 3.206
Elk 0.918 0.159 10.290 9.364 12.978 57.410 9.957
Gratiot 2.493 0.290 0.714 0.000 2.315 86.864 10.108
Gull 0.236 0.082 19.237 2.597 22.105 41.643 14.418
Higgins 0.756 0.104 18.961 0.350 11.805 60.544 8.340
Houghton 0.470 0.261 47.663 0.078 10.835 26.614 14.810
Imp 0.109 0.000 4.473 0.000 0.131 95.329 0.066
Littlefield 0.060 0.049 10.480 1.168 11.934 41.966 34.452
Mullett 1.067 0.212 18.683 1.027 11.634 57.289 11.367
Paw Paw 0.1 14 0.074 41.055 4.973 8.780 27.366 17.826
Round 0.174 0.112 2.269 2.546 10.190 51.676 33.319
Torch 1.320 0.165 11.226 4.185 16.683 60.357 7.550
Whitmore 0.006 0.001 29.160 0.497 17.923 47.804 4.616
Witch 0.163 0.008 10.278 0.000 6.422 79.279 4.021

 

271

 

APPENDIX B

METHODS FOR ACQUISITION OF WATERSHED CHARACTERISTTC DATA

272

 

APPENDIX B. METHODS FOR ACQUISITION OF WATERSHED
CHARACTERISTICS DATA

The following methods were used in the acquisition of the watershed
characteristics data used in Chapters 4, 5, and 6. ArcView v. 3.2, and Arclnfo v. 8.2
run on a UNIX platform were used for all geographic information systems work.

Delineating watersheds using Shuttle Radar Topography Mission (SRTM) digital
elevation data

1. Determine what county the lake is in.
a. Download the lake coverage from the Michigan DNR spatial data
library, and open it in ArcView to get a good sense of where the lake is.
2. Download SRTM DEM data from USGS enhanced download site:
http://seamless.usgs.gov/viewer.htm
a. Make sure the 30 m DEM is checked on the right — this is the data you
want. Some areas (near Detriot) don’t have 30 m DEMs. Then use 60
m DEM instead, but note the different resolution.
b. Zoom into the area around the lake
0 Select area to download - here try to guess the size of the watershed -
if too small an area is downloaded it will have to be done again, but the
larger the area the longer everything takes.
(1. Download only the 30 m DEM ﬁle, extract in vvinzip. FTP to rurix
machines.
3. Rename all ﬁles and folders in lowercase. Renaming can be in WS-FTP, or
before FTPing to Unix machines.
4. If the ﬁle names have 8 numbers, rename the ﬁrst 3 numbers to “dat” so that
the results is, for example, dat52369. Leave the last 5 numbers as they were.
(I have no idea why this is necessary - but when it was all numbers it didn’t
seem to work).
5. Build the statistics ﬁle with buildsta command in GRID:
GRIT): buildsta dat28789
6. Project the data in GRID with the project command - use cubic convolution:
Usage: (*) PROJECT (<grid>, {projection_ﬁle}, {NEARESTI
BILINEAR | CUBIC}, {out_cellsize}, {x_register, y__register})

*To get to the GRID module type grid at the Arc> prompt.

Grid: mul_elev = project(dat28397,#,cubic)

273

 

 

iititttti*tttt*#******¥***¢*#********¢$****###***#

* The INPUT projection has been deﬁned. "‘
#80808088##0##080800888#0808088*******#*******tt**
Use OUTPUT to deﬁne the output projection and END

to ﬁnish.

Project: output

Project: projection oblique_mercator

Project: datum nar_c

Project: units meters

Project: spheroid grsl980

Project: parameters

Projectiontype<l |2>: 2

Scale factor at the projection's center [ 1.00000 ]: 0.99960
Longitude of the projection's center (DMS) [ 0 0 0.000 ]: -86 0

0.000

Latitude of the projection's center (DMS) [ 0 0 0.000 ]: 45 18
33.000

Azimuth at the projection's center [ 0.00000 ]: 337.25556
False easting (meters) [ 0.00000]: 254673149600
False northing (meters) [ 0.00000 ]: -4354009.8160
Project: end

7. At this point I usually open this coverage in ArcView with a ﬁle for the lakes
of that county (Available at DNR spatial data library) to make sure that I
projected right. If there is a typo, the DEM will not appear on top of the lake
in ArcView.

8. In Arcview, look at the DEM and check for missing data (will appear as black
grid cells). IF data are missing, you MUST put data into these spots — or else
the water will ﬂow into the missing data. See separate ﬁle to determine how to
do this.

9. Go back to Are Info. Fill the sinks: Usage: FILL <in_grid> <out_grid> {SINK
I PEAK} {z_limit} {out_dir_glid}

GRIT): ﬁll mul_elev mul_elevf sink

10. Compute ﬂowdirection in GRID:

Usage: (1) F LOWDIRECTION (<surface_grid>, {o_drop_grid}, {NORMAL
| F ORCE})
mul_fdir = ﬂowdirection(mul_elevf)

ll. Compute ﬂowaccumulation (only to help determine point of highest
accumulation to delineate watershed from)

Usage: (F) F LOWACCUMULATTON (<dir_grid>, {weight _grid})
GRID: mul_facc = ﬂowaccumulation(mul_fdir)

12. Open Arcview and ﬁnd point with highest ﬂow accumulation within the lake.
You will have to change the scale — I usually use standard deviation, in units of
'/4 deviation.

274

13. Create shapeﬁle with points to delineate watershed:
a. view>new theme
i. feature type: point
ii. mulout.shp

b. Click on point button, then pick points with high ﬂow accum around
lake. Arc will delineate the areas that ﬂow into these points. May be
one or many for a lake. Always try one ﬁrst.

c. theme>stop editing

14. Are info — goto lake/mullett/ directory.
shapearc mulout mulout
Shapearc changes it from a shp to a coverage. shapearc shp-ﬁle coverage

15. Go into Grid (type “grid”)

16. mulshed = watershed(mul_fdir,selectpoint(mul_fdir,mulout,inside))
selectpoint says to select all the grid points inside of mulout for the grid

coverage mul_fdir. Watershed says use the ﬂow direction grid stating at the
point given and delineate the area that ﬂoes into the points given.

17. Open this grid up in arcview to determine if the area covers what appears to be
the whole watershed. If not, reopen shapeﬁle and edit it. Go into arc and kill
mulout all and kill mulshed all. Then repeat steps 5-7. IF the watershed
touches the edge of the DEM or comes very near it, go back to step one and
download a larger area.

18. IF more than one point was used to delineate the watershed, change to one
grid:

- create a text ﬁle reclass.txt with the text: 1 500 : 3 where 3 is an arbitrary #.
- GRID> mulshedg = reclass(mulshed,reclass.brt)

19. Change to poly
ARC> gridpoly mulshedg mulshedv

20. IF one or several points will not deﬁne the watershed, use a polygon.
Shapeﬁle > new theme > polygon
Draw a polygon arormd lake, save and close.

In arcview, transform it to a grid Theme>convert to grid

Then skip to watershed delineation step.

mulched = watershed(mul_fdirmuloutg)

If necessary, add onto the polygon using the append new polygon button.

NOTES:

Projections - we project everything into Michigan Georef. You can tell what

projection a ﬁle is in by (1) Opening it in Arcview and opening up something else you

know is in MI Georef and seeing if they line up, and (2) in Are, type describe

coveragename, and it will give you info on the projection. The datum that we type in

275

 

to switch the SRTM data to Michigan Georef is not real datum, but a code for ArcInfo
that allows it to switch between those two particular datums. The actual Michigan

Georef datum is NAD83 (see the help menu for more info about this)

I often check the watershed again the previously delineated ﬁle — for small lakes they
usually don’t have a watershed delineated just for that lake, but some of the edges

should match up. There is it UP and LP ﬁle already on the UNIX machines.

IF there is another lake in the watershed, I remove the watershed of that lake from the
overall watershed. [rationalez lakes act as sediment traps, most metals are on
sediments. Therefore, the land upstream from a lake within the watershed of interest
will not contribute to the sediment of the lake of interest. This will be signiﬁcant for
some lakes (e.g., Mullett, Torch) some of this will be a judgment call.]. To do this:
delineate the watershed of the secondary lake in the same way. Merge the watersheds
of the primary and secondary lakes (the secondary watershed should be within the
primary) - both will have to be coverages or shape ﬁles to do this, it can be done in
ArcView. Then select the portion of the merged watershed that is just the primary
watershed with the secondary removed, and convert to shapeﬁle (again, arcview).

Then change this to a coverage (shapearc command in Arcinfo).

276

 

Creation of a watershed with sample lake removed (watershed_nl)
For many of the analyses, it is desirable to have a coverage or shape ﬁle of the
watershed without the lake present. For example, soils data in the watershed, land
cover not including the lake, slope of the watershed — all of these characteristics of the

watershed should not include the actual lake.

This watershed_nl ﬁle will also need to be used to clip grids for ﬂow weighted land
cover and average slope calculations. The grid clip command does not work well if
the clip shape (in this case, the watershed_nl ﬁle) is too complex. Therefore, it is
necessary to try clipping a grid alter the watershed_nl ﬁle is created to determine if
this ﬁle will work. If not, sometimes the shape of the lake has to be simpliﬁed (see
Cass Lake). Grid clip is most likely not to work when the watershed_nl ﬁle is made of
more than one polygon (e.g., if there are 2 places where the lake touches the

watershed).

1. Lake must touch the watershed edge so that the lake is not an “island polygon”.
a. If the lake does NOT touch the edge of the watershed, edit the shape
ﬁle of the lake and add a polygon to create a small channel ﬁ'om the
edge of the lake to the boundary of the watershed. It is preferable to do
this where the lake is closest to the watershed boundary, and/or where
there is a river outflow ﬁom the lake to the watershed boundary.
Typically, the lake should touch or be very close to the watershed
boundary.
In Arcview: Open shp ﬁle for the lakes in that county.
Select the sampled lake, and create a new shape ﬁle. (e.g., cass_lake.shp)
Union this with the watershed polygon. (Using geoprocessing wizard)
Select the watershed without the lake, and make this a new shapeﬁle.
(Theme>convert to shapeﬁle)
In arcinfo Shapearc this to get a coverage of the watershed without the lake.
a. ARC: shapearc shapeﬁle coveragename
7. Clean the new coverage

Shit?!"

9‘

277

a. ARC: clean coveragename
8. Check to see if gridclip will work

a. Go to arcinfo, grid module

b. Grid: gridclip gridtobeclipped outgrid cover clipcoverage

c. For example - to clip the IFMAP landcover grid for Cadillac Lake:

Grid: gridclip ifmap\lp cadillac\cad_ifmap cover cadillac\oad8hed_nl

d. *NOTE that the grid module does NOT like the path ..\ to go back up a
directory, therefore you MUST be in the lowest common directory of
the ﬁles that you want to use.

6. Open the clipped grid in Arcview and see if the landcover ﬁle was
clipped properly. If it did not work, there will active grid cells either in
the lake or outside the watershed, and often no active grid cells within a
part of the watershed.

f. If the grid was clipped correctly, continue using this watershed_nl ﬁle,
if it was not clipped correctly, modify this ﬁle or the lake ﬁle until the
gridclip will work. Then use this new watershed_nl ﬁle for all
appropriate analysis. Although modiﬁcations to make the gridclip
work add some inaccuracy to the ﬁle, it is necessary to be consistent
through all data acquisition.

