.
.erifl.el:x .
.Kikufifluy
viii!
5 E.

La iv. évlifv \ré
1.... n. .nflELHnszmhflumfi

xs.

5...... ..

x 9.
\ \‘i
I . .

E. 3... a...
. 3.18.:
.. «$.39:
F 3. 52d
.- - Mt &
Wang...

\ I.“ .1
u. .w; m}: .s -
,. . 1&4.
a aha. .. Mr. 3 . .....13..z :HS‘. 1 .
\ I... E... .u.§.4... 1 (3.. .3... . iii-Wu...

n .14 .11th . . . 9
. . ... ’fit!.11lil}[flg1ulli d

 

 

LIBRARY
Michigan State
University

 

This is to certify that the
thesis entitled

EVALUATING STREAM HABITAT IN NORTHERN MICHIGAN:

IMPLICATIONS FOR CONSERVING ARCTIC GRAYLING
(THYMALLUS ARC TICUS)

presented by

Ralph w. Tingley III

has been accepted towards fulfillment
of the requirements for the

MS. degree in Fisheries and Wildlife

 

 

Data”? Jnf-rw: va

Major Professor’ 3 Signature
5 / I .2. / / 0

Date

 

MSU is an Affinnative Action/Equal Opportunity Employer

 

 

 

..-.-.—.—.—.-.--.-.--.-.-»-.-.-.—.—.-.---- a..- -o-l-l-|-A-A-A_-

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

 

DATE DUE DATE DUE DATE DUE

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

5/08 K:IProj/Acc&Pres/ClRC/DateDue.indd

 

EVALUATING STREAM HABITAT IN NORTHERN MICHIGAN: IMPLICATIONS
FOR CONSERVING ARCTIC GRAYLIN G (T H YMALL US ARC TIC US)

By

Ralph W. Tingley III

A THESIS
Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of
MASTER OF SCIENCE
Fisheries and Wildlife

2010

ABSTRACT

EVALUATING STREAM HABITAT IN NORTHERN MICHIGAN: IMPLICATIONS
FOR CONSERVING ARCTIC GRAYLING (TH YMALL US ARC TIC US)

By

Ralph W. Tingley III
The Arctic grayling was the dominant salmonid species in Michigan’s Lower Peninsula
but was extirpated from the state in 1936 for reasons theorized to include overfishing,
competition with other salmonids, and habitat degradation. Several unsuccessful
reintroductions have been attempted, but locations selected for reintroductions were
based on remoteness, lack of competitors, and ability to support other trout species, not
on habitat characteristics specific to grayling. The goal of this thesis is to consider
relationships between landscape predictors and stream habitat and use this information to
identify streams that could support grayling. I conducted a literature review to identify
historical and current information describing habitat requirements of Arctic grayling. I
then examined landscape relationships to in-stream habitat using various multivariate
analytical techniques to assess which landscape conditions may control habitat in my
study region. I then developed an assessment tool to rate habitat at multiple spatial
scales, incorporating predicted landscape effects on habitat and modeled reach data. The
Little Manistee, Pine, White, Pere Marquette, Sturgeon, and Sucker Rivers were the
highest rated stream segments, and critical in-stream habitat characteristics specific to
Arctic grayling were present in these segments. I believe that suitable habitat for Arctic
grayling remains within Michigan that can potentially be considered for the expansion of
the range restricted Montana subpopulation, though biological limitations may exist. I

also believe that my approach can be applied to help restore species in other regions.

To my family and friends.

iii

ACKNOWLEDGEMENTS

This research would not have been possible without the financial support of Schrems
West Michigan Chapter of Trout Unlimited, Paul H. Young Chapter of Trout Unlimited,
Red Cedar Fly Fishers, The Department of Fisheries and Wildlife, The Graduate School,
Dr. Douglas Schemske, and Dr. William Taylor. I would also like to thank Gaylord
Alexander, Glen Eberly, and Robb Smith for personal communications and for supplying

me with historical information. Thank you for believing in my work.

