NATURAL LANDSCAPE DRIVERS OF TOTAL PHOSPHORUS CONCENTRATIONS
IN MICHIGAN LAKES
By
Alison Marie Keener

A THESIS
Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of
Geography – Master of Science
2013

ii

ABSTRACT
NATURAL LANDSCAPE DRIVERS OF TOTAL PHOSPHORUS CONCENTRATIONS
IN MICHIGAN LAKES
By
Alison Marie Keener
Paleolimnological reconstructions have demonstrated an inherent variability in
phosphorus concentrations among lakes prior to anthropogenic landscape disturbances.
However, the natural drivers of this variability are less frequently studied than agricultural or
urban phosphorus loading to freshwaters, and are not as well understood. Therefore, this
research examined pre-settlement phosphorus concentrations of 48 inland lakes in Michigan in
conjunction with pre-settlement land cover data, hydrogeomorphic features, and geographic
variables, to better understand these natural landscape drivers. This multi-step process involved
an analysis of diatom assemblages from lacustrine sediment cores, relative dating of core
samples using Ambrosia pollen, a native land cover analysis in GIS, and finally, regression
modeling of the data. The results of OLS regression modelling suggest that the variability in
pre-settlement phosphorus concentrations of my dataset is best explained by native wetland
cover. Moreover, native wetlands at a local (100 m buffer) scale had a stronger relationship with
pre-settlement phosphorus concentrations than native wetlands within the entire lake catchment.
The strength of this relationship improved further when wetlands were modeled with maximum
lake depth. My analyses show that as the amount of pre-settlement phosphorus concentrations in
a lake increased as wetlands near the lake shore increased. Furthermore, as lake depth decreases,
the effect of wetlands on lake phosphorus concentrations increases. However, because my study
was only able to explain 34% of the variability in pre-settlement phosphorus concentrations,
further research is warranted to better understand this natural variability.
ii

ACKNOWLEDGEMENTS

First and foremost, I would like to express my deepest gratitude to my advisor, Dr.
Catherine Yansa for her mentorship and guidance throughout the duration of this project. Thank
you for encouraging me to pursue my own research interests and for sharing your valuable
knowledge in regards to pollen, writing techniques, and life in general. I also would like to thank
my committee members, Dr. David Lusch and Dr. Patricia Soranno for their input and guidance
in every stage of project development. A huge thanks goes to Dr. Sarah Hession for consulting
in the statistical analyses of this study; without your help this journey would have been a lot
more difficult.
This research would not have been possible without the data contributed by Paul Garrison
and Dr. Patricia Soranno. Additionally, Emi Fergus deserves credit for laying the groundwork
for this research, back in 2008. I would also like to thank the Environmental Protection Agency
for making their data publically available for independent analyses, as well as Dr. Jan Stevenson
and Jason Zalack for providing me with access to the sediment sample.
Special thanks to my family and friends who have provided necessary distractions and
have been great supporters over the course of this research project. I would like to acknowledge
my parents, John and Mary, for their constant love and support, for teaching me to love science,
and for reminding me that there is always time to waterski. Annie and Chris, thanks for being
there for me in the eleventh hour by helping with final reviews and edits. Last but not least,
there are not enough words to show my appreciation to Joenick. Thank you for bringing me
coffee, sweatpants, and constant words of encouragement. You’ve been my rock during these
last few years and I wouldn’t have been able to do it without you!

iii

TABLE OF CONTENTS

LIST OF TABLES ......................................................................................................................... vi
LIST OF FIGURES ..................................................................................................................... viii
1. INTRODUCTION .......................................................................................................................1
1.1
Cultural Eutrophication and Phosphorus Regulations ...................................................1
1.2
Natural Variations in Lake Phosphorus Concentrations ................................................3
1.3
Problem Statement .........................................................................................................5
1.4
Research Objectives .......................................................................................................6
1.5
Outline of Thesis ............................................................................................................7
2. LITERATURE REVIEW ............................................................................................................8
2.1
Background ....................................................................................................................8
2.2
Phosphorus Cycle...........................................................................................................9
2.2.1 Phosphorus Retention ..................................................................................................10
2.2.2 Phosphorus Release .....................................................................................................11
2.3
Lake Productivity .........................................................................................................12
2.3.1 Trophic Classification of Lakes ...................................................................................12
2.3.2 Eutrophication. ............................................................................................................14
2.4
Phosphorus Management and Regulations ..................................................................15
2.5
Paleolimnological Reconstructions ..............................................................................17
2.5.1 Diatoms as Proxy Indicators of Water Chemistry .......................................................18
2.5.2 Pollen as a Proxy Indicator of Landscape Disturbances ..............................................22
2.6
Lake-Landscape Interactions .......................................................................................22
2.6.1 Hydrogeomorphic Features ..........................................................................................23
2.6.2 Phosphorus Dynamics in Wetland Ecosystems ...........................................................26
2.6.3 Phosphorus Yields from Forested Ecosystems ............................................................27
2.6.4 Proximity to Surface Waters ........................................................................................28
3. STUDY AREA ..........................................................................................................................30
3.1
Study Area Description ................................................................................................30
3.1.1 Climate and Vegetation................................................................................................30
3.1.2 Glacial History .............................................................................................................32
3.1.3 Hydrology ....................................................................................................................33
3.1.4 Euro-American Settlement...........................................................................................34
3.2
Study Lake Selection and Characteristics ....................................................................36
3.2.1 Lake Datasets ...............................................................................................................36
3.2.2 Lake and Native Vegetation Catchment Characteristics .............................................39
4. METHODS ................................................................................................................................42
4.1
Lacustrine Core Sample Collection .............................................................................42
4.2
Diatom-Inferred Phosphorus Modeling .......................................................................44
iv

4.2.1
4.2.2
4.3
4.3.1
4.3.2
4.4
4.4.1
4.4.2
4.5
4.5.1
4.5.2
4.5.3

Soranno/Garrison Methods ..........................................................................................44
EPA Methods ...............................................................................................................45
Relative “Dating” of Sediment Using Pollen...............................................................46
Background ..................................................................................................................46
Laboratory Methods .....................................................................................................47
Spatial Data Processing................................................................................................49
Acquisition and Description of Data............................................................................50
Data Analysis ...............................................................................................................52
Statistical Data Analysis ..............................................................................................53
Summary of Potential Predictor Variables ..................................................................54
Initial Variable Selection .............................................................................................57
Model Selection ...........................................................................................................58

5. RESULTS AND INTERPRETATIONS ...................................................................................61
5.1
Results of Diatom Ordinations and Analyses ..............................................................61
5.1.1 Results of Soranno/Garrison Dataset Analysis ............................................................61
5.1.2 Results of EPA Dataset Analysis .................................................................................63
5.1.3 General Discussion of Diatom-Inferred Data ..............................................................65
5.2
Pollen Analysis Results................................................................................................65
5.2.1 Correlation of Ambrosia with Depth............................................................................70
5.2.2 Correlation of Ambrosia with Total Phosphorus .........................................................71
5.2.3 Phosphorus Reconstructions from Study Lakes ..........................................................73
5.3
GIS Analysis Results ...................................................................................................73
5.4
Statistical Data Analysis Results and Discussion ........................................................76
5.4.1 Bivariate Regression Results and Discussion ..............................................................76
5.4.2 Model Selection Results ..............................................................................................79
5.3.5 Final OLS Model Interpretation...................................................................................81
6. DISCUSSION ............................................................................................................................83
6.1
Discussion of Model Results .......................................................................................83
6.2
Nutrient Policy Implications ........................................................................................85
6.3
Lake Management Implications ...................................................................................88
6.4
Future Research Directions ..........................................................................................89
7. CONCLUSION ..........................................................................................................................92
APPENDICES ...............................................................................................................................94
APPENDIX A ..........................................................................................................................95
APPENDIX B ..........................................................................................................................96
APPENDIX C ..........................................................................................................................99
APPENDIX D ........................................................................................................................110
REFERENCES ............................................................................................................................113

v

LIST OF TABLES

Table 2.1:

Trophic lake types, as suggested by Naumann (1919) (modified from Carlson and
Simpson 1996) .......................................................................................................13

Table 3.1:

Study area information for the 48 lakes in the final dataset ..................................38

Table 4.1:

Potential landscape parameters hypothesized to explain variations in presettlement phosphorus concentrations (μg/L) .......................................................55

Table 5.1:

Summary statistics of reconstructed phosphorus (inferred from diatom
assemblage) and observed phosphorus (measured in water column) from the
Soranno/Garrison lake dataset (n=31). Diatom-inferred total phosphorus
concentrations from the bottom (DI-TP Bottom) and top (DI-TP Top) core
intervals are in units μg/L ......................................................................................63

Table 5.2:

Summary statistics of reconstructed phosphorus (inferred from diatom
assemblage) and observed phosphorus (measured in water column) from the EPA
lake dataset (n=42). Diatom-inferred total phosphorus concentrations from the
bottom (DI-TP Bottom) and top (DI-TP Top) core intervals are in units μg/L…65

Table 5.3:

Results of the pollen analyses for 29 bottom-core samples from the
Soranno/Garrison dataset. The italicized rows located at the bottom of the table
are representative of non-pre-settlement conditions ..............................................67

Table 5.4:

Results of the pollen analyses for 34 bottom-core samples from the EPA dataset.
The italicized rows located at the bottom of the table are representative of nonpre-settlement conditions .......................................................................................68

Table 5.5:

Summary statistics of reconstructed phosphorus (inferred from diatom
assemblage) and observed phosphorus (measured in water column) from the “presettlement” lake dataset (n = 48). Diatom-inferred total phosphorus
concentrations from the bottom (DI-TP Bottom) and top (DI-TP Top) core
intervals are in units μg/L ......................................................................................73

Table 5.6:

Reclassification of pre-settlement land cover categories, as captured by the GLO
surveys (Anderson 1976).......................................................................................74

Table 5.7:

Summary of reclassified native land covers in the 100 m buffer and local
catchment area .......................................................................................................75

Table 5.8:

Pearson’s Correlation Coefficients (r) of the potential predictor variables with
pre-settlement phosphorus concentrations .............................................................78

Table 5.9:

Results of model-fitness testing from seven OLS regressions...............................79
vi

Table 5.10:

Results of diagnostic testing for spatial autocorrelation, conducted in GeoDa .....81

Table 5.11:

Final OLS model (#7) ............................................................................................82

Table A.1:

List of lakes excluded from final dataset ...............................................................95

Table B.1:

Correlation matrix of all potential predictor variables. ..........................................96

vii

LIST OF FIGURES

Figure 1.1:

Map of the inland lakes in Michigan including the EPA Level III Ecoregion
boundaries (modified from USEPA 2013a). For interpretation of the references to
color in this and all other figures, the reader is referred to the electronic version of
this thesis..................................................................................................................4

Figure 2.1:

An example of a common ordination approach to transfer function development,
redundancy analysis (RDA), to assess the significance, strength, and direction of
environmental variables as predictors of diatom species composition from a
calibration dataset (modified from Juggins et al. 2013) ........................................20

Figure 2.2:

Vollenweider’s (1968) phosphorus loading plot classifies lakes into trophic states
-2 -1
based on mean depth (m) and annual phosphorus loads (gPm yr ). The lines
marking the thresholds between eutrophic and oligotrophic lake conditions are
suggested to correspond to phosphorus concentrations of 20 μg/L and 10 μg/L,
respectively (modified from Chapra 1997) ............................................................25

Figure 3.1:

Map of pre-settlement forests in Michigan as interpreted from the General Land
Office (GLO) public land surveys between 1816-1856 (Comer et al. 1995) ……31

Figure 3.2:

Map of pre-settlement lakes, rivers, and wetlands in Michigan as interpreted from
the General Land Office (GLO) public land surveys between 1816-1856 (Comer
et al. 1995) .............................................................................................................33

Figure 3.3:

Map of pre-settlement land cover in Michigan as interpreted from the General
Land Office (GLO) public land surveys between 1816-1856(Comer et al. 1995) ....
................................................................................................................................35

Figure 3.4:

Map of study lakes and the forest tension zone (after Hupy and Yansa 2009a).
Fourteen lakes are located south of the forest tension zone’s center mark; thirtyfour lakes are located north of the forest tension zone ..........................................40

Figure 3.5:

Land cover in local catchment of lakes south of the forest tension zone ..............41

Figure 3.6:

Land cover in local catchment of lakes north of the forest tension zone ...............41

Figure 4.1:

A “top-bottom” approach to core sampling (left) yields two sediment samples,
representative of pre-settlement and modern conditions. The numerous sediment
intervals from “stratigraphic” sampling (right) provide a continuous analog of
lake conditions (modified from USEPA 2010)......................................................43

Figure 4.2:

Flow chart of GIS methods. A dotted line around a box indicates a task
performed, whereas a solid line identifies a coverage or product..........................51

viii

Figure 5.1:

Canonical correspondence analysis (CCA) ordination diagram of speciesenvironment relationship in the 187 lake calibration dataset generated by the
WDNR with the forward-selection option and Monte Carlo permutation tests in
CANOCO (version 4.1) (ter Braak and Šmilauer 1998; Garrison pers. comm,
2012) ......................................................................................................................62

Figure 5.2:

Correlation plot between the Ambrosia counts and the depth (cm) of the bottomcore interval (r= -0.120) .........................................................................................71

Figure 5.3:

A scatter plot of the bivariate correlation testing results between diatom-inferred
total phosphorus (DI-TP) in the bottom sediment cores and the total Ambrosia
pollen count from lakes in the Soranno/Garrison and EPA datasets (r = 0.3635).
An ‘X’ denotes sediment intervals that were interpreted to be representative of
pre-settlement conditions. Conversely, an ‘O’ denotes post-settlement sediment
intervals because Ambrosia-type pollen comprises > 3% of the total pollen spectra
................................................................................................................................72

Figure C.1:

Histogram of the dependent variable, showing the distribution of pre-settlement
phosphorus concentrations for the 48 lakes. ..........................................................99

Figure C.2:

Univariate and bivariate testing results for latitude. The histogram (left) shows the
distribution of the variable and the scatter plot (right) shows the strength of the
relationship with pre-settlement phosphorus (n=48) ...........................................100

Figure C.3:

Univariate and bivariate testing results for longitude. The histogram (left) shows
the distribution of the variable and the scatter plot (right) shows the strength of the
relationship with pre-settlement phosphorus (n=48) ...........................................100

Figure C.4:

Univariate and bivariate testing results for native deciduous forest (%) in the 100
m buffer. The histogram (left) shows the distribution of the variable and the
scatter plot (right) shows the strength of the relationship with pre-settlement
phosphorus (n=48) ...............................................................................................101

Figure C.5:

Univariate and bivariate testing results for native deciduous forest (%) in the local
catchment. The histogram (left) shows the distribution of the variable and the
scatter plot (right) shows the strength of the relationship with pre-settlement
phosphorus (n=48). ..............................................................................................101

Figure C.6:

Univariate and bivariate testing results for native evergreen forest (%) in the 100
m buffer. The histogram (left) shows the distribution of the variable and the
scatter plot (right) shows the strength of the relationship with pre-settlement
phosphorus (n=48). ..............................................................................................102

ix

Figure C.7:

Univariate and bivariate testing results for native evergreen forest (%) in the local
catchment. The histogram (left) shows the distribution of the variable and the
scatter plot (right) shows the strength of the relationship with pre-settlement
phosphorus (n=48). ..............................................................................................102

Figure C.8:

Univariate and bivariate testing results for native mixed forest (%) in the 100 m
buffer. The histogram (left) shows the distribution of the variable and the scatter
plot (right) shows the strength of the relationship with pre-settlement phosphorus
(n=48)...................................................................................................................103

Figure C.9:

Univariate and bivariate testing results for native mixed forest (%) in the local
catchment. The histogram (left) shows the distribution of the variable and the
scatter plot (right) shows the strength of the relationship with pre-settlement
phosphorus (n=48). ..............................................................................................103

Figure C.10: Univariate and bivariate testing results for native total forest (%) in the 100 m
buffer. The histogram (left) shows the distribution of the variable and the scatter
plot (right) shows the strength of the relationship with pre-settlement phosphorus
(n=48)...................................................................................................................104
Figure C.11: Univariate and bivariate testing results for native total forest (%) in the local
catchment. The histogram (left) shows the distribution of the variable and the
scatter plot (right) shows the strength of the relationship with pre-settlement
phosphorus (n=48). ..............................................................................................104
Figure C.12: Univariate and bivariate testing results for native wetland cover (%) in the 100 m
buffer. The histogram (left) shows the distribution of the variable and the scatter
plot (right) shows the strength of the relationship with pre-settlement phosphorus
(n=48)...................................................................................................................105
Figure C.13: Univariate and bivariate testing results for native wetland cover (%) in the local
catchment. The histogram (left) shows the distribution of the variable and the
scatter plot (right) shows the strength of the relationship with pre-settlement
phosphorus (n=48). ..............................................................................................105
Figure C.14: Univariate and bivariate testing results for native grassland cover (%) in the 100
m buffer. The histogram (left) shows the distribution of the variable and the
scatter plot (right) shows the strength of the relationship with pre-settlement
phosphorus (n=48). ..............................................................................................106
Figure C.15: Univariate and bivariate testing results for native grassland cover (%) in the local
catchment. The histogram (left) shows the distribution of the variable and the
scatter plot (right) shows the strength of the relationship with pre-settlement
phosphorus (n=48). ..............................................................................................106

x

Figure C.16: Univariate and bivariate testing results for wetland/lake ratio in the 100 m buffer.
The histogram (left) shows the distribution of the variable and the scatter plot
(right) shows the strength of the relationship with pre-settlement phosphorus
(n=48)...................................................................................................................107
Figure C.17: Univariate and bivariate testing results for wetland/lake ratio in the local
catchment. The histogram (left) shows the distribution of the variable and the
scatter plot (right) shows the strength of the relationship with pre-settlement
phosphorus (n=48). ..............................................................................................107
Figure C.18: Univariate and bivariate testing results for maximum lake depth (m). The
histogram (left) shows the distribution of the variable and the scatter plot (right)
shows the strength of the relationship with pre-settlement phosphorus (n=48).
..............................................................................................................................108
Figure C.19: Univariate and bivariate testing results for lake area (ha). The histogram (left)
shows the distribution of the variable and the scatter plot (right) shows the
strength of the relationship with pre-settlement phosphorus (n=48). ..................108
Figure C.20: Univariate and bivariate testing results for lake/catchment ratio. The histogram
(left) shows the distribution of the variable and the scatter plot (right) shows the
strength of the relationship with pre-settlement phosphorus (n=48). ..................109
Figure C.21: Univariate and bivariate testing results for lake perimeter. The histogram (left)
shows the distribution of the variable and the scatter plot (right) shows the
strength of the relationship with pre-settlement phosphorus (n=48). ..................109
Figure D.1:

Moran’s I results; nearest neighbor= 2. ...............................................................110

Figure D.2:

Moran’s I results; nearest neighbor= 4. ...............................................................111

Figure D.3:

Moran’s I results; nearest neighbor= 9. ...............................................................112

xi

1. INTRODUCTION
1.1 Cultural Eutrophication and Phosphorus Regulations
Phosphorus is a naturally occurring nutrient required by all primary producers for cell
maintenance (Harper 1992; Smith and Smith 2001; Wetzel 2001). However, excess phosphorus
loading from anthropogenic sources, termed cultural eutrophication, is consistently ranked by the
United States Environmental Protection Agency (EPA) as one of the leading causes of toxic algal
blooms, oxygen depletion, and an attendant loss of aquatic biodiversity (USEPA 2000; Herlihy
et al. 2013). In addition to these direct ecosystem impairments, degraded water resources
increase the costs of water treatment, reduce aesthetic tourism and property values, and place
unneeded strain on commercial and/or recreational fisheries (Seppälä et al. 2004).
The role of phosphorus as a limiting nutrient in freshwater lakes was first recognized by
th

limnologists in the mid-20 century (Hutchinson 1973; Schindler 1977; Juggins et al. 2013). In
an effort to correct these impairments and degradations of lakes and streams, effluents are now
limited from discrete conveyances under the jurisdiction of the National Pollutant Discharge
Elimination System (NPDES) program, a cornerstone of the 1972 Clean Water Act (CWA)
(Clean Water Act 1972). The NPDES regulation has reduced the impact of point source
pollutant discharges, but has only partially lessened the effects of cultural eutrophication
(Svendsen et al. 1995). In fact, an analysis of a subset of lakes from the National Eutrophication
Survey (NES) indicated the number of eutrophic lakes actually increased from 16% in 1972 to
26.9% of lakes in 2007 (USEPA 2009). Indisputably, point source pollution control alone is not
enough to achieve nutrient reduction goals for water bodies (Svendsen et al. 1995).
It is now widely recognized that the majority of cultural eutrophication in freshwater
lakes has emanated from agricultural and urban runoff contributions (Carpenter et al. 1998;
1

Reynolds and Davies 2001). Unfortunately, these nonpoint source inputs of phosphorus from
overland flow are inherently more difficult to monitor and regulate in comparison to point source
pollutants (Carpenter et al. 1998). Furthermore, because of the decentralized nature of nonpoint
phosphorus sources, it is challenging to distinguish between natural and anthropogenic inputs.
Moreover, due to natural variations in ambient levels of total phosphorus, nutrient management
strategies now acknowledge that standardized nutrient regulations may not be appropriate to
achieve water restoration goals (USEPA 2000).
Rather, current regulations and management efforts focus on establishing numeric
nutrient criteria (enrichment limits, typically for phosphorus and nitrogen) for geographic regions
based on water parameters representative of ‘pristine’ waters (USEPA 2000). Several methods
have been developed to identify lake phosphorus concentrations as they were in undisturbed
conditions. These methods include the use of reference lakes, expert judgments, and
paleolimnological reconstructions of lake conditions prior to Euro-American settlement in the
th

early to mid-19 century (USEPA 2000; Herlihy et al. 2013). Reference lakes are “minimally
impacted” sites representative of undisturbed water quality conditions in a given region. While
this method has its merits, it has been argued that anthropogenic impacts are so pervasive and
ubiquitous in the environment that truly undisturbed aquatic systems no longer exist (Herlihy et
al. 2013).
Therefore, paleolimnological reconstructions of pre-settlement lake conditions are a
valuable alternative approach to establishing numeric phosphorus criteria in the absence of
undisturbed sites (Dixit et al. 1999; Herlihy et al. 2013). This method employs the use of fossil
diatom assemblages from lake sediment cores as proxies to infer past water conditions, and is
particularly well-suited to moist-temperate environments where phosphorus is the limiting
2

nutrient and lake sediments are well preserved (Cohen 2003). Furthermore, the coupling of presettlement, diatom-reconstructed phosphorus concentrations with pre-settlement land cover and
hydrogeomorphic (HGM) data can provide a better understanding to the natural sources of
phosphorus loading.
1.2 Natural Variations in Lake Phosphorus Concentrations
It is well known that lake water quality is reflective of the watershed characteristics that
control the transport of materials transferred by surface runoff (Soranno et al. 1996). The
relationship between lake nutrients and the gradient of urban and/or agricultural intensities in a
watershed has been clearly demonstrated over the last few decades (Benoit and Fizaine 1999;
Berka et al. 2001; Cuffney et al. 2000). There has been comparatively little research examining
the natural variations in phosphorus concentrations because very few long-term datasets exist
and it is rare to examine an undisturbed watershed today (Fritz et al. 1993). In many cases, the
paleolimnological record is the only means of documenting pre-settlement trends in water quality
data within the ecological context needed to separate natural lake variability from anthropogenic
stress (Smol 2002). Exploring the spatial complexity of natural phosphorus variation, reflective
of differences in vegetation and hydrogeomorphology within a region, will contribute to a better
understanding of the combined effects of the surrounding landscape on nutrient and water
inflow, a precursor to the establishment of more accurate and attainable nutrient criteria as
management goals (Zhang et al. 2012; Herlihy et al. 2013).
These complex relationships are best approached from a landscape limnology perspective
which is defined as, “the spatially explicit study of lakes, streams, and wetlands as they interact
with freshwater, terrestrial, and human landscapes to determine the effects of pattern on
ecosystem processes across temporal and spatial scales” (Soranno et al. 2010, p. 442). In other

3

words, the effects of a landscape on surface water chemistry should not be presumed consistent
among regions, but rather, these relationships may be influenced by cross-scale interactions
(Fergus et al. 2011). Additionally, water chemistry, in a landscape limnology context, requires
the incorporation of hydrogeomorphic (HGM) features such as basin morphometry and
hydrologic connectivity (Bremigan and Soranno 2008; Fergus et al. 2011).
Therefore, many management strategies rely on landscape-based ecological
classifications to improve the tractability of thousands of lakes within the United States (Soranno
et al. 2010). For example, the EPA’s Level III Ecoregions divides Michigan into five different
ecological units (Figure 1.1), each characterized by local interactions of vegetation, climate, soils

V

EPA Level III Ecoregions
I = S. Michigan/N. Indiana Drift
Plains
II = Eastern Corn Belt Plains
III = Huron/Erie Lake Plains
IV = North Central Hardwood
Forests
V = Northern Lakes and Forests

IV

V
III
I
II

Figure 1.1: Map of the inland lakes in Michigan including the EPA Level III Ecoregion
boundaries (modified from USEPA 2013a). For interpretation of the references to color in
this and all other figures, the reader is referred to the electronic version of this thesis.
4

and topography (USEPA 2013a). Each unit in Michigan denotes a group of terrestrial and
aquatic ecosystems that are assumed to respond similarly to management actions because of
comparable landscape characteristics such as forest species composition, remnant glacial
landforms, and land use capabilities.
1.3 Problem Statement
Although Michigan has approximately 11,000 inland lakes, there is little understanding
of how these lakes have changed since Euro-American settlement in the mid-1800s. The
majority of lake phosphorus research is associated with modeling the effects of human land
use/land use change activities to predict and explain nutrient patterns in lakes after EuroAmerican settlement. Mostly, these studies focus on individual lakes and specific nonpoint
sources such as fertilizer runoff or septic systems leakages. Therefore, much of our
understanding of lake ecology is derived from systems already highly modified by anthropogenic
activities. There are comparatively few studies that examine the variability in pre-settlement
lake phosphorus concentrations in the context of natural landscape variables in the United States.
However, as demands on aquatic ecosystems continue to escalate, the knowledge of these
baseline conditions and natural variability in total phosphorus concentrations is increasingly
recognized of prime importance for effective lake ecosystem management (Smol 1992).
Although long-term records of previous lake conditions are often unavailable, proxy indicators
can be used to reconstruct these pre-settlement patterns and provide insight into natural drivers of
ecosystem functioning (Smol 1992; Reavie et al. 1995). This reference information can be used
in informed ecosystem management to define restoration goals, determine restoration potential,
and evaluate the success of restoration efforts (White and Walker 1997; Rhemtulla and
Mlandenoff 2007; Stein et al. 2010).