NOTES:
- There will be some error in method because the DNR lake ﬁles will not be
identical to where the lakes are located on the DEMs or land cover ﬁles.

Filling in no data cells
Grid: outgrid = con (isnull (ingridl), 0, ingridl)

The above expression says that if (CON) the cell value on ingridl equals NODATA
(ISNULL), then assign 0 to it (true_expression = 0); if it does not (it is a valid value),
then get the input value back (false__expression = ingridl ). To perform the reverse and
set cells with speciﬁc valid values to NODATA (to mask out cells), use the SETNULL
function.

However, we don’t want to set nodata cells to 0, we want to set them to a reasonable
elevation. So, pick a value slightly lower than those around the missing data point (so
that it will get ﬁlled). Usually the missing data is in the lake anyway, so the exact
elevation is unimportant.

Grid: crb_elev2 = con (isnull (crb_elev),169,crb_elev)

278

Determining watershed population density 1990-1870
This method determines the population density in the watershed with the assumption
that if the land is state owned or underwater people will not be living there. Township
census data are used. To determine the “livable” area in each township census area,
the area of each census polygon that is completely or partially in the watershed is
determined, then the area of each census polygon within the watershed is determined.
Then the area owned and the area in lakes within each census polygon is determined,
then the area within each census polygon within the watershed is determined. This
allows the calculation of what % of the livable land in a census polygon is in the
watershed. The assumption is that if 5% of the livable land of a census polygon is in

the watershed, then 5% of the people live in the watershed.

The calculations are as follows.

For EACH census area:

Total livable area = total area — lake area — owned area

Total livable in watershed (in that census area) = total watershed — lake in watershed —
owned in wshed

% livable of census area that is in watershed = livable in watershed / total livable

For each census area: people in watershed = total people in census area " % livable of
census area in watershed.

Sum this for all of the census areas, then divide by the size of the watershed to get
population density (people/km2). It seems best to use the size of the watershed
without the sampled lake to determine population density, since this is the land
size of the watershed. Whether or on the sampled lake is included in the
watershed area will make a signiﬁcant difference for lakes with small watersheds.

279

 

1. Determine lake area within all of the township census areas
a. Copy or download lake shape ﬁles to the correct lake directory.
b. In arcview open mi _pop4 and lake.shp
c. Anaysis>Tabulate areas:

i.

ii.
iii.
iv.

V.

Row theme: lake. shp

Row ﬁeld: Lake_type

Column theme: poly (from mi _pop4)
Column ﬁeld: poly#

Save resulting table as tab delimited text.

2. Determine lake area within the watershed
a. Union watershed and mr _pop4

i.

In arcview: Arctools>command tools> analysis>overlay>
union, mi _pop4 and watershed

b. Open new union coverage in arcview, open the polygon, not the region
c. Tabulate area:

i.

ii.
iii.
iv.

V.

Row theme: lake. shp

Row ﬁeld: Lake_type

Column theme: lakeunion

Column ﬁeld. lakeunion#

Save resulting table as tab delimited text.

(1. Manually relate which union polygon 18 in which original mi_pop4
polygon by using arcview and seeing which polygon numbers
correspond. Write this down.

3. Determining area owned in mi_pop4

3. Download or copy state ownership ﬁles for each county, open in
Arcview

b. Theme>Query: sm'f_cn >= 1 or fee_cnt >= 1, new set

c. Theme> convert to shapeﬁle

d. Analysis> tabulate area:

i.
ii.
iii.
iv.

V.

Row theme: lakeown. shp

Row ﬁeld: county

Column theme: poly

Column ﬁeld: poly#

Save resulting table as tab delimited text.

4. Determine area owned in watershed
a. Open lakeunion
b. Analysis> tabulate area:

i.
ii.
iii.
iv.
v.

Row theme: lakeown. shp

Row ﬁeld: county

Column theme: lakeunion

Column ﬁeld: lakeunion#

Save resulting table as tab delimited text.

280

 

5. Determine area ofeach mi_pop4 that is in the watershed

a.

Analysis> tabulate area:
1. Row theme: lakeunion. shp
ii. Row ﬁeld: lakeunion#
iii. Column theme: watershed
iv. Column ﬁeld: watershed_ id
v. Save resulting table as tab delimited text.

6. Excel calculations: areas calculated are the areas that are owned, and the areas
that are lake, and therefore the remaining area must be calculated.

a.

b.

FTP all .txt ﬁles to disk, open in Excel

Copy ﬁles into calculation ﬁle (See area calculations with
ownership.xls), transpose data when copying so that the poly number is
a column.

From the watershed area, copy the lakeunion#. These are the polygons
ﬁom union in the watershed.

Go to arcview and determine what the poly# are for the union#. These
polygons are the only ones of importance, and other poly#s can be
ignored.

If there are multiple union polygons in one mi _pop4 polygon, make
only one row and add all lake and union data for that one mim_pop4
polygon.

Total area can be collected ﬁ'om the population data, matching the
correct poly# (NOT union#!)

Lake area and owned area: match poly#, include lake and island area
(most islands are small, we assume no one lives on them)

Total livable area = total area — lake area - owned area

Watershed area, lake area in watershed and owned in watershed: match
union#

Total livable in watershed = total watershed — lake in watershed —
owned in wshed

% livable in watershed = livable in watershed / total livable

 

281

Population densities for 2000

This method is the same concept as the historical population densities, except that
everything must be recalculated because the census bocks are different.

1. Downloaded Tigerline census tract data ﬁ'orn DNR spatial data library
httpzllwwwmcgj.state.mi.u8/mgdl/?2_rction=thm
2. Determined which census tracts were included in each watershed
Collected total population information for each of these census tracts from the
American F actﬁnder on the US Census website:
hgpzﬂfactﬁndencensus.gov/servlet/CTGeoSearchByListServlet?ds name=DEC 2000
SF4 U& lang:——en&_ts=79955755602
3. Since census tracts had changed, the dasymetric mapping and all other parts of the
population density calculations have to be redone.
4. First tract00_miv2ashp was transformed into a polygon coverage using shapearc
tract00_miva tract00 polygon (this can be found on UND( machine)
5. Determine lake area within all of the township census areas
a. Copy or download lake shape ﬁles to the correct lake directory.
b. In arcview open tractOO (open entire coverage, not polygon) and lake.shp
c. Anaysis>Tabulate areas:
i. Row theme. lake. shp
ii. Row ﬁeld: Lake_type
iii. Column theme: polygon
iv. Column ﬁeld: tract_ lbl (these are the actual census tract numbers)
v. Save resulting table as tab delimited text (county lake)
6. Determine lake area within the watershed
a. Union watershed and tract00
i. In arcINFO: Type Arctools. Then command tools>
analysis>overlay> union, tract00 (..\tract00) and watershed
b. Open new union coverage in arcview, open the m not the region
0 Tabulate area:
i. Row theme: lake.shp
ii. Row ﬁeld: Lake_type
iii. Column theme: lakeunion
iv. Column ﬁeld: lakeunion#
v. Save resulting table as tab delimited text (county lakeshed).
d. For all polygons within the watershed (and ONLY within the watershed),
manually determine in Arcview which cadunion# number(s) are in each

 

census tract.
7. Determining area owned in tract00 if necessary. This is completed for year 1-4
lakes.
a. Download or copy state ownership ﬁles for each county, open in Arcview
b. Theme>Query: surf_cn >= 1 or fee_cnt >= 1, new set
c. Theme> convert to shapeﬁle
d. Analysis> tabulate area:
282

 

i. Row theme: lakeown.8hp
ii. Row ﬁeld: county
iii. Column theme: polygon
iv. Column ﬁeld: tract_ lbl
v. Save resulting table as tab delimited text (county own).

8. Determine area owned rn watershed
a. Open laketmion
b. Analysis> tabulate area:

i. Row theme: lakeown. shp
ii. Row ﬁeld: county
iii. Column theme: lakeunion
iv. Column ﬁeld. lakeunion#
v. Save resulting table as tab delimited text (county ownshed).

9. Determine area of each mi_pop4 that 18 in the watershed
a. Analysis> tabulate area:

i. Row theme: polygon
ii. Row ﬁeld: tract_lbl
iii. Column theme: watershed
iv. Column ﬁeld: watershed_id
v. Save resulting table as tab delimited text.

10. Determine the total area of the census tracts

a. Using the select tool, select the census tracts included in the watershed
b. Go to the table and move highlighted records to the top.
c. Export these records as tab delimited text (tract areas).

11. Excel calculations: areas calculated are the areas that are owned, and the areas
that are lake, and therefore the remaining area must be calculated.
a. FTP all .txt ﬁles to disk, open in Excel
b. Copy ﬁles into calculation ﬁle (See area calculations with ownership.xls),

C.

no

transpose data when copying so that the poly number is a column.

Paste tract area, lake area, owned area, and watershed area into the
appropriate columns for the proper census tracts. For lake areas, use
lake + island when there is island area (assume that there are no islands
big enough for people to be living on)

Fill in the appropriate lakeunion#(s) for each of the census tracts (this
relationship was determined manually in step 6d). Using these
numbers, paste the appropriate lakeshed, and ownshed data into the
columns.

Total livable area = total area — lake area — owned area

Watershed area, lake area in watershed and owned in watershed: match
union#

Total livable in watershed = total watershed — lake in watershed — owned
in wshed

% livable in watershed = livable in watershed / total livable

283

 

Area of land cover in total watershed
The area or percentage land cover in the watershed can be calculated two ways - with
or without the study lake. Whether or not the lake is included will signiﬁcantly
inﬂuence the percentage land cover, as the lake will cover very different proportions
of the watershed. If the lake is included, then all percentage land covers will be
signiﬁcantly inﬂuenced by the lake to watershed: area ratio. Because of this, all
calculations for the Yohn dissertation were completed using the watershed without the
lake. The instructions below will be for this method, but the percentages for the entire
watershed may be determined by using the entire watershed coverage instead of the

watershed no lake coverage.

Data are acquired at the most detailed level (Level 3 or 4) and then the coarser land
cover categories are calculated ﬁom these data (Level 1). Because the classiﬁcations
are different for the two time periods, the Level 1 data is the most appropriate for
comparisons, but the acquisition of Level 3 data allows for more detailed analyses if

desired.

1. Open necessary land cover ﬁle(s) and watershed without lake ﬁle (watershed_nl)
a. For IF MAP ﬁles, open lower or upper peninsula
b. For 1978 MIRIS ﬁles, open lands cover for all counties necessary for that
watershed.
2. IF the watershed is in more than one watershed, it is easier if the county land cover
ﬁles are merged before analysis
a. Use the merge command in geoprocessing wizard in ArcView.
3. Using analysis>tabulate area in ArcView, determine the Level 3 label area of land
cover.
a. For the 1978 landcover area, use 30 m for the resolution - this is the same
resolution of the DEM.
4. Import as tab-delimited text into Excel.

284

5. Using the codes listed below, determine the level 1 land cover for each time
period.
6. Sum the total area for each watershed, and calculate percentage Level 1 land cover
for each watershed.
a. The total area for each watershed should be very close to the area of the
watershed_nl coverage. If it is not similar, there is some error.
NOTES:

For the UP IFMAP landcover the classiﬁcation is numbers and not text
descriptions. See table below to match land cover codes with descriptions.
IFMAP classiﬁcation system places lowland forest as wetlands, while the MIRIS
classiﬁcation places lowland forest as forests. To make the two datasets more
consistent, lowland forests and upland forests in the IFMAP coverage were

combined to create a “forest” category.