I would like to thank my advisors Dr. Dana Infante and Dr. Douglas Schemske, for all of
their support and assistance. I would also like to thank Dr. Lizhu Wang and Dr. Gary
Roloff for their role as committee members. Additionally, I offer my sincere gratitude to
the numerous individuals who have helped me in the completion of field work and GIS
analysis as well as offering personal and professional support: Jared Ross, Darren
Thombrugh, Daniel Wieferich, Arthur Cooper, Benjamin Felt, Seth Hunt, Amy
Schueller, Jon Deroba, Tracy Kolb, and Andrea Needham. To my Mother, Father, Bea,

Rob and my entire family: Thank you for all of your support and love.

iv

TABLE OF CONTENTS

LIST OF TABLES ........................................................................................................... viii
LIST OF FIGURES ............................................................................................................ x
INTRODUCTION ............................................................................................................... 1

Literature Cited ........................................................................................................ 4
CHAPTER 1

ARCTIC GRAYLING (THYMALLYS ARC TIC US): HISTORICAL AND NEW
PERSPECTIVES FROM ACROSS NORTH AMERICA WITH IMPLICATIONS FOR

MANAGEMENT IN MICHIGAN.

Introduction .............................................................................................................. 5

The Arctic grayling
Classification ................................................................................................ 6
Historical and current range ......................................................................... 7
Morphology .................................................................................................. 8
Habitat requirements .................................................................................... 9
Diet and feeding ......................................................................................... 12
Migration .................................................................................................... 13
Spawning .................................................................................................... 16
The decline of the Arctic grayling in Michigan ......................................... 16
Overexploitation ........................................................................................ l 7
Interspecific competition ........................................................................... 18
Habitat loss and degradation ...................................................................... 19
Previous recovery efforts in Michigan ....................................................... 21
Montana recovery efforts: past and present ............................................... 23
Canadian recovery efforts .......................................................................... 24
United States legal status ........................................................................... 25
Research needs ........................................................................................... 27

Literature Cited ...................................................................................................... 35

CHAPTER 2

EFFECTS OF MULTIPLE SCALES OF LAND COVER ON NORTHERN MICHIGAN
STREAM HABITAT: IMPLICATIONS FOR UNDERSTANDING CONTROLS ON

‘ STREAM SYSTEMS.

Introduction ............................................................................................................ 41

Methods
Study area ................................................................................................... 45
Stream habitat data and collection summary ............................................. 46
Landscape data ........................................................................................... 47
Data analysis .............................................................................................. 48

Results

Landscape and in-stream habitat conditions .............................................. 53

Patterns of in-stream habitat conditions ..................................................... 54
Relationships between landscape and habitat variables ............................. 55
The CSA model .......................................................................................... 56
CSA results ................................................................................................ 57
Effects of landscape on flow stability ........................................................ 57
Effects of landscape variables on in-stream habitat ........................... 58
Discussion
Overview .................................................................................................... 60
Physical landscape controls on stream habitat ........................................... 61
General effects of land cover at different spatial scale: watershed versus
riparian zone ............................................................................................... 63
Specific effects of network and local riparian wetlands on habitat ........... 65
Utility of CSA and RDA ............................................................................ 67
Implications ................................................................................................ 68
Literature Cited ...................................................................................................... 86

CHAPTER 3

DEVELOPMENT AND APPLICATION OF A RATING SYSTEM TO IDENTIFY
AND EVALUATE POTENTIAL ARCTIC GRAYLING (THYMALL US ARC TIC US)
HABITAT IN MICHIGAN STREAMS

Introduction ............................................................................................................ 91
Methods
Study area ................................................................................................... 95
Habitat data collection and summary ......................................................... 96
Landscape and modeled reach data ............................................................ 97
Fish data and index of biotic integrity ....................................................... 99
Overview of the assessment tool ................................................................ 99
Scoring stream reaches ............................................................................ 101
Rating stream segments ........................................................................... 102
Identification of in-stream habitat and validation of assessment tool ...... 103
Results
Landscape conditions ............................................................................... 104
Break points of variables included in the habitat scoring system ............ 105
Scored stream reaches .............................................................................. 109
Rated stream segments ............................................................................. 109
Evaluation of grouped stream segments .................................................. 110
Identification of in-stream habitat critical to Arctic grayling life history
characteristics ........................................................................................... 1 1 1
Presence/absence of other species ........................................................... 112
Discussion
Overview .................................................................................................. 113
Habitat assessment ................................................................................... 114
High rated stream reaches versus former reintroduction locations .......... 116
In-stream habitat of high rated segments ................................................. 117
Consideration of biological factors .......................................................... l 19

vi

The use of thresholds ............................................................................... 120