5

1.4 Research Objectives
The objective of this paper is to measure the pre-Euro-American landscape-lake
phosphorus relationships Michigan. I hope that this knowledge can be applied to better-informed
management plans and policy decisions that regulate anthropogenic phosphorus loading to lakes.
I used fossil diatom assemblages, recovered from the bottoms of short sediment cores taken from
48 lakes distributed throughout Michigan, as paleoindicators of pre-settlement lake phosphorus
concentrations. Native vegetation data (obtained from the Michigan Natural Features
th

Inventory’s compilation of 19 century Public Land Office surveys (Comer et al. 1995)) and
hydrogeomorphic lake features were included as predictor variables in a regression model that
describes the variability of diatom-inferred phosphorus concentrations among the 48 lakes.
Similar research has been done before by others, however prior work tends to be sitespecific, reconstructing pre-settlement changes of a single lake based on diatom assemblages
from several stratigraphic levels of a sediment core, dated using radioisotopes

210

Pb and

137

Cs.

Unlike these intensive studies of individual lakes, the aim of this thesis is to assess broad spatial
patterns in landscape composition and correlated trends in lake phosphorus concentrations in
Michigan.
My research will therefore address the following questions:
1. Which land cover features are the most effective predictors (higher statistical
explanatory power) of phosphorus concentrations in lakes, prior to Euro-American
settlement?
2. Are land cover metrics combined with hydrogeomorphic features and/or geographic
variables better predictors of phosphorus variability than land cover alone?

6

3. Can land cover within 100 m of the lake shore better account for the variability in
pre-settlement phosphorus concentrations than land cover for the entire lake
catchment?
1.5 Outline of Thesis
This thesis is presented in six chapters. Chapter One introduces the topic and presents the
research questions in more depth. Chapter Two is a literature review that primarily includes a
discussion of existing studies of relevance to this research. The main topics presented in the
chapter are eutrophication, the phosphorus cycle, management and regulations, and landscape
limnology. Chapter Three includes a description of the study lakes and relevant physical
parameters within the state of Michigan. This chapter also includes an outline of the site
selection process. The fourth chapter presents a synopsis of the materials and methods used in
this thesis. This work was conducted in several stages both in the field and in the laboratory.
First, an overview of the sediment core extraction and the diatom analysis from 2007 is
presented, followed by a description of the pollen analysis, GIS methods, and lastly the statistical
modeling used to explore the relationship between phosphorus concentrations and landscape
variables. Chapter Five presents the results of the methods used, while Chapter Six discusses the
major findings and also discusses further research needed if we are to understand the complexity
in landscape-water interactions. Finally, a conclusion for the entire study is presented in Chapter
Seven.

7

2. LITERATURE REVIEW
This chapter will review the major scientific findings related to phosphorus and the
physiographic factors that influence nutrient loading in freshwater lakes. In the first sections, I
will provide essential background information regarding the phosphorus cycle and lake
productivity. I will then discuss how paleolimnology can provide data useful to inform and
supplement existing water regulations and management efforts. The latter half of this chapter is
devoted to an assessment of the complex array of landscape factors that have been used as
predictors of water chemistry.
2.1 Background
Phosphorus is a naturally occurring element derived from eroded sedimentary rocks, and
required by all organisms for cell maintenance (Harper 1992; Smith and Smith 2001; Wetzel
2002). Due to the inherently low availability of phosphorus in freshwater ecosystems, as
compared to terrestrial environments, it typically functions as the growth limiting nutrient for
primary productivity in lakes, ponds, and streams (Harper 1992; Kalf 2002; Richardson and
Vaithiyanathan 2009). However, under high levels of phosphorus loading, algae blooms and
enhanced rooted macrophyte (submerged aquatic plant) growth can compromise the utility of
surface waters for drinking, recreation, fisheries, and industry (Hasler 1947; Sharpley et al. 1994;
Wetzel 2001). Therefore, control of nutrient pollution from point and nonpoint sources has been
at the forefront of regulatory agencies since the 1970s (Sharpley et al. 1994). Unfortunately,
setting goals for nutrient reduction is complicated by conflicting stakeholder interests and
inherent regional watershed differences (Sharpley et al. 1994). To establish more realistic and
impartial lake restoration targets based on natural, background phosphorus concentrations,
paleolimnological techniques can be used (USEPA 2000).
8

2.2 Phosphorus Cycle
Many quantitative studies have been conducted within the last 60 years to determine the
fate of various phosphorus fractions in freshwater ecosystems, often with the overarching goal of
identifying major threats to water quality (Johnston 1991; Boers et al. 1993; Reddy and Flaig
1995; Dong et al. 2011). However, measuring the mobility and bioavailability of phosphorus in
aquatic ecosystems has proved challenging because cycling is strongly regulated by a myriad of
complex biogeochemical mechanisms (Richardson 1985; Richardson and Vaithiyanathan 2009).
Indeed, phosphorus is an extremely active element that can change from particulate to dissolved
forms as it undergoes biotic and abiotic transformations in the environment (Richardson and
Vaithiyanathan 2009). Therefore, most water quality studies refer to measures of total
phosphorus, an aggregate term for both particulate and dissolved phosphorus content in a water
sample (Pathak et al. 2012).
Particulate phosphorus compounds bound to soil sediments and/or organic matter often
dominate the total phosphorus budget (Svendsen et al. 1995; Reynolds and Davies 2001). This
fraction of phosphorus enters aquatic systems primarily through stream bank erosion or nonpoint
source runoff with a high content of suspended sediments (Williams 1971; Harper 1992; Wetzel
2001). Although it typically constitutes a smaller fraction of total phosphorus, dissolved
inorganic phosphate (sometimes referred to as orthophosphate or soluble reactive phosphate)
represents the portion of total phosphorus immediately available for biological uptake (Sharpley
et al. 1994; Wetzel 2001; Mitsch and Gosselink 2007). Once considered an inconsequential
transport process, subsurface and surface runoff containing dissolved phosphorus leached from
agricultural soils is now recognized as an environmentally significant source of anthropogenic
phosphorus enrichment of surface waters (Sims et al. 1998; Djodjic et al. 2004). Even though
9

soil leaching rates can be high following rainfall or snowmelt events, dissolved phosphorus is
more commonly sourced from industrial inflows or sewage effluents (Holtan et al. 1988; Harper
1992; Dorioz et al. 1997; Wetzel 2001; Kalf 2002; Djodjic et al. 2004). The amount and
composition of dissolved versus particulate phosphorus fractions are strongly regulated by
retention and release processes in aquatic ecosystems (Svendson et al. 1995).
2.2.1 Phosphorus Retention
The retention capacity of an aquatic ecosystem is the ability of that system to remove and
retain phosphorus from the water column through biogeochemical processes (Reddy et al. 1999).
Research on aquatic ecosystems has repeatedly attributed phosphorus retention to mechanisms
such as adsorption, sedimentation, and vegetative assimilation (Harper 1992; Reddy et al. 1999;
Wetzel 2001; Devesa-Rey et al. 2009). Thus, the extent of ecosystem productivity is in part
regulated by these biotic and abiotic mechanisms.
Abiotic adsorption to organic and inorganic soil constituents (e.g., clay particles, organic
matter, and ferric/aluminum oxides) renders phosphorus insoluble and thus unavailable to plants
and microconsumers (Wetzel 2001; Mitsch and Gosselink 2007). Therefore, elevated
concentrations of dissolved phosphorus can be indicative of low adsorption capacity (Kleeberg et
al. 2007). This adsorption capacity is largely determined by pedogenic soil properties; for
example, highly weathered, acidic soils dominated by large quantities of sesquioxides readily
adsorb phosphorus (Sollins et al. 1988). Although this mechanism substantially limits its
3-

bioavailability, adsorbed phosphate (PO4 ) can become soluble and remobilized under anoxic
conditions (Khalid et al. 1977; Scheffer 1998).

10

Sedimentation is a fairly permanent retention mechanism that occurs as organic matter
and inorganic particles settle out of the water column, potentially reducing turbidity and algae
growth by immobilizing total phosphorus (Wetzel 2001). Aquatic sediments, however, have a
finite ability to retain phosphorus. The extent of phosphorus immobilization in benthic substrate
is greatly influenced by the composition, temperature, bioturbation, and the oxidation-reduction
potential of the sediment-water interface (Khalid et al. 1977; Johnston 1991; Reynolds and
Davies 2001; Devesa-Rey et al. 2009; Richardson and Vaithiyanathan 2009).
Phosphorus is also retained through biotic assimilation by autotrophic organisms, such as
phytoplankton, macrophytes and periphyton into microbial (Reddy et al. 1999). In general,
phosphorus storage through assimilation is more short-term than the other retention mechanisms,
and is regulated by the autotrophic organism’s lifespan (Reddy et al. 1999; Lund 2000; Cheng
2011). Furthermore, the efficiency of biotic assimilation varies with site-specific circumstances
such as the species of plants and the plant development stage (Reddy et al. 1999).
2.2.2 Phosphorus Release
Under certain physiochemical and biological conditions, phosphorus bound to sediments
can be released back into the overlying water column and made available, as dissolved
phosphorus, to primary producers (Johnston 1991; Dong et al. 2011). The resuspension and
entrainment rates of benthic phosphorus depend on the chemical bonding nature of the
sediments, oxidation-reduction (redox) conditions, and mixing from biota or wind/wave action
(Engstrom and Wright 1984; Kalf 2002; Dong et al. 2011). For example, lakes with anoxic
hypolimnetic waters have demonstrated elevated rates of phosphorus release from benthic
sediments. Phosphorus can also be released, or desorbed, from suspended sediments in the

11

photic zone of the epilimnion, where it is readily available for phytoplankton uptake (Kalf 2002).
Additionally, the partial senescence of submerged aquatic macrophytes can also cause nutrient
leaching as foliage matures (Wetzel 2001). However, it is more common that dissolved
phosphorus is liberated from decomposing detritus (Reynolds and Davies 2001). Combined,
these processes ultimately determine the productivity and trophic status of lakes.
2.3 Lake Productivity
2.3.1 Trophic Classification of Lakes
European limnologists Einar Naumann and August Thienemann laid the foundation for
th

the ecological classification of lakes shortly after the turn of the 20 century (Hasler 1947;
International Symposium on Eutrophication 1969; Hutchinson 1973; Harper 1992). Naumann
(1919) was the first limnologist to introduce the concept of oligotrophic and eutrophic waters in
terms of visually observed phytoplankton productivity in lakes during the summer months
(International Symposium on Eutrophication 1969; Hutchinson 1973; Wetzel 2001). Conceding
to Naumann’s terminology, Thienemann (1921) characterized lake trophic conditions as a
function of the hypolimnetic oxygen regime during summer stratification based on observed
benthic organism assemblages (Wetzel 2001). Thienemann found oligotrophic lakes with
oxygenated waters to display great diversity in benthic fauna, whereas eutrophic lakes were
prone to oxygen depletion and low biological diversity (Harper 1992; Wetzel 2001). It was
because of these complimentary ideas that Naumann and Thienemann eventually collaborated to
develop a general lake classification system reflective of trophic conditions that Naumann
referred to as “asymmetric”, or disharmonious (Table 2.1; Hutchinson 1973; National Research

12

Council 1996). However, as the number of recognized lake types increased, so did the
complexity and ambiguity of the terminology (Carlson and Simpson 1996).
Table 2.1: Trophic lake types, as suggested by Naumann (1919) (modified from Carlson
and Simpson 1996).
Trophic Classification
Oligotrophy
Eutrophy
Acidotrophy
Alkalitrophy
Argillotrophy
Siderotrophy
Dystrophy

Characteristics
Lakes with low production associated with low
phosphorus and nitrogen
Lakes type with high production, associated with
high phosphorus and nitrogen
Lake type with low production, with pH values
less than 5.5
Lake type with low production, associated with
high calcium concentrations
Lake type with low production, associated with
high clay turbidity
Lake type with low production, associated with
high iron content
Lake type with low production, associated with
high humic color

Regardless, lake trophic classifications are useful for communicating technical
information to stakeholders, given the escalating demands on aquatic resources. Therefore,
adaptations of the original, qualitative lake classification system still persist today (Hutchinson
1973). Numerous indices of trophic state have been proposed that reflect the general level of
primary productivity, quantitatively defined by total phosphorus concentrations, chlorophyll-a,
and/or Secchi disk depth (Carlson and Simpson 1996; Wetzel 2001; USEPA 2007a). Although
there is still some disagreement about the precise trophic boundaries in temperate freshwater
lakes, total phosphorus generally ranges from < 4 μg/L in unproductive, oligotrophic waters to
>100 μg/L in highly productive, hypereutrophic waters (Carlson 1977; Rast and Lee 1978; Ney
1996; Wetzel 2001). These trophic conditions are required to be monitored and reported to the
EPA under Section 314 of the Clean Water Act (Clean Water Act 1972).
13

2.3.2 Eutrophication
Eutrophication is a natural part of the successional development of aquatic ecosystems
that often signifies the intermediate and terminal stages of lake ontogeny (Wetzel 2001; Brenner
and Escobar 2009). Under eutrophic conditions, rates of photosynthetic productivity are
relatively high and the accumulation of organic biomass and sediments can eventually result in a
reduction in water depth or even a complete infill of the lake basin (Wetzel 2001). The
timeframe required for autochthonous and allochthonous sediments to accumulate and cause
natural lake senescence can be thousands of years, long surpassing the duration of most
ecological investigations (Wetzel 2001; Brenner and Escobar 2009; Mitraki 2012).
Although eutrophication is a natural part of lake succession, anthropogenic nutrient
enrichment has accelerated changes that would have otherwise taken centuries or millennia to
occur (Sawyer 1947; Harper 1992). This acceleration of primary production, aptly referred to as
cultural eutrophication, is consequential of nonpoint source pollution from agricultural drainage
and urban effluents (Richardson and Vaithiyanathan 2009). There have been ample experiments
to date that have demonstrated these associations, particularly as they pertain to phosphorusbased fertilizers, which increased in use by commercial agricultural operations from 0.5 million
metric tons in 1945 to 10.3 million metric tons in 1993 (e.g., Puckett 1995; Liu et al. 2010; Liu et
al. 2012).
th

Yet up until the mid-20 century, lake fertilization studies were few in number and
mostly offered only circumstantial evidence that was difficult to interpret because experimental
controls were lacking (Hasler 1947). Carbon dioxide production from bacterial, aerobic
respiration was even hypothesized to be the primary regulator of algal growth for a period in the
14

late-1960s (Kuenzel 1969; Harper 1992). However, several projects involving deliberate lake
fertilization as a surrogate for nonpoint runoff during the early 1970s conclusively identified
total phosphorus as the limiting nutrient in temperate, freshwater aquatic systems (Dillon and
Rigler 1974; Schindler 1974; Smol 2002).
The most recognized of these studies to unequivocally accredit eutrophication to excess
phosphorus came from a series of controlled, whole-lake experiments in Northwestern Ontario’s
Experimental Lakes Area (Schindler 1974). This experiment manipulated the chemical input of
several adjacent lakes and examined the outcome for biological reactions (Schindler 1974;
National Research Council 1996). Noxious algal blooms were observed in the lake with high
phosphorus influents, while the lake with only carbon and nitrogen inputs remained relatively
unchanged (Schindler 1974). This significant limnological breakthrough disproved speculations
that hypothesized carbon or nitrogen to be the limiting nutrient for primary production in most
freshwater lakes, and instead identified phosphorus as the most important nutrient (Schindler
1977). As a result, phosphorus has arguably become the most intensively studied element in
fresh waters (Vollenweider 1968; Chapra and Reckhow 1979; Wetzel 2001; Håkanson and
Boulion 2002), contributing data useful for the management and regulation of this limiting
nutrient.
2.4 Phosphorus Management and Regulations
In systems degraded by cultural eutrophication, the reduction of phosphorus can produce
beneficial results such as the elimination of nuisance algal blooms that otherwise clog the filters
at waste water treatment facilities and cause odors in treated waters (Juggins et al. 2013).
Phosphorus reduction in eutrophic lakes has also rehabilitated fish communities that were

15

previously limited in biodiversity (Makarewicz and Bertram 1991; Smith et al. 1999; Ludsin et
al. 2001; Wetzel 2001; Juggins et al. 2013). For example, nutrient abatement programs
implemented as part of the Great Lakes Water Quality Agreement of 1972 restored the severely
degraded Lake Erie basin system to its naturally oligotrophic condition (Ludsin et al. 2001). As
a result, the higher oxygen levels in the hypolimnion aided the return of sensitive benthic
macroinvertebrate species and increased fish diversity (Nürnberg 1995; Ludsin et al. 2001).
Furthermore, enhanced light penetration improved prey detection capabilities, causing an
increase in piscivorous fish abundance (Persson et al. 1991; Miner and Stein 1993; Ludsin et al.
2001).
Nevertheless, some scholars argue that the over-reduction of nutrients can cause an
imbalance in an aquatic ecosystem’s food web structure, ultimately leading to unproductive and
declining fisheries (Ney 1996; Ashley et al. 1997; Larkin and Slaney 1997; Stockner et al. 2000).
Stockner et al. (2000, p.7), for example, even suggest the reintroduction of phosphorus in a
“carefully controlled and ecologically sensitive way to restore sufficient fisheries production
levels.” Furthermore, nutrient removal can stem the natural ontogeny of lakes to become
shallower and more productive as they senesce (Stockner et al. 2000). This dichotomy of
stakeholder interests in ecosystem services can make it challenging for watershed organizations
and/or local governments to develop water quality strategies that uphold the objectives of the
Clean Water Act (CWA 1972).
To address the need for impartial benchmarks by which to gage water pollution
prevention and control measures under CWA jurisdiction, the EPA developed a national nutrient
strategy to encourage states to adopt numeric nutrient criteria that define concentrations limits in
lakes and reservoirs (USEPA 2000). However, setting nutrient criteria for phosphorus has been
16

difficult, in part, because there is a great deal of variability in inherent phosphorus levels and in
the manifestation of eutrophication in freshwater systems throughout the country (USEPA 1998).
For this reason, a single comprehensive concentration number is not appropriate for managing
nutrient pollution (USEPA 1998). It is thus necessary to determine natural limnological
conditions, prior to human influences, to effectively manage lake systems; however, long-term
data that can provide this anecdotal water quality information prior to 1950 are generally lacking
(Kalff 2002).
Therefore, considerable attention has been given to reconstructing past nutrient levels
from lake sediments profiles (Smol 2002). However, because of diagenetic physiochemical
changes that occur after deposition, it is inaccurate to simply analyze lacustrine sediments for an
archived chemical record (Engstrom and Wright 1984; Smol 2002). Rather, paleolimnological
proxies can be used to reconstruct pristine baseline limnological conditions that function as the
benchmarks for management and restoration goals in disturbed watersheds.
2.5 Paleolimnological Reconstructions
Sedimentary chemical profiles from lacustrine cores are limited in their ability to
reconstruct pre-settlement phosphorus loading because retention in sediments is strongly
regulated by geochemical processes that alter the sorption capacity of soluble reactive
phosphorus (Carignan and Flett 1981; Anderson et al. 1993; Wetzel 2001; Dong et al. 2011).
Additionally, the possibility of post-depositional mobility of phosphorus in lacustrine sediments
had long been proposed by many (Williams 1971; Bloesch 1976), but was first conclusively
proven by Carignan and Flett (1981). Their study provided evidence of an upward migration in
dissolved phosphorus from the reduced zone in lake sediments toward the oxidized sediment-

17

water interface. Consequentially, interpretation of dated sedimentary phosphorus may yield
erroneous records of past lake conditions.
Although misinterpretations of sedimentary phosphorus obviate its usefulness as a record
of past ambient lake chemistry, other methods have been developed to reconstruct nutrient
concentrations based on biological proxy data from the sediment record (Anderson et al.1993;
Garrison and Fitzgerald 2005). These proxies are not mobile in sediments to the same extent that
sedimentary phosphorus is in the lacustrine sediment profile. The most common
paleolimnological indicators for pre-settlement water quality reconstructions are fossil
diatomaceous algal assemblages (Schindler 1987; Anderson et al. 1993; Reavie et al. 1995).
2.5.1 Diatoms as Proxy Indicators of Water Chemistry
Diatoms are small, rapidly-reproducing algae (Class Bacillariophyceae) found abundantly
in fresh and saline waters. Species are relatively easily distinguished by distinct morphological
features in the frustule (cell wall) structure (Smol and Stoermer 2010). These intricate, siliceous
frustules of diatoms are well preserved in most lake sediments and can be used to decipher longterm ecological changes (Hall and Smol 1992; Battarbee 2000; Smol and Stoermer 2010). Fossil
diatom assemblages are exceptionally useful for quantifying pre-settlement lake conditions to a
high degree of certainty because each species has optimal ecological preferences and specific
abiotic constraints (Fritz et al. 1991; Dixit et al. 1992). These paleolimnological proxies are
well-suited for studies of lake eutrophication, in particular, because certain diatom taxa have
demonstrated a rapid response to changes in nutrient concentrations (Hall and Smol 1999).
Prior to the late 1980s, diatom records in lacustrine sediments were subjectively
interpreted using a relatively small number of indicator species to qualitatively describe trophic
18

status (Brumgan 1978; Ennis et al. 1983; Hall and Smol 1992; Anderson et al. 1993). For
example, a diatom assemblage dominated by Stephanodiscus revealed eutrophic conditions,
whereas an abundance of Cyclotella was indicative of oligotrophic conditions (Anderson et al.
1993). Fortunately, the ability to make quantitative, statistically-based inferences of historic
phosphorus concentrations has strengthened within the last few decades due to the development
of multivariate ordination techniques and transfer functions for diatom data (ter Braak 1986;
Birks 1998; Hall and Smol 1992; Anderson et al. 1993; Reavie et al. 1995; Smol 2002).
Before a transfer function can be developed to infer pre-settlement limnological
conditions from fossil diatom assemblages, the environmental optima of taxa must be estimated
(Anderson et al. 1993; Smol 2002). This information can be gathered from pre-existing lake
surveys, but more frequently involves using surface calibration sets (also known as ‘training
sets’) from a range of lakes, under the assumption that ecological preferences of taxa have not
changed over time (Fritz et al. 1999; Stevenson and Pan 1999; Smol 2002). To produce a
calibration set, diatom assemblages are collected from sediments at the surface-water interface
(i.e., the uppermost ~1-2 cm) of a calibration lake core (Smol 2002). Under normal
sedimentation rates, this interval represents the last two or three years of accumulation within a
lake (Smol 2002). Water samples are also collected at the time of sediment coring and analyzed
for environmental variables such as total phosphorus, total nitrogen, and pH and used in
conjunction with the diatom data (Dixit et al. 1992). Finally, ordination techniques are applied to
summarize the calibration dataset.
Direct and indirect gradient analyses are the most commonly used ordination techniques
to relate latent (theoretical) environmental variables to diatom growth and assemblage
distributions in a calibration dataset (ter Braak 1986; Smol 2002; Smol and Stoermer 2010). An
19

indirect gradient analysis is a linear, two-step approach involving ordination and environmental
gradient identification to reveal overall patterns in species variation within a dataset (ter Braak
1986; Smol 2002). Essentially, variations of diatom species and water chemistry variables are
explored independently using techniques such as detrended correspondence analysis (DCA) and
principal components analysis (PCA), respectively (Figure 2.1; Battarbee et al. 1999).