285

 

Table l. MIRIS land cover data codes.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Level 3 Level 2 Level 1
Multi-Famin-Medium to High Rise Residean Urban
MuIti-Famin-Low Rise

Single Family, Duplex

Mobile Home Park

Commercial, Services, and

Institutional (Level 2) Commercial, Services, Institutional

Central Business District

ShoppiprLCenter, Mall

Neighborhood Business

Institutional

Industrial (Level 2) Industrial

Industrial Park

Transportation, Communication,

and Utilities (L2) Transportation, communications utilities

Air Transportation

Water Transportation

Road Transportation

Utilities, Waste Disposal

Open Pit Extractive

Wells

Outdoor Recreation Open Land and other

Cemeteries

Crepland, Rotation, and Cropland, Rotation, and Permanent

Permanent Pasture Pasture Agriculture
Orchards, Vineyards, and

Ornamental Orchards, Vineyards, and Ornamental

Conﬁned feeding Conﬁned feeding

Permanent Pasture Permanent Pasture

Other Agricultural Land Other Agricultural Land

Herbaceous Rangeland Herbaceous Rangeland Nonforested
Shrub Rangepland Shrub Rangeland

Pine or Oak Openipg (Savanna) Pine or Oak Opening (Savanna)

Northern Hardwood Deciduous Forested
Central Hardwood

Aspen, Birch

Lowland Hardwood

Pine Coniferous

Other Upland Conifer

Lowland Conifer

Christmas Tree Plantation

Streams and Waterways Streams and Waterways Water
Lakes Lakes

Reservoirs Reservoirs

 

286

 

 

 

Table 1 (con’t)

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Wooded Wetland Forested wetlands Wetlands
Shrub/Scrub Wetland
Aquatic Bed Wetland Nonforested wetlands
Emergent Wetland
Flats
Beaches and Riverbanks Beaches and Riverbanks Barren
Sand Other than Beaches Sand Other than Beaches
Table 2. 2001 IF MAP land cover data codes.
Value Level 4 or 3 Level 2 Level 1
0 Background
1 Low Intensity Urban Low Intensity Urban Urban
2 High Intensity Urban High intensity urban
3 Airports
4 Roads l Paved
5 Non-rpgetated Farmland Herbaceous agriculture Agriculture
6 Row Crops
7 Forage Crops I Non-tilled herbae
Non-herbaceous
9 Orchards / Vineyards I Nursery agriculture
10 Herbaceous Openland No level 2 Upland openland
Upland Shrub/ Low-density
12 trees
13 Parks / Golf Courses
14 Northern Hardwood Association Upland deciduous forest Upland forest
15 Oak Association
16 Aspen Association
17 Other Upland Deciduous
18 Mixed Upland Deciduous
19 Pines Upland coniferous forest
20 Other Upland Conifers
21 Mixed Upland Conifers
22 Upland Mixed Forest Upland mixed forest
23 Water Water Water
24 Lowland Deciduous Forest Lowland forest Wetland
25 Lowland Coniferous Forest
26 Lowland Mixed Forest
27 Floating Aquatic Nonforested wetlands
28 Lowland Shrub
29 Emerggnt Wetland
30 Mixed Non-Forest Wetland
Bare/Sparsely
31 Sand / Soil No level 2 vegetated
32 Exposed Rock
35 Other Bare / Sparsely Vpgetated

 

 

 

 

 

* Upland and lowland forests (italics) were reclassiﬁed into a “forest” group.
Lowland forests were NOT included in wetlands.

287

 

 

 

 

 

Flow path weighted land cover

Flow path weighted land cover is a measure of the amount of land cover within the

watershed, but is weighted by the inverse distance to the lake along the ﬂow path.

Data were calculated using the inverse distance, and the inverse distance squared.

Additionally, both data using the actual value calculated and the percentage (of values

for all land cover types) were calculated. The use of percentage compensates for the

size of the watershed.

The ﬂow path weighted value for each land cover was determined by:

>195”?pr

Clip ﬂow direction grid, calculate ﬂow length

Clip land cover grid

Assign all of one land use to 1, rest to 0

Calculate 1/ ﬂow length

Multiply l/ﬂow length by land cover # (0 or 1)

Sum

Do this for all desired land uses (level 1, excluding water and barren)

The detailed method follows.

2001 IFMAP data method

1.

Create a ﬂow length grid:
g. Clip the SRTM ﬂowdirection grid with watershed without lake ﬁle
Grid: gridclip cad_fdir cadshed_fdir cover cadshed_nl
i. Must do this from a common directory- it will not take ..\
h. Calculate ﬂowlength:
Grid: cad_ﬂow = ﬂowlength(cad8hed_fdir,#,down8tream)
Create a land cover grid for land cover of choice (IF MAP) (ag, for, urban, wet,
up)
i. Clip the IFMAP grid in a similar fashion to the ﬂowdirection grid
Grid: gridclip ifmap\lp cadillac\cad_'rfmap cover
cadillac\cadshed_nl
j. Open this in arcview and assign all cells with the appropriate landcover
a,“l” and the rest a “0”.
i. Analysis>map query
ii. For urban: cad_ifmap >= 1 and cad_ifmap <= 4 (see below for
correct code numbers)
iii. Convert this to a grid:
1. Theme>convert to grid, save as abbreviate of land cover
type, urb, ag, for, up, wet

288

 

 

3. Take the inverse and inverse squared of the ﬂowlength grid (80 cells closest to
the lake have the highest values)
k. Grid: ﬂi = l /cad_ﬂow
1. Grid: ﬂi2 = 1 / sqr(cad_ﬂow)
4. Multiply the inverted ﬂowlength grid and the selected landcover grid. Do this
for each land cover, and for ﬂow inverted and ﬂow inverted squared.
m. Grid. urb_ ﬂi= urb‘cad_ ﬂi
n. Grid: urb_ﬂ12=urb * ﬂi2
5. Sum the resultant grid for the ﬁnal ﬂow weighted land cover, and then show
the results. Total and tota12 can be any words that aren’t commands.
a. In GRID:
Grid: docell
:: total += urb_ﬂi
:: tota12 += urb_fli2
:: end
Running... 100%
Grid: show total
12.97288418311
Grid: show tota12
0.3155395542706
" The scalars (total and tota12) are not preserved in the Arclnfo memory —
it is critical to type the “show” commands right after the docell.
6. Numbers produced ﬁom show will have to be cut and paste into Excel (x-win,
copy to clipboard)

IFMAP codes
Urban: 1-4

Agp'culmturezi 6217, 9

Upland opgpland1:0,12, 13
F OREST (this will include lowland forest which they deﬁne as wetland but I deﬁne as

forest): 14-22 ,24-26
Water: 23
Wetland: 27-30
Barren: 3 1,32,35

Method for 1978 landcover.

1. Open 1978 landcover ﬁle that has been clipped with the watershed no lake.
2. Convert to grid:

Theme>convert to grid

Save this in the 781u ﬁle

Output grid extent: same as landcover shapeﬁle

Cell size: 30 m.

Field area for cells: label 1

Join feature attributes to grid? No.

rep .e-p 9'1»

289

 

 

 

3. Query this grid to create a new grid ﬁle for each landcover type
a. Analysis>map query
b. S_value = land cover of interest. . .
c. Convert the map query coverage to a grid ﬁle, save as the name of the
land cover in the 78lc folder
4. Using data manager in are view, copy the inverted ﬂow (ﬂi) and inverted ﬂow
squared (ﬂi2) to the 78 land cover folder.
5. Follow directions from above (directions 5-7)

Soil erodibility — k factor
K factor is standardized to a 9 % slope, therefore K should be greater of slope >9%

and less if slope <9%. Determine both k-factor and clope corrected k-factor (k-factor

* slope/9).

 

Rock ﬁagments reduce erodibility (rain hits rocks instead of erodible soil). K-factor
includes rock fragments, Kf-factor removes them. If there is a difference, the Kf-
factor will be greater than the k-factor. Often they are the same. Collect data on both

to ﬁnd difference. K factor is probably more appropriate than Kf-factor.

Map units are the area on the map. These map units are broken down into
components. The % of each component within the map unit is in the comp table under
comppct. Components are broken down into depth layers —- I am only interested in the
surface layer. So, each component has a sequence number (seqnum) that is listed in
both the layer and comp databases. Use the surface layer for each comp. Then calc a
weighted average of the k-factor for each map unit. Then determine the weighted
average k-value of the watershed from the relative areas of the map units in the

watershed.

290

 

To use slopes ﬁ'om the STATGO database:

1)

2)

3)

4)
5)

Calculate area of MUID within each watershed without lake.
a. The statsgo coverage was converted to grid to speed the area tabulation.
b. Tabulate areas: lakeshedv, lakeshedv#, statsgo_grd, MUID. Export as tab
delimited.
In excel: For all MU IDs of interest, go to layer, and for each sequence number get
the k-factor values for the top layer.
Then, using the sequence number to correlate, go to camp and ﬁnd the %comp of
each sequence number. Use this to do a weighted average of the k-factor values.
This will give the average k-factor for each MUID. Do the same for slope, using
the average of the min and max slope.
Multiply k-factor "‘ slope/9
The do a weighted average with all the MUIDs in the watershed, using the % of
the watershed that is covered by each MU II). EXCLUDE the area of water ﬁ'om
the total area within the watershed, since there will be no credibility for this.

When kfact or kffact were missing for some MU ID layers, that values was estimated

item the average of the other layers. These are noted in RED on the page of k-fact

calculations

y—a
O

H

Determining slope of the watershed

Clip the ﬁlled elevation ﬁle with the watershed_no lake vector ﬁle.

Calculate slope:
a. Grid: cad_slope = slope(shed_elevf,degree)

Then describe this in the grid module and use the result from the mean value.
a. Grid: describe cad_slope

Creating buﬂ’ers around lake:

. Open lake ﬁle - polygon or shp ﬁle

Units MUST be set for in ArcView to create buffers (must do this every time you
reopen ArcView)

a. Go to View> properties

b. Set map units and distance units as meters

. IF the lake ﬁle includes more than one polygon, use geoprocessing wizard to

dissolve it. Find a parameter (e.g., ID) that is common to all polygons and use that
to dissolve. If there is not a common parameter, edit the table and make the ID the
same for all polygons. There is no need to append any of the data
Go to Theme> create buffer

a. Choose The Features of a theme, and choose the lake ﬁle, next

291

 

 

we:

. Open the watershed_nl clipped iﬁnap ﬁle for that lake (MUST use this and not the

b. Choose At a speciﬁed distance (100 m, or 10000 m), and make sure
distance units are meters, next
c. Choose outside polygon, and choose to save it in a new theme (buffm for
100m temporary and buﬁkm for 10 km).
For bufﬁn: This will create a shape ﬁle that includes the lake and the extra 100m.
The goal is to get a shape ﬁle only 100 m wide.
a. Use Geoprocessing wizard (V iew>geoprocessing wizard) and union buffrnt
and lake ﬁles. Save as buffimion
b. Select the outer ring, and save as shape ﬁle bufﬁn_nl. This should be just
the outer ring.
For buftkm: do a similar thing if the entire watershed is not included in the 10 km
boundary — that is, create the buffer no lake shape ﬁle. This is necessary for the
soils data l"
* Most watersheds are <10 km, and using this buffer is probably unnecessary. -

 

To calculate land cover in buﬁer

total ifmap coverage in case the buffer goes outside of the watershed)
Open the buffer

 

. Analysis, tabulate area with lake_ifmap value, and buffm and ID.

a. It will be better to use buffm and buffkm rather than buffm_nl because the
lake is already rrrissing. Using buffm will make sure everything near shore
will be included, and hopefully reduce error.