Regional separation .................................................................................. 121
Management Implications and Future Needs
Conservation of the Arctic grayling ......................................................... 121
The assessment method ............................................................................ 123
Literature Cited .................................................................................................... 160
APPENDICES
APPENDIX A: Correlations among landscape variables .................................... 168

vii

LIST OF TABLES

Table 1.1. Locations, age, and number stocked during most recent Arctic grayling
reintroduction attempt (modified from Nuhfer 1992). Streams marked with an asterisk
were known to contain the species historically .................................................................. 30
Table 2.1. Mean, range and standard deviation of habitat variables used in analyses...
............................................................................................................................................ 72
Table 2.2. Mean, range and standard deviation of landscape variables used in
analyses .............................................................................................................................. 73

Table 2.3. PCA results of 22 habitat variables. Four axes explained 65.05% of the
variation in habitat data. Bold values indicate variable weights with an absolute value

greater than 0.6 that were used to interpret individual axes .............................................. 74
Table 2.4. Pearson pairwise correlations for landscape variables used in CSA.
Variables with correlations higher than an absolute value of 0.2 are shown in bold and
were included in the model ................................................................................................ 75

Table 2.5. Fit statistics for CSA models predicting habitat variables from landscape

factors. Statistics include Chi-square goodness of fit (X2), root mean square error of
approximation (RMSEA), Tucker-Lewis Index (TLI) and Normed Fit Index (NFI). Fit of
each model is indicated by “yes” or “no.” ......................................................................... 76

Table 2.6. Standardized total effects generated through CSA of landscape factors on
10/90 flow ratio. Significant effects are shown in bold (P<0.05) ...................................... 77

Table 2.7. ‘ Standardized total effects generated through CSA of landscape variables
on individual habitat variables. Effects that are statistically significant (P<0.05) are
shown in bold ..................................................................................................................... 78

Table 2.8. Standardized direct effects of landscape variables generated through CSA
on individual habitat variables. Effects that are statistically significant (P< 0.05) are

shown in bold ..................................................................................................................... 80
Table 3.1. Scores applied to each stream reach over all five filters. Two indicates
suitable, 1 is marginal, while 0 is considered unsuitable for Arctic grayling .................. 126
Table 3.2. Variables included in the stream segment rating tool. The reason for
including each variable is described briefly, and the sources of data and literature used to
create suitability scorings are listed ................................................................................. 127

viii

Table 3.3. Length of connected stream segments and the re-ranked score given based
on individual reach habitat score ..................................................................................... 128

Table 3.4. Mean, range and standard deviation (SD) of variables used as filters in the
assessment tool throughout the study region ................................................................... 129

Table 3.5. Final rating of a selection of stream segments including top ranked
segments, segments where reintroductions were performed, and a collection of other
segments where stream field data were available. River segments marked with * are
those that historically contained Arctic grayling and a voucher specimen. Those marked
with ** are streams where a historical record or written account of Arctic grayling occurs

but a voucher specimen does not exist ............................................................................. 130
Table 3.6. Percent of geomorphic units averaged across sites in top-ranked stream
segments ........................................................................................................................... 13 1

Table 3.7. Rifer substrate composition and embeddedness averaged over all riffles
within each high rated stream segment ............................................................................ 132

Table 3.8. Mean and maximum depth, average bank stability, average in stream
cover and overhanging vegetation for all pools in high rated stream segments .............. 133

Table 3.9. Mean and maximum depth, average bank stability, and average in-stream
cover for all runs in high rated stream segments. The average percent of left and right
transect points (stream margins) in which silt was the dominate substrate type is also
presented .......................................................................................................................... 134