Figure 2.1: An example of a common ordination approach to transfer function
development, redundancy analysis (RDA), to assess the significance, strength, and
direction of environmental variables as predictors of diatom species composition
from a calibration dataset (modified from Juggins et al. 2013).
20

Conversely, direct gradient analysis allows both datasets to be analyzed together
(Battarbee et al. 1999). This form of multivariate regression identifies the best fitting
environmental variable to the diatom species data using a response model in which the
abundance of any species corresponds to the value of each latent environmental variable (ter
Braak 1986; Smol 2002; Smol and Stoermer 2010). For example, a canonical correspondence
analysis (CCA) is a commonly used direct gradient analysis technique developed by ter Braak in
1986. Diagrams generated by CCA use arrows of varying length and orientation to signify the
relative importance of the relationship between diatom community composition and the
measured environmental variables of the calibration lakes (ter Braak 1986; Battarbee et al. 1999;
Smol 2002).
The results of these techniques can provide valuable information for lake management,
however if the sediment samples corresponding to the diatom assemblages are not dated (either
with absolute or relative dating techniques), the inferred phosphorus concentrations could be
misinterpreted. For example, in a recent publication, Bachmann et al. (2013) estimated the
extent of trophic changes that have occurred from anthropogenic activities by examining diatominferred phosphorus concentrations for 240 lakes from the EPA’s National Lake Assessment
(NLA). They concluded that total phosphorus in these lakes decreased by 14% since EuroAmerican settlement. However, this determination was made under the assumption that the
bottom-core sediments represent a time prior to Euro-American settlement, when in fact; the
EPA only conducted a rudimentary, and somewhat subjective, evaluation of bottom-core sample
ages (USEPA 2010). Therefore, these results, which could have serious implications to nutrient
criteria establishment, are indefensible because of the unconfirmed time period represented by
the reconstructed phosphorus concentrations.
21

2.5.2 Pollen as a Proxy Indicator of Landscape Disturbances
In addition to diatom fossil interpretations, a considerable amount of information can be
gained from allochthonous materials in the sediment profile (Smol 2002). Pollen grains
deposited from the vegetational community surrounding a lake and preserved in lake sediments
offer a high-resolution archive of past vegetation (Smol 2002). The identification and
enumeration of grains to the genera or species level allows palynologists to reconstruct trends in
vegetation succession and/or document landscape changes associated with anthropogenic
activities. Moreover, an increase in Ambrosia artemisiifolia (common ragweed) pollen
percentages has been used as a chronostratigraphic marker of agriculture, development, and
deforestation in lake catchments (Reeder and Eisner 1994; Smol 2002; Gerber et al. 2011).
Ambrosia is a herbaceous plant native to temperate North America that quickly flourishes
in disturbed habitats such as roadsides, urban areas, and agricultural clearings (Smol 2002;
Gerber et al. 2011). In the Great Lakes region, several sediment core studies have found a
distinct peak in Ambrosia pollen corresponding to

14

C dates at 150 ± 50 years B.P., the

approximate time frame of European land clearance in the Midwestern United States and
adjacent Canada (McCarthy and McAndrews 1988; Weninger and McAndrews 1989; Reeder
and Eisner 1994; Finkelstein et al. 2005). For this reason, Ambrosia pollen can serve as a
paleoenvironmental proxy in place of written analogs of landscape changes during EuroAmerican settlement.
2.6 Lake-Landscape Interactions
Numerous studies have shown that the allochthonous materials in a lake are reflective of
the land use/land cover composition in the local catchment. Proportions of impervious surface,
22

agriculture, and lake shore development are some of the most widely used predictors of
phosphorus concentrations because these are the anthropogenic land uses known to strongly
influence water chemistry (Wang et al. 1997; Paul and Meyer 2001). Although land cover is
often used as a direct linkage to water quality, not all of the variation in lake phosphorus
concentration can be explained by these effects alone (Bremigan and Soranno 2008). The
chemistry and biology of inland lakes can also be naturally influenced by hydrogeomorphic
(HGM) features such as lake morphometry, geology, and hydrologic flowpaths (Bremigan and
Soranno 2008; Martin et al. 2011).
2.6.1 Hydrogeomorphic Features
HGM features are natural aspects of the landscape, minimally impacted by anthropogenic
activities, and can therefore be used as a foundation upon which to measure changes in the
environment (Bremigan and Soranno 2008; Martin et al. 2011). Spatial variations of these
physio-chemical landscape factors are hypothesized to influence the rate at which phosphorus
enters a lacustrine environment, at both local and regional scales (Soranno et al. 1996;
Schnurrenberger 2002; Veith et al. 2003; Wagner et al. 2011). For example, phosphorus
mobilization processes at a regional scale are largely affected by climatic and geologic factors
such as precipitation, lake alkalinity, and soil chemistry (Wagner et al. 2011). However, in localscale models, primary productivity has been predicted by basin morphometry (i.e., lake depth
and surface area) that influence water residence time and internal loading in lentic (still) surface
waters (Wetzel 2001; Kalff 2002; Wagner et al. 2011).
Water residence time (WRT), a function of surface hydrology and lake depth, is defined
as the average time required to drain a basin from the outflow if the inflow was removed (Wetzel

23

2001). This metric is often used as a rough measure of how quickly water quality and planktonic
communities will respond to changes in phosphorus loading (Wetzel 2001; Kalff 2002).
Generally, if surface runoff from the local catchment is rich in nutrients, organic matter, and/or
pollutants, lakes with a shorter WRT will have greater water clarity (Reavie et al. 1995).
Lake depth is another important HGM feature that can influence biological and chemical
parameters through the regulation of solar energy distribution, thermal stratification, and lake
circulation patterns (Kalff 2002; Wetzel 2001). For example, summer thermal stratification of
deep lakes (> 10 m) in temperate regions can isolate the epilimnion from substrate interaction
and prevent the resuspension of sediments in the water column (Hutchinson 1973; Scheffer
1998). Deep lakes generally maintain thermal stratification following periods of fall and spring
turnover because the metalimnetic thermocline acts as an effective barrier to convection currents
and wind-induced circulation (Wetzel 2001; Kalf 2002). Without mixing, the cold temperatures
(~4C) of the hypolimnia promote oxygen solubility and phosphorus retention in benthic
sediments (Kalf 2002). Anoxic hypolimnia most commonly occur when lakes are shallow ( < 3
m) and have a WRT of < 2 years (Kalf 2002).
Shallow lakes have a higher ratio of sediment surface to water column, resulting in
increased susceptibility to internal nutrient loading via resuspension of sediments (Scheffer
1998; Dong et al. 2011). Håkanson (1996), for example, found a significant positive relationship
between lake depth and Secchi depth because deeper lakes had less resuspension of materials in
the water column. Frequent mixing by wind and wave action is more likely to occur in lakes that
are shallow relative to their surface area (Osgood 1988; Detenbeck et al. 1993). This
replenishment of phosphorus into the water column has, in many cases, delayed the reduction of
in-lake phosphorus concentrations (Jeppesen 2005). Furthermore, Vollenweider (1968) and
24

others have demonstrated that consideration of total phosphorus loading and mean depth alone
can be a fairly accurate metric to predict lake eutrophication (Volleinweider 1968; Dillon and
Kirchner 1975; Chapra 1997). As shown in Figure 2.2, shallow lakes become eutrophic at lower
rates of phosphorus loading, in comparison to deep lakes.

-2 -1

PHOSPHORUS LOAD (gPm yr )

10

EUTROPHIC

MESOTROPHIC

OLIGOTROPHIC

0.1
1

10

100

1000

MEAN DEPTH (m)
Figure 2.2: Vollenweider’s (1968) phosphorus loading plot classifies lakes into
-2 -1
trophic states based on mean depth (m) and annual phosphorus loads (gPm yr ).
The lines marking the thresholds between eutrophic and oligotrophic lake
conditions are suggested to correspond to phosphorus concentrations of 20 μg/L
and 10 μg/L, respectively (modified from Chapra 1997).

25

2.6.2 Phosphorus Dynamics in Wetland Ecosystems
Wetlands are transitional ecosystems commonly located at the interface between
terrestrial and aquatic ecosystems. They are defined by prolonged periods of saturation that
result in the establishment of organic or mineral hydric soils and hydrophytic vegetation (Tiner
1999; Mitsch and Gosselink 2007). Often gleying and redoximorphic features (mottling) of
hydric soils reflect the poor drainage and reductive conditions common to wetlands (Etherington
1983; Tiner 1999; Mitsch and Gosselink 2007). Due to their strategic landscape location,
wetlands have received considerable attention for their potential to sequester both sediments and
nutrients (Tiner 1997). In fact, among the many valued ecosystem services provided by wetlands
(i.e., waterfowl habitat, flood mitigation, and groundwater recharge), perhaps none are as highly
regarded as phosphorus retention (Mitsch and Gosselink 2007; Stein et al. 2010).
Numerous studies have demonstrated the ability of wetlands to reduce water quality
impairments related to excess nutrient loading (Johnston 1991; Reddy et al. 1999). In fact, much
of the published research within the last two decades has largely focused on the use of
constructed wetlands as a nutrient sink in the wastewater treatment process. For example, one
study found total phosphorus removal in a constructed wetland to vary between 40 to 60%,
depending on inflow loading (Vymazal 2007).
Although wetlands are hypothesized to be natural nutrient sinks that filter out
contaminants and improve water quality, some studies show the amount of total phosphorus lost
from wetlands to lakes or rivers is significantly greater than that from terrestrial systems
(Richardson 1985; Devito et al. 2000; Richardson and Vaithiyanathan 2009). For example,
Richardson (1985) suggested phosphorus retention efficacy may be overestimated because the

26

biogeochemical mechanisms controlling phosphorus transformation and translocation in
wetlands remain poorly understood.
For example, emergent macrophytes in wetland systems can effectively assimilate and
store phosphorus during the growing season, and even during the fall/winter months, albeit with
decreased uptake rates (Reddy and Flaig 1995). However, a relatively large amount of dissolved
phosphorus is returned back into the water column as a result of initial leaching or detrital
decomposition (Reddy and Flaig 1995; Reddy et al. 1999). Additionally, accumulations of
phosphorus in poorly and very poorly drained soils with anaerobic conditions have a high
potential to become remobilized (Benhart 1994; Carpenter et al. 1998). Therefore, phosphorus
storage in wetland biomass is usually short-term, particularly when compared to the long-term
storage ability of upland terrestrial ecosystems (Reddy et al. 1999).
2.6.3 Phosphorus Yields from Forested Ecosystems
The distribution and abundance of tree species reflects broad spatial patterns in soil
texture and climate (Harmon 2009). Generally, sandy-textured soils in cold climate regimes (i.e.,
Spodosols) are well-tolerated by coniferous trees, which produce a thick, acidic O horizon that
fosters slow decomposition rates (Barnes and Wagner 1981; Harmon 2009). The eastern
deciduous forest tree species that dominate southern Lower Michigan’s forest composition, such
as American beech, red maple, and tulip poplar, tend to require sandy loam soils and longer
growing seasons (Barnes and Wagner 1981; Yahner 2000; Harmon 2009). Regardless of the
forest classification, the relatively coarse-textured forest soils enhance precipitation and
snowmelt infiltration, resulting in minimal surface runoff and nutrient contributions from upland
ecosystems (Fisher and Binkley 2000; Miller et al. 2005). Therefore, forested buffers, narrow

27

bands of arboreal vegetation surrounding lakes and streams that reduce inputs of solar radiation
(by shading), sediments, and nutrients, are one of the strategies to mitigate phosphorus runoff
from agricultural and urban areas (Osborne and Wiley1998; Devito et al. 2000; Baker 2006).
Unsurprisingly, the rates of dissolved and particulate phosphorus retention from upland
ecosystems to aquatic ecosystems have been shown to decrease following tree harvesting or
other periods of rapid landscape denudation (Omernik et al. 1981; Filippelli and Souch 1999;
Reynolds and Davies 2001). This is, in part, due to the presence of long-chained, hydrophobic,
organic molecules released from decaying or incinerated detritus and micro-organisms that can
reduce the infiltration capacity of soils and increase runoff into surface waters (Reynolds and
Davies 2001). Although the hydrophobicity of soils is spatially variable, it is most prominent
after prolonged dry periods or episodes of intense wildfires (Doerr and Shakesby 2011). When
hydrophobicity is low, densely forested areas with non-hydrophobic soils appear to retain
phosphorus through biomass accumulation and sorption to clay sediments (Reynolds and Davies
2001; Kramer et al. 2006).
2.6.4 Proximity to Surface Waters
Simple measures of watershed composition are often evaluated to explain variability in
water quality parameters, however there are uncertainties regarding the appropriate scale for
these analyses (Soranno et al. 1996; Cifaldi 2002; Gémesi et al. 2011). Although they are less
considered and more difficult to quantify, empirical evidence suggests that knowledge of spatial
variations in land cover could more accurately explain source contributions of phosphorus
loadings (Alexander 2002; Gémesi et al. 2011). Therefore, spatially explicit measures of
landscape configuration and proximity are becoming more common in models (Gergel and

28

Turner 2002; Fraterrigo and Downing 2008; Gémesi et al. 2011). An evaluation of these
variables in Michigan will help to achieve a greater understanding of lake-landscape interactions
as a whole.

29

3. STUDY AREA
3.1 Study Area Description
The 48 inland lakes in this study are geographically distributed throughout Michigan.
Given the expansive extent of the study area across several ecoregions, it is necessary to first
discuss the spatial diversity of landscape patterns that are largely a factor of climatic controls and
glacial history. I will then address the process of study lakes selection and the broad
characteristics of the dataset.
3.1.1 Climate and Vegetation
Michigan is geographically situated between 41.5 N to 48.5 N, the approximate
halfway point between the equator and the North Pole. Given this large range in latitude, as well
as Michigan’s proximity to the Great Lakes, temperature and precipitation fluctuate within the
state (Andresen and Winkler 2009). The annual mean temperature, based on 1971-2000 climate
data, ranges from approximately 45 F in the Lower Peninsula to approximately 40 F in the
Upper Peninsula (from 1971-2000) (NOAA National Climate Data Center 2002). Colder
temperatures prevail in more northerly regions, resulting in a shorter growing season.
Historically, annual mean temperatures in the Midwest were approximately 6.6 cooler in 1900
than modern recorded temperatures, according to data obtained from the CRUMET3 data set
(Brohan et al. 2006; Andresen et al. 2012).
Total annual precipitation tends to decrease from west to east as proximity to prevailing
westerly winds over the Great Lakes decreases (Andresen and Winkler 2009). This gradient of
moisture across Michigan, known as the “lake effect,” enhances winter snowfall and moderates
temperatures near the eastern shore of Lake Michigan and the southeast shore of Lake Superior.
30

The result is a mesic and mild microclimate within 80 km of these Great Lakes that can support
temperate hardwood tree species, such as red oak and sugar maple during winters in the
otherwise frigid Upper Peninsula (Kenoyer 1933; Harmon 2009).
The study region lies within the northernmost extent of the Eastern Deciduous Forest
biome and near the southernmost extent of the Canadian Boreal Forest biome (Figure 3.1;
Harman 2009). Climate is the broad-scale influence of predominant vegetation in Michigan,
which shifts from eastern broadleaf forest to mixed coniferous-deciduous forest along a
transitional area in the central-Lower Peninsula referred to as the forest tension zone (Figure 3.1;
Medley and Harmon 1998; Anderson 2005; Harmon 2009; Hupy and Yansa 2009a, 2009b).

Figure 3.1: Map of pre-settlement forests in Michigan as interpreted from
the General Land Office (GLO) public land surveys between 1816-1856
(Comer et al. 1995).
31

The tension zone is an ecotonal area that marks a general climatic boundary between
cooler and warmer mean annual temperatures and the decreased number of growing degree days
as latitude increases and the influence of the polar jet stream becoming more predominant
(Eichenlaub et al. 1990; Andresen and Winkler 2009; Hupy and Yansa 2009a). Localized
patterns in vegetation, however, are strongly related to variations in soil types resulting from
repeated glaciations (Blewett et al. 2009; Andresen et al. 2012).
3.1.2 Glacial History
Thick glacial deposits and complex sedimentology and landforms are reflective of
advancing and retreating continental ice that covered Michigan during the Late Wisconsin
glaciation (Wetzel 2001; Blewett et al. 2009). The Laurentide ice sheet began to recede from the
southernmost part of the Lower Peninsula of Michigan about 17,000 cal (calendar) years ago and
had vacated the northern Upper Peninsula by 10,000 cal yr BP (Blewett et al. 2009). Glacial
sediments vary in thickness across the state, ranging from 365 m in the interlobate region of
Osceola County (Lower Peninsula) to none at exposed bedrock in the driftless regions of the
Upper Peninsula (Blewett et al. 2009; Schaetzl 2009). Similarly, soil textures fluctuate across
the state in close reflection to the glacial parent material (e.g., till, outwash, or loess) in which
they formed (Schaetzl 2009).
These variations in soil textures and drainage classes are known to support structurally
and functionally different vegetation communities (Henne 2006). Furthermore, the interaction of
spatial edaphic variations with climatic gradients (latitudinal, as well as lake-affected areas vs.
inland locales) creates a mosaic of different forest community types in Michigan (Albert 1995,
Henne 2006). For example, coarse-textured sandy soils south of the forest tension zone mostly

32

support broadleaved forests, but coarse-textured sandy soils north of the tension zone are
predominately associated with coniferous needleleaf forests (Harmon 2009; Schaetzl 2009).
3.1.3 Hydrology
The last glacial recession not only shaped Michigan’s soil and vegetation patterns, but
also left a mosaic of interacting lakes, rivers, and wetlands (Figure 3.2). Kettle lakes commonly
formed in interlobate areas and outwash plains when chunks of ice detached from a retreating

Figure 3.2: Map of pre-settlement lakes, rivers, and wetlands in Michigan as interpreted
from the General Land Office (GLO) public land surveys between 1816-1856
(Comer et al. 1995).

33

glacial lobe became buried in outwash and eventually melted to form depressions which filled
with water. These lakes are discernible by deep basins with irregular bathymetry and are often
geographically isolated from streams (Wetzel 2001; Tiner 2003; Wolfson 2009). Many of the
shallower kettle lakes eventually filled with plant debris and sediments during periods of
stabilizing drainage patterns and falling lake levels, becoming depressional wetlands (Tiner
2003).
Some lakes and wetlands in Michigan have other origins. Oxbow lakes, for example,
form when alluvial deposits cutoff river meander bends. Additionally, karst lakes are known to
form where caverns in highly soluble carbonate bedrock have collapsed to form water-filled
depressions, a phenomenon common to the eastern Upper Peninsula and northern Lower
Peninsula (Wolfson 2009). The abundance of surface water in Michigan makes it an ideal region
for land cover-water quality studies, however much of these natural resources have been greatly
altered since Euro-American settlement in the mid-1800s.
3.1.4 Euro-American Settlement
Michigan’s native landscape, as interpreted from the General Land Office (GLO) surveys
of 1816-1856 (Figure 3.3), consisted of approximately 29% mixed forest, 27% deciduous forest,
15% evergreen forest, 23% wetlands, 3% grasslands, 2% lakes and streams, and ~ 1% barren
land (Comer et al. 1995). The impetus of the surveys was to facilitate the sale of available land
to the general public by defining the location and description of parcels in a systematic manner
(Thomas 2009). Initial observations from deputy surveyors often reported the land as “poor and
brushy” or “swampy and tedious to survey” (Thomas 2009). Regardless of this inhospitable
impression, pioneer farmers were attracted to Southern Michigan for the fertile prairie soils and
potentially arable land below the hardwood forests (Lewis 2009).
34

Figure 3.3: Map of pre-settlement land cover in Michigan as interpreted from
the General Land Office (GLO) public land surveys between 1816-1856 (Comer
et al. 1995).
By 1860, the landscape was drastically altered by Euro-American colonization that
culminated in widespread deforestation, wetland drainage, and agricultural expansion. For
example, industrial lumbering of the northern Lower Peninsula and the Upper Peninsula’s
expansive coniferous forests began servicing the rapid expansion of cities such as Chicago by
1840 (Lewis 2009). Farther south, much of the original oak-openings/prairie lands in
southeastern Lower Michigan were converted to oak forests through fire suppression practices
(Albert 1995; Comer et al. 1995). Statewide, an estimated 50% of the native wetlands were
ultimately drained and converted to agricultural land (Albert 1995; Comer et al. 1995; Lewis
2009). An understanding of the natural variability of pre-settlement vegetation, landforms, soils
35

and hydrology was a critical requirement for selecting lakes in this study, in order to capture the
spatial variability of these natural features that have since been anthropogenically altered, to
varying degrees.
3.2 Study Lake Selection and Characteristics
In this thesis research, lacustrine sediment core data and water chemistry measurements
were compiled from two broader sources: the EPA’s 2007 National Lake Assessment (NLA) and
an associated project managed by Dr. Patricia Soranno of Michigan State University (MSU) and
Paul Garrison of the Wisconsin Department of Natural Resources (WDNR) in 2007. Diatom
assemblages from bottom-core sediments of 48 lakes were used to reconstruct phosphorus
concentrations, theoretically representative of undisturbed lake conditions. Verification of the
‘pre-settlement age’ of the bottom ~2 cm of core sediments (maximum depth of core bottoms
ranged from 23 to 69 cm from the sediment-water interface) was essential to this research;
therefore initial lake selection from the NLA and Soranno/Garrison datasets was based on access
to these sediment samples and by a pollen method employed to make this determination, which
is detailed in Chapter 4 of this thesis.
3.2.1 Lake Datasets
The EPA’s 2007 NLA sampled sediment diatoms and physiochemical parameters of
1,028 lakes throughout the contiguous United States using a randomized, probability-based site
selection method (USEPA 2007a, 2009; Stevenson et al. 2013). This statistical sampling
approach was leveraged to randomly select study lakes from the USGS National Hydrography
Dataset (NHD) if they met a set of predetermined criteria. The main factors in selection were
size, depth, and equal geographic representation of lakes throughout the United States. More
36

specifically, freshwater lakes or reservoirs were required to be > 4 ha in size, >1m deep, and
spatially distributed across the lower 48 states (USEPA 2007a, 2009). Accessibility and safety
were also taken into consideration. The final compilation of target lakes included seepage lakes,
drainage lakes, and even constructed impoundments (USEPA 2009).
Fifty-four Michigan lakes were sampled as part of the 2007 NLA, however diatominferred phosphorus analyses of sediment core-bottoms were not performed for twelve of these
lakes. For the remaining forty-two lakes, a set of criteria was used by the EPA to establish
confidence in core-bottoms representing pre-disturbance conditions (USEPA 2010). In general,
long cores and/or natural lakes with presumably low sedimentation rates were awarded a high
reliability ranking (USEPA 2010). Based on these criteria, the lab results from six lakes were
excluded from this thesis study because the sediment cores yielded uncertain data or the lake was
a constructed impoundment. Lastly, two of the lakes were duplicated in the Soranno/Garrison
dataset and therefore not included in my initial dataset that was subjected to pollen analysis
(Appendix A). Therefore, a total of thirty-four lakes from the EPA’s 2007 NLA were considered
for potential use in this study.
The lacustrine sediment samples and water chemistry data collected in 2007 by Mr. Paul
Garrison and Dr. Patricia Soranno came from 31 inland lakes located throughout Michigan.
Sample sites were originally chosen to represent a range of catchment and lake morphometry
values and hydrologic settings. Similar to the NLA dataset, lakes are each at least 1 m deep, a
minimum of 4 ha in area, and they represent both seepage and drainage hydrological settings (P.
Soranno pers. comm. 2012).

37

Of the 31 lakes in the Soranno/Garrison dataset, two of the lake sediment samples were
not available to test for relative age and were therefore excluded from this study (Appendix A).
The remaining 29 lakes from the Soranno/Garrison dataset were analyzed to verify presettlement conditions using Ambrosia pollen as a relative dating method (described in Chapter
4). Based on the palynological dating method used to verify pre-settlement representation of
core-bottoms, diatom-inferred phosphorus and hydrogeomorphic data from a total of 48 study
lakes (compiled from the 2007 NLA and Soranno/Garrison data sources) were used in the final
dataset that is the basis of this thesis research (Table 3.1).

Table 3.1: Study area information for the 48 lakes in the final dataset.
Lake Name

County

Coldwater
Bartholomew
Clark
Campbell
Eagle
Pleasant
Warner
Chemung
Mill
Campau
Murray
Stoner
Pine
Whitefish
Blue
Big Star
Eight Point
Big
Cadillac
Mitchell
Bear
Au Sable

Branch
Branch
Jackson
Kalamazoo
Kalamazoo
Jackson
Barry
Livingston
Oakland
Kent
Kent
Kent
Kent
Montcalm
Mecosta
Lake
Clare
Osceola
Wexford
Wexford
Manistee
Ogemaw

Latitude Longitude

41.83
41.88
42.12
42.32
42.33
42.39
42.47
42.58
42.74
42.83
43.03
43.15
43.22
43.33
43.62
43.83
43.84
43.87
44.24
44.24
44.43
44.43

-84.98
-84.93
-84.32
-85.47
-85.32
-84.35
-85.53
-83.85
-83.31
-85.45
-85.38
-85.48
-85.47
-85.54
-85.28
-85.95
-85.07
-85.20
-85.43
-85.47
-86.16
-83.92
38

North/South
of FTZ*

South
South
South
South
South
South
South
South
South
South
South
South
South
South
North
North
North
North
North
North
North
North

Max
Lake
Depth
(m)
28.05
17.07
15.50
11.10
9.45
15.85
11.20
21.10
9.00
15.24
19.51
1.50
6.80
16.46
15.00
7.62
7.70
25.91
8.23
6.71
7.20
15.00

Surface
Area (ha)

647.77
35.30
232.86
59.59
29.88
114.14
16.02
120.09
11.22
38.29
140.62
30.02
20.59
206.77
88.65
365.26
159.99
86.15
481.65
1089.16
768.20
105.58

Table 3.1 (cont’d)
Big Blue
Kalkaska
Big Twin
Kalkaska
Starvation
Kalkaska
West Twin
Montmorency
Hanley
Antrim
Deer
Charlevoix
Blomgren
Dickinson
McDonald
Delta
Ottawa
Iron
Duck
Gogebic
Imp
Gogebic
Lotto
Marquette
Little
Marquette
Witch
Marquette
Dewey
Marquette
Bobcat
Gogebic
St. Kathryn
Iron
Howe
Alger
Fence
Baraga
Teal
Marquette
Gogebic
Gogebic
Keewaydin
Baraga
Craig
Baraga
Hall
Houghton
Mud
Houghton
Bailey
Keweenaw
*FTZ = Forest Tension Zone

44.81
44.82
44.85
44.88
45.08
45.16
45.87
46.04
46.08
46.20
46.22
46.26
46.27
46.28
46.30
46.36
46.38
46.43
46.48
46.51
46.47
46.60
46.62
46.99
47.13
47.46

-84.90
-84.96
-84.95
-84.35
-85.26
-84.98
-87.80
-86.81
-88.77
-89.22
-89.08
-88.09
-87.36
-88.01
-87.78
-89.67
-88.72
-87.08
-88.19
-87.62
-89.58
-88.12
-88.18
-88.76
-88.32
-88.10

North
North
North
North
North
North
North
North
North
North
North
North
North
North
North
North
North
North
North
North
North
North
North
North
North
North

27.40
24.38
14.33
8.53
8.23
6.71
1.80
2.20
27.43
7.60
26.21
5.00
15.24
28.96
2.50
5.49
9.10
1.00
16.00
9.75
7.00
7.00
7.62
14.00
3.20
1.00

44.24
85.20
79.18
535.13
37.44
197.84
31.73
11.82
217.42
261.42
38.82
43.39
187.16
87.57
20.65
38.37
65.05
36.23
93.91
202.60
5280.38
52.59
57.18
6.23
69.28
77.70

3.2.2 Lake and Native Vegetation Catchment Characteristics
Of the 48 lakes in the dataset, 14 are located south of the forest tension zone’s center
mark and 34 lakes are located north of this ecotone (Figure 3.4). Comparisons of the presettlement land cover in both the northern and southern local lake catchments are shown in
Figures 3.5 and 3.6. Unsurprisingly, deciduous forests comprise the majority (55%) of the presettlement land cover in local lake catchments south of the forest tension zone while mixed
coniferous-deciduous forests comprise the majority (68%) of the pre-settlement land cover in
39

local lake catchments north of the forest tension zone. The percentage of wetlands in the local
catchments of the northern and southern lakes is approximately equal.