To calculate k factor in buffer

First calculate areas of MUID in each of the buffers, using the no lake buffers.

a. Tabulate area: Statsgo_grd, S_value; bufﬁn_nl, ID
Put these areas into the already prepared sheet to calculate k values — this will be
the same as calculating it for the entire watershed, only with different areas.
This does create some error — if the 100m buffer is outside the watershed, it will
still be included. However, I think this error will be minor in most cases. If it is
not, then will have to gridclip the statsgo_grd with the lakeshed_nl ﬁle, and then
tabulate area.

 

292

To calculate average slope in buﬂ'er

1. To calculate the average slope in the buffer area, it is necessary to clip the
lake_slope ﬁle (slope calculated on lake_fdir clipped with watershed_nl) with the

buffer ﬁle.
a. Shape arc the buffm ﬁle: shapearc buffm buffm
b. Clean buﬂ‘m
c. Grid: gridclip lake_slope lake_s10pem cover buffm
d. Describe lake_s10pem
The average value is the average slope for that portion of the watershed

Plotting lat/long coordinates in ArcView in Michgan Georef projection
This method can be used not only to plot locations of data points, but can also be used
to associate data with those points. This method was used in plotting lake sampling
locations (and inputting metal inputs and watershed characteristics with these points),
Toxic Release Inventory, and sulfate deposition.
1. Make table in Excel with latitude, longitude (lat/long MUST be in decimal
degrees), and any other variables of interest. Make sure that all column headers

have no spaces or odd characters. There may be a header character limit. Save
table as delimited text.

2. Arcview:
Add table, DELIMITED TEXT
table.txt
Open a view
View>add event theme
X ﬁeld longitude
Y ﬁeld latitude

3. Theme> convert to shapeﬁleﬁlename
4. Got to Are and use shapearc shapeﬁle coverage
5. Deﬁne projection of coverage - tell Arc that this coverage is in geographic
projection.
Arc: proj ectdeﬁne cover coverage
Deﬁne Projection
Project: projection geographic
Project: units dd
Project: datum wgs84
Project: spheroid wgs84
Project: parameters

293

6. Now project into MI georef using the same info as for the SRTM data, except do
not do this in grid, do it in arc.
Arc: project cover coverage projected_coverage

t*t****¢$*tt*¥*$#itttttttt*ttttttttttittt**t*t¢***

* The INPUT projection has been deﬁned. *
##***t*#******¥$****tt‘tt*tttt**#$¢**t#**t*t*¢t#*t
Use OUTPUT to deﬁne the output projection and END

to ﬁnish.

Project: output

Project: projection oblique_mercator

Project: datum nar_c

Project: units meters

Project: spheroid grsl980

Project: parameters

Projectiontype<l |2>: 2

Scale factor at the projection's center [ 1.00000 ]: 0.99960
Longitude of the projection's center (DMS) [ 0 0 0.000 ]: -86 0 0.000
Latitude of the projection's center (DMS) [ 0 0 0.000 ]: 45 18 33.000

Azimuth at the projection's center [ 0.00000 ]: 337.25556
False casting (meters) [ 0.00000 ]: 2546731.49600
False northing (meters) [ 0.00000 ]: 43540098160
Project: end

Toxic Release Inventory
The EPA Toxic Release Inventory (FRI) can be used to determine if there are point
sources of metals within the watershed. Data were acquired from the EPA TRI
Explorer, plotted in ArcView, and each watershed was checked visually to see if there
were any TRI sites within it.
To acquire TRI data:
1. Go to TRI explorer: http://www.epa_.gov/trie)g)lorer/
2. Choose Facilities, Michigan, Metal and metal category compounds.

3. Some sites will not have Lat/long listed — for those sites within the counties of
interest click on the location to get more information, and take lat/long from this.

294

Sulfate deposition

- Data were collected from the National Atmospheric Deposition Program (N ADP
ht_tp:l/nadp.sws.uiuc.edu_/) for sites within Michigan and ﬂour sites in adjacent states
that were near the MI border. Data from all time periods were collected, but there

were only enough sites available for interpolation ﬁom 1990-2000 (and more recent).

Latitude / longitude and average sulfate deposition (kg/ha/y) for 1990-2000 were input
into ArcView (see Plotting lat/long coordinates in ArcView in Michgan Georef
projection). Used inverse distance weighting in ArcView (power of 2), and used

information tool to determine approximate value for each watershed.

295

 

APPENDIX C

INDUCTIVELY COUPLED MASS SPECTROMETER ANALYTICAL METHODS

296

APPENDIX C. INDUCTIVELY COUPLED MASS SPECTROMETER
ANALYTICAL METHODS

ICP-MS
Making standards and other preparatory lab work
Introduction

A typical suite of elements that are analyzed in water and digested sediments
(sample preparation details given in another document) includez, “’25 Mg, 27A1, 29 Si,
3‘1), 39K, “Ca, "Ti, ”V, ”Cr, “Fe, ”Mn, ”Co, “Ni, 63C“, “Zn, 75 As, ”Sr, ”Mo, 114C (1,
120Sn, 133Ba, me, and 238U. These include some of the common, important
environmental elements and the ﬁrst row transition metals. Metals may be added or
removed, depending on the project. However, the more elements that are analyzed,
the higher the detection limits will be; there is a compromise between number of
elements and time spent by the quadrupole in the settings for each isotope. Ca and Fe
may be analyzed on the AA, or diluted and run on the ICP-MS. Lithium and boron
will have very high detection limits due to contamination by samples analyzed using
laser ablation. Phosphorous has high background concentrations, and the lowest
concentration in the standard curve should not be less that 200 ppb.

Elements present in high concentrations (1003 ppm), like calcium, have severe
effects on the response of the instrument, causing matrix effects, especially in digested
sediment samples. The calcium may increase the background concentration for some
elements (e. g. 80Se, 88Sr), or suppress the signal for other elements. Because of this, if
Ca concentrations are above 10 ppm, a concentration similar to that of the samples

should be added to the standards to provide a similar matrix. High concentrations of

sodium will also cause signiﬁcant matrix effects. Because of potential matrix effects,

297

and to determine which elements might be of interest, a full mass scan should be
preformed on representative samples before a new experiment is carried out. Then
concentrations of Ca should be determined by using a higher dilution or using the AA.
These concentrations can be used to matrix match the standards. Standards have only
been matrix matched with calcium successfully, attempts to matrix match with iron
failed — it was better to run the samples with no matrix match than with iron.

In order to analyze for mercury, use a mercury analyzer and not the ICP-MS.

The standards made for the ICP-MS must be multi-elemental (if carrying out

multi-element analysis), however, not all the elements should cover the same

 

concentration ranges (e. g. arsenic concentrations may be in the low ppb, while iron
may be in the ppm range). Therefore elements are grouped according to their
concentration ranges, with each group representing a different standard curve. For

example:

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Table 1. Final concentrations of standards in pb.
STD STD STD STD STD STD STD
Blank 1 2 3 4 5 6 7
V, Cr, Co, Mo,
Cd, U, Sn, Se 0 0.03 0.05 0.07 0.1 0.2 0.5 1
Ti, Ni, Cu, As,
Pb, Sr, Sc, RD, 0 0.1 0.2 0.5 1 5 10 25
Ba
K, Na, Zn, Mn 0 10 20 30 40 60 80 100
P, Si 0 25 50 100 300 500 750 1000
STD STD STD
8 9 10
V, Cr, Co, Mo,
Cd, U, Sn, Se 5 1° 25
Ti, Ni, Cu, As,
Pb, Sr, Sc, Rb, 50 100 200
Ba
K, Na, Zn, Mn 150 200 500
P, Si 1500 2000 3000

 

 

 

 

 

 

298

 

 

 

Concentrations of ICP-MS standards (ppb to low ppm range) are much lower
than the purchased original standards (1,000 ppm). Because of this it is not possible to
dilute ﬁom the original standards to the ICP-MS standards. Therefore solutions are
made, and these solutions are used to make the standards. A solution is a lower
concentration solution of all the elements in one group (e.g. a 1 ppm solution of V, Cr,
Co, Mo, Cd, U, Sn and Se). This solution is then diluted to different extends to make
the different standards.

To obtain the greatest accuracy, the concentration range of the standards will
differ for each type of sample, and therefore will likely be different for each project.
The standards given in this method are merely examples of standards for the Inland
Lakes project.

To choose concentration ranges for the standards it is best to ﬁrst do a
preliminary analysis of some samples. By looking at the minimum and maximum
concentration for each of the elements, elements should be grouped into several ranges
of concentration (e.g. one group may be 0.2 to 10 ppb, while another may be 10 to 500
ppb —- see above for example groupings). Things to consider: (1) The smaller the
range of concentrations the more accurate the curve is (a curve from 0.2-10 ppb will
be much better for analyzing a 2 ppb solution than a cm've ﬁom 0.2-100 ppb) -
however - the more concentration groups of elements, the more work it is to make the
standards. (2) 0.03 ppb is reaching near detection limits and/or the concentrations in
the blank — this varies for each element and the clean methods used — but as a general
guide, low concentrations should not go below 0.03 ppb unless discussed with the lab

manager: (3) The upper limit of the mass spectrometer is reached by saturating the

299

 

detector, which occurs as low as 1000 ppb for some elements, and as high as >15 ppm
in other element. Check notes on speciﬁc elements for a better idea, but keep
standards below the saturation levels of the elements.

Five standards plus a blank is the minimum number of standards that should be
used. Currently, 7 to 10 standards are typical. More points allow for greater
ﬂexibility in the range of concentration in the metal without loosing accuracy, and
therefore it is often worth the extra work in preparing the standards. There should be
i more points near the low concentration end of the curve than the high (see example
above). For more examples of standard curves, see the lab manager or the Inland
Lakes lab book.

From each of these concentration groupings at least one solution is made. This
solution contains all the elements in the group and is needed because the purchased
standards are too high to dilute directly to the ppb range. The dilutions are typically
done in 100 mL volumetric ﬂasks, and the volume that can be accurately pipetted
limits the size of the dilution. The solution should be transfer to a Nalgene bottle soon
after it is prepare to avoid contamination from the volumetric ﬂask. As a general rule,
volumes ~<50 uL begin to be less accurate, and we cannot pipette <10 uL. Also, the
larger pipettors (> 1 mL) often leave signiﬁcant volume in the pipette tip and should
not be used to make lCP-MS standards. Because of these constraints, it is sometime
necessary to use two solutions for one group of elements (e.g. 1 ppm and 100 ppb),
either because the low concentration is too low to reach in two dilutions (e.g. 0.05

ppb), or the range of concentrations is too great (e.g. 1-500 ppb).