Table 3.10. Presence of fish species that could act as competitors or predators of the
Arctic grayling within high rated stream segments. Those species marked with as “X”
were present in sampling efforts. Fish assemblage data were obtained from the Michigan
Department of Natural Resources (DNR) Fish Collection System and Michigan River
Inventory databases (Seelbach and Wiley, 1997) ............................................................ 135

Table A. 1. Correlations among landscape variables .................................................. 168

ix

LIST OF FIGURES

Figure 1.1. Historical ranges of the three major Arctic grayling (T hymallus arcticus)
populations in North America. (Modified from Vincent 1962 and Scott and Crossman
1973) ................................................................................................................................. 31

Figure 1.2. Catchments historically inhabited by the Arctic grayling in Michigan.
Catchments were based on records from Vincent (1962) and voucher specimen locations
from Bailey et al. (2004) ................................................................................................... 32

Figure 1.3. Shoreline of the Au Sable River after heavy logging in the late 1800’s.
Much of the riparian zone was cleared, limiting terrestrial input into the stream and
eliminating shading. Photo courtesy Lovells Township Historical Society and Glen
Eberly ................................................................................................................................. 33

Figure 1.4. A log drive on the Au Sable River during the 1800’s. Rivers were a
useful means of log transportation, however, this had detrimental effects on Michigan
streams including loss of riparian zone vegetation, increased sedimentation, increased
flashiness, breaks in connectivity, and alteration of substrate composition. Photo
courtesy Lovells Township Historical Society and Glen Eberly ....................................... 34

Figure 2.1. Study sites locations and major river basins in Michigan. Forty five sites
were located in the Lower Peninsula and twenty were located in the Upper Peninsula....81

Figure 2.2. Model developed for CSA used to evaluate direct effects of landscape
factors on stream habitat and indirect effects through stream flow. Error terms are
included for endogenous variables and curved arrows linking landscape factors indicate
correlations ......................................................................................................................... 82

Figure 2.3. Sampling sites plotted over A. the first two axes generated by PCA of
habitat variables named “Riffle and coarse substrate” and “Stream stability” respectively
and B. The third and fourth axes named “Stream stability” and “Percent pools and
overhanging vegetation.” Note that two of the axes, responses to flow stability and
stream size, show separation of study sites by peninsula .................................................. 83

Figure 2.4. Percent of total explained variance by RDA predicting habitat variables
from groups of landscape variables. A total of 41 .2% of the variance within habitat
variables was explained by all groups of landscape variables ........................................... 84

Figure 2.5. Plot of the first two axes from redundancy analysis (RDA) with vectors
representing landscape features and points representing physical habitat variables. Gray
text represents landscape variables while habitat variables are shown in black ................ 85

Figure 3.1. Major river basins of Michigan historically inhabited by the Arctic
grayling (Vincent 1962) with streams from which voucher specimens were collected

indicated (Bailey et a1. 2006) ........................................................................................... 136
Figure 3.2. Comparison of modeled July mean temperatures with measured July
means for 26 sites in northern Michigan. Modeled temperatures tended to be
overestimated for 2009 .................................................................................................... 137
Figure 3.3. Habitat assessment tool used to rate stream segments within Michigan as

the most likely to support a population of Arctic grayling. I organized the assessment
over 3 scales, the network catchment, network riparian zone, and reach inclusive of 5
filters including drainage area, catchment landscape condition, riparian zone landscape
condition, reach condition, and temperature. After the cumulative scoring of a reach, I
rescored based on total connectivity and habitat score then rated a stream segment on
total length of unsuitable, marginal, and suitable reaches ............................................... 138

Figure 3.4. Brook trout abundance in stream reaches of my study region plotted
against drainage area. As drainage area increases, brook trout abundance decreases, with

a marked decrease in numbers occurring at 40 km2 ....................................................... 139

Figure 3.5. 10/90 ratio versus network catchment coarse geology (%) for all reaches
in Michigan. Reaches with coarse geology percentages less than 50% generally have
10/90 ratios greater than 8; reaches with coarse geology between 50 and 80% show a
decrease in minimum 10/90 ratio values, and those values greater than 80% have a
minimum value of about 2. .............................................................................................. 140