Figure 3.4: Map of study lakes and the forest tension zone (after Hupy and Yansa
2009a). Fourteen lakes are located south of the forest tension zone’s center mark;
thirty four lakes are located north of the forest tension zone’s center mark.

40

41

4. METHODS
4.1 Lacustrine Core Sample Collection
As discussed in Chapter 3, the lake sediment samples analyzed in this thesis came from
two sources, however, the water data collection methods and analysis techniques performed on
these two sets of samples are comparable and of high quality. In both cases, the deepest part of
each lake was determined using bathymetric maps or a hand-held sonar unit and sampling
locations were recorded with a GPS device. The methods used to core the sediments of the
sampled lakes are also similar.
Sediment cores from the Soranno/Garrison dataset were extracted from each lake in
August and September 2007 using a gravity corer with a barrel that fits a 6.8-cm diameter plastic
tube that can be replaced to take multiple cores. The length of sediment cores varied between the
lakes sampled, ranging from 28 to 69 cm below the sediment-water interface. That same
summer, EPA personnel collected sediment cores from lakes included in the NLA using a
modified K-B gravity corer, a piece of equipment specially designed for collecting the topmost
lake sediments (Dixit et al. 1999; USEPA 2007a). The core lengths recovered by the NLA
project varied from 23 to 64.5 cm between sites in Michigan based on substrate permeability
(Dixit et al. 1999; USEPA 2007a).
Of the sediment cores taken from 63 lakes in Michigan by WDNR research scientists and
EPA personnel, only the topmost 1-2 cm section (“top,” capturing the sediment-water interface)
and the bottommost 1-3 cm (“bottom”) section of each sediment core were kept and analyzed the
rest of the core was discarded. The “top” of each sediment core, thus, is interpreted as modern
time (the last 2-3 years) and the “bottom” represents an earlier timeframe, the age of which is
undetermined until the sediments are dated using absolute or relative techniques. This “top42

bottom” approach to sediment sampling (Figure 4.1) can be used in paleolimnological studies as
an alternative to stratigraphic core analyses when the goal is to compare current and past water
quality parameters (Cumming et al. 1992; Dixit et al. 1999; Smol 2002; USEPA 2010).
Although a centimeter-by-centimeter analysis of the entire sediment core can provide more
detailed information for temporal trends in lake water conditions, these studies are often costly
and time-consuming, and therefore unrealistic for large-scale studies with many lakes (Smol
2002). Rather, sediment samples from two discrete points in time, before (bottom) and after
(top) major human impact, provide enough information to determine the degree of change, as
exemplified by the results obtained in this thesis.

depth

Sediment-water
interface interval

Regular interval
sampling every 2 cm

Bottom-core
sediment interval

Figure 4.1: A “top-bottom” approach to core sampling (left) yields two
sediment samples, representative of pre-settlement and modern conditions.
The numerous sediment intervals from “stratigraphic” sampling (right)
provide a continuous analog of lake conditions (modified from USEPA
2010).

In addition to the sediment core collection, both the Soranno/Garrison and the EPA
projects collected water column chemistry measurements at the deepest part of each lake to be
used in the development of calibration data sets. Specifically, the physical data obtained
43

included water depth, temperature, and transparency, which was measured with a Secchi disk.
The collected chemical/biological parameters included total phosphorus, total nitrogen,
alkalinity, specific conductance, chlorophyll, and pH.
4.2 Diatom-Inferred Phosphorus Modeling
4.2.1 Soranno/Garrison Methods
The top and bottom intervals of sediment cores collected from 31 lakes in the
Soranno/Garrison dataset were kept chilled until delivery to a Wisconsin Department of Natural
Resources (WDNR) laboratory in Monona, Wisconsin, at which point diatom analyses were
conducted under the direction of Paul Garrison. Samples were placed into tall beakers and
cleaned with strong oxidizing agents (i.e., hydrogen peroxide and potassium dichromate) to
remove organic matter (van der Werff 1956; Garrison and LaLiberte 2010). The cleaned
samples were dried and mounted onto glass slides to be counted under a microscope with an oil
immersion objective (1400X) until a total of 400 diatom frustules were identified and
enumerated (Garrison and LaLiberte 2010).
Water column chemistry and modern diatom assemblages in the uppermost 1-2 cm of the
lacustrine cores from 31 lakes in Michigan and 152 lakes in Wisconsin were collated to create a
surface-sediment calibration set. Relationships between modern diatom assemblages and
limnological variables in the calibration lakes were examined using a canonical correspondence
analysis (CCA) in the statistical program CANOCO (ter Braak and Šmilauer 1998). This
multivariate statistical technique is appropriate when diatom species have a unimodal response
curve with respect to environmental gradients (ter Braak 1986; Birks 1998). The tested
limnological variables included lake surface area, maximum depth, mean depth, alkalinity,
44

summer mean phosphorus, summer mean nitrogen, summer mean Secchi depth, and water color
(Garrison pers. comm. 2012).
4.2.2 EPA Methods
The top and bottom sections of the EPA sediment cores were stored on ice until delivery
to an algal ecology laboratory located in the Department of Zoology at Michigan State
University in East Lansing, Michigan, where diatom analyses were performed by Dr. Jan
Stevenson and his assistants. They prepared samples for analysis following a standard digestion
procedure, and transferred the prepared material to glass microscope slides (USEPA 2007b). A
minimum of 500-600 valves per sample were identified and enumerated by diatom taxonomists
under research quality microscopes at 1000X magnification. The relative abundances of all taxa
were calculated in order to reconstruct changes in the study lakes (USEPA 2007b; Stevenson et
al. 2013).
A calibration set was developed by Stevenson and colleagues using modern diatom
assemblages from the uppermost core sediments and the chemical variables measured in the
water column of each lake (USEPA 2010; Stevenson 2013). The calibration set was examined
with a detrended correspondence analysis (DCA) to determine if a linear-based or an unimodalbased ordination was appropriate for the environmental gradient based on the observed diatom
species data (Smol 2002; USEPA 2010). Unimodal statistical methods were selected because the
length of the environmental gradient was > 2.5 standard deviation units (SD) (USEPA 2010). A
gradient length less than 2.5 SD would have indicated a relatively narrow range of variation in
the diatom assemblage for which a linear, principal components analysis (PCA) approach would
have been appropriate.

45

4.3 Relative “Dating” of Sediment Using Pollen
4.3.1 Background

Stratigraphic levels of a sediment core are commonly dated using radioisotopes
and

210

Pb

137

Cs in site-specific reconstructions of limnological changes (e.g., Garrison and Wakeman

2000; Umbanhowar et al. 2003; Moser et al. 2010). In fact, producing

210

Pb and

137

Cs curves

and fitting them to the known dates of events (i.e., banning of leaded gasoline beginning in 1973
and atomic bomb detonations in the 1950s and 1960s) is the only method able to determine an
“absolute date” of historic-age sediments. Radiocarbon dating is not reliable, because the
youngest age for this method is ~ 200 cal yr BP, due to the longer half-life of this isotope (Bjӧrk.
and Wohlfarth 2001). However, analyzing sediment profiles for

210

Pb and

137

Cs in regional

studies involving many lakes is not feasible as these are very costly dating procedures
(Umbanhowar et al. 2003). Moreover, this technique requires the analysis of

210

Pb and other

isotopes from many stratigraphic levels per core to assign ages for these levels, which is
impossible in this study given that only top and bottom samples were retained.
Although sediment samples in this study were not dated using radioisotopes, I employed
a pollen-based, relative dating method identical to those used elsewhere in paleoenvironmental
studies. Fossil pollen preserved in lake sediments are useful proxy indicators of dates relative to
known historical events that altered land cover composition (e.g., Euro-American settlement) in
temperate watersheds of North America (McAndrews 1988; Blais et al. 1995; Dixit et al. 1999).
Events such as land clearance and agricultural expansion, in particular, are correlated with a
distinct “rise” in Ambrosia (ragweed) pollen (Finkelstein et al. 2005; Jaques et al. 2008). Most
46

studies that use this technique compare continuous, stratigraphic levels of sediment core pollen
assemblages to find the interval of abrupt Ambrosia increase, that has shown to correspond to
210

Pb dates of 150±50 yr BP (Reeder and Eisner 1994). Such studies from the Midwestern U.S.

have documented Ambrosia percentages to increase from around 5% (prior to wide-spread
disturbances) to around 18% at the time of settlement (Grimm 1983). This relative dating is
sufficient in comparative limnological studies where only top and bottom sediment core intervals
are examined. However, relative dating methods cannot replace “absolute” dating techniques
that provide calibrated ages for studies that require high resolution, detailed reconstructions.
Therefore, I analyzed pollen grains preserved in the bottom 1-3 cm interval of each lake
core to verify that it has conservatively low (1-3%) values of Ambrosia pollen, which would be
expected for sediments dating before Euro-American land clearance and agriculture (Grimm
1983). Grimm (1983) had a higher cutoff value (5%), however, his site was farther west at the
forest-prairie boundary in Minnesota, a region with naturally higher Ambrosia percentages.
Therefore, I hypothesized that if any bottom-core sediment from the sampled lakes contained
>3% Ambrosia pollen values, then that sample did not “date” to pre-settlement times, but rather
came from sediments deposited after Euro-American settlement and would be excluded from the
statistical analysis of diatom-inferred phosphorus data.
4.3.2 Laboratory Methods
Pollen analysis of 63 lake core-bottoms (34 from the EPA dataset and 29 from the
Soranno/Garrison dataset) was conducted at Michigan State University’s Palynology Laboratory
to determine the abundance of Ambrosia pollen. A multi-step sample preparation process was
used that concentrated pollen grains through chemical and physical removal of extraneous matter
47

(Faegri and Iverson 1975). Centrifuging in swinging buckets at 2900 rpm concentrated the
samples after each step.
3

First, a consistent volume (approximately 1cm ) of each sediment sample was treated
with 10% hydrochloric acid (HCl) to dissolve calcium carbonates. This step was repeated as
necessary for reactive, marl sediments until carbonates were effervescently removed. Next,
sodium pyrophosphate (Na4P2O7) was added and the test tubes were placed in a hot bath to
suspend clay particles and prevent coagulation during centrifuging. Samples were then sieved
through 125 μm metal screens to remove coarse inorganic and organic particles from the fine
particles that include pollen grains. In the next step, samples were treated with 10% KOH and
placed in a hot bath to remove humic acids. After the KOH was decanted, the samples were
made acidic through a 10% HCl rinse and chemically treated with 48% hydrofluoric acid (HF) in
a hot water bath to dissolve siliceous matter. The HF treatment was followed immediately by
another 10% HCl rinse to remove the dissolved silicates (i.e., colloidal silicates and
silicoflourides). Samples were then dehydrated with glacial acetic acid (C2H4O2) before adding
an acetolysis mixture of concentrated sulfuric acid (H2SO4) and acetic anhydrite ((CH3CO)2O).
During acetolysis, samples were boiled for 1.5 - 2 minutes to clean pollen exines of cellulose and
fine organic residue. This reaction was immediately neutralized with glacial acetic acid. Finally,
the samples were dehydrated using ethyl alcohol (ETOH) and tertiary-butanol ((CH3)3COH),
and then transferred to a 1-dram vial along with non-volatile silicon oil.
The concentrated pollen specimens were mounted on glass slides with additional silicone
oil and examined using a LEICA DMLB light microscope at 400x magnification. One hundred

48

pollen grains were enumerated per slide and classified into four categories: Ambrosia, deciduous
taxa, coniferous taxa, and maize (corn). Ambrosia was distinguishable from other tricolporate
(C3P3) pollen types by its short spines and small, oblate grains (Moore and Webb 1978).
Abundance was expressed as a percentage of the total pollen count per sample.
4.4 Spatial Data Processing
Geographic Information Systems (GIS) is an integrated platform for spatial analysis and
modeling that provides a framework for the visualization and management of geographic data
(Maguire et al. 2005). The emergence of GIS within the last few decades has revolutionized
natural resource management and monitoring of land use/land cover changes across spatial and
temporal scales, in part because it is well developed for simultaneously depicting discrete and
continuous fields (Turner et al. 2001; Gergel and Turner 2002; Maguire et al. 2005; Johnson and
Host 2010). For example, it is now possible to detect and quantitatively analyze vegetation
changes within a delineated area or measure the amount of runoff within urban boundaries using
a digital mapping program. Furthermore, high quality geospatial data, such as digital elevation
models (DEMs), surface hydrology networks, and mosaicked digital raster graphs (DRGs), are
now widely available from government agencies, free of charge.
The native land cover, local lake catchment, and inland lake data used in this analysis
were acquired from the Michigan Center for Geographic Information (CGI) and the MSU
Fisheries and Wildlife Department. Spatial analyses of land cover data were performed in
ArcGIS 10 software using basic analytical tools (ESRI 2011). All spatial data were defined and
projected into a common coordinate system, North_American_1983 Hotine Oblique Mercator.

49

4.4.1 Acquisition and Description of Data
A flow-chart of the data acquired and analysis performed in GIS is shown in Figure 4.2.
First, I imported a vector shapefile of the Michigan’s native landscape circa mid-1800s into
ArcMap. This file contained pre-settlement vegetation details such as the species composition of
forests and the location of wetlands in Michigan, as documented in the original General Land
Office (GLO) survey notes of 1816-1856 (Comer et al. 1995; Albert et al. 2008). The Michigan
Natural Features Inventory (MNFI) earlier had translated these field notes alongside expert
interpretation of soils, climate, geology, and landforms into a digital map of over 74,000
polygons with attribute data such as the location, area, and type of native land cover in Michigan
(Comer et al. 1995; Kost et al. 2007; Albert et al. 2008). Polygon land cover attributes were
classified into 31 unique categories based on the dominant vegetative community or the type of
barren land. For simplification purposes, I reclassified these categories into seven groupings to
reflect Anderson’s Level I and II framework of a national land use and land cover classification
system (Anderson et al. 1976).

50

Figure 4.2: Flow chart of GIS methods. A dotted line around a box indicates a task
performed, whereas a solid line identifies a coverage or product.

51

Next, I imported a vector shapefile of Michigan inland lake boundaries and selected the
48 lakes represented in my dataset from the attribute table. Geographic coordinates were used as
supplementary identifiers because the abundance of lakes in Michigan has generated many repeat
names. The selected lakes were then exported using ‘Data Export’ to create a new shapefile
containing only the study lakes. Three study lakes were not included in the original shapefile
were digitized based on existing watershed delineations and DRG contour lines and merged with
the new study lakes shapefile. I then edited the lake polygon vertices, if necessary, to reflect the
hydrologic boundaries in the native land cover layer; this was to account for perimeter changes
that may have occurred since pre-settlement.
4.4.2 Data Analysis
Pre-settlement land cover data surrounding each study lake was quantified in the context
of a 100-m buffer and the local lake catchment to be used in a spatial comparison of the
influence of land cover on lake phosphorus at different scales. First, I used buffer analysis to
delineate a shoreline corridor at a fixed-width of 100 m around each lake polygon boundary. I
then clipped the land cover data with the 100 m buffers. The attribute data were extracted from
the clipped land cover data and percentages of land cover classes were calculated by dividing the
area of each cover type by the total land cover area in the 100 m buffer (excluding the lake
itself).
Next, I imported a shapefile containing local lake catchment boundaries for inland lakes
in Michigan. A lake catchment encompasses the area on the landscape that drains to the focal
lake, including land that drains into inflowing streams. These catchment boundaries were
delineated in GIS by creating or readjusting nodes of existing line features compiled from the

52

USGS DRG, MDNR 24k watershed boundary lines, and the Institute for Fisheries Research
(IFR) 100k lake catchment boundary lines. Similar to the 100 m buffer extent, I clipped the land
cover with the local catchment delineations for each lake. Finally, I imported these data into an
Excel spreadsheet and calculated the proportions of total land cover for each study lake.
4.5 Statistical Data Analysis
Quantifying the relationship between spatial landscape patterns and key biophysical
drivers is often of primary interest in ecological research (Wiens 2002). Multiple regression
analysis is a multivariate statistical technique commonly applied in these investigations of
complex ecological system dynamics because it allows for the determination of a single
dependent response variable against several explanatory variables (Clark and Hosking 1986;
Devito et al. 2000; Webster et al. 2008). Furthermore, multiscale analyses based on multiple
regressions can test the explanatory power of a set of variables at different scales, thus allowing
ecological responses to be examined in regards to habitat configuration and spatial landscape
patterns (Turner et al. 2001). The goal of regression modelling is to find the explanatory values
that minimize the difference between predicted and observed values of the dependent variable.
The mathematical equation for a linear regression equation takes the following form:
y= Xβ + e
Where:
y is n observations of a dependent variable;
X is a n by k matrix of observations on predictor variables
β is the standardized regression coefficient for X; and
e is the observed error term.

53

Developing a regression model with low error terms that explains landscape contributions
to lake phosphorus can be challenging because of regional variability in interactions among
biological, geochemical, and climatic variables. Regardless, models are important tools in
landscape limnology, as it is necessary to simplify these interactions to define problems more
precisely and concepts more clearly (Turner et al. 2001). Therefore, simple linear and multiple
regressions were used in this study to explore the relationship between diatom-inferred
phosphorus concentrations and pre-settlement landscape variables.
My research used a quasi-experimental (non-randomized) design to examine these
relationships. Unlike true-experimental designs where the minimum sample size required is
controllable, these designs can be used when it is not logistically feasible to conduct a
randomized, full manipulation of the dataset parameters (Harris et al. 2004). Therefore, a power
analysis, which normally determines the size of the sample needed, was not performed because
the data used in this research was predetermined and collected by others.
4.5.1 Summary of Potential Predictor Variables
Exploratory data analysis of potential predictor (independent) variables in conjunction
with the dependent variable (pre-settlement, diatom-inferred total phosphorus concentrations) is
a way to maximize the amount of information extracted from a dataset, formulate new
hypotheses, and identify the most important predictor variables (Behrens and Yu 2003).
Automated selection procedures, a method of exploratory analysis that relies on various
algorithms to iteratively identify predictor variables, can be useful when the researcher has no
basis for formulating a planned analysis (Marona et al. 2006; Hession and Moore 2011).
However, this technique has many pitfalls, including multicollinearity and confounding between

54

predictor variables, which can yield inaccurate results (Hession and Moore 2011). To avoid
confounding, I specified hypotheses-driven independent variables relevant to the dependent
variable, based on the existing breadth of knowledge, before running my analysis. This method
allowed the correlation among independent variables to be evaluated and described before
developing the regression model. Thus, the first step to predicting patterns in past phosphorus
concentrations was to identify an initial set of potential predictor variables, a priori. The
following land cover, hydrogeomorphic (HGM) and geographic variables were evaluated prior to
the final model selection procedure (Table 4.1).

Table 4.1: Potential landscape parameters
hypothesized to explain variations in pre-settlement
phosphorus concentrations (μg/L).
Potential Predictor Variables
% Native Land Cover
Deciduous forest*
Coniferous forest*
Mixed forest*
Total forest*
Wetland*
Grassland*
HGM Features
Wetland/lake ratio*
Maximum lake depth
Lake area
Lake perimeter
Lake area/depth ratio
Catchment/lake ratio
Geographic Variables:
Latitude
Longitude
Forest Tension Zone
*tested at 100 m buffer and local
catchment scale

55

Considerable attention has been given to statistical modeling of land cover effects on
water quality, although it is often done in light of anthropogenic drivers that affect phosphorus
export rather than natural landscape factors. Many studies that examine nutrient export suggest
sediment accumulation and phosphorus loading monotonically change relative to the amount of
agricultural land in a watershed (Dillon and Kirchner 1975; Gémesi et al. 2011). However, some
studies do give merit to natural controllers of phosphorus variation. For example, Allen et al.
(1997) demonstrated that nutrient yields and sedimentation rates decreased proportionally to
forested land cover in modeled watersheds. Therefore, I hypothesize that lakes with greater
proportions of aggregated forest cover (deciduous, coniferous, and mixed) will have lower presettlement phosphorus concentrations.
Additionally, there is much debate as to whether wetlands act as nutrient sinks or sources
(Richardson 1985). Much of the published research within the last two decades has largely
focused on the use of constructed wetlands as a nutrient sink in the wastewater treatment process.
However, although wetlands can remove and store large amounts of nutrients, they can also
release nitrogen and phosphorus to the adjacent aquatic ecosystems. Once a nutrient enters a
wetland, it is stored, altered, or discharged, depending on the complex biotic and abiotic
mechanisms involved. My hypothesis is that in the absence of anthropogenic phosphorus
contributions (i.e., agricultural and urban runoff), wetland ecosystems provide the essential
nutrients needed to sustain primary productivity in lakes.
In developing the regression model, I hypothesized that phosphorus concentrations had a
negative linear relationship with lake depth, based on existing research involving this abiotic
variable. For example, in a series of lake management models, Håkanson (1996) demonstrated a
significant positive relationship between lake depth and Secchi depth, an indicator of
56

limnological conditions. This was because sediments in deeper lakes (with more water beneath
the wave base) were less likely to be resuspended in the water column. The resuspension of
sediments also causes a resuspension of benthic phosphorus, adsorbed to the soil particles
(Wetzel 2001). For this reason, total phosphorus concentrations often correspond with water
transparency, as measured by Secchi depth (Carlson and Simpson 1996; Kleeberg et al. 2007).
Precipitation and temperature can affect the rate of weathering of phosphorus-containing
rocks, as well as the erosion and accumulation of sediments (Chuman et al. 2013). Therefore,
three parameters reflective of climate were included as potential predictor variables of
phosphorus variability: latitude, longitude, and relative location to the forest tension zone (FTZ)
in Michigan. Relative location to the FTZ, a categorical variable, was included in addition to
latitude and latitude to capture the approximate divergence in the number of degree growing days
between northern and southern Michigan (Andresen et al. 2012). Lakes in areas south of the
FTZ, were coded ‘0,’ whereas lakes north of the FTZ were coded ‘1’.
4.5.2 Initial Variable Selection
The a priori prediction that these aforementioned independent variables would explain
some of the variability in diatom-inferred phosphorus concentrations among the study lakes was
tested using bivariate, exploratory data analyses in the statistical computing package, R (R Core
Team 2013). Simple linear correlation coefficients (r) measured the degree and direction of
associations between the predictor variables and the diatom-inferred phosphorus concentrations.
Values of r can range from -1 (perfect negative correlation) to +1 (perfect positive correlation),
with increasing significance as the value approaches ± 1 (Clark and Hosking 1986).

57

The calculation of correlation coefficients was repeated for each potential predictor
variable to obtain a set of correlations, output in the form of a correlation matrix (Appendix B).
The correlation matrix was used to assess collinearity, the degree of independence and similarity
in the variability among the independent variables (Hunsaker and Levine 1995). More
importantly, the potential predictor variables were correlated against the dependent variable,
historic diatom-inferred phosphorus. The results of these correlation analyses were used to
determine which variables explained the most variance in pre-settlement phosphorus as an aid to
model selection.
4.5.3 Model Selection
Linear regressions based on ordinary least squares (OLS) were performed in R using presettlement diatom-inferred phosphorus concentrations as the dependent variable (y) and the
predictor variables identified by the selection process in Section 4.5.2. An OLS equation seeks
to predict the value of a dependent variable (y) based on the value of one or more independent
variables (x) by minimizing the sum of the squared differences (residual errors) between the
observed and predicted values of the dependent variable (Overmars et al. 2003). Seven linear
models were run for the predictor variables identified with bivariate, correlation testing to
determine which set of theoretically relevant variables provided the most satisfactory model.
These standard linear regression models were based on the assumption that observations
were statistically independent of one another (Overmars et al. 2003). However, if this data
assumption is incorrect, and spatial autocorrelation does exist within the dependent variable or
the residual errors, the regression equation would likely yield biased results (Anselin 1998;
Hession and Moore 2011). Therefore, I evaluated spatial dependency with Moran’s I, a

58

conventional measure of autocorrelation that compared the value of a pre-settlement, diatominferred phosphorus concentration at a given location to the value of diatom-inferred phosphorus
at all other lake locations (Anselin 1998). Similar to Pearson’s Product Correlation Coefficient,
the significance level of Moran’s I increases as the value approaches -1 or +1. I used this
measure to identify the number of nearest neighbors and weights matrix that yielded the most
spatial autocorrelation. Using the number of nearest neighbors and the weights matrix identified
in the Moran’s I step, I ran the model in the statistical platform, GeoDa to administer further
diagnostic tests for spatial autocorrelation. Both methods used to evaluate spatial autocorrelation
revealed that spatial regression was not necessary (further detail provided in Chapter Five), and
an OLS regression models was appropriate for explaining the variation in my dataset. These
OLS regression models imply spatial independence amongst pre-settlement lake phosphorus
concentrations.
Finally, the candidate OLS models that incorporated historic lake phosphorus data and
landscape parameters were evaluated by minimizing Akaike’s information criterion (AIC) in R.
This method of model selection, which uses maximum likelihood measures of the fit of a model
(Burnham and Anderson 2004), can be defined as:
AIC= 2k-2ln(L)
Where:
k is the number of parameters in the model;
and L is the maximum likelihood.

59

This statistical model selection technique replaces the mean squared error of prediction as
a measure of model “badness” (Akaike 1981). In the model selection process, lower AIC values
indicate a better model. The final model, best fit to the data variability, was an aspatial OLS
regression with the relatively lowest AIC values of all models tested.