300

Nitric acid is added to each standard to both prevent precipitation and to create
a similar matrix to the sample. Nitric acid will enhance the ionization of different
elements, therefore the concentration of nitric acid in the standards should be as close
as possible to the concentration in the sample. All solutions should be prepared in 2%
HNO3.

Samples in the ICP-MS vial include 1 mL of internal standard and 1 mL of
sample. The internal standard is used to account for drift in the instrument, and for
different instrument response to different matrixes. In other words, if one sample was
run multiple times, the area (actual area beneath the peak) may change over time. In
order to correct for this, we run an element as an internal standard at a constant
concentration in every sample. The theory is that the element acting as the internal
standard will behave in the instrument in exactly the same manner as the elements of
interest, and can be used to correct for any drift or effects of matrix. The response is
calculated for each element for each sample as the area of the element for that sample
divided by the area of the internal standard for the sample. The response (rather than
the area) is used to calculate the standard curves and concentrations. Unfortunately,
elements behave differently in the instrument, and it is impossible to have internal
standards that behave exactly the same as the elements of interest, making it desirable
to minimize the drift and matrix effects in the instrument. Traditionally, indium (1 ls1n)
and bismuth (209m) have been used as internal standards (In for low masses, Bi for
208Pb and 2380). Currently, beryllium (isotope 9), scandium (45), germanium (74), and
yttrium (89) are also being used. For further discussion on internal standards, see

below.

301

 

 

Digested sediments are run at a dilution of 1 :10 (not including the internal
standard dilution) so that most elements do not saturate the detector (e. g. 1 mL internal
standard + 900 uL 2% nitric acid + 100 uL sample). Additionally, the matrix effects
may be too signiﬁcant for concentrated digested sediment samples. Water samples
can often be run without dilution.

Three rinses of 2% nitric acid are rim before and after each set of standards.
For some elements, the high standard may contaminate the next sample if the sample

has a low concentration. Rinses are used to eliminate this possibility. It is a good idea

 

to check for samples that have high concentrations of some elements to make sure that

the next sample is not contaminated.

 

The ICP-MS should be conditioned to the type of sample before the run (e.g.
digested sediments have a very different matrix than the solutions used to tune the
machine). 1-4 vials should be prepared with 3 times the normal volume of a sample
(including internal standard) to use as the conditioning solution (e.g., 3 mL internal
std, 3 mL sample). If the instrument has just been cleaned and the samples have a
heavy matrix (e.g., digested sediments), it is advisable to use more than one

conditioning solution.

 

Concentrations are determined by the standards run before and after each
group of ~ 30 samples. Therefore the run should be organized as follows: standards, 2
rinses, SRM 1643d, 2 rinse, ~30 samples, 3 rinses, another set of standards, 3 rinses,
~30 samples, 3 rinses, standards. Typically runs have up to ~60 samples, though

longer runs can be done.

302

 

Procedure

All ICP-MS work should be scheduled with the lab manager (currently Lina Patino
patinol@msu.edu, 432-3100 ext 138).

This is an example method from digested sediments —- actual concentrations should be
varied dependant on the type of sample.

‘Be sure to check pipettors before use!!

*Use only Ultrapm'e nitric acid and Nanopure water ﬁom the ICP-MS lab

*All ICP-MS vials should be hot acid washed

* All solutions are made in 2% nitric acid.

1. Prepare solutions for standards:
a. Add 2% nitric acid and appropriate volumes of 1,000 ppm standards to 100 mL
volumetric ﬂasks. For example:

Table 2. Dilutions to create solutions for ICP-MS standards
Conc ppb volume to 100 mL

 

 

 

 

 

Solution A 500 50 uL of 1000 ppm std
Solution A2 10 2 mL of A
Solution B 10 ppm 1 mL of 1000 ppm std
Solution 32 100 1 mL of B

 

Solution C 10 ppm 1 mL of 1000 ppm std
Solution D 100 ppm 10 mL of 1000 ppm std

 

 

 

 

 

 

2. Prepare standards from solutions:

a. Add 2% nitric acid, appropriate matrix match, and appropriate amounts of each
solution to the standards, and ﬁll to the line with the 2% HNO3. A blank should be
prepared with 2% I-INO3 and matrix match. Matrix match currently only includes Ca,
but may be needed with high concentrations of Na. Calcium should be added to the
standards in concentrations similar to what is found in the samples.

303

 

 

Table 3. Volume of solution added in uL to create ICP-MS standards.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Conc STD STD STD STD STD STD STD
ppb 1 2 3 4 5 6 7
Solution A 500 100 200
Solution A2 10 300 500 700 1000 2000
Solution B 10000 100 250
Solution 82 100 100 200 500 1000 5000
Solution C 10000 100 200 300 400 600 800 1000
Solution D 100000 25 50 100 300 500 750 1000
STD STD
STD 8 9 10
Solution A 1000 2000 5000
Solution A2
Solution B 500 1000 2000
Solution 82
Solution C 1500 2000 5000
Solution D 1500 2000 3000
Table 4. Final concentrations of ICP-MS standards.
STD STD STD STD STD STD STD
Blank 1 2 3 4 5 6 7
V, Cr, Co, Mo,
Cd, U, Sn, Se 0 0.03 0.05 0.07 0.1 0.2 0.5 1
Ti, Ni, Cu, As,
Pb, Sr, Sc, Rh, 0 0.1 0.2 0.5 1 5 10 25
Ba
K, Na, Zn, Mn 0 10 20 30 40 60 80 100
P, Si O 25 50 100 300 500 750 1000
STD STD STD
8 9 10
V, Cr, Co, Mo,
co, u, sh, Se 5 1° 25
Ti, Ni, Cu, As,
Pb, Sr, Sc, Rb, 50 100 200
Ba
K, Na, Zn, Mn 150 200 500
P, Si 1500 2000 3000

 

 

 

 

 

 

 

 

 

3. Standards may be stored in acid washed Nalgene bottles, and should be
reﬁigerated to prevent evaporation. Do NOT store ANY solution in volumetric
ﬂasks or any glass container.

4. Prepare internal standard (20 ppb In, Bi, Sc, Y; 100 ppb Ge; 1000 ppb Be in 2%
nitric acid): (there is a 1 L bottle of this in the hood. Make more when necessary.
ALWAYS use the same batch for a run, internal standard concentrations should be
exactly the same in all solutions for a run.)

a. Dilute 1000 uL of 1,000 ppm standard In, Bi, Sc, Y to 100 mL in a volumetric
ﬂask with 2% HNO3 to make a 10 ppm solution.

304

 

inﬁ—_ _ — m, ‘ ‘_,‘

b. In a 1 L volumetric add:
i. 28 mL Optima HNO3
ii. 2 mL of 10 ppm In, Bi, Y, Sc solution
iii. 100 uL of 1,000 ppm Ge
iv. 1000uLof1,000pmee

All ICP-MS vials should be taped with less than one full rotation at the very top of

the vial (if there is excess tape it is harder to get the vials into the auto sampler).

Pipette samples:

a. For standards: pipette 1 mL of internal standard and 1 mL of standard

b. For samples: pipette 1 mL of internal standard and a total of 1 mL of straight
or diluted sample (e. g. 100 uL sample + 900 uL water for a 1:10 dilution).

c. For rinses: pipette 2 mL of 2% HNO3.

d. For SRMs: pipette 1 mL internal standard and 1 mL of aqueous SRM(1643d). r
Liquid SRMs should be run even if digested SRMs are run. The SRMs are
used for quality control. The digested SRMs do not represent total digestions
and it is harder to compared obtained values with known concentrations.

Prepare a conditioning solution:

a. To an ICP-MS vial add 3 mL of internal standard and 3 mL of appropriately k

 

diluted sample (if samples are not diluted, add 3 mL of sample: if a 1:10
dilution is used, add 30 p1. of sample and 2700 uL of water)
Organize samples in the test tube holder:
a. one set of blank and standards
b. 2 rinses
SRM
2 rinses
~ 30 sample
3 rinses
blank and standards” ..etc
Conditioning solution may be set aside on the rack
List down a column in Excel the labels of all the samples (including standards) 1n
order. This rs cut and pasted into the ICP-MS software to provide the labels on the
data that is returned. This ﬁle should be e-mailed to the lab manager or given to
her on a Zip disk. Also send or give the lab manager a chart of the standards used.

we was» 9.0

10. When the data are returned, there may be columns for area, response,

11.

concentration and ﬂags. The area is the actual counts by the mass spec, the

response is the area normalized to the internal standard, and the concentration is in

ppb. The ﬂags apply only to the standards. Each standard curve is examined, and

if any points are anomalous, they are not considered in the calculation of the

concentrations. Standards that are not used are ﬂagged.

Check for samples that saturated the detector (therefore concentrations are not

valid):

a. Select the entire “area” column in Excel and format for scientiﬁc notation.

b. Any sample with an area x108 has satruated the detector and the concentration
is not valid.

305

12. Graph the internal standard areas (indium area as y, arbitrary sample number as x).
Internal standard should remain relatively constant throughout the run. If it does
not, notify the lab manager. Rinses will have much lower concentrations because
they have no internal standard, these samples can be ignored.

13. Carefully look through the data for any unusual points. Rerun ALL anomalous
samples — especially Cu, Ni, Al, Fe, these elements are in the lab and the
instrument.

ICP notes

This section includes notes and observations from analysis of digested
sediment solutions. T

This instrument seems to have signiﬁcant matrix issues, and line splitting is
frequently a problem, especially for the low masses. In order to combat this, I have
tried adding addition internal standards, matrix matching, and conditioning for longer.
Thoughts on these attempts are below:

 

Internal standards:

The current suite of internal standards that we run is:
20 ppb “’rn, 209m, ”Y, “Se
100 ppb 74Ge
1000 ppb 9Be

I have found that frequently, especially for majors and the low masses, that
other internal standards than In and Bi are necessary to correct for line 5 litting.
However, the success is not complete. Some elements, especially 39K, 3 P and 29Si,
are not satisfactorily ﬁxed by any of the internal standards. Unfortunately, it seems
that very little is consistent, and each element for each run needs to be evaluated to see
which internal standard is best.

My method involves:

- Running an SRM, and if possible, the same sample at the beginning and end of the
run.

- I quantify ﬁrst with In and Bi (Bi seems to work really well, so I don’t mess with Pb
and U), and remove points like normal. Any lines that are split so that all points are
<15% deviation I do not try to improve.

306

 

- Next I look at the other internal standards and see the shift in that internal standard
over the run. First I check to see if it looks like that element is in the samples (much
higher areas in the samples than the standards). Next I try to choose the next best
internal standard to try. For example, if I have a lot of line splitting, and Ge looks
similar to In, I would not try Ge next because it probably won’t improve things. I
would try an internal standard that had changed more over the run than In.

- Then I quantify with what I guess is the next best internal standard. IF the standard
line appears to be better or similar to In, then I remove necessary points etc., to create
a good line. IF the line appears worse that In (more splitting), I just ignore it and
move on.

- I continue this process, quantifying with each of the internal standards, and only
ﬁxing up those lines that appear to be less split that with any of the other internal
standards (this is just to save time...)