60

5. RESULTS AND INTERPRETATIONS
This chapter presents the results derived from the basic spatial and statistical analyses of
this study. First, phosphorus concentrations representative of pre-settlement conditions were
inferred from transfer functions applied to diatom assemblages in the bottom interval of
lacustrine sediment cores. Relative dating of sediments using Ambrosia-type pollen verified that
the concentrations are reflective of undisturbed lake conditions, prior to Euro-American
settlement. Quantitative measures of pre-settlement land cover, as interpreted from GLO survey
data, were calculated and analyzed at two scales to be used as predictor variables in statistical
modeling. Lastly, the relationship between potential predictor variables and diatom-inferred
phosphorus concentrations in lakes prior to Euro-American land clearance and settlement was
examined using bivariate and multivariate regression analyses.
5.1 Results of Diatom Ordinations and Analyses
5.1.1 Results of Soranno/Garrison Dataset Analysis
Results of the ordination analyses conducted by the WDNR on the Soranno/Garrison
calibration dataset revealed that distributions of diatom taxa in surface sediments were most
correlated with total phosphorus and alkalinity (Garrison pers. comm. 2012). The CCA
ordination diagram (Figure 5.1) produced in CANOCO (version 4.1) by the WDNR presents a
graphical summary of the patterns of correlation between diatom species variation and water
chemistry variables from the calibration set. The vector (arrow) length represents the
proportional strength of correlation between the observed environmental variables and diatom
assemblages. This information was used to reconstruct diatom-inferred phosphorus
concentrations from the bottom core assemblages. Error estimates in inferences were derived by
the WDNR using the bootstrap error estimation approach in the statistical computing program
61

WACALIB version 3.3 (Line et al. 1994). The root mean squared error of prediction (RMSEp)

1.0

for summer mean phosphorus is log 0.29 μg/L (Garrison pers. comm. 2012).

TOTAL P

COLOR

ALK
ALK

TOTAL N

-1.0

-1.0

SECCHI

-1.0

1.0

Figure 5.1: Canonical correspondence analysis (CCA) ordination diagram of speciesenvironment relationship in the 187 lake calibration dataset generated by the WDNR with
the forward-selection option and Monte Carlo permutation tests in CANOCO (version
4.1) (ter Braak and Šmilauer 1998; Garrison pers. comm, 2012).

62

A summary of the observed and inferred phosphorus concentrations from the top
(modern) and bottom (pre-settlement) sediment core samples of the 31 lakes (Soranno/Garrison
dataset, prior to relative dating) are shown in Table 5.1. The mean of reconstructed phosphorus
concentrations of the 31 lake core-bottom samples was 11.61 μg/L. These bottom-core samples
are representative of oligotrophic lake conditions (reconstructed phosphorus ranged from 5.52
μg/L to 17.92 μg/L). Reconstructed diatom-inferred phosphorus increased by a mean of by 0.47
μg/L from the core bottom to the core top.

Table 5.1: Summary statistics of reconstructed phosphorus (inferred from diatom assemblage) and
observed phosphorus (measured in water column) from the Soranno/Garrison lake dataset (n=31).
Diatom-inferred total phosphorus concentrations from the bottom (DI-TP Bottom) and top (DI-TP
Top) core intervals are in units μg/L.

Mean
Min
Max
Median
Range

DI-TP Bottom
11.61
5.52
17.92
10.94
12.40

DI-TP Top Observed TP
12.08
13.26
3.67
2.02
44.39
32.88
11.50
11.02
40.72
30.86

Δ DI-TP Top
& Bottom
0.47
-7.63
26.47
-0.76
34.10

|Δ| Inferred TP
& Observed TP
4.83
0.28
19.21
3.03
18.93

5.1.2 Results of EPA Dataset Analysis
The results of the EPA’s CCA ordination (and associated Monte Carlo permutation test)
of the 42 sampled lakes (undated samples) suggested statistically significant relationships
(p = 0.001) between the surface sediment diatom species of the national calibration dataset and
total phosphorus (μg/L), total nitrogen (μg/L), conductivity (μS/cm at 25 C), and pH (USEPA
2010a). The total phosphorus transfer function developed by EPA operatives using weighted
averaging partial-least-squares (WA-PLS) models provided a robust reconstructive relationship.

63

Evaluations of transfer function strength, characterized by squared correlation coefficient (r

2

=0.74), involved a comparison of diatom-inferred total phosphorus from the surface sediments to
total phosphorus measured from the water column. Moreover, jackknife (leave-one-out) cross2

validation was used for more realistic error estimation (r jacknife=0.67; RMSEp=0.36 log
(μg/L)) (USEPA 2010).
Unlike the results of the 31 Soranno/Garrison dataset, mean diatom-inferred phosphorus
concentrations from the 42 EPA lakes decreased from the bottom (27.34 μg/L) to the top (18.49
μg/L) of the cores (Table 5.2). At first glance, these findings could be misinterpreted to mean
that natural levels of phosphorus in Michigan lakes were higher prior to Euro-American
settlement. In actuality, these relatively high phosphorus concentrations results could indicate
that the bottom-core depths are not representative of natural, pre-settlement conditions. In fact,
the range of diatom-inferred phosphorus concentrations from the undated sediment core-bottoms
of the EPA lakes (2.29 μg/L to 286.03μg/L) is considerably greater than the range from the
Soranno/Garrison core-bottoms (5.52 μg/L to 17.92 μg/L) (Table 5.2). This variability in the
EPA data more likely implies that bottom cores were not taken from a great enough depth, and
so the diatom assemblages in the sediments are not reflective of pre-settlement conditions, but
are of historic times when anthropogenic impacts were greater. Fortunately, the relative dating
technique based on Ambrosia pollen that was used in this study, and discussed in Section 5.2,
was able to substantiate this hypothesis.

64

Table 5.2: Summary statistics of reconstructed phosphorus (inferred from diatom assemblage)
and observed phosphorus (measured in water column) from the EPA lake dataset (n=42).
Diatom-inferred total phosphorus concentrations from the bottom (DI-TP Bottom) and top (DITP Top) core intervals are in units μg/L.
Δ DI-TP Top |Δ| Inferred TP
DI-TP Bottom DI-TP Top Observed TP & Bottom
& Observed TP
Mean
27.34
18.49
14.36
-8.85
4.13
Min
2.29
2.88
3.00
-274.97
0.03
Max
286.03
118.22
65.00
35.53
86.22
Median
13.26
10.65
8.50
-1.32
4.83
Range
283.74
115.34
62.00
310.50
86.19

5.1.3 General Discussion of Diatom-Inferred Data
Water chemistry reconstructions from calibration datasets require sufficient knowledge
about the factors controlling relative dominance of a diatom species. However, despite strong
correlations between measured environmental variables and surface sediment assemblages,
calibration datasets may not include all the significant limnological factors driving patterns in
species composition (Fritz et al. 1999). It is also possible to introduce error into diatom
inferences from potential within-lake variability of sediment diatom assemblages or taxonomic
inconsistencies between the top and bottom core sediments (Hall and Smol 1999). Comparison
of diatom-inferred epilimnetic total phosphorus and water column total phosphorus are therefore
necessary to determine the predictive ability of the model despite these obstacles that can hinder
accurate inferences of pre-settlement total phosphorus.
5.2 Pollen Analysis Results
I analyzed all available and suitable bottom-core samples from the Soranno/Garrison and
EPA datasets for Ambrosia pollen, which, when in high abundance, marks the approximate

65

timeframe of Euro-American settlement in the Midwest. Of the 85 bottom-cores samples (54
EPA and 31 Soranno/Garrison), 22 were excluded from the pollen analysis for various reasons
including: lab data for sediment sample did not exist, the sediment samples were unavailable for
relative dating (i.e., discarded or misplaced), low confidence levels in the accuracy of the
diatom-inferred phosphorus data reflecting pre-settlement conditions (this index was unique to
the EPA samples), or the sample was duplicated in both Soranno/Garrison and EPA datasets
(Appendix A). The results of my palynological investigations of the 63 Soranno/Garrison and
EPA bottom-core samples are shown in Table 5.3 and Table 5.4.
Ambrosia was found in 75% of the samples. The amount of Ambrosia pollen per
sediment sample ranged from 0 to 24%, with a mean of approximately 3%. In addition to the 22
excluded lakes, I eliminated 15 lakes from the final dataset of study lakes because the bottomcore sediments contained ≥ 4% of Ambrosia-type pollen, which is interpreted as above the
natural (pre-settlement) value of ragweed in the forests of Michigan. Based on this criterion, I
found 48 (out of 63) lake core-bottoms to be representative of pre-settlement conditions. In all
samples, total arboreal deciduous and coniferous pollen percentages exceeded that of Ambrosia.
Lake samples north of the FTZ were dominated by pollen grains of a mix of deciduousconiferous species, whereas the pollen spectra in samples south of the FTZ in Michigan were
comprised mainly of deciduous species. Although attempts were made to count maize pollen as
further evidence of Euro-American settlement and agricultural expansion, none were found, but
this is not unexpected as corn pollen is poorly dispersed and is rarely recovered in pollen studies
of lakes (McAndrews and Turton 2010).

66

Table 5.3: Results of the pollen analyses for 29 bottom-core samples from the Soranno/Garrison dataset. The italicized rows
located at the bottom of the table are representative of non-pre-settlement conditions.
Lake Name

County

Bartholomew
Bear
Big Blue
Big Lake
Big Star
Bobcat
Cadillac
Campau
Coldwater
Craig
Deer
Duck
Eagle
Hanley
Imp
Mitchell
Little
Murray
Ottawa
Pleasant
Starvation
Teal
Twin
West Twin
Whitefish
Witch
Reeds
Portage
Townline

Branch
Manistee
Kalkaska
Osceola
Lake
Gogebic
Wexford
Kent
Branch
Baraga
Charlevoix
Gogebic
Kalamazoo
Antrim
Gogebic
Wexford
Marquette
Kent
Iron
Jackson
Kalkaska
Marquette
Kalkaska
Montmorency
Montcalm
Marquette
Kent
Manistee
Montcalm

Bottom Core
DI-TP (μg/L)
14.63
9.84
11.30
10.78
10.04
14.52
10.25
10.53
10.83
9.89
12.06
10.40
12.55
14.03
5.52
15.52
10.14
12.90
10.02
11.11
7.92
13.00
9.81
7.80
10.17
12.55
17.92
13.06
10.94

Sample Depth (cm)
44-46
48-50
30-31
43-45
67-69
48-50
33-35
32-34
36-38
54-56
46-48
50-52
31-33
37-40
49-51
40-42
50-52
46-48
45-47
28-30
39-41
50-52
46-48
46-48
32-34
55-57
38-40
40-42
56-58

Ambrosia
1
2
0
0
0
1
1
2
2
1
1
1
2
1
0
0
0
1
0
1
0
2
0
0
0
0
10
10
11

68
67

Deciduous

Coniferous

86
45
56
52
36
51
31
70
91
28
48
44
84
69
28
37
65
72
47
10
46
29
44
9
67
64
58
52
46

13
53
44
48
64
48
68
28
7
71
51
55
14
30
72
63
35
27
53
89
54
69
56
91
33
36
32
38
43

Table 5.4: Results of the pollen analyses for 34 bottom-core samples from the EPA dataset. The italicized rows located at the
bottom of the table are representative of non-pre-settlement conditions.
Lake Name
Au Sable
Bailey
Blomgren
Blue
Campbell
Chemung
Clark
Dewey
Eight Point
Fence
Gogebic
Hall
Howe
Kathryn
Keewaydin
Lotto
McDonald
Mill
Mud
Pine
Stoner
Warner
Saddle
Big
Bogie
Crooked
Martin

County
Ogemaw
Keweenaw
Dickinson
Mecosta
Kalamazoo
Livingston
Jackson
Marquette
Clare
Baraga
Gogebic
Houghton
Alger
Iron
Baraga
Marquette
Delta
Oakland
Houghton
Kent
Kent
Barry
Van Buren
Baraga
Oakland
Emmet
Newaygo

DI-TP (μg/L)
2.92
11.98
41.18
6.32
2.29
8.34
9.46
5.99
6.89
11.64
14.71
12.63
14.79
19.54
9.83
10.56
15.87
9.03
12.72
4.81
16.28
32.23
16.04
14.03
9.07
5.11
18.01

Sample Depth (cm)
41.5-42.5
28-30
27-32
44-45
53.5-64.5
37.3-38.3
36-39
59-61
54.5-55.5
56-61
26-28
34-35
56-58
56-57
37-39
40-42
58-61
41.5-42.5
23-25
38.5-39.5
52-53
54.5-55.5
49-50
31-33
54-55
33.5-34.5
31.5-32.5
68

Ambrosia
0
0
2
0
2
2
0
0
0
3
1
2
0
0
0
0
1
0
2
1
1
0
8
4
14
9
24

Deciduous
63
35
37
41
85
81
91
15
36
36
50
56
65
37
38
47
52
76
36
53
73
84
69
31
74
56
41

Coniferous
37
65
61
59
13
17
9
85
63
61
49
42
35
63
62
53
47
24
62
46
26
16
23
65
12
35
35

Table 5.4 (cont’d)
Muskegon
PereMarquette
Round
Tallman
Tims
Upper Scott
Wyckoff

Muskegon
Mason

70.25
65.33

46-47
26.5-27.5

15
8

78
70

7
22

Van Buren
Mason
Jackson
Allegan
Oceana

8.35
12.69
7.08
21.40
50.33

37.5-38.5
42-43
49-51.5
34.5-35.5
36.5-37.5

11
8
17
14
13

78
25
76
48
61

11
67
7
38
26

69

5.2.1 Correlation of Ambrosia with Depth
The abundance of Ambrosia pollen from my analysis was statistically examined for
trends that could provide insight for future paleolimnological investigations. First, I tested the
correlation between sediment core-bottom depths and Ambrosia pollen counts to determine if a
high percentage of Ambrosia in the pollen spectra generally corresponded to relatively short lake
cores. This would provide some explanation as to the why 35% of sediment cores from the EPA
dataset and 10% of the sediment cores from the Soranno/Garrison dataset were not representative
of pre-settlement conditions.
However, the results of simple linear correlation test (Figure 5.2) indicated a low
association between the two variables (r = -0.120). Regardless, the negative correlation
indicates a slight tendency for Ambrosia counts to decrease as the core depth increases. This
suggests that the deeper cores are generally more successful in representing lake sediments that
correspond to a time period before anthropogenic landscape disturbances. Although the
direction of this relationship was expected, I anticipated a stronger correlation between bottomcore depth and the percentage of Ambrosia pollen. It is possible that variations in sedimentation
rates, due to regional or localized differences in erosion or lake productivity, is confounding
what might otherwise be a simple and linear relationship.

70

20
15
10
0

5

Ambrosia Pollen Count

30

40

50

60

Core Depth

Figure 5.2: Correlation plot between the Ambrosia counts and the
depth (cm) of the bottom-core interval (r = -0.120).

5.2.2 Correlation of Ambrosia with Total Phosphorus
I also conducted a bivariate correlation test to determine statistical relationship between
Ambrosia pollen counts and pre-settlement total phosphorus concentrations from the bottom-core
samples. The relationship between Ambrosia and phosphorus concentrations (Figure 5.3) was
stronger (r = 0.3635) than the relationship with sediment core depth (r = -0.120). Additionally,
the relationship was positive, which means that sediments with relatively higher percentages
Ambrosia pollen had relatively higher total phosphorus concentrations. This result was expected,
as phosphorus concentrations are known to be higher in lakes of disturbed watersheds (such as
those initiated by Euro-American settlement) (e.g., Rast and Lee 1978; Omernik et al. 1981).

71

Figure 5.3: A scatter plot of the bivariate correlation testing results between diatom-inferred total phosphorus (DI-TP) in the bottom
sediment cores and the total Ambrosia pollen count from lakes in the Soranno/Garrison and EPA datasets (r = 0.3635). An ‘X’
denotes sediment intervals that were interpreted to be representative of pre-settlement conditions. Conversely, an ‘O’ denotes postsettlement sediment intervals because Ambrosia-type pollen comprises > 3% of the total pollen spectra.

72

5.2.3 Phosphorus Reconstructions from Study Lakes
The summary statistics of the final dataset of pre-settlement sediment samples from 48
lakes (Table 5.5) are noticeably different from the original EPA data. This is because 34% of the
EPA phosphorus reconstructions were eliminated from the final dataset due to > 3% Ambrosia
pollen in the bottom-core samples. The mean of the final dataset is 15.54 μg/L less than the
mean of the EPA samples alone. However, the mean of the 31 Soranno/Garrison data (11.61
μg/L) more closely resembles the final dataset.

Table 5.5: Summary statistics of reconstructed phosphorus (inferred from diatom assemblage) and
observed phosphorus (measured in water column) from the “pre-settlement” lake dataset (n=48).
Diatom-inferred total phosphorus concentrations from the bottom (DI-TP Bottom) and top (DI-TP
Top) core intervals are in units μg/L.

Mean

DI-TP Bottom
11.84

DI-TP Top
10.64

Observed TP
11.24

Δ DI-TP Top
& Bottom
-1.20

|Δ| Inferred TP &
Observed TP
4.93

Min

2.29

2.88

2.02

-25.04

0.03

Max

41.18

29.65

37.00

24.84

20.86

Median

10.67

9.95

9.96

-1.02

2.96

Range

38.88

26.77

34.98

49.88

20.83

5.3 GIS Analysis Results
The reclassification of native land cover from the MNFI shapefile resulted in the
following seven categories: “Deciduous Forest,” “Evergreen Forest,” “Mixed Forest,”
“Wetland,” “Surface Water,” “Grassland,” and “Barren” (Tables 5.6). Based on the total area of
the 48 lake catchments in this study, approximately 155,411 square kilometers of native land
cover in Michigan’s Upper and Lower Peninsulas were analyzed. The GIS analysis from this
73

Table 5.6: Reclassification of pre-settlement land cover categories, as captured by
the GLO surveys (Anderson 1976).
Original Land Cover Description
Land Cover Reclassification
Aspen-Birch Forest
Deciduous Forest
Original
Land
Cover
Description
Land
Cover Reclassification
Beech-Sugar Maple Forest
Aspen-Birch
Forest
Deciduous Forest
Black Oak Barren
Beech-Sugar
Maple
Forest
Mixed Oak Forest
Black
Oak Barren
Oak-Hickory
Forest
Mixed
Oak
Forest
Sugar Maple-Basswood Forest
Oak-Hickory
Forest Birch Forest
Sugar Maple-Yellow
Sugar
Maple-Basswood
Forest
Hemlock-White Pine Forest
Evergreen Forest
Sugar
Maple-Yellow
Birch Forest
Jack Pine-Red
Pine Forest
Hemlock-White
Evergreen Forest
Pine Barrens Pine Forest
Jack
Pine-Red PineForest
Forest
Spruce-Fir-Cedar
Pine
Barrens
White Pine-Red Pine Forest
Spruce-Fir-Cedar
Forest
Beech-Sugar Maple-Hemlock
Forest
Mixed Forest
White
Pine-Red
Pine
Forest
Hemlock-Yellow Birch Forest
Beech-Sugar
Maple-Hemlock
Forest
Mixed Forest
Mixed Pine-Oak
Forest
Hemlock-Yellow
Birch
Forest
Oak/Pine Barrens
Mixed
Forest Forest
Sugar Pine-Oak
Maple-Hemlock
Oak/Pine
Barrens Hardwood Forest
White Pine-Mixed
Sugar
WhiteMaple-Hemlock
Pine-White OakForest
Forest
White
Pine-Mixed
Hardwood
Forest
Black Ash Swamp
Wetland
White
CedarPine-White
Swamp Oak Forest
Black
Swamp
Wetland
MixedAsh
Conifer
Swamp
Cedar
MixedSwamp
Hardwood Swamp
Mixed
Conifer
Muskeg/Bog Swamp
Mixed
Swamp Marsh
ShrubHardwood
Swamp/Emergent
Muskeg/Bog
Wet Prairie
Shrub
Swamp/Emergent Marsh
Lake/River
Surface Water
Wet
Prairie
Grassland
Grassland
Lake/River
Surface Water
Mixed Oak Savanna
Grassland
Grassland
Exposed Bedrock
Barren
Mixed
Oak
Savanna
Sand Dune
Exposed Bedrock
Barren
Sand Dune
Table 5.6: Reclassification of pre-settlement land cover categories, as captured by
th
the indicated
GLO surveys
research
that (Anderson
at the time 1976).
of the 19 century surveys, forests were the dominant land
cover at both the 100 m buffer and local catchment (lakeshed) spatial extent (Table 5.7). In local
lake catchments, the aggregation of forest classes comprised approximately 82% of the native
74

land cover composition, compared to approximately 70% of the total land cover within the 100
m buffer. Conversely, wetland cover is at a greater percentage (nearly 29%) in the 100 m buffer
than in the local catchment (about 17%). Furthermore, wetlands constituted > 50% of the 100 m
buffer of 13 of the study lakes, but only 1 of the local catchments were composed of >50%
wetlands. The greater percentage of native wetland cover within the100 m buffer of lakes is not
surprising, as wetlands are known to develop in shallow water or areas associated with high
water tables within near proximity to lakes, rivers, or streams (Tiner 2003; Mitsch and Gosselink
2007).
Table 5.7: Summary of reclassified native land covers in the 100 m buffer and local
catchment area.
Land Cover

% in 100 m Buffer

% in Local Catchment

Deciduous
19.2%native land covers in 12.1%
Table
5.7: Forest
Summary of reclassified
the 100 m buffer and local
Mixed Forest
36.4%
catchment
area.
63.6%
Coniferous Forest
15.0%
6.1%
Wetland
28.7%
17.2%
Grassland
0.8%
1.0%
Barren
0%
>1%

Although GIS has revolutionized the way ecologists monitor landscape changes across
spatial and temporal scales, positional data interpreted from survey notes should be approached
with caution (Turner et al. 2001; Gergel and Turner 2002). For example, the GLO surveyors
typically recorded landscape observations at limited locations such as along township and section
lines. Moreover, the experience and overall thoroughness of deputy surveyors is at times
reflected in the precision of the GLO mapping project (Thomas 2009). Potential inconsistencies
between USPLS sections of the GLO native land cover map may be caused by other reasons than
surveyor error. For example, mapping of wetlands varied with the season of the field work, or
whether it was a wet or dry year (Thomas 2009). As a result, the positional accuracy of land
75

cover delineations from in the MNFI digital map may be within an estimated 80-186 m of their
actual position (MNFI 2000; Thomas 2009). Unfortunately, coarse resolution of digital maps
adds an inevitable degree of uncertainty to any landscape mapping project, particularly when
remotely sensed data or aerial imagery is unavailable. Regardless of these caveats, the presettlement digital dataset has been used by scientists and researchers investigating the impact of
landscape-scale changes in vegetation patterns (e.g., Hupy and Yansa 2009b).
5.4 Statistical Data Analysis Results and Discussion
I used an a priori model approach to potential predictor variable selection as an
alternative approach to automated model selection procedures (e.g., forward, backward, and
stepwise regressions) to avoid inaccuracies introduced by multicollinearity and confounding
between the predictor variables. Automated selection procedures that rely on various algorithms
for decision making can yield unreasonable results when there are missing values in the potential
covariates (Burnham and Anderson 1998; Hession and Moore 2012). Rather, the inherent
limitations of automated procedures were avoided through careful selection of hypothesis-driven
predictor variables. Only the predictor variables with a significant correlation to historic
phosphorus concentrations were selected and evaluated in an OLS regression model.
5.4.1 Bivariate Regression Results and Discussion
After ranking the % land cover variables in the 100 m buffer and local catchment
according to their correlation coefficients (r), I found that the relationship between prehistoric
% wetland in the 100 m buffer demonstrated the strongest correlation (r = 0.269; Table 5.8) with
pre-settlement phosphorus concentrations. I then derived a variable that was the combination of
land cover and a hydrogeomorphic feature (lake area) in an attempt to reduce multicollinearity
76

that inevitably occurs when using land cover percentages as variables. I used wetland area within
the 100m buffer and local catchment because wetland percentage showed the highest correlation
with phosphorus out of all the native land cover variables. The derived variables were the ratio of
wetland area to lake area, at two spatial scales. These variables describe the amount of wetlands
relative to the size of the lake, which I hypothesized would be more accurate in capturing the
influence of land cover on phosphorus variability. I defined wetland/lake ratio as the total
wetland area (ha) (in 100 m buffer and local catchment) divided by the total lake area (ha).
Interestingly, the correlation between wetland/lake ratio in the 100 m buffer and presettlement phosphorus concentrations (r = 0.565) was stronger than the correlation of phosphorus
with % wetlands (r = 0.269). Furthermore, bivariate analysis between phosphorus and the
wetland/lake ratio in the local catchment revealed an even weaker correlation (r = 0.230). This
suggests that land cover within near proximity (100 m) to the lake shore may be able to better
account for the variability in pre-settlement phosphorus concentrations than land cover in the
entire lake catchment.
Highly correlated variables in regression equations can cause mathematical corruption of
a model (Van Sickle 2003). As shown in the correlation matrix produced from bivariate
correlation analysis (Appendix B), the forest tension zone (FTZ) variable is highly correlated
with both latitude and longitude (r=0.826 and -0.531, respectively). Therefore, to reduce
collinearity in the models and to filter out weak correlations with phosphorus, I excluded the
FTZ variable (r = 0.004) from the model selection procedure. In addition to multicollinearity
introduced from redundant variables, land cover percentages are not mutually independent of one
another because by definition, they must total 100%. For example, a high wetland percentage in
a 100 m buffer inevitably means that land cannot be forested, and as a result, the percentage of
77

forest in a 100 m buffer is low. For this reason, I chose only one land cover metric (at both
spatial scales), wetland/lake ratio, based on correlation coefficient values, to include in the OLS
regression analysis (Table 5.8).
Table 5.8: Pearson’s Correlation Coefficients (r) of the potential
predictor variables with pre-settlement phosphorus concentrations.
Potential Predictor Variable
Spatial Scale (if applicable)
%
Native
CoverCorrelation
100 m
Buffer
Local
Catchment
Table
5.8:Land
Pearson’s
Coefficients
(r)
of the
potential
predictor
Deciduous
variables
forest
with pre-settlement0.03
phosphorus concentrations.
0.066
Coniferous forest
-0.146
-0.185
Mixed forest
-0.088
0.01
Total forest
-0.198
-0.132
Wetland
0.269
0.131
Grassland
-0.03
0.055
HGM Features
Wetland/lake ratio
0.565
0.230
Maximum lake depth
---0.248
Lake area
--0.042
Lake perimeter
---0.051
Lake area/depth ratio
0.074
Catchment/lake ratio
--0.039
Geographic Variables:
Latitude
--0.101
Longitude
---0.21
Forest Tension Zone
--0.004

Scatter plots of these bivariate analyses between diatom-inferred phosphorus and the
continuous predictor variables visually demonstrated strength and direction of relationships
(Appendix C). The simple linear regression line on each plot is a graphical representation of the
correlation coefficients (r) as shown in Table 5.8. However, the results from these bivariate
correlation analyses alone are not sufficient to meaningfully declare a level of significance to the
relationship between the predictors and the dependent variable (Van Sickle 2003). Therefore,
multiple linear regressions were used to further explore trends in pre-settlement phosphorus
78

variability. The five predictor variables used individually, and then combined, in the model
selection process were latitude, longitude, maximum depth, wetland/lake ratio in the local
catchment, and wetland/lake ratio in a 100 m buffer.
5.4.2 Model Selection Results
Individual predictor variables and combinations of predictor variables (selected based on
methods described in Section 5.4.1) were evaluated for fit in seven OLS regression models based
2

2

on AIC and adjusted R values (Table 5.9). The adjusted R values are an indicator of how well

Table 5.9: Results of model-fitness testing from seven OLS regressions.
Model ID

Model Name

AIC

Adjusted R

2

Notes

Table
5.9:lat
Results of model-fitness317.6
testing from-0.0113
seven OLS regressions.
1
Latitude, in decimal degrees
315.93
0.02317
2
Longitude, in decimal degrees
long
Maximum depth as defined in
315.04
0.04127
3
max_d
original data collection
Area of total wetlands in the
315.19
0.03812
local catchment divided by the
4
wetland_ratio_LC
area of lake
Area of total wetlands in the
299.64
0.3044
100 m buffer divided by the
5
wetland_ratio_100
area of lake
Wetland:lake ratio combined
wetland_ratio +
301.24
0.2948
6
with maximum lake depth
max_d
Wetland:lake ratio combined
with maximum lake depth and
wetland_ratio +
298.7
0.3438
an interaction term of
7
max_d +
wetland:lake ratio x maximum
wetland_ratio*max_d
lake depth

the OLS model fits the data; values close to 1.0 indicates that the model accounts for nearly all of
the variance in the independent variables. AIC values in this study were meaningful and

79

comparable between models because the dependent variable was not log or square roottransformed.
Model #7 qualifies as a good-fitting, parsimonious regression model with a (relatively)
2

large R and only three predictor variables which keeps the interpretation simple. Models #1, 2,
2

3, and 4 yielded poor results, with adjusted R values < 0.1. Models #5 and 6 were similar in
2

adjusted R values to #7, however the AIC values were higher, indicating less fit with the data.
Therefore, I chose Model #7 as the final OLS model best fit to explain the variability in pre2

settlement phosphorus concentrations because it had the highest adjusted R and the lowest AIC.
This model can explain 34.4% of the variability in pre-settlement phosphorus.
Although it is tempting to interpret this model as a final model, OLS does not take into
account spatial autocorrelation. If spatial autocorrelation does exist, but is not accounted for, the
estimated coefficients (β) and the standard errors (SE) of this model would be biased. It is
therefore necessary to test for spatial autocorrelation within the best-fit OLS model, Model #7,
and consequently reject or accept a spatial regression model. The tests for spatial dependence,
conducted in GeoDa using Moran’s I and three nearest neighbor matrices (2, 4, and 9), showed
the nearest neighbor configuration of nn=2 to have the strongest correlation with reconstructed
phosphorus concentrations in my dataset (Appendix D). However, even at nn=2, the Moran’s I
value was low. Furthermore, diagnostic testing of Model #7 residuals confirmed that spatial
autocorrelation was not present in the dataset (Table 5.10). This conclusion was reached because
the p-values of the tested spatial regression models were not significant.