- After all the data are in Excel, I look at each element and determine which internal
standard is the best to use. In general, I try to use as few different internal standards as
possible (i.e., if 89Y is just slightly better than llsIn for one element only, I would
chose consistency and use mm). 115In and 74Ge are the most consistently useful, but
sometimes other elements are necessary for the low masses. See discussion below on
the speciﬁc elements. The way I judge which element is “best” is by looking at (1) the
SRM - if the values of the elements are way off, then I become concerned.
Unfortunately, the SRM has a different matrix than the samples. .. Note that the SRM
1643d has Bi and Be in it, so can’t be used to evaluate these. (2) Line splitting and %
error in standards. I try to reduce the error in the standards and have as low a % as
possible. (3) Drift - if I ran a sample at the beginning and end of the run, I look to see
how much the concentration changed. This can be pretty signiﬁcant sometimes. (4)
Consistency among internal standards. Occasionally, one internal standard will give
really different sample concentrations, maybe due to the presence of the internal
standard in the standard (e.g., Sc and Y in UP lakes), or due to I have no idea. But, if
one internal standard gives really different results, I don’t trust it.

- Typically, I take the columns of sample name, % deviation, ﬂags and concentrations
and put results from all of the internal standards on one sheet and compare them. If
“sln looked good originally (no line splitting) I won’t bother with any of this. Or, if
an internal standard didn’t seem to improve any of the line splitting I will not ever
bring it into Excel.

9Be: Seems to be more affected by matrix than 1151n. Occasionally proves to be
useful, but doesn’t consistently ﬁx the low masses.

4sSc: Has been useful in getting the standard lines to not split, but we seen to have Se
in our samples. Areas will suddenly jump to 200% with samples — not a good sign. . .

307

 

 

 

 

 

 

“Ge: Often provides an improvement over “sln for the lower masses, but not great for
the lowest masses (e.g., Al, Mg, etc).

89Y: Seems to be more affected by matrix than llsIn. This sometimes can be
advantageous in bringing the split lines together. May be Y in Upper Peninsula Lakes.

llSIn: I still trust this one the most, but it really isn’t working with the low masses
these days. Works great for 114Cd. . .

209Bi: Can get very high % deviations during the run, but seems to do a good job w/ Pb
and U (though still some line splitting with U). I have not tried to replace Bi, and
always use it for these masses (above method does not apply). Could try thallium.

Conditioning

Recently, due to the large amount of line splitting, I have increased the amount
of conditioning that we do. I have conditioned with up to 6 x 6mL of conditioning
solution, but it is hard to tell how much it helps, since we still have line splitting. The
instrument clearly runs better the second day of a similar matrix, so it seems that hours
of conditioning are what the instrument really needs. Unfortunately, the tune tends to
drift as the instrument cleans itself, and I’m not sure that it is worth running the tune
until the drift stops, as this may just make the actual run drift more as the instrument
gets dirty again. I still recommend at least 4 conditioning solutions (6mL ea) when the
instrument is clean. This could be tested... And I don’t rinse for a long time between
the conditioning and actual run.

Matrix matching

Matrix matching with Ca still seems necessary. If you want data for Sr, it is
necessary to run a blank without Ca. Do NOT include this blank in the line, run it as
an analyte, we will use the data later in Excel to calculate corrected Sr cone.

Ca is the only element I so far have had success matrix matching with. I tried
matrix matching Witch Lake with Fe (60 ppm), but that lead to poor results in
elements such as 114Cd. Samples I reran at the beginning and end of the core had
drifted considerably. I reran the lake with no matrix match, and it ran ﬁne, so Fe
should not be used to matrix match. Na might be necessary to match with, but I’ve
never had enough Na to ﬁnd out.

It seems that the instrument is not intensely sensitive to exact amount of Ca,
just that there is some. I have run lakes with a large variation of Ca concentrations
and just used a value in the middle and gotten good results. I just look at the entire
core and visually estimate an average Ca value, and pick a nice round number. I
recommend matrix matching >10 ppm Ca, but this is a guess, not experimentally

proven.

308

 

 

 

Other

Overall, the amount of line splitting that has been occurring recently has
resulted in a lot of lost data. My last thought to try is to let background go >400 so
that baking will not be necessary so often. There seems to be the worst splitting after
baking or after cones / torch has been cleaned. This should perhaps be experimented
with. Additionally, running several days in a row will probably help things, especially
if done without baking.

ICP-MS element notes

7Li: Interference from laser, high backgron
llB: interference from laser, high background

24Mg: Saturates 5000 ppb, not at 2000
25 Mg: Corresponds well with 24Mg, does not saturate at 5000 ppb

27Al: Saturates at 5000 ppb, not 2000. Good curve to 15 ppb (6-13-02 LS)
29Si: Tend to get split lines, ok after ~200 ppb. Try 288i.

31P: Tend to get split lines, ok > 200 ppb. Have run 30 ppm without saturating on old
instrument. High ionization energy.

39K: Saturates near 1 ppm. Splits easily.

44Ca: Does not saturate at 15 ppm, use instead of 40Ca

458c: Low end poor, only good > 10 ppb

“Ti: or. > 10 ppb, but does not run well at all at low end. Much line splitting.
5 IV: good to 0.05 ppb

52Cr: good

S4Fe: does not saturate at 5000 ppb. Low end ok, ~0.1 ppb.

5 5 Mn: Saturates detector near 1 ppm

59Co: ﬁne, but does not always am well at low end. Splitting issues.

6"Ni: good >o.2 ppb. Ni in cones, watch for odd high pts.

309

 

63 Cu: watch blanks and spikes, Cu in instrument
66an runs ﬁne, watch for contamination from latex gloves.

70Ge: consistently non-linear on old instrument
7“Ge: ﬁne, but cone diff from 7°Ge

75As: good
78Se: OK, also try 80Se (interference by 40Ar, watch background)

85Rb: great

88Sr: ﬁne curve, but Ca matrix match also adds Sr. Need to correct for this or Sr cone
will be artiﬁcially low.

95Mo: ﬁne

mCd: ﬁne

”’In: internal std
120Sn: very good lines

138Ba: very good lines
208sz good

209Bi: internal std
238U: great
For more ICP info:

Montaser, Akbar (ed.) (1998) Inductively coupled plasma mass spectrometry. Wiley-
VCH, Inc. New York. 964pp.

310

 

 

 

BIBLIOGRAPHY

311

 

BIBLIOGRAPHY

Alfaro-De la Torre, MC. and Tessier, A., 2002. Cadmium deposition and mobility in
the sediments of an acidic oligotrophic lake. Geochim. Cosmochim. Acta 66,
3549-3562.

Appleby, P.G. and Oldﬁeld, F., 1978. The calculation of lead-210 dates assuming a
constant rate of supply of unsupported 2"’Pb to the sediment. Catena 5, 1-8.

Auer, M.T., Johnson, N.A., Penn, MR. and Effler, S.W., 1996. Pollutant sources,
depositional environment, and the surﬁcial sediments of Onondaga Lake, New
York. J. Environ. Qual. 25, 46-55.

Balogh, S.J., Engstrom, D.R., Alemendinger, J .E., Meyer, ML. and Johnson, K.D.,
1999. History of mercury loading in the Upper Mississippi River reconstructed
ﬁ'om sediments of Lake Pepin. Environ. Sci. Technol. 33, 3297-3302.

Basnyat, P., Teeter, L.D., Flynn, K.M. and Lockaby, B.G., 1999. Relationships
between landscape characteristics and nonpoint source pollution inputs to
coastal estuaries. Environmental Management 23, 539-549.

Benoy, GA. and Kalaff, J ., 1999. Sediment accumulation and Pb burdens in
submerged macrophyte beds. Limnology and Oceanography 44, 1081-1090.

Berner, E.K. and Berner, RA., 1996. Global environment : water, air, and
geochemical cycles. Prentice Hall, Upper Saddle River, N.J., xiv, 376 pp.

Birch, L., Hanselmann, K.W. and Bachofen, R., 1996. Heavy metal conservation in
Lake Cadagno sediments: historical records of anthropogenic emissions in a
meromictic alpine lake. Water Resources 30, 679-687.

Blais, J .M. and Kalff, J ., 1995. The inﬂuence of lake morphometry on sediment
focusing. Limnology and Oceanography 40, 582-588.

Bollhofer, A. and Rosman, K.J.R., 2001. Isotopic source signatures for atmospheric
lead: The Northern Hemisphere. Geochim. Cosmochim. Acta 65, 1727-1740.

Boyle, J.F., Rose, N.L., Bennion, H., Yang, H. and Appleby, P.G., 1999.
Environmental impacts in the J ianghan Plain: evidence from lake sediments.
Water, Air, Soil Pollut. 112, 21-40.

Brown, E.T., Callonnec, LL. and German, CR, 2000. Geochemical cycling of redox-

sensitive metals in sediments ﬁom Lake Malawi: A diagnostic paleotracer for
episodic changes in mixing depth. Geochim. Cosmochim. Acta 64, 3515-3523.

312

 

 

Bruland, K.W., Bertine, K., Koide, M. and Goldberg, ED, 1974. History of metal
pollution in Southern California costal zone. Environ. Sci. Technol. 8, 425-
432.

Callender, E. and Rice, K.C., 2000. The urban environmental gradient: Anthropogenic
inﬂuences on the spatial and temporal distributions of lead and zinc in the
environment. Environ. Sci. Technol. 24, 232-23 8.

Callender, E. and vanMetre, P.C., 1997. Reservoir sediment cores show US lead
declines. Environ. Sci. Technol. 31, A424-A428.

Catallo, W.J., Schlenker, M., Gambrell, R.P. and Shane, B.S., 1995. Toxic chemicals
and trace metals ﬁom urban and rural Louisiana lakes: recent historical proﬁles
and toxicological signiﬁcance. Environ. Sci. Technol. 29, 1439-1446.

Cole, K.L., Engstrom, D.R., Futyma, R.P. and Stottlemyer, R., 1990. Past
Atmospheric Deposition of Metals in Northern Indiana Measured in a Peat
Core ﬁ'om Cowles Bog. Environmental Science & Technology 24, 543-549.

Comeleo, R.L., Paul, J.F., August, P.V., Copeland, J ., Baker, C., Hale, SS. and
Latimer, R.W., 1996. Relationships between watershed stressors and sediment
contamination in Chesapeake Bay estuaries. Landscape Ecology 11, 307-319.

Cooper, DC. and Morse, J .W., 1998. Biogeochemical controls on trace metal cycling
in anoxic marine sediments. Environ. Sci. Technol. 32, 327-330.

Crusius, J. and Anderson, R.P., 1995. Sediment focusing in six small lakes inferred
from radionuclide proﬁles. Journal of Paleolimnology 13, 143-155.

Davis, B.M., 1976. Erosional rates and land use history in southern Michigan.
Environmental Conservation 3, 139-148.

Davis, J .C., 1986. Statistics and Data Analysis in Geology. John Wiley & Sons, New
York.

Davis, MB. and Ford, MS, 1982. Sediment focusing in Mirror Lake, New
Hampshire. Limnology and Oceanography 27, 137-150.

Dearing, J.A., Hakansson, H., Leidberg-Jonsson, B., Persson, A., Skansjo, S.,
Widholm, D. and El-Daoushy, F ., 1987. Lake sediments used to quantify the
erosional response to land use change in southern Sweden. Oikos 50, 60-78.

Detenbeck, N.E., Johnston, CA. and Niemi, G.J., 1993. Wetland Effects on Lake
Water-Quality in the Minneapolis St-Paul Metropolitan-Area. Landscape
Ecology 8, 39-61.