80

Table 5.10: Results of diagnostic testing for spatial autocorrelation, conducted in GeoDa.
DIAGNOSTICS FOR SPATIAL DEPENDENCE (nn=2)
TEST
MI/DF
Test-statistic value
Table
5.10: Results of diagnostic testing
for spatial autocorrelation,
conducted inp-value
GeoDa.
Moran's I (error)

-0.223

-1.5792292

0.1142835

Lagrange Multiplier (lag)

1

1.7645536

0.1840576

Robust LM (lag)

1

0.3261957

0.5679078

Lagrange Multiplier (error)

1

2.3995222

0.1213723

Robust LM (error)

1

0.9611643

0.3268937

Lagrange Multiplier (SARMA)

2

2.7257179

0.2559281

5.3.5 Final OLS Model Interpretation
Since spatial autocorrelation was not present in my data, I was able to accept OLS Model
#7 as the best fit model. In this model, the intercept, wetland/lake ratio, and the interaction term
of wetland/lake ratio*max lake depth were statistically significant (Table 5.11). Although
maximum lake depth did now show statistical significance, it was influential in the relationship
between wetlands and lake phosphorus concentrations. As shown in Table 5.11, the interaction
term actually has a negative estimated coefficient (β), which implies that as maximum depth
decreases (the lake becomes shallower), the influence of wetland within the 100 m buffer has a
greater effect on pre-settlement lake phosphorus concentrations

81

Table 5.11: Final OLS model (#7).
Estimated
Coefficient (β)
Intercept
9.41262
Table 5.11: Final OLS model (#7).
Wetland_ratio100
12.2749
Max_d
0.05978
Wet_ratio100*max_d
-0.7294

Std. Error
(SE) t-value
p-value
1.67794
5.61
1.26E-06
2.64739
4.637
3.17E-05
0.11485
0.52
0.6053
0.34933
-2.088
0.0426

Independent Variables

82

6. DISCUSSION
It would be easy to argue that all biotic and abiotic characteristics of a watershedhydrology, geology, topography, climate, land cover- contribute to phosphorus dynamics and
primary productivity in aquatic ecosystems. However, overfitting from the addition of too many
parameters can cause overly complex statistical models that seldom produce meaningful results.
It is therefore important to consider AIC values, which impose a penalty to increasing the
number of estimated parameters in a regression model. Therefore, simplified models that
quantify the strength of correlation between phosphorus and spatial patterns in land cover and/or
abiotic variables (e.g., morphometry and latitude) can be used to provide insight into complex
dynamics in landscape limnology (Soranno et al. 1996).
6.1 Discussion of Model Results
Past nutrient-related research has focused primarily on agriculture as a single predictive
variable that causes a synchronous shift in phosphorus loading to aquatic ecosystems (e.g.
Osborne and Wiley 1988; Shapley et al. 1994; Sims et al. 1998). However, my research has
demonstrated that natural factors contributed to pre-settlement phosphorus variability in lakes in
Michigan. This suggests that anthropogenic factors alone may not adequately explain modern
spatial heterogeneity in lake phosphorus concentrations.
The results of my regression models support the distance decay theory, central to
geographical studies, which implies that near things are more related than distant things
(Montello and Sutton 2006). As such, it was not surprising to find that wetlands within a 100 m
buffer of lakes had a more significant relationship to reconstructed phosphorus concentrations
than wetlands in the local catchment. This metric of proximity likely represents a degree of
hydrologic connectivity that facilitates an exchange of materials between terrestrial and aquatic
83

networks. Although the proximity of land covers to surface waters is not as frequently
investigated as land cover composition (primarily because these spatially explicit measures are
often more difficult to quantify), my results indicate this parameter is important to understanding
phosphorus variability (Osborne and Wiley 1988; Soranno et al. 1996, Gémesi et al. 2011).
Although wetlands within the 100 m buffer showed higher correlation with diatominferred phosphorus than any other land cover metric, it is necessary to recognize the effect of
collinearity between the native land cover variables. For example, a greater wetland/lake ratio
inevitably corresponds to a lower forest/lake ratio, although this metric was not derived. From
this standpoint, it could be argued that not only does the increase of wetlands yield higher
phosphorus concentrations, but the corresponding decrease in forests results in greater observed
phosphorus concentrations. Other studies have demonstrated forested areas to be important in
the physical retention of fine sediments from upslope erosion (e.g., Omernik et al. 1981;
Filippelli and Souch 1999; Reynolds and Davies 2001). Therefore, it is necessary to
acknowledge that not only is the increase of wetlands influencing the phosphorus concentrations,
but also the decrease in forest cover.
However, land cover and proximity alone did not produce the most successful model for
explaining pre-settlement phosphorus variability. The model best fit to describe this variability
also included maximum lake depth and an interaction term (wetland/lake ratio in the 100 m
buffer X maximum lake depth). Based on these findings, I was able answer the following
research question: “Are land cover metrics combined with hydrogeomorphic features and/or
geographic variables better predictors of water quality than land cover alone?” My results
demonstrated that the strength of the relationship between phosphorus and land cover was
greatly improved when combined with hydrogeomorphic features (lake area and lake depth).
84

While wetland cover alone was able to explain some of the variability in pre-settlement
phosphorus, the strength of this relationship increased (as determined by a significant, ≥3,
decrease in AIC values) when lake depth was incorporated in the model. I speculate that this
results from the combination of phosphorus additions from adjacent wetlands and resuspended
lake sediments. Lake sediments typically contain greater concentrations of phosphorus than the
overlying water column and, in shallow lakes, wind and wave disturbances can resuspend these
sediments. In deep waters where sediments are more likely left undisturbed, phosphorus
becomes buried and is essentially “removed” from the system. The exception to this is if deep
lakes develop anoxic conditions in their hypolimnion and consequentially release phosphorus
into the overlying waters. However, internal phosphorus loading such as this is more often the
result of cultural eutrophication and soil erosion associated with shoreland development
(Garrison and Wakeman 2000).
Maximum depth, was thus an important variable in this study. Although other studies
have incorporated mean lake depth, this metric was not available for all 48 lakes in my dataset. I
was, however, able to test mean depth for 29 of the lakes (from the Soranno/Garrison dataset)
against pre-settlement phosphorus. The results of the correlation were weaker (r = -0.099) than
the results from the maximum depth bivariate correlation testing (r = -0.227) of those same 29
lakes.
6.2 Nutrient Policy Implications
Under the jurisdiction of the CWA, states are required to establish a set of water quality
standards that define watershed goals and pollution limits for surface waters within their
administrative authorities (Clean Water Act 1972). Designated uses, antidegradation policies,

85

and water quality criteria are the three major, interrelated components of these water quality
standards that must be reported to and approved by the EPA. However, efforts to establish
numeric nutrient criteria (a subset of water quality criteria) specific to phosphorus have been
stymied by substantial aquatic ecosystem variability, and only eight states or territories have
adopted site-specific standards for some or all of their lakes (Obreza et al. 2011; USEPA 2013b).
Ideally, nutrient criteria would be site-specific for every lake based natural background
levels inferred from paleolimnological reconstructions. However with nearly 70,000 lakes > 4 ha
in the continental United States, this is an unrealistic goal. It is thus necessary to develop a
classification scheme to condense the number of different ecosystems types into more
manageable facets (Martin et al. 2011; Herlihy et al. 2013). Numeric nutrient criteria have been
derived to reflect regional patterns in lake conditions, minimally-disturbed reference lakes,
diatom-inferred phosphorus reconstructions, lake morphometry, and/or designated usage of the
water body (e.g., recreational or ecological) (WDNR 2007; Heiskary and Wilson 2008; Soranno
et al. 2011).
In 2001, the Michigan Department of Environmental Quality (MDEQ) began developing
numeric phosphorus criteria based on the relationship between total phosphorus and biological
responses (namely, fish and algae), however the Michigan Legislature withdrew the MDEQ’s
rule-making authority in 2006 (MDEQ 2013). As of now, Michigan has only a set narrative
nutrient criterion (a statement of water quality goals) and a total phosphorus limit of 1 μg/L for
effluents from point source discharges (MDEQ 2013). Methods for quantifying nutrient criteria
for Michigan lakes and streams are currently being developed by a team of scientists at Michigan
State University (MDEQ 2013).

86

Given the substantial range of regional geographic patterns and diversity of lake types in
Michigan, a single, statewide nutrient criterion for total phosphorus concentrations is not
appropriate for adoption (Soranno et al. 2008). Although aquatic ecosystems should not be
treated as if they are all the same, it is impractical to devise individual management strategies for
each system. Rather, the methods described in this study (namely, the comparison of diatominferred phosphorus to lake morphometry and pre-settlement land cover) can be applied to the
nutrient criteria development process. Unlike the commonly used approach of defining reference
site conditions, the paleolimnological approach described in this thesis is guaranteed to quantify
nutrient conditions in lakes as they were before significant human disturbances began (Herlihy et
al. 2013). However, this method can only be applied to natural lakes and is less applicable in
arid regions (e.g., the southwestern United States) that are dominated by constructed
impoundments. Additionally, given the underlying uncertainties introduced through inferring
phosphorus levels from fossil diatom assemblages and interpreting GLO survey notes, using
minimally-disturbed reference sites is certainly a defensible approach to establishing nutrient
criteria (Herlihy et al. 2013).
One of the significant findings of this study was the demonstrated importance of dating
the lake-core sediments of paleolimnological reconstructions of pre-settlement phosphorus
concentrations. For example, I determined that 15 of 63 sediment samples analyzed were not
truly representative of undisturbed conditions;12 of these samples were collected as part of the
EPA’s NLA in 2007. This finding alone is meaningful enough to refute the conclusions drawn
in a recent, controversial publication by Bachmann et al. (2013) that used data from the same
dataset. Therefore, their deduction that phosphorus concentrations in lakes were 14% higher

87

during pre-settlement conditions, is not entirely based on data reconstructed from undisturbed
lake conditions.
6.3 Lake Management Implications
Most inland lakes in Michigan have highly sought-after real estate along their shore lines.
However, the majority of these residential developments have a constructed sea-wall that
abruptly marks the boundary between terrestrial and aquatic ecosystems, thus altering the natural
cycle of allochthonous materials such as sediments and nutrients. Meanwhile, lake associations
and foundations spend thousands of dollars annually to stock sport fish in support of local
recreational fishing (Goodman 1991). It could simply be that the rate at which these fish are
harvested is greater than the rate at which they are produced. However, it is also possible that
these anthropogenically altered lakes are not as well-suited to productive fisheries as they might
have been if left undisturbed. This, I hypothesize, is due in part to the substantial decrease in
wetlands around inland lake shores.
My research demonstrated that wetlands in the 100 m buffer (Model #5) were
significantly related (p< 0.005) to reconstructed phosphorus concentrations, and could explain
approximately 30% of the variability in the study lakes. When this variable was combined with
maximum lake depth (Model #7), the explanatory strength increased to 34%. These results
revealed that wetlands in an undisturbed environment were capable of influencing pre-settlement
lake phosphorus concentrations to a degree. Although most studies of phosphorus are in regards
to cultural eutrophication processes and over-enrichment of freshwaters, this element is
fundamental to normal food web functioning when in manageable concentrations. Thus, I
speculate that the re-emergence of hydrologically-connected wetlands could improve fish

88

productivity by restoring a natural nutrient balance that stimulates primary productivity in
oligotrophic lakes to a reasonable level. This, of course, would greatly benefit lakes associations
that invest tremendous amounts of money annually to stock fish in order to appease the
recreational fishing industry interests.
The point here is not to say that lake managers should enhance waters with phosphorus as
some have suggested (Stockner et al. 2000), because biomanipulation almost always has
unforeseen negative repercussions. It is also important to recognize that excess phosphorus from
cultural eutrophication is undeniably detrimental to surface waters for a myriad of interrelated
reasons (e.g., nuisance algae blooms, dissolved oxygen depletion, and toxin-producing
cyanobacteria). Rather, I argue that the restoration of native land covers within a 100 m buffer
of lakes could yield satisfactory results for a diverse group of stakeholders. Poor shore habitat
was identified by the EPA’s NLA as a leading stressor affecting the biological health of lakes.
(USEPA 2010). While it would be unreasonable and infringing to suggest an outlaw of all
shoreline development, the construction of wetlands at public or association/foundation-owned
parks would be beneficial and economical in the long-run. It is also sensible to encourage new
residential developments to adopt natural lakeshores comprised of native wetland vegetation.
6.4 Future Research Directions
Several shortcomings exist with the final regression model in this thesis. First, it is
possible that important independent variables may have been excluded from the model. For
example, I excluded soil texture and drainage variables as potential predictor variables to reduce
the introduction of multicollinearity into the model selection process. Many studies have made
reference to the tendency of vegetation patterns to covary with edaphic patterns in soil texture

89

and drainage class (e.g., Yahner 2000; Harmon 2009). For example, coarse-textured sandy soils
with poor water-retaining capacity provide optimal growing medium for coniferous tree
communities such as pine barrens because these species require less water than their deciduous
counterparts (Yahner 2000; Kost et al. 2007). In contrast, poorly drained soils with extended
periods of saturation are optimal locations for hydric vegetation growth, characteristic of wetland
ecosystems (Mitsch and Gosselink 2007). Regardless of the confounding potential of these
edaphic variables, the relationships here are worth exploring in GIS. Using soil patterns, rather
than native land cover patterns, could potentially improve the model results. These results of
such an analysis could be more applicable to establishing nutrient criteria because unlike native
vegetation, soil texture has remained relatively unchanged by Euro-American settlement.
Furthermore, when I expanded the examination of wetlands near the lake perimeters,
versus the entire lake catchment, I found a spatial trend in patterns of wooded vs non-wooded
wetlands. I noticed that 74% of the wetlands within the local catchment were wooded (i.e., black
ash swamp, cedar swamp, mixed conifer swamp, and mixed hardwood swamp), and the
remaining 26% were non-wooded wetlands (bog, emergent marsh, and wet prairie). Within a
100 m buffer, the percentage of non-wooded wetlands was slightly higher (33%), when
compared to the entire local lake catchment. However, at both spatial scales, the total area of
wooded wetlands is twice that of non-wooded wetlands. This may suggest that there is some
underlying mechanism driving phosphorus loading from these land cover types, which warrants
further investigations.
Future studies should consider expanding upon hydrogeomorphic features in terms of
interactions with land cover. Of these hydrogeomorphic features, water residence time could be
an important predictor of pre-settlement phosphorus variability. Although the data was not
90

available for this study, modern studies of phosphorus loading could easily incorporate this
variable with knowledge of lake volume and watershed hydrology. While it is unlikely that
residence time alone could explain the remaining ~60% variability in phosphorus concentrations,
the ultimate model results would likely increase in fit if such a variable was considered in
conjunction with wetland cover.
Finally, a significant proportion of the variation in diatom-inferred phosphorus was not
explained by the model, illustrating that the landscape factors controlling water chemistry are
complex and remain poorly understood. The use of a larger and more diverse lake dataset would
likely increase the power of the explanatory data analysis. Although there is some debate as to
the minimum number of observations required to generate significant results, large sample sizes
are generally more reliable. Nonetheless, the findings of this study can improve our
understanding of natural mechanisms influencing lake phosphorous.

91

7. CONCLUSION
The many inland lakes of Michigan are ecologically and economically valuable resources
that can increase property values, and provide recreational opportunities, wildlife habitat, and
drinking water sources. Anthropogenic impacts on lakes have increased during recent decades
due to intensification of agriculture practices and water consumption. Excess nutrients in surface
runoff from agricultural and urban are now recognized as one of the leading causes of water
body impairments in the United States (Interlandi et al. 2003; USEPA 2010). It is therefore in
our best interest to protect these precious water reserves, in part through the establishment of
numeric nutrient criteria that define phosphorus pollution limits for freshwaters.
Lake sediment cores containing diatom fossil assemblages can be interpreted to
determine natural, pre-settlement nutrient concentrations in a lake before anthropogenic
perturbation (Battarbee 2000; Wetzel 2001). Determining lake conditions prior to EuroAmerican settlement can facilitate in defining practical restoration goals, as it would be
unrealistic to expect a reduction in phosphorus beyond natural, ambient concentrations.
Furthermore, the evaluation of inferred phosphorus concentrations with lake morphometry and
land cover data can be used to establish nutrient criteria as an alternative to minimally-disturbed
reference conditions.
The primary objective of this research was to understand the interaction between in-lake
phosphorus concentrations and land cover/HGM variables, as it existed before human
disturbances. I compiled data from two pre-existing datasets containing reconstructed
phosphorus concentrations from lake sediment cores collected in 2007. A pollen-based relative
dating technique was used to verify that the bottom-core sediments were representative of pre-

92

settlement lake conditions. The phosphorus reconstructions for 48 lakes in Michigan were
supplemented by pre-settlement land cover metrics at two spatial scales: a 100 m buffer and local
catchment. Finally, hydrogeomorphic features and geographic variables were incorporated in the
statistical modeling to test their relative importance to predicting phosphorus variability.
The findings of this study indicated that, when considering land cover percentages alone,
native wetlands were able to best explain some pre-settlement phosphorus variability in my
dataset. However, wetlands in a 100 m buffer, when combined with lake area (wetland/lake
ratio), was the predictor variable with the strongest correlation with total phosphorus.
Furthermore, results of OLS Model #7 suggested that wetlands are influenced by lake depth such
that as depth decreases, the effect of wetlands on lake phosphorus increases. Although this
model applied pre-settlement data, these interactions likely still occur and should be carefully
considered when establishing realistic nutrient criteria.

93

APPENDICES

94

APPENDIX A
Table A.1: List of lakes excluded from final dataset.
Lake Name
Data Source
Rational
Belleville
EPA
No lab data for core
Besser
EPA
No lab data for core
Appendix
A:
Lakes
excluded
from
final
dataset.
Bridge
EPA
No lab
data for core
Clark
Co.
EPA
No
lab
data for core
Appendix A: Lakes excluded from final dataset.
Donnell
EPA
No lab data for core
Ford
EPA
No lab data for core
Forestville
EPA
No lab data for core
Pine (2)
EPA
No lab data for core
Silver
EPA
No lab data for core
Stony Creek
EPA
No lab data for core
Sullivan
EPA
No lab data for core
West
EPA
No lab data for core
Brighton
EPA
Man-made lake
Hi-Land
EPA
Man-made lake
Palmer
EPA
Man-made lake
Loon
EPA
Uncertainty in representation of pre-settlement conditions
Squaw
EPA
Uncertainty in representation of pre-settlement conditions
Thornapple
EPA
Uncertainty in representation of pre-settlement conditions
Deer
EPA
Duplicated in Soranno/Garrison dataset
Ottawa
EPA
Duplicated in Soranno/Garrison dataset
Big
EPA
> 3% Ambrosia, therefore not pre-settlement conditions
Bogie
EPA
> 3% Ambrosia, therefore not pre-settlement conditions
Crooked
EPA
> 3% Ambrosia, therefore not pre-settlement conditions
Martin
EPA
> 3% Ambrosia, therefore not pre-settlement conditions
Muskegon
EPA
> 3% Ambrosia, therefore not pre-settlement conditions
Pere Marquette EPA
> 3% Ambrosia, therefore not pre-settlement conditions
Round
EPA
> 3% Ambrosia, therefore not pre-settlement conditions
Saddle
EPA
> 3% Ambrosia, therefore not pre-settlement conditions
Tallman
EPA
> 3% Ambrosia, therefore not pre-settlement conditions
Tims
EPA
> 3% Ambrosia, therefore not pre-settlement conditions
Upper Scott
EPA
> 3% Ambrosia, therefore not pre-settlement conditions
Wyckoff
EPA
> 3% Ambrosia, therefore not pre-settlement conditions
Big Chub
Soranno/Garrison > 3% Ambrosia, therefore not pre-settlement conditions
Hess
Soranno/Garrison > 3% Ambrosia, therefore not pre-settlement conditions
Portage
Soranno/Garrison > 3% Ambrosia, therefore not pre-settlement conditions
Reeds
Soranno/Garrison > 3% Ambrosia, therefore not pre-settlement conditions
Townline
Soranno/Garrison > 3% Ambrosia, therefore not pre-settlement conditions

95

APPENDIX B

Table B.1: Correlation matrix of all potential predictor variables.
Appendix B: Correlation matrix of all potential predictor variables.
DI-TP FTZ Latitude Longitude Max_D Lake Area Perimeter Lake area/depth
1.000 0.004
0.101
-0.210 -0.248
0.042
-0.051
0.074
DI-TP
Appendix
B: Correlation matrix
of
all
potential
predictor
variables.
FTZ
1.000
0.826
-0.531 -0.171
0.121
0.142
0.160
Latitude
1.000
-0.816 -0.282
0.106
0.080
0.169
Longitude
1.000
0.212
-0.213
-0.174
-0.254
Max_D
1.000
-0.087
-0.029
-0.183
Lake Area
1.000
0.940
0.989
Perimeter
1.000
0.908
Lake area/depth
1.000

96

Table B.1 (cont’d)
Catchment/Lake Area % Wet_LC % Dec_LC % Mix_LC % Ever_LC % For_LC % Grass_LC
0.039
0.131
0.066
0.010
-0.185
-0.132
0.055
DI-TP
(Appendix B cont’d)
FTZ
0.067
0.143
-0.678
0.479
0.280
0.181
-0.470
Latitude
0.114
0.238
-0.448
0.256
0.228
0.048
-0.377
Longitude
-0.115
-0.239
0.167
-0.124
-0.041
-0.014
0.319
Max_D
0.292
-0.190
0.107
0.141
-0.210
0.212
-0.113
Lake Area
-0.057
0.057
-0.154
0.150
-0.018
0.011
-0.086
Perimeter
-0.069
0.137
-0.126
0.061
0.068
-0.006
-0.159
Lake area/depth
-0.063
0.062
-0.178
0.153
0.004
0.004
-0.080
Catchment/Lake
1.000
-0.022
-0.048
0.097
-0.066
0.027
-0.014
Area
% Wet_LC
1.000
-0.027
-0.227
-0.136
-0.782
0.050
% Dec_LC
1.000
-0.696
-0.297
-0.062
0.140
% Mix_LC
1.000
-0.342
0.380
-0.321
% Dec_LC
1.000
0.165
-0.133
% For_LC
1.000
-0.660
% Grass_LC
1.000

97

Table B.1 (cont’d)
% Wet_100 % Dec_100 % Mix_100 % Ever_100 % For_100 % Grass_100 WR100 WRLC
0.269
0.030
-0.088
-0.146
-0.198
-0.030
0.565
0.242
DI-TP
(Appendix
B
cont’d)
FTZ
-0.140
-0.507
0.394
0.307
0.215
-0.374
-0.067
0.031
Latitude
-0.086
-0.279
0.222
0.169
0.125
-0.279
0.023
0.123
Longitude
0.062
0.118
-0.182
-0.034
-0.136
0.281
-0.004 -0.092
Max_D
-0.227
0.084
0.245
-0.168
0.268
-0.130
-0.313 -0.238
Lake Area
-0.117
-0.153
0.188
0.037
0.109
-0.064
-0.179 -0.076
Perimeter
-0.148
-0.145
0.087
0.172
0.095
-0.116
-0.267 -0.091
Lake area/depth
-0.113
-0.157
0.182
0.043
0.102
-0.056
-0.149 -0.051
Catchment/Lake Area
0.052
0.036
-0.053
-0.009
-0.037
-0.044
0.031
0.042
% Wet_LC
0.449
-0.034
-0.294
-0.074
-0.471
0.076
0.428
0.654
% Dec_LC
0.050
0.807
-0.586
-0.338
-0.169
0.062
-0.004
0.007
% Mix_LC
-0.101
-0.654
0.896
-0.242
0.265
-0.255
-0.114 -0.143
% Dec_LC
-0.244
-0.054
-0.317
0.878
0.238
-0.105
-0.163 -0.132
% For_LC
-0.506
0.025
0.399
0.124
0.636
-0.622
-0.521 -0.516
% Grass_LC
0.278
0.004
-0.272
-0.144
-0.459
0.907
0.322
0.038
% Wet_100
1.000
-0.208
-0.310
-0.273
-0.832
0.260
0.702
0.511
% Dec_100
1.000
-0.569
-0.173
0.187
-0.092
-0.132 -0.027
% Mix_100
1.000
-0.273
0.463
-0.215
-0.290 -0.248
% Ever_100
1.000
0.266
-0.115
-0.227 -0.140
% For_100
1.000
-0.463
-0.691 -0.458
% Grass_100
1.000
0.291
0.005
WR100
1.000
0.526
WRLC
1.000

98

APPENDIX C

10 15 20 25
5
0

Number of Lakes

Figure C.1: Histogram of the dependent variable, showing the distribution of presettlement phosphorus concentrations for the 48 lakes.