313

Dillion, PJ. and Evans, RD., 1982. Whole-lake lead burdens in sediments of southern
Ontario, Canada. Hydrobiologia 91, 121-130.

Door, J .A. and Eschman, DR, 1970. Geology of Michigan. University of Michigan
Press, Ann Arbor, 476 pp.

Edgington, D.N. and Robbins, J .A., 1976. Records of lead deposition in Lake
Michigan sediments since 1800. Environ. Sci. Technol. 10, 266-274.

Eisenreich, S.J., 1980. Atmospheric input of trace metals to Lake Michigan. Water,
Air, Soil Pollut. 13, 287-301.

Eisenreich, S.J., Capel, P.D., Robbins, J .A. and Bourbonniere, R, 1998. Accumulation
and diagenesis of chlorinated hydrocarbons in lacustrine sediments. Environ.
Sci. Technol. 23, 1116-1126.

Eisenreich, S.J., Metzer, N.A. and Urban, N.A., 1986. Response of atmospheric lead
to decreased use of lead in gasoline. Environ. Sci. Technol. 20, 171-174.

Eisenreich, SJ. and Strachan, W.M.J., 1992. Estimating atmospheric deposition of
toxic substances in the Great Lakes - an update, Canada Centre for Inland
Waters, Burlington, Ontario, January 3l-Febuary 2, 1992.

Encyclopedia_Britannica_Premium_Service, Michigan, 2003.
<http://www.britannica.com/needmore>

Evans, RD. and Dillon, P.J., 1982. Historical changes in anthropogenic lead fallout in
southern Ontario, Canada. Hydrobiologia 91 , 131-137.

Fritz, S.C., Kingston, J .C. and Engstrom, D.R., 1993. Quantitative trophic
reconstruction from sedimentary diatom assemblages: a cautionary tale.
Freshwater Biology, 1-23.

Garrels, RM. and Thompson, M.E., 1962. A chemical model for sea water at 25C and
one atmosphere total pressure. American Joru'nal of Science 260, 57-66.

Gatz, DR, 1975. Pollutant aerosol deposition into southern Lake Michigan. Water Air
Soil Poll 5, 239-251.

Gatz, D.F., Bowersox, V.C. and Su, J ., 1989. Lead and cadmium loadings to the Great
Lakes ﬁ'om precipitation. J. Gt. Lakes Res. 15, 246-264.

314

 

Gergel, S.E., Turner, M.G., Miller, J .R., Melack, J .M. and Stanley, EH, 2002.
Landscape indicators of human impacts to riverine systems. Aquatic Sciences
64, 1 18-128.

Golden, K.A., Wong, C.S., Jeremiason, J .D., Eisenreich, S.J., Sanders, M.G.,
Hallgren, J ., Swackhammer, D.L., Engstrom, DR. and Long, D.T., 1993.
Accumulation and preliminary inventory of organochlorines in Great Lakes
sediments. Wat. Sci. Technol. 28, 19-31.

Goodarzi, F ., Sanei, H. and Duncan, W.F., 2001. Monitoring the distribution and
deposition of trace elements associated with a zinc-lead smelter in the Trail
area, British Columbia, Canada. J. Environ. Monit. 3, 515-525.

Gottgens, J.F., Rood, B.E., Delﬁno, J .J. and Simmers, B., 1999. Uncertainty in
paleoecological studies of mercury in sediment cores. Water, Air, Soil Pollut.
110, 313-333.

Gotz, R. and Lauer, R., 2003. Analysis of sources of dioxin contamination in
sediments and soils using multivariate statistical methods and neural networks.
Environmental Science & Technology 37, 5559-5565.

Graney, J .R., Halliday, A.N., Keeler, G.J., Nriagu, J .O., Robbins, J .A. and Norton,
S.A., 1995. Isotopic record of lead pollution in lake sediments from the
northeastern United States. Geochim. Cosmochim. Acta 59, 1715-1728.

Griffith, DA. and Amrhein, CG, 1997. Multivariate statistical analysis for
geographers. Prentice Hall, Upper Saddle River, N.J., xix, 345 pp.

Griffith, J .A., 2002. Geographic techniques and recent applications of remote sensing
to landscape-water quality studies. Water Air Soil Poll 138, 181-197.

Hair, J .F ., 1998. Multivariate data analysis. Prentice Hall, Upper Saddle River, N.J.,
730, 12 pp.

Harrington, J .M., Laforce, M.J., Rember, W.C., Fendorf, SE. and Rosenzweig, R.P.,
1998. Phase associations and mobilization of iron and trace elements in Coeur
d'Alene Lake, Idaho. Environ. Sci. Technol., 650-656.

Harrison, RR. and Winchest, J .W., 1971. Area-Wide Distribution of Lead, Copper,
and Cadmium in Air Particulates from Chicago and Northwest Indiana. Atrnos.
Environ. 5, 863-&.

Hermanson, M.H., 1998. Anthropogenic mercury deposition to arctic lake sediments.
Water, Air, Soil Pollut. 101.

315

Hewitt, AD. and Reynolds, C.M., 1990. Dissolution of metals ﬁom soils and
sediments with a microwave-nitric acid digestion technique. Atomic
Spectroscopy 11, 1425-1436.

Heyvaert, A.C., Reuter, J .E., Slotton, D.G. and Goldman, CR, 2000.
Paleolimnological reconstruction of historical atmospheric lead and mercury
deposition at Lake Tahoe, California-Nevada. Environ. Sci. Technol. 34, 3588-
3597.

Hoff, R.M., Strachan, W.M.J., Sweet, C.W., Chan, C.H., Shackleton, M., Bidleman,
T.F., Brice, K.A., Burniston, D.A., Cussion, S., Gatz, D.F., Harlin, K. and
Schroeder, W.H., 1996. Atmospheric deposition of toxic chemicals to the
Great Lakes: A review of data through 1994. Atrnos. Environ. 30, 3505-3527.

Hunsaker, GT. and Levine, D.A., 1995. Hierarchical Approaches to the Study of
Water-Quality in Rivers. Bioscience 45, 193-203.

Hurley, P.J., Benoit, J .M., Babiarz, C.L., Shafer, M.M., Andren, A.W., Sullivan, J .R.,
Hammond, R. and Webb, D.A., 1995. Inﬂuences of watershed characteristics
on mercury levels in Wisconsin rivers. Environ. Sci. Technol. 29, 1867-1875.

Iskander, I.K. and Keeney, D.R., 1974. Concentration of heavy metals in sediment
cores from selected Wisconsin Lakes. Environ. Sci. Technol. 8, 165-170.

Jaquet, J .-M., Davaud, E., Rapin, F. and Vemet, J.-P., 1982. Basic concepts and
associated statistical methodology in the geochemical study of lake sediments.
Hydrobiologia 91.

Johnson, L.B., Richards, C., Host, GE. and Arthur, J.W., 1997. Landscape inﬂuences
on water chemistry in Midwestern stream ecosystems. Freshwater Biology 37,
193-205.

Johnson, MG. and Nicholls, K.H., 1988. Temporal and spatial trends in metal loads to
sediments of Lake Simcoe, Ontario. Water, Air and Soil Pollution 39, 337-354.

Jones, K.B., Neale, A.C., Nash, M.S., Van Remortel, R.D., Wickham, J .D., Riitters,
K.H. and O'Neill, R.V., 2001. Predicting nutrient and sediment loadings to
streams from landscape metrics: A multiple watershed study from the United
States Mid-Atlantic Region. Landscape Ecology 16, 301-312.

Joreskog, K.G., Klovan, J .E. and Reyment, R.A., 1976. Geological factor analysis.
Methods in Geomathmatics. Elsevier Scientiﬁc Publishing Company, New

York, 178 pp.

316

Kada, J. and Heit, M., 1992. The inventories of anthropogenic Pb, Zn, As, Cd and the
radionuclides 137Cs and excess me in lake sediments of the Adirondack
region, USA. Hydrobiologia 246, 231-241.

Kang, W., Trefry, J .H., Nelsen, T.A. and Wanless, HR, 2000. Direct atmospheric
inputs versus runoff ﬂuxes of mercury to the lower Everglades and Florida
Bay. Environ. Sci. Technol. 34, 4058-4063.

Kemp, A.L.W. and Thomas, R.L., 1976. Cultural impacts on the geochemistry of the
sediments of Lakes Ontario, Erie and Huron. Water, Air and Soil Pollution 5,
469-490.

Kerfoot, W.C., Lauster, G. and Robbins, J .A., 1994. Paleolimnological study of
copper mining around Lake Superior: Artiﬁcial varves ﬁ'om Protage Lake
provide a high resolution record. Limnology and Oceanography 39, 649-669.

Kerfoot, W.C. and Robbins, J .A., 1999. Nearshore regions of Lake Superior: Multi-
element signatures of mining discharges and a test of Pb-210 deposition under
conditions of variable sediment mass ﬂux. J. Gt. Lakes Res. 25, 697-720.

Khim, J., Villeneuve, D., Kannan, K., Lee, K., Snyder, S., Koh, C. and Giesy, J ., 1999.
Alkylphenols, polycyclic aromatic hydrocarbons (PAHs), and organochlorines
in sediment ﬁ'om Lake Shihwa, Korea: Instrumental and bioanalytical
characterization. Environ Toxicol Chem 8, 2424-243 2.

Klovan, J .E., 1975. R- and Q-Mode factor analysis. In: R.B. McCammon (Editor),
Concepts in geostatistics. Springer-Verlag, New York.

Kneebone, RE. and Hering, J .G., 2000. Behavior of arsenic and other redox sensitive
elements in Crowley Lake, CA: A reservoir in the Los Angeles Aqueduct
system. Environmental Science & Technology 34, 4307-4312.

Koelmans, A.A., 1998. Geochemistry of suspended and settling solids in two
freshwater lakes. Hydrobiologia 364, 15-29.

Kolak, J.J., Long, D.T., Beals, T.M., Eisenreich, SJ. and Swackhamer, D.L., 1998.
Anthropogenic inventories and historical and present accumulation rates of
copper in Great Lakes sediments. Appl. Geochem. 13, 59-75.

Kolak, J.J., Long, D.T., Kerfoot, W.C., Beals, TM. and Eisenreich, S.J., 1999.

Nearshore versus offshore copper loading in Lake Superior sediments:
Implications for transport and cycling. J. Gt. Lakes Res. 25, 611-624.

317

Lamoureux, S., 1999. Spatial and interannual variations in sedimentation patterns
recorded in nonglacial varved sediments from the Canadian High Arctic.
Journal of Paleolimnology 21, 73-84.

Landis, M.S., Vette, A.F. and Keeler, G.J., 2002. Atmospheric mercury in the Lake
Michigan basin: Inﬂuence of the Chicago/Gary urban area. Environ. Sci.
Technol. 36, 4508-4517.

Lantzy, R.J. and Mackenzie, F.T., 1979. Atmospheric trace metals, global cycles and
assessment of man's impact. Geochim. Cosmochim. Acta 43, 511-525.

Long, D.T., Eisenreich, SJ. and Swackhammer, D.L., 1995. Metal Concentrations in
Sediments of the Great Lakes, Great Lakes Protection Fund.

Long, D.T. and Saleem, Z.A., 1974. Hydrogeochemistry of carbonate groundwaters of
an urban area. Water Resources Research 10, 1229-1238.

MacDonald, D.D., Ingersoll, CG. and Berger, T.A., 2000. Development and
evaluation of consensus-based sediment quality guidelines for ﬁeshwater
ecosystems. Archives of Environmental Contamination and Toxicology 39, 20-
3 1.

Manta, D.S., Angelone, M., Bellanca, A., Neri, R. and Sprovieri, M., 2002. Heavy
metals in urban soils: a case study from the city of Palermo (Sicily), Italy. Sci.
Total Environ. 300, 229-243.

McKee, J .D., Wilson, T.P., Long, D.T. and Owen, R.M., 1989. Pore water proﬁles and
early diagenesis of Mn, Cu and Pb in sediments from large lakes. J. Gt. Lakes
Res. 15, 68-83.

Mecray, E.L., King, J .W., Appleby, P.G. and Hunt, A.S., 2001. Historical trace metal
accumulation in the sediments of an urbanized region of the Lake Champlain
watershed, Burlington, Vermont. Water Air Soil Poll 125, 201-230.

Michigan Center for Geographic Information, 2002. Statewide Layer: Michigan
geographic framework. Michigan Center for Geographic Information, Lansing,
MI.

Michigan Department of Natural Resources, 1992. MIRIS base data. MDNR, Lansing,
MI.

Michigan Department of Natural Resources, 1999. 1978 Land cover / use. MDNR,
Lansing, MI.

318

Michigan Department of Natural Resources, 2000. State ownership. MDNR, Lansing,
MI.

Michigan Department of Natural Resources, 2003. IF MAP/GAP land cover. MDNR
Forest, Mineral, and Fire Management Division, Lansing, MI.

Muhlbaier, J. and Tisue, G.T., 1981. Cadmium in the Southern Basin of Lake-
Michigan. Water Air Soil Poll 15, 45-59.

National Aeronautics and Space Administration, 2002. Shuttle Radar Topography
Mission (SRTM) Spatial Dataset. US. Geological Survey, Sioux Falls, SD.

National Atmospheric Deposition Program (NRSP-3)/National Trends Network, 2003.
NADP Program Ofﬁce, Illinois State Water Survey, 2204 Griﬂith Drive,
Champaign, IL 61820.

Nriagu, J .0. (Editor), 1978. The Biogeochemistry of Lead in the Environment, 1A.
Elsevier, New York, 422 pp.

Nriagu, J .O., 1989. A Global Assessment of Natural Sources of Atmospheric Trace-
Metals. Nature 338, 47-49.

 

Oldﬁeld, F. and Appleby, P.G., 1984. Empirical testing of 210Pb-dating models for
lake sediments. In: E.Y. Haworth and J .W.G. Lund (Editors), Lake Sediments
and Environmental History. University of Minnesota Press, Minneapolis, pp.
93-124.

Outridge, P.M., Hermanson, M.H. and Lockhart, W.L., 2002. Regional variations in
atmospheric deposition and sources of anthropogenic lead in lake sediments
across the Canadian Arctic. Geochim. Cosmochim. Acta 66, 3521-3531.

Paul, J .F ., Comeleo, R.L. and Copeland, J ., 2002. Landscape metrics and estuarine
sediment contamination in the mid-Atlantic and southern New England
regions. J. Environ. Qual. 31, 836—845.

Perkins, S.M., Filippelli, GM. and Souch, C.J., 2000. Airborne trace metal
contamination of wetland sediments at Indiana Dunes National Lakeshore.
Water Air Soil Poll 122, 231-260.

Qu, W.C., Dickman, M. and Wang, S.M., 2001. Multivariate analysis of heavy metal
and nutrient concentrations in sediments of Taihu Lake, China. Hydrobiologia
450, 83-89.

Ramanathan, A.L., Subramanian, V. and das, B.K., 1996. Sediment and heavy metal
accumulation in the Cauvery basin. Environ. Geol. 27, 155-163.

 

319

Reimann, C. and de Caritat, P., 1998. Chemical Elements in the Environment:
F actsheets for the Geochemist and Environmental Scientist. Springer-Verlag,
Berlin.

Rice, K.C., 1999. Trace-element concentrations in streambed sediment across the
conterminous United States. Environ. Sci. Technol. 33, 2499-2504.

Richards, C., Johnson, LB. and Host, GE, 1996. Landscape-scale inﬂuences on
stream habitats and biota. Canadian Journal of Fisheries and Aquatic Sciences
53, 295-311.

Robbins, J .A., 1982. Stratigraphic and dynamic effects of sediment reworking by
Great Lakes zoobenthos. Hydrobiologia 92, 611-622.

Sanchez-Cabeza, J .A., Ani-Ragolta, I. and Masque, P., 2000. Some considerations of
the Pb-210 constant rate of supply (CRS) dating model. Limnology and
Oceanography 45, 990-995.

Sanei, H., Goodarzi, F. and Van der Flier-Keller, E., 2001. Historical variation of
elements with respect to different geochemical fractions in recent sediments
ﬁ'om Pigeon Lake, Alberta, Canada? J. Environ. Monit. 3, 27-36.

Sanudo-Wilhelmy, SA. and Gill, G.A., 1999. Impact of the Clean Water Act on the
levels of toxic metals in Urban Estuaries: the Hudson River estuary revisited.
Environ. Sci. Technol. 33, 3477-3481.

Shahin, U., Yi, S.M., Paode, RD. and Holsen, T.M., 2000. Long-term elemental dry
deposition ﬂuxes measured around Lake Michigan with an automated dry
deposition sampler. Environmental Science & Technology 34, 1887-1892.

Simonetti, A., Gariepy, C. and Carignan, J ., 2000a. Pb and Sr isotopic compositions of
snowpack ﬁ'om Quebec, Canada: Inferences on the sources and deposition
budgets of atmospheric heavy metals. Geochim. Cosmochim. Acta 64, 5-20.

Simonetti, A., Gariepy, C. and Carignan, J ., 2000b. Pb and Sr isotopic evidence for
sources of atmospheric heavy metals and their deposition budgets in
northeastern North America. Geochim. Cosmochim. Acta 64, 3439-3452.

Simpson, S.J., Long, D.T., Geisey, J .P. and Fett, J.D., 2000. Inland lakes sediment
trends: sediment analysis results for ﬁve Michigan Lakes (MI/DEQ/SWQ-
01/030), Department of Environmental Quality, East Lansing, MI.
http://www.michigan.gov/deq/0,1607,7-1 35-33 1 3_3 686_3 728-3 23 65—,00.html

320

 

 

 

Soranno, P.A., Hubler, S.L., Carpenter, SR. and Lathrop, RC, 1996. Phosphorus
loads to surface waters: A simple model to account for spatial pattern of land
use. Ecological Applications 6, 865-878.

Souch, C.J., Filippelli, G.M., Dollar, N., Perkins, S. and Mastalerz, M., 2003.
Accumulation rates of airborne heavy metals in wetlands. Phys. Geogr. 23, 21-
43.

Spiethoﬁ', HM. and Hemond, HR, 1996. History of toxic metal discharge to surface
waters of the Aberjona Watershed. Environ. Sci. Technol. 30, 121-128.

spss, 2000. SYSTAT 10, Chicago.

Stenson, H. and Wilkinson, L., 2000. Factor Analysis. In: SPSS Inc. (Editor),
SYSTAT 10 manual, Statistics 1, Chicago, pp. 327-364.

Stewart, 8.1., 1994. The seasonal home location decision process: toward a dynamic
model. Ph.D. Thesis, Michigan State University, East Lansing, MI, 156 pp.

Sweet, C.W., Weiss, A. and Vermette, S.J., 1998. Atmospheric deposition of trace
metals at three sites near the Great Lakes. Water Air Soil Poll 103, 423-439.

Tuncer, G., Tuncel, G. and Balkas, T.I., 2001. Evolution of metal pollution in the
Golden Horn (Turkey) sediments between 1912 and 1987. Marine Pollution
Bulletin 42, 350-360.

US. Department of Agriculture, 1994. State Soil Geographic (STATSGO) database
for Michigan. US. Department of Agricultrue, Natural Resources
Conservation Service, Fort Worth, Texas.

Walling, DE. and Qingping, H., 1992. Interpretation of caesium-137 proﬁles in
lacustrine and other sediments: the role of catchment-derived inputs.
Hydrobiologia 235, 219-230.

Wayland, K.G., Long, D.T., Hyndman, D.W., Pijanowski, B.C., Woodhams, SM. and
Haack, S.K., 2003. Identifying relationships between baseﬂow geochemistry
and land use with synoptic sampling and R-mode factor analysis. J. Environ.
Qual. 32, 180-190.

Wetzel, R.G., 2001. Limnology : lake and river ecosystems. Academic, San Diego,
Calif. London, xvi, 1006 pp.

Wilkinson, L., Engelman, L., Corter, J. and Coward, M., 2000. Cluster Analysis,
Systat 10, Statistics I. SPSS Inc., Chicago, pp. 53-85.

321

 

 

Winchester, J .W. and Nifong, GD, 1971. Water pollution in Lake Michigan by trace
elements from pollution aerosol fallout. Water, Air, Soil Pollut. 1, 50-64.

Wong, C.S., Sanders, G., Engstrom, D.R., Long, D.T., Swackhammer, BL. and
Eisenreich, S.J., 1995. Accumulation, inventory, and diagenesis of chlorinated
hydrocarbons in Lake Ontario sediments. Environ. Sci. Technol. 29, 2661-
2672.

Yohn, SS. and Long, D.T., in preparation. The importance of the multi-element
approach for interpreting sediment histories.

Yohn, S.S., Long, D.T., Fett, JD. and Patino, L.C., 2004. Regional versus local
inﬂuences on lead and cadmium loading in the Great Lakes region. Appl.
Geochem. 19, 1157-1175.

Yohn, S.S., Long, D.T., Fett, J.D., Patino, L.C., Giesy, J .P. and Kannan, K., 2002a.
Assessing environmental change through chemical-sediment chronologies
from inland lakes. Lakes & Reservoirs: Research and Management 7, 217-230.

Yohn, S.S., Long, D.T., Giesy, J.P., Fett, JD. and Kannan, K., 2001. Inland lakes
sediment trends: sediment analysis results for two Michigan Lakes.
MI/DEQ/WD-02/115, Department of Environmental Quality, East Lansing.

Yohn, S.S., Long, D.T., Giesy, J .P., Scholle, L.K., Patino, L.C., Fett, J .D. and Kannan,
K., 2002b. Inland lakes sediment trends: sediment analysis results for ﬁve
Michigan Lakes. MI/DEQ/WD-03/052, Department of Environmental Quality,
East Lansing.

Yohn, S.S., Long, D.T. and Parsons, M.J., in preparation-a. Lead in the environment:
evaluating sources and pathways over time using lake sediments.
Environmental Science & Technology.

Yohn, S.S., Long, D.T. and Winkler, J .A., in preparation-b. The use of multivariate
statistics for the interpretation of sediment chronologies. Limnology and
Oceanography: Methods.

Yohn, S.S., Parsons, M.J., Long, D.T., Giesy, J.P., Scholle, L.K. and Patino, L.C.,

2003. Inland lakes sediment trends: sediment analysis results for six Michigan
Lakes, Michigan Department of Environmental Quality, East Lansing, MI.

322