0

10

20

30

40

Pre-Settlement Phosphorus Concentrations (ug/L)

99

41

43

45

40
30
20

r = 0.101

10

10
5
0

Number of Lakes

15

Pre-Settlement TP (ug/L)

r = 0.101

47

42

Latitude (decimal degrees)

44

46

Latitude (decimal degrees)

Figure C.2: Univariate and bivariate testing results for latitude. The histogram (left) shows the
distribution of the variable and the scatter plot (right) shows the strength of the relationship with
pre-settlement phosphorus (n=48).

-90

-88

-86

-84

40
20

30

r = -0.210

10

Pre-Settlement TP (ug/L)

10
5
0

Number of Lakes

15

r = -0.210

-89

Longitude (decimal degrees)

-87

-85

Longitude (decimal degrees)

Figure C.3: Univariate and bivariate testing results for longitude. The histogram (left) shows the
distribution of the variable and the scatter plot (right) shows the strength of the relationship with
pre-settlement phosphorus (n=48).
100

0 20 40 60 80
% Native Deciduous Forest Cover
in 100 m Buffer

40
30
20

r = 0.030

10

Pre-Settlement TP (ug/L)

20
5 10
0

Number of Lakes

30

r = 0.030

0 20 40 60 80
% Native Deciduous Forest Cover
in 100 m Buffer

Figure C.4: Univariate and bivariate testing results for native deciduous forest (%) in the 100 m
buffer. The histogram (left) shows the distribution of the variable and the scatter plot (right)
shows the strength of the relationship with pre-settlement phosphorus (n=48).

0 20 40 60 80
% Native Deciduous Forest Cover
in Local Catchment

40
30
20
10

Pre-Settlement TP (ug/L)

20
15
10
5
0

Number of Lakes

25

r = 0.066

0 20 40 60 80
% Native Deciduous Forest Cover
in Local Catchment

Figure C.5: Univariate and bivariate testing results for native deciduous forest (%) in the local
catchment. The histogram (left) shows the distribution of the variable and the scatter plot (right)
shows the strength of the relationship with pre-settlement phosphorus (n=48).
101

0
20 40 60 80
% Native Evergreen Forest
in 100 m Buffer

40
30
20

r = -0.185

10

Pre-Settlement TP (ug/L)

30
20
10
0

Number of Lakes

r = -0.185

0
20 40 60 80
% Native Evergreen Forest Cover
in 100 m Buffer

Figure C.6: Univariate and bivariate testing results for native evergreen forest (%) in the 100 m
buffer. The histogram (left) shows the distribution of the variable and the scatter plot (right)
shows the strength of the relationship with pre-settlement phosphorus (n=48).

0 20 40 60 80
% Native Evergreen Forest
in Local Catchment

40
30
20

r = -0.146

10

Pre-Settlement TP (ug/L)

30
20
10
0

Number of Lakes

40

r = -0.146

0 20 40 60 80
% Native Evergreen Forest Cover
in Local Catchment

Figure C.7: Univariate and bivariate testing results for native evergreen forest (%) in the local
catchment. The histogram (left) shows the distribution of the variable and the scatter plot (right)
shows the strength of the relationship with pre-settlement phosphorus (n=48).
102

40
30
20

r = -0.088

10

Pre-Settlement TP (ug/L)

20
15
10
5
0

Number of Lakes

25

r = -0.088

0 20 40 60 80
% Native Mixed Forest Cover
in 100 m Buffer

0 20 40 60 80
% Native Mixed Forest Cover
in 100 m Buffer

Figure C.8: Univariate and bivariate testing results for native mixed forest (%) in the 100 m
buffer. The histogram (left) shows the distribution of the variable and the scatter plot (right)
shows the strength of the relationship with pre-settlement phosphorus (n=48).

0 20 40 60 80
% Native Mixed Forest Cover
in Local Catchment

40
30
20

r = 0.010

10

Pre-Settlement TP (ug/L)

15
10
5
0

Number of Lakes

20

r = 0.010

0 20 40 60 80
% Native Mixed Forest Cover
in Local Catchment

Figure C.9: Univariate and bivariate testing results for native mixed forest (%) in the local
catchment. The histogram (left) shows the distribution of the variable and the scatter plot (right)
shows the strength of the relationship with pre-settlement phosphorus (n=48).
103

40
30
20

r = -0.198

10

Pre-Settlement TP (ug/L)

15
10
5
0

Number of Lakes

r = -0.198

0 20 40 60 80
% Native Forest Cover
in 100 m Buffer

0 20 40 60 80
% Native Forest Cover
in 100 m Buffer

Figure C.10: Univariate and bivariate testing results for native total forest (%) in the 100 m
buffer. The histogram (left) shows the distribution of the variable and the scatter plot (right)
shows the strength of the relationship with pre-settlement phosphorus (n=48).

0 20 40 60 80
% Native Forest Cover
in Local Catchment

40
30
20

r = -0.132

10

Pre-Settlement TP (ug/L)

10 15 20 25
5
0

Number of Lakes

r = -0.132

0 20 40 60 80
% Native Forest Cover
in Local Catchment

Figure C.11: Univariate and bivariate testing results for native total forest (%) in the local
catchment. The histogram (left) shows the distribution of the variable and the scatter plot (right)
shows the strength of the relationship with pre-settlement phosphorus (n=48).
104

40
30
20

r = 0.269

10

Pre-Settlement TP (ug/L)

15
10
5
0

Number of Lakes

r = 0.269

0 20 40 60 80
% Native Wetland Cover
in 100 m Buffer

0 20 40 60 80
% Native Wetland Cover
in 100 m Buffer

Figure C.12: Univariate and bivariate testing results for native wetland cover (%) in the 100 m
buffer. The histogram (left) shows the distribution of the variable and the scatter plot (right)
shows the strength of the relationship with pre-settlement phosphorus (n=48).

0
5 10 15 20 25
% Native Wetland Cover
in Local Catchment

40
30
20

r = 0.131

10

30
20
10
0

Number of Lakes

40

Pre-Settlement TP (ug/L)

r = 0.131

0
5
10 15 20
% Native Wetland Cover
in Local Catchment

Figure C.13: Univariate and bivariate testing results for native wetland cover (%) in the local
catchment. The histogram (left) shows the distribution of the variable and the scatter plot (right)
shows the strength of the relationship with pre-settlement phosphorus (n=48).
105

40
30
20

r = -0.03

10

30
20
10
0

Number of Lakes

40

Pre-Settlement TP (ug/L)

r = -0.030

0
10
20
30
40
% Native Grassland Cover
in 100 m Buffer

0
10
20
30
% Native Grassland Cover
in 100 m Buffer

Figure C.14: Univariate and bivariate testing results for native grassland cover (%) in the 100 m
buffer. The histogram (left) shows the distribution of the variable and the scatter plot (right)
shows the strength of the relationship with pre-settlement phosphorus (n=48).

0
20
40
60
% Native Grassland Cover
in Local Catchment

40
30
20

r = 0.055

10

30
20
10
0

Number of Lakes

40

Pre-Settlement TP (ug/L)

r = 0.055

0
20
40
60
% Native Grassland Cover
in Local Catchment

Figure C.15: Univariate and bivariate testing results for native grassland cover (%) in the local
catchment. The histogram (left) shows the distribution of the variable and the scatter plot (right)
shows the strength of the relationship with pre-settlement phosphorus (n=48).
106

0.0

40
30
20

r = 0.565

10

Pre-Settlement TP (ug/L)

30
20
10
0

Number of Lakes

40

r = 0.565

1.0
2.0
Wetland/Lake Ratio
in 100 m Buffer

0.0

0.5 1.0 1.5 2.0
Wetland/Lake Ratio
in 100 m Buffer

Figure C.16: Univariate and bivariate testing results for wetland/lake ratio in the 100 m buffer.
The histogram (left) shows the distribution of the variable and the scatter plot (right) shows the
strength of the relationship with pre-settlement phosphorus (n=48).

0

5 10 15 20 25
Wetland/Lake Ratio
in Local Lake Catchment

40
30
20

r = 0.131

10

Pre-Settlement TP (ug/L)

30
20
10
0

Number of Lakes

40

r = 0.131

0

5
10 15 20
Wetland/Lake Ratio
in Local Catchment

Figure C.17: Univariate and bivariate testing results for wetland/lake ratio in the local
catchment. The histogram (left) shows the distribution of the variable and the scatter plot (right)
shows the strength of the relationship with pre-settlement phosphorus (n=48).
107

0
0

5 10

20

40
30
20

r = -0.230

10

5

10

Number of Lakes

15

Pre-Settlement TP (ug/L)

r = -0.230

30

0

Maximum Lake Depth (m)

5

10

20

30

Maximum Lake Depth (m)

Figure C.18: Univariate and bivariate testing results for maximum lake depth (m). The
histogram (left) shows the distribution of the variable and the scatter plot (right) shows the
strength of the relationship with pre-settlement phosphorus (n=48).

0

2000

4000

6000

40
30
20

r = 0.042

10

Pre-Settlement TP (ug/L)

40
30
20
10
0

Number of Lakes

r = 0.042

0

Lake Area (ha)

2000

4000

Lake Area (ha)

Figure C.19: Univariate and bivariate testing results for lake area (ha). The histogram (left)
shows the distribution of the variable and the scatter plot (right) shows the strength of the
relationship with pre-settlement phosphorus (n=48).
108

0

200

400

40
30
20

r = 0.039

10

Pre-Settlement TP (ug/L)

40
30
20
10
0

Number of Lakes

r = 0.039

600

0

Lake/Catchment Ratio

200

400

600

Lake/Catchment Ratio

Figure C.20: Univariate and bivariate testing results for lake/catchment ratio. The histogram
(left) shows the distribution of the variable and the scatter plot (right) shows the strength of the
relationship with pre-settlement phosphorus (n=48).

0 20000

50000

40
30
20

r = -0.0512

10

Pre-Settlement TP (ug/L)

30
20
10
0

Number of Lakes

r = -0.051

0

20000

50000

Lake Perimeter (m)
Lake Perimeter (m)
Figure C.21: Univariate and bivariate testing results for lake perimeter. The histogram (left)
shows the distribution of the variable and the scatter plot (right) shows the strength of the
relationship with pre-settlement phosphorus (n=48).
109

APPENDIX D

Figure D.1: Moran’s I results; nearest neighbor= 2

110

Figure D.2: Moran’s I results; nearest neighbor= 4

111

Figure D.3: Moran’s I results; nearest neighbor= 9

112

REFERENCES

113

REFERENCES

Albert, D. A., 1995. Regional Landscape Ecosystems of Michigan, Minnesota, and Wisconsin: A
Working Map and Classification. General Technical Report NC-178. St. Paul, Minn.:
U.S. Dept. of Agriculture, Forest Service, North Central Forest Experiment Station.
Albert, D. A., Cohen, J. G., Kost, M. A., Slaughter, B. S., and Enander. H. D., 2008. Distribution
Maps of Michigan’s Natural Communities. Michigan Natural Features Inventory, Report
No. 2008-01, Lansing, MI.
Allen, J. A., Erickson, D. L., and Fay, J., 1997. The influence of catchment land use on stream
integrity across multiple spatial scales. Freshwater Biology 37: 149-161.
Anderson, J. R., Hardy, E. E., Roach, J. T., and Witmer, R. E.. 1976. A Land Use and Cover
Classification for Use with Remote Sensor Data. Geological Survey Professional Paper
974. USGS.
Anderson, N. J., Rippey, B., Gibson, C. E., 1993. A comparison of sedimentary and diatominferred phosphorus profiles: implications for defining pre-disturbance nutrient
conditions. Hydrobiologia 253: 357-366.
Anderson, B. J., 2005. The Historical Development of the Tension Zone Concept in the Great
Lakes Region of North America. The Michigan Botanist 44: 127-138.
Andresen, J. A., and Winkler, J. A., 2009. Weather and Climate, p. 288-314. In Schaetzl, R.,
Darden, J., and Brandt, D. [eds.], Michigan Geography and Geology. Pearson, New York,
New York.
Andresen, J., Hilberg, J., Kunkel, K., 2012: Historical Climate and Climate Trends in the
Midwestern USA. In: U.S. National Climate Assessment Midwest Technical Input
Report. Winkler, J., Andresen, J., Hatfield, J., Bidwell, D., and Brown, D., coordinators.
Available from the Great Lakes Integrated Sciences and Assessments (GLISA) Center,
http://glisa.msu.edu/docs/NCA/MTIT_Historical.pdf.
Anselin, L., 1988. Spatial Econometrics: Methods and Models. Kluwer Academic Publishers,
Dordrecht, The Netherlands.
Ashley, K. A., Thompson, L. C., Lasenby, D. C., McEachern, L, Smokorwski, K. E., and
Sebastian, D., 1997. Restoration of an interior lake ecosystem: the Kootenay Lake
fertilization experiment. Water Quality Research Journal of Canada 32:192-212.
Bachmann, R. W., Hoyer, M. V., and Canfield, D. E., 2013. The extent that natural lakes in the
United States of America have been changed by cultural eutrophication. Limnological
Oceanography 58(3): 945-950.
114

Barnes, B. V. and Wagner, W. H., 1981. A Guide to the Trees of Michigan and the Great Lakes
Region University of Michigan Press, Ann Arbor, Michigan.
Battarbee, R. W., Charles, D. F., Dixit, S. S., and Renberg, I., 1999. Diatoms as indicators of
surface water acidity. In Stoermer, E. F. and Smol, J. P. [eds], The Diatoms: Applications
for the Environmental and Earth Sciences. Cambridge University Press, Cambridge.
Battarbee, R. W., 2000. Palaeolimnological approaches to climate change, with special regard to
the biological record. Quaternary Science Reviews 19: 107-124.
Behrens, J. T. and Yu, C. H., 2003. Exploratory Data Analysis. Handbook of Psychology. 1: 3364.
Benoit, M. and Fizaine, G. 1999. Quality of water in forest catchment areas. Revue Forestiere
Francaise 50: 162-172.
Berka, C., Schreier, H., and Hall, K., 2001. Linking water quality with agricultural
intensification in a rural watershed. Water, Air, and Soil Pollution 127: 389-401.
Birks, H. B., 1998. Numerical tools in paleolimnology-progress, potentialities, and problems.
Journal of Paleolimnology. 20: 307-332.
14

Bjӧrk, S. and Wohlfarth, B., 2001. C Chronostratigraphic Techniques in Paleolimnology. In
Tracking Environmental Change Using Lake Sediments, 2001 [Eds] Last, W. M. and
Smol, J. P. Kluwer Academic Publishers.
Blais, J. M., Kalff, J., Cornett, J. R., and Evans, D. R., 1995. Evaluation of
sediments using stable Pb, Ambrosia pollen, and
169-178.

137

210

Pb dating in lake

Cs. Journal of Paleolimnology 13:

Blewett, W. L., Lusch, D. P., Schaetzl, R. J., 2009. The Physical Landscape: A Glacial Legacy,
p. 249-273. In Michigan Geography and Geology, Schaetzl, R., Darden, J., and Brandt,
D. [eds.], Pearson, New York, New York.
Bloesch, J. 1976. Sedimentation rates and sediment cores in two Swiss lakes of different trophic
states, In Interactions between sediments and freshwater: Proceedings of an International
Symposium, Golterman, H. L. [ed.], Amsterdam, the Netherlands.
Boers, P. C. M., Cappenberg., T. E., and van Raaphorst, W., 1993. The third international
workshop on phosphorus in sediments: Summary and synthesis. Hydrobiologia. 253: xi–
xviii.
Bremigan, M. T. and Soranno, P. A., 2008. Hydrogeomorphic features mediate the effects of
land use/cover on reservoir productivity and food webs. Limnology and Oceanography
53(4): 1420-1433.
115

Brenner, M. and Escobar, J., 2009. Ontogeny of Aquatic Ecosystems. p. 456-461. In
Encyclopedia of Inland Waters: Volume 1, Likens, G. E. [ed.],Oxford: Elsevier.
Brohan, P., Kennedy, J. J., Harris, I., Tett, S. F. B., and Jones, P. D., 2006. Uncertainty estimates
in regional and global observed temperature changes: A new data set from 1850. Journal
of Geophysical Research: Atmospheres. 111 (D12).
Brumgan, R. B., 1978. Human disturbance and historical development of Linsley Pond. Ecology
59: 19-36.
Burnham K. P. and Anderson, D. R., 1998. Model Selection and Inference: A Practical
Information-Theoretic Approach. Springer Verlag: New York.
Burnham, K. P. and Anderson, D. R., 2004. Multimodel Inference: Understanding AIC and BIC
in Model Selection. Sociological Methods and Research 33(2): 261-304.
Carignan, R., and Flett, R. J., 1981. Postdepositional mobility of phosphorus in lake sediments.
Limnoogy and Oceanography 26(2): 361-366.
Carlson, R. E. 1977. A trophic status index for lakes. Limnology and Oceanography 22: 361-369.
Carlson, R. E. and Simpson, J. 1996. Coordinator’s Guide to Volunteer Lake Monitoring
Methods. North American Lake Management Society.
Carpenter, S. R., Caraco, N. F., Correll, D. L., Howarth, R. W., Sharpley, A. N., and Smith, V.
H., 1998. Nonpoint Pollution of Surface Waters with Phosphorus and Nitrogen,.
Ecological Applications 8(3): 559-568.
Chapra, S. C. and Reckhow, K., 1979. Expressing the phosphorus loading concept in
probabilistic terms. Journal of Fisheries Research Board of Canada 36(2): 225-229.
Chapra, S. C., 1997. Surface Water-Quality Modeling. WCB McGraw-Hill, Boston.
Chuman, T., Hruška, J., Oulehle, F., Gürtlerová, P., and Majer, V., 2013. Does stream water
chemistry reflect watershed characteristics? Environmental Monitoring and Assessment
185: 5683-5701.
Clark and Hosking, 1986. Statistical Analysis for Geographers, John Wiley and Sons.
Clean Water Act, 1972, 33 U.S.C.§1251 et seq.
Cohen, A. S., 2003. Paleolimnology: The History and Evolution of Lake Systems. Oxford
University Press, Oxford.
Comer, P. J., Albert, D. A., Wells, H. A., Hart, B. L., Raab, J. B., Price, D. L., Kashian, D. M.,
Corner, R. A., and Schuen, D. W., 1995. Michigan’s Native Landscape as Interpreted
from the General Land Office Surveys 1816-1856. Report to the USEPA Water Division
and the Wildlife Division, Michigan Department of Natural Resources. Michigan Natural
Features Inventory, Lansing, MI.
116

Cuffney, T. F., Meador, M. R., Porter, S. D., and Gurtz, M. E., 2000. Responses of physical,
chemical, and biological indicators of water to a gradient of agricultural land use in the
Yakima River, Washington. Environmental Monitoring and Assessment 64: 259-270.
Cumming, B. F., Smol, J. P., Kingston, J. C., Charles, A. F., Birks, H. J. B., Camburn, K. E.,
Dixit, S. S., Uutala, A. J., and Selle, A. R., 1992. How much acidification has occurred in
Adirondack region (New York, USA) lakes since pre-industrial times? Canadian Journal
of Fisheries and Aquatic Sciences 49: 128-141.
Devesa-Rey, R., Iglesias, M. L., Díaz-Fierros, F., and Barral, M. T., 2009. Total Phosphorus
Distribution and Bioavailability in the Bed Sediments of an Atlantic Basin (Galicia, NW
Spain): Spatial Distribution and Vertical Profiles. Water, Air, & Soil Pollution. 200: 341352.
Devito, K. J., Creed, I. F., Rothwell, R. L., Prepas, E. E., 2000. Landscape controls on
phosphorus loading to boreal lakes: implications for the potential impacts of forest
harvesting. Canadian Journal of Fisheries and Aquatic Sciences 57: 1977-1984.
Dillon, P. J. and Rigler, F. H., 1974. The phosphorus-chlorophyll relationship in lakes,
Limnology and Oceanography 19: 767-773.
Dillon, P. J. and Kirchner, W. B., 1975. The Effects of Geology and Land Use on the Export of
Phosphorus from Watersheds. Water Research 9: 134-148.
Dixit, S. S., Smol, J. P., Kingston, J. C., and Charles, D. F., 1992. Diatoms: Powerful Indicators
of Environmental Change. Environmental Science and Technology 26: 22–33.
Dixit, S. S., Smol, J. P., Charles, D. F., Hughes, R. M., Paulsen, S. G., Collins, G. B., 1999.
Assessing water quality changes in the lakes of northeastern United States using sediment
diatoms. Canadian Journal of Fisheries and Aquatic Sciences 56: 131-152.
Djodjic, F., Bӧrling, K., and Bergstrӧm, L., 2004. Phosphorus Leaching in Relation to Soil Type
and Soil Phosphorus Content. Journal of Environmental Quality. 33: 678-684.
Doerr, S. H. and Shakesby, R. A., 2011. Soil water repellency. In Handbook of Soil Science (2
edition), Huang, P., Levi, G. [ed], Taylor and Francis.

nd

Dong, L., Yang, Z., and Liu, X., 2011. Phosphorus fractions, sorption characteristics, and its
release in the sediments of Baiyangdian Lake, China. Environmental Monitoring and
Assessment 179: 335-345.
Dorioz, J. M., Pelletier, J. P., and Benoit, P., 1997. Physio-chemical properties and
bioavailability of particulate phosphorus of various origin in a watershed of Lake Geneva
(France). Water Resources 32(2): 275-286.
Eichenlaub, F. L., Harman, J. R., Nurnberger, F. V., and Stolle, H. J., 1990. The
Climatic Atlas of Michigan. Notre Dame: University of Notre Dame Press.

117

Engstrom, D. and Wright, H. E., 1984. Chemical stratigraphy of lake sediments. In Lake
Sediments and Environmental History, Haworth, E. and Lund, J. [eds], University of
Minnesota Press, Minneapolis, 11-67.
Ennis, G. L., Northcote, T. G., and Stockner, J. G., 1983. Recent trophic changes in Kootenay
Lake, British Columbia, as recorded by fossil diatoms. Canadian Journal of Botany 61:
1983-1992.
ESRI (Environmental Systems Research Institute), 2011. ArcGIS Desktop: Release 10.
Redlands, CA.
Etherington, J.R., 1983. Wetland Ecology. Edward Arnold, London.
Faegri, K. and Iverson, J., 1975. Textbook of Pollen Analysis. Copenhagen, Denmark:
Munksgaard.
Fergus, C. E., Soranno, P. A., Cheruvelil, K. S., and Bremigan, M. T., 2011. Multiscale
landscape and wetland drivers of lake total phosphorus and water color. Limnology and
Oceanography 56(6): 2127-2146.
Filippelli, G. M. and Souch, C., 1999. Effects of climate and landscape development on the
terrestrial phosphorus cycle. Geology 27(2): 171-174.
Finkelstein, S. A., Peros, M. C., and Davis, A. M, 2005. Late Holocene paleoenvironmental
change in a Great Lakes coastal wetland: integrating pollen and diatom datasets. Journal
of Paleolimnology 33: 1-12.
Fisher, R. F. and Binkley, D., 2000. Ecology and Management of Forest Soils, Wiley, New
York.
Fraterrigo, J. M. and J. A. Downing., 2008. The Influence of Land Use on Lake Nutrients Varies
with Watershed Transport Capacity. Ecosystems 11 (7): 1021-1034.
Fritz, S. C., Juggins, S., Battarbee, R. W., and Engstrom, D. R., 1991. Reconstruction of past
changes in salinity and climate using a diatom-based transfer function. Nature 352: 706708.
Fritz, S. C., Kingston, J. C., and Engstrom, D. R., 1993. Quantitative trophic reconstruction from
sedimentary diatom assemblages: a cautionary tale. Freshwater Biology 30: 1-23.
Fritz, S. C., Cumming, B. F., Gasse, F., and Laird, K. R., 1999. Diatoms as indicators of
hydrologic and climatic change in saline lakes. p. 41-72. In The Diatoms: Applications
for the Environmental and Earth Sciences. Stoermer, E. F. and Smol, J. P., [eds],
Cambridge University Press, Cambridge.
Garrison, P. J. and Wakeman, R. S., 2000. Use of paleolimnology to document the effect of lake
shoreland development on water quality. Journal of Paleolimnology 24: 369-393.
Garrison, P. J. and Fitzgerald, S. A., 2005. The role of shoreland development and commercial
cranberry farming in a lake in Wisconsin, USA. Journal of Paleolimnology 33: 169-188.
118

Garrison, P. J. and LaLiberte, G. D., 2010. The importance of water level changes and shoreline
development in the eutrophication of a shallow, seepage lake. Proceedings of the
Academy of Natural Sciences of Philadelphia 160: 113-126.
Gémesi, Z., Downing, J. A., Cruse, R. M., and Anderson, P. F., 2011. Effects of Watershed
Configuration and Composition on Downstream Lake Water Quality. Journal of
Environmental Quality 40: 517-527.
Gerber, E., Schaffner, U., Gassmann, A., Hinz, H. L., Seier, M., and Müller-Scharer, H., 2011.
Prospects for biological control of Ambrosia artemisiifolia in Europe: learning from the
past. Weed Research 51: 559-573.
Gergel, S. and Turner, M. 2002. Landscape Ecology: A Practical Guide to Concepts and
Techniques. Springer, New York, NY.
Goodman, B., 1991. Keeping Anglers Happy Has A Price : Ecological And Genetic Effects Of
Stocking Fish. Bioscience 41: 294-299.
Grimm, E. C., 1983. Chronology and dynamics of vegetation change in the prairie-woodland
region of southern Minnesota, U.S.A. New Phytologist 93: 311-350.
Håkanson, L., 1996. Predicting important lake habitat variables from maps using modern
modeling tools. Canadian Journal of Fisheries and Aquatic Sciences 53(1): 364-382.
Håkanson, L. and Boulion, V. V., 2002. The Lake Foodweb: Modeling Predation and
Abiotic/Biotic Interactions. Backhuys Publishers, Leiden.
Hall, R. I. and Smol, J. P., 1992. A weighted-averaging regression and calibration model for
inferring total phosphorus concentration from diatoms in British Columbia (Canada)
lakes. Freshwater Biology 27: 417-434.
Hall, R. I. and Smol, J. P., 1999. Diatoms as Indicators of lake eutrophication, p 128-168. In The
Diatoms: Applications for the Environmental and Earth Sciences. Stoermer, E. F. and
Smol, J. P, [eds], Cambridge University Press, Cambridge.
Harmon, J. R., 2009. Surface Water Resources, p. 330-345. In Michigan Geography and
Geology. Schaetzl, R., Darden, J., and Brandt, D. [eds.], Pearson, New York, New York.
Harper, D., 1992. Eutrophication of Freshwaters. Chapman and Hall, London.
Harris, A. D., Bradham, D .D., Baumgarten, M., Zuckerman, I. H., Fink, J. C., and Perencevicch,
E. N., 2004. The Use and Interpretation of Quasi-Experiemental Studies in Infections
Diseases. Antimicrobial Resistance 83: 1586-1591.
Hasler, A. D., 1947. Eutrophication of Lakes by Domestic Drainage. Ecology 28(4): 383-395.
Henne, P. D., 2006. Spatial and Temporal Variation in Lake-Effect Snow Control Vegetational
Distributions in the Great Lakes Region. PhD Dissertation. University of Illinois, Urbana.

119

Herlihy, A. T., Kamman, N. C., Sifneos, J. C., Charles, D. Enache, M. D., and Stevenson, R. J.,
2013. Using multiple approaches to develop nutrient criteria for lakes in the
conterminous USA. Freshwater Science 32 (2): 367-384.
Hession, S. L. and Moore, N. 2011. A spatial regression analysis of the influence of topography
on monthly rainfall in East Africa. International Journal of Climatology31: 1440-1456.
Holtan, H., Kamp-Nielsen, L., and Stuanes, A. O., 1988. Phosphorus in soil, water and sediment:
an overview. Hydrobiologia 170: 19-34.
Hunsaker, C. T. and Levine, D. A., 1995. Hierarchical Approaches to the Study of Water
Quality in Rivers. BioScience 45(3): 193-203.
Hupy, C. M. and Yansa, C. H., 2009a. Late Holocene Vegetation History of the Forest Tension
Zone in Central Lower Michigan, USA. Physical Geography 30(3): 205-235.
Hupy, C. M. and Yansa, C. H. 2009b. The Last 17,000 Years of Vegetation History, p. 91-105.
In Schaetzl, R., Darden, J., and Brandt, D. [eds.], Michigan Geography and Geology.
Pearson, New York, New York.
Hutchinson, G. E., 1973. Eutrophication: The scientific background of a contemporary practical
problem. American Scientist 61(3): 269-279.
International Symposium on Eutrophication. 1969. Eutrophication: Causes, Consequences,
Correctives; Proceedings of a Symposium. Washington: National Academy of Sciences.
Johnston, C. A., 1991. Sediment and Nutrient Retention by Freshwater Wetlands: Effects on
Surface Water Quality. Critical Reviews in Environmental Control 21: 491–565.
Johnson, L. B. and Host, G. E., 2010. Recent developments in landscape approaches for the
study of aquatic ecosystems. The Journal of the North American Benthological Society
29(1): 41-66.
Juggins, S., N. J., Anderson, J. M., Hobbs, J. M. R., and Heathcote, A. J., 2013. Reconstructing
epilimnetic total phosphorus using diatoms: statistical and ecological constraints. Journal
of Paleolimnology 49: 373-390.
Kalff, J., 2002. Limnology: Inland Water Ecosystems. Prentice-Hall, Inc., Upper Saddle River,
NJ.
Kenoyer, L. A., 1933. Forest distribution in south-west Michigan as interpreted from the original
survey. Proceedings of the Michigan Academy of Science, Arts, and Letters. 19: 107111.
Khalid, R. A., Patrick, W. H., and DeLaune, R. D., 1977. Division S-2: Soil Chemistry. Soil
Science Society of America Journal 41: 305-310.

120

Kleeberg, A., Hupfer, M., and Gust, G., 2007. Phosphorus Entrainment Due to Resuspension in a
Lowland River, Spree, NE Germany- A Laboratory Microcosm Study. Water, Air, and
Soil Pollution 183: 129-142.
Kost, M. A., Albert, D. A., Cohen, J. G., Slaughter, B. S., Schillo, R. K., Weber, C. R., and
Chapman, K. A., 2007.Natural Communities of Michigan: Classification and Description.
Michigan Natural Features Inventory, Lansing, MI. Report Number 2007-21.
Kuenzel, L. E., 1969. Bacteria, carbon dioxide and algal blooms. Journal of the Water Pollution
Control Federation 41: 1737-1747.
Larkin, G. A. and Slaney, P. A., 1997. Implications of trends in marine-derived nutrient influx to
south coastal salmonid production. Fisheries 22(11): 16-24.
Lewis, K. E. 2009. The Settlement Experience, p. 412-429. In Michigan Geography and
Geology. Schaetzl, R., Darden, J., and Brandt, D. [eds.], Pearson, New York, New York
Line, J. M., ter Braak, C. J. F., and Birks, H. J. B., 1994. WACALIB version 3.3; A computer
program to reconstruct environmental variables from fossil assemblages by weighed
averaging and to derive sample-specific errors of prediction. Journal of Paleolimnology 10:
147-152.
Liu, W., Zhang, Q., and Liu, G., 2010. Lake eutrophication associated with geographic location,
lake morphology and climate in China. Hydrobiologia 644: 289-299.
Liu, W., Zhang, Q., and Liu, G., 2012. Influences of watershed landscape composition and
configuration on lake-water quality in the Yangtze River basin of China. Hydrological
Processes 26: 570-578.
Ludsin, S. A., Kershner, M. W., Blocksom, K. A., Knight, R. L., and Stein, R. A., 2001. Life
After Death in Lake Erie: Nutrient Controls Drive Fish Species Richness, Rehabilitation.
Ecological Applications 11(3): 731-746.
Maguire, D. J., Batty, M., and Goodchild, M. F., 2005. GIS, Spatial Analysis, and Modeling,
ESRI Press, Redlands, CA.
Makarewicz, J. C. and Bertram, P., 1991. Evidence for the Restoration of the Lake Erie
Ecosystem. BioScience 41(4): 216-223.
Marona, R. A., Martin, R. D., and Yohai, V. J., 2006. Robust statistics: Theory and methods.
Wiley, New York.
Martin, S. L., Soranno, P. A., Bremigan, M. T., and Cheruvelil, K. S., 2011. Comparing
Hydrogeomorphic Approaches to Lake Classification, Environmental Management 48:
957-974.
McAndrews, J. H., 1988. Human disturbance of North American forests and grasslands: The
fossil pollen record. p. 673-697, In Vegetation History. Huntley, B. and Webb, T. III.
[eds.] Kluwer Academic Publishers.
121

McAndrews J. H., and. Turton, C. L., 2010. Fungal spores record Iroquoian and Canadian
agriculture in sediment of Crawford Lake, Ontario, Canada. Vegetation History and
Archaeobotany 19: 495-501.
McCarthy, F. M. G. and McAndrews, J. H., 1988. Water Levels in Lake Ontario 4230-2000
years B.P.: Evidence from Grenadier Pond, Toronto, Canadian Journal of
Paleolimnology 1: 99-113.
Medley, K. E. and Harmon, J. R., 1988. Growing season temperature and a Midwestern
vegetation transition. The East Lakes Geographer 23: 130-136.
Miller, W. W., Johnson, D. W., Denton, C., Verburg, P. S. J., Dana, G. L., and Walker, R. F.,
2005. Inconspicuous Nutrient Laden Surface Runoff from Mature Forest Sierran
Watersheds. Water, Air, and Soil Pollution 163: 3-17.
Miner, J. G. and Stein, R. A., 1993. Interactive influence of turbidity and light on larval bluegill
(Lepomis macrochirus) foraging. Canadian Journal of Fisheries and Aquatic Sciences
50: 781-788.
Mitraki, C., 2012. Ontogeny and Littoral Structure of Lakes Created on Phosphate Mined Lands
of Central Florida. PhD Dissertation, University of South Florida.
Mitsch, W. J. and Gosselink, J. G., 2007. Wetlands. John Wiley & Sons, Inc. Hoboken.
Montello, D. R. and Sutton, P. C., 2006. An Introduction to Scientific Research Methods in
Geography. Sage Publications, Thousand Oaks, CA.
Moore, P. D. and Webb, J. A. 1978. An Illustrated Guide to Pollen Analysis. John Wiley and
Sons, New York.
Moser, K. A., Mordecai, J. S., Reynolds, R. L., Rosenbaum, J. G., and Ketterer, M. E., 2010.
Diatom changes in two Uinta mountain lakes, Utah, USA: responses to anthropogenic
and natural atmospheric inputs. Hydrobiologia.
Naumann, E., 1919. Nägra synpunkte angående planktons ŏkôlogi. Med. Sarskild hänsyn till
fytôplankton. Svensk. Bot. Tidskr. 13: 129-158.
Ney, J. J., 1996. Oligotrophication and Its Discontents: Effects of Reduced Nutrient Loading on
Reservoir Fisheries. American Fisheries Society Symposium 16: 285-295.
NOAA National Climatic Data Center, 2002. Michigan Monthly Normals of Temperature,
Precipitation, and Heating and Cooling Degree Days, 1971-2000.
Nürnberg, G. K., 1995. The Anoxic Factor, a Quantitative Measure of Anoxia and Fish Species
Richness in Central Ontario Lakes. Transactions of the American Fisheries Society
124(5): 677-686.

122

Omernik, J. M., Abernathy, A. R., and Male, L. M., 1981. Stream nutrient levels and proximity
of agricultural and forest land to streams: Some relationships. Soil and Water
Conservation Society 36(4): 227-231.
Osborne, L. L. and Wiley, M. J., 1988. Empirical relationships between land use/ cover and
stream quality in an agricultural watershed. Journal of Environmental Management 26: 927.
Overmars, K. P., Koning, G. H. J., and Veldkamp, A., 2003. Spatial autocorrelation in multiscale land use models. Ecological Modeling 164: 257-270.
Pathak, P., Stine, R., Hershey, A., Whalen, S., Nelson, E., and Liu, Z., 2012. Landscape Controls
over Major Nutrients and Primary Productivity in Arctic Lakes. Cartography and
Geographic Information Systems 39(4): 187-198.
Paul, M. J. and Meyer, J. L., 2001. Streams in the urban landscape. Annual Review of Ecology
and Systematics 32: 333–365.
Persson, L., Diehl, S., Johansson, L., Anderson, G., and Hamrin, S. F., 1991. Shifts in fish
communities along the productivity gradient of temperate lakes--patterns and the
importance of size-structure interactions. Journal of Fish Biology 38: 281-293.
Puckett, L. J., 1995. Identifying the Major Sources of Nutrient Water Pollution. Environmental
Science and Technology 29(9): 408-414.
R Core Team, 2013. R: A language and environment for statistical computing. R Foundation for
Statistical Computing, Vienna, Austria. URL http://www.R-project.org/.
Rast, W. and Lee, G. F., 1978. Summary analysis of the North American (U.S. portion) OECD
eutrophication project: nutrient loading-lake response relationships and trophic state
indices. U.S. Environmental Protection Agency Publication (EPA-60013-78-008),
Corvallis, Oregon.
Reavie, E. D., Hall, R. I., and Smol, J. P., 1995. An expanded weighted-averaging model for
inferring past total phosphorus concentrations from diatom assemblages in eutrophic
British Columbia (Canada) lakes, Journal of Paleolimnology 14: 49-67.
Reddy, K. R. and Flaig, E. G., 1995. Phosphorus dynamics in the Lake Okeechobee watershed,
Florida. Special Issue. Ecological Engineering. 5: 121–331.
Reddy, K. R., Kadlec, R. H., Flaig, E., and Gale, P. M. 1999. Phosphorus Retention in Streams
and Wetlands: A Review. Critical Reviews in Environmental Science and Technology 29:
83-146.
Reeder, B. C. and Eisner, W. R., 1994. Holocene Biogeochemical and Pollen History of a Lake
Erie, Ohio, Coastal Wetland. Ohio Journal of Science 94(4): 87-93.

123

Reynolds, C. S. and Davies, P. S., 2001. Sources and bioavailability of phosphorus fractions in
freshwaters: a British perspective. Journal of Biological Reviews 76: 27-64.
Rhemtulla, J. M. and Mlandenoff, D. J., 2007. Why history matters in landscape ecology.
Landscape Ecology 22: 1-3.
Richardson, C. J., 1985. Mechanisms Controlling Phosphorus Retention Capacity in Freshwater
Wetlands. Science 228: 1424-1427.
Richardson, C. J. and Vaithiyanathan, P., 2009. Biogeochemical Dynamics II: Cycling and
Storage of Phosphorus in Wetlands. In Maltby, E. and Barker, T. [eds], The Wetlands
Handbook. Blackwell Publishing Ltd.
Sawyer, C. N., 1947. Fertilisation of lakes by agricultural and industrial drainage. New England
Water Works Association 61: 109-127.
Schaetzl, R. J. 2009., Soils, p. 315-329. In Michigan Geography and Geology. R. Schaetzl, J.
Darden, and D. Brandt [eds.], Pearson, New York.
Scheffer, M., 1998. Ecology of Shallow Lakes. Chapman and Hall, London.
Schindler, D. W., 1974. Eutrophication and Recovery in Experimental Lakes: Implications for
Lake Management. Science 184 (4139): 897-899.
Schindler, D. W., 1977. Evolution of phosphorus limitation in lakes: natural mechanisms
compensate for deficiencies of nitrogen and carbon in eutrophied lakes. Science 195:
260–262.
Seppälä, J., Knuuttila, S., and Silvo, K., 2004. Eutrophication of Aquatic Ecosystems: A New
Method for Calculating the Potential Contributors of Nitrogen and Phosphorus.
International Journal of Life Cycle Assessment 9(2): 90-100.
Sharpley, A. N., Chapra, S. C., Wedepohl, R., Sims, J. T., Daniel, T. C., and Reddy, K. R., 1994.
Managing Agricultural Phosphorus for Protection of Surface Waters: Issues and
Opinions. Journal of Environmental Quality 23: 437-451.
Sims, J. T., Simard, R. R., and Joern, B. C., 1998. Phosphorus Loss in Agricultural Drainage:
Historical Perspective and Current Research. Journal of Environmental Quality 27: 277293.

Smith, V. H., Tilman, G. D., and Nekola, J. C., 1999. Eutrophication: impacts of excess nutrient
inputs on freshwater, marine, and terrestrial ecosystems. Environmental Pollution 100:
179-196.
Smith, R. L. and Smith, T. M., 2001. Ecology and Field Biology. Benjamin Commings, San
Francisco.
Smol, J., 1992. Paleolimnology: an important tool for effective ecosystem management. Journal
of Aquatic Ecosystem Health 1: 49-58.
124

Smol, J., 2002. Pollution of Lakes and Rivers: A Paleolimnological Perspective. Oxford
University Press Inc., New York.
Smol, J. and Stoermer, E., 2010. The Diatoms: Application for Environmental and Earth Science.
Cambridge University Press, New York.
Sollins, P., Robertson, G. P., and Uehara, G., 1988. Nutrient mobility in variable- and
permanent-charge soils. Biogeochemistry 6: 181-199.
Soranno, P. A., Hubler, S. L., Carpenter, S. R., and Lathrop R. C., 1996. Phosphorus Loads to
Surface Waters: A Simple Model to Account for Spatial Pattern of Land Use. Ecological
Applications 6(3): 865-878.
Soranno, P. A., Cheruvelil, K. S., Stevenson, R. J., Rollins, S. L., Holden, S. W., Heaton, S. and
Torng, E. K., 2008. A framework for developing ecosystem-specific nutrient criteria:
Integrating biological thresholds with predictive modeling. Limnology and
Oceanography 53(2): 773-787.
Soranno, P. A., Cheruvelil, K. S. Webster, K. E, Bremigan, M. T., Wagner, T., and Stow, C. A.,
2010. Using landscape limnology to classify freshwater ecosystems for multi-ecosystem
management and conservation. BioScience 60(6): 440-454.
Soranno, P. A., Wagner, T., Martin, S. L., McLean, C., Novitski, L. N., Provence, C. D., and
Rober, A. R., 2011. Quantifying regional reference conditions for freshwater ecosystem
management: A comparison of approaches and future research needs. Lake and Reservoir
Management 27(2): 138-148.
Stein, E. D., Dark, S., Longcore, T., Grossinger, R., Hall, N., and Beland, M., 2010. Historical
Ecology as a Tool for Assessing Landscape Change and Informing Wetland Restoration
Priorities. Wetlands. 30: 589-601.
Stevenson, R. J. and Pan, Y. 1999. Assessing environmental conditions in rivers and streams
with diatoms, p 11-40. In: The Diatoms: Applications for the Environmental and Earth
Sciences [Eds] Stoermer, E. F. and Smol, J. P, Cambridge University Press, Cambridge.
Stevenson, R. J., Zalack, J. T., and Wolin, J., 2013. A multimetric index of lake diatom condition
based on surface-sediment assemblages. Freshwater Science 32(3): 1005-1025.
Stockner, J. G., Rydin, E., and Hyenstrand, P., 2000. Cultural Oligotrophication: Causes and
Consequences for Fisheries Resources. Fisheries 25(5): 7-14.
Svendsen, L. M., Kronvang, B., Kristensen, P., and Grǣsbøl, P., 1995. Dynamics of Phosphorus
Compounds in a Lowland River System: Importance of Retention and Non-point
Sources. Hydrological Processes 9: 119-142.
ter Braak, C. J. F., 1986. Canonical correspondence analysis: a new eigenvector technique for
multivariate direct gradient analysis. Ecology 67: 1167-1179.
125

ter Braak, C. J. F. and Šmilauer, P., 1998. CANOCO reference manual and user’s guide to
CANOCO for windows: software for canonical community ordination (version 4).
Microcomputer Power, Ithaca, N.Y.
Thienemann, A., 1921. Seetypen. Naturwissenschaften. 18: 1-3.
Thomas, M. O. 2009. The United States Public Land Survey System, p. 430-445. In Michigan
Geography and Geology. R. Schaetzl, J. Darden, and D. Brandt [eds.], Pearson, New
York.
Tiner, R. W., 1999. Wetland Indicators: A Guide to Wetland Identification, Delineation,
Classification, and Mapping. Lewis Publishers, Boca Raton.
Tiner, R. W., 2003. Geographically Isolated Wetlands of the United States. Wetlands 23(3): 494516.
Turner, M. G, Gardner, R. H., and O’Neill, R. V., 2001. Landscape Ecology in Theory and
Practice. Springer, New York, NY.
Umbanhowar, C. E., Engstrom, D. R., and Bergman, E. C., 2003. Reconstructing Eutrophication
and Phosphorus Loading for Lake Volney, Minnesota: Combining Lake Sediments and
Land-Use History to Establish ‘Natural’ Baselines for Management and Restoration.
Lake and Reservoir Management. 19(4): 364-372.
USEPA (US Environmental Protection Agency), 1998. National Strategy for the Development of
Regional Nutrient Criteria. EPA 822-R-98-002. Office of Water, US Environmental
Protection Agency, Washington, DC.
USEPA (US Environmental Protection Agency), 2000. Nutrient Criteria Technical Guidance
Manual: Lakes and Reservoirs. EPA 822-B00-001. Office of Water, US Environmental
Protection Agency, Washington, DC.
USEPA (US Environmental Protection Agency), 2007a. Survey of the Nation’s Lakes: Field
Operations Manual. EPA 841-B-07-004. Office of Water, US Environmental Protection
Agency, Washington, DC.
USEPA (US Environmental Protection Agency), 2007b. Survey of the Nation’s Lakes:
Laboratory Manual. EPA 841-B-06-005. Office of Water, US Environmental Protection
Agency, Washington, DC.
USEPA (US Environmental Protection Agency), 2009. National Lakes Assessment: A
Collaborative Survey of the Nation’s Lakes. EPA 841-R-09-001a. Office of Water, US
Environmental Protection Agency, Washington, DC.
USEPA (US Environmental Protection Agency), 2010. National Lakes Assessment:. A
Collaborative Survey of the Nation’s Lakes and Technical Appendix EPA 841-R-09001/a. Office of Water, US Environmental Protection Agency, Washington, DC.
126

USEPA (US Environmental Protection Agency), 2013a, Level III ecoregions of the continental
United States: Corvallis, Oregon, U.S. EPA Accessed September 2013,
http://www.epa.gov/wed/pages/ecoregions/level_iii_iv.htm.
USEPA (US Environmental Protection Agency), 2013b, State Development of Numeric Criteria
for Nitrogen and Phosphorus Pollution. Accessed September 2013,
http://cfpub.epa.gov/wqsits/nnc-development.
van der Werff, A., 1956. A new method of concentrating and cleaning diatoms and other
organisms. Verhandlungen der Internationalen Vereinigung für Theoretische und
Angewandte Limnologie. 12: 276-277.
Van Sickle, J., 2003. Analyzing Correlations Between Stream and Watershed Attributes. Journal
of the American Water Resources Association 39(3): 717-726.
Veith T. L., Srinivasan, M. S., and Gburek, W. J., 2003. Process representation in watershedscale hydrologic models: an evaluation in an experimental watershed. In Paper Read at
First Interagency Conference on Research in the Watersheds, October 27–30 2003,
Benson, AZ.
Vollenweider, R. A., 1968. The scientific basis of lake eutrophication, with particular reference
to phosphorus and nitrogen as eutrophication factors. Technical Report. OECD, Paris.
Vymazal, J., 2007. Removal of nutrients in various types of constructed wetlands. Science of the
Total Environment 380: 48-65.
Wagner, T., Soranno, P., Webster, K., and Spence Cheruvelil, K., 2011. Landscape drivers of
regional variation in the relationship between total phosphorus and chlorophyll in lakes.
Freshwater Biology 56: 1811-1824.
Wang, L., Lyons, J., Kanehl, P., and Gatti. R., 1997. Influences of watershed land use on habitat
quality and biotic integrity in Wisconsin streams. Fisheries 22: 6-12.
Webster, K. E., Soranno, P. A., Cheruvelil, K. S., Bremigan, M. T., Downing, J. A., Vaux, P. D.,
Asplund, T. R., Bacon, L. C., and Connor, J., 2008. An empirical evaluation of the
nutrient-color paradigm for lakes. Limnology and Oceanography 53(3): 1137-1148.
Weninger, J. M. and McAndrews, J. H., 1989. Late Holocene Aggradation in the Lower Humber
River Valley, Toronto, Ontario. Canadian Journal of Earth Sciences 26: 1842-1849.
Wetzel, R. G., 2001. Limnology: Lake and River Ecosystems. Elsevier Academic Press, San
Diego.
WDNR (Wisconsin Department of Natural Resources), 2007. Eutrophication Management
Strategy. Accessed June 2013, http://water.epa.gov/scitech/swguidance/standards/
criteria/nutrients/upload/wi_ncp_080007.pdf.
127

White, P. S. and Walker, J. L., 1997. Approximating nature’s variation: selecting and using
reference information in restoration ecology. Restoration Ecology 5: 338-349.
Williams, R. B, 1971. The chemical composition of water from land drains at Saxmundham and
Woburn, and the influence of rainfall upon nutrient losses. Report of the Rothamsted
Experimental Station for 1970 Part 2: 36-67.
Wolfson, L. 2009. Surface Water Resources, p. 206-222. In Michigan Geography and Geology.
Schaetzl, R., Darden, J., and Brandt, D. [eds.], Pearson, New York.
Yahner, R. H. 2000. Eastern Deciduous Forest: Ecology and Wildlife Conservation, 2
University of Minnesota Press, Minneapolis.

nd

edition.

Zhang, T., Soranno, P. A., Spence-Cheruvelil, K., Kramer, D. B., Bremigan, M. T., and
Ligmann-Zielinska, A., 2012. Evaluating the effects of upstream lakes and wetlands on
lake phosphorus concentrations using a spatially-explicit model. Landscape Ecology 27:
1015-1030.

128