Figure 3.6. Fish Index of Biotic Integrity (IBI) total score versus forest in the network
catchment (%) across Michigan. Less than 30% forest is related to increased numbers of
sites with low scores. At 30 to 60%, the number of low scores decreases, and few sites

with greater than 60% forest have an IBI less than 40 .................................................... 141

Figure 3.7. Fish Index of Biotic Integrity (IBI) total score versus urban land use in the
network riparian zone (%) across Michigan. At 6%, sites show marked decrease in IBI
score, with few values above 70 present. At 15%, only one value exists above 60 ....... 142

Figure 3.8. Fish Index of Biotic Integrity (IBI) total score versus agriculture in the
network riparian (%) across Michigan. Sites with agricultural values of 10 to 40% show
a decline in IBI score overall, and at 40%, values of agriculture drop sharply ............... 143

Figure 3.9. Visual metric of stream flow stability versus network riparian zone

wetlands (%). Sites scored 8 or higher first appear with more than 25% wetlands, and
sites with less than 4 do not occur at about 25% wetlands .............................................. 144

xi

Figure 3.10. Fish Index of Biotic Integrity (IBI) total score plotted against 10/90 flow
ratio in Michigan. IBI scores decrease at a 10/90 ratio of 20 ......................................... 145

Figure 3.11. Percent of stream reaches scoring a 0, 1 or 2 at each hierarchical filter.
Note that only reach condition and temperature could score a 0. Landscape variables
seem to be overall suitable across segments, while drainage area and reach condition
(10/90 ratio, 90/50 ratio, and gradient) seem to be the most limiting filters ................... 146

Figure 3.12. Rating of individual stream reaches before application of connectivity.
The majority of high ranked reaches occur in the Lower Peninsula, while in the Upper
Peninsula the number of high scoring reaches is lower ................................................... 147

Figure 3.13. Frequency distribution of final stream segment rankings grouped by 0.10
changes in segment score. Black lines indicate breaks between groupings. Sites were
grouped by final rating score into top (1 .50-1 .70), high (1.30-1.49), marginal (1.10—1.29),
low (0.90-1.09), and lowest (0.00-0.89). Stream segments with 0.00 values were not
included within this figure ............................................................................................... 148

Figure 3.14. Final rating of stream segments across Michigan. The majority of high
rankings streams occurs in the Lower Peninsula ............................................................. 149

Figure 3.15. Descriptive statistics for stream segments grouped by final rating, with
statistics for reintroduction sites also shown. Horizontal bars represent mean values,
boxes represent the lower and upper quartiles, vertical bars represent the minimum and
maximum scores, and asterisks represent outliers. The top ranked group of segments has
the highest mean of forest land cover. Reintroduction sites are shown to have a wide
range in forest land cover, but the mean falls above 60%, the percentage determined to be
suitable for Arctic grayling .............................................................................................. 150

Figure 3.16. Descriptive statistics for stream segments grouped by final rating, with
statistics for reintroduction sites also shown. Horizontal bars represent mean values,
boxes represent the lower and upper quartiles, vertical bars represent the minimum and
maximum scores, and asterisks represent outliers. All values fall in the suitable range of
urban land use (less than 5%) within my study region .................................................... 151

Figure 3.17. Descriptive statistics for stream segments grouped by final rating, with
statistics for reintroduction sites also shown. Horizontal bars represent mean values,
boxes represent the lower and upper quartiles, vertical bars represent the minimum and
maximum scores, and asterisks represent outliers. Reintroduction sites have the lowest
mean agricultural percentage, while the top ranked site have the highest average, yet is
below 10%, the break point between good and suitable ranges ...................................... 152

Figure 3.18. Descriptive statistics for stream segments grouped by final rating, with

statistics for reintroduction sites also shown. Horizontal bars represent mean values,
boxes represent the lower and upper quartiles, vertical bars represent the minimum and

xii

maximum scores, and asterisks represent outliers. All mean values are above 25%, but
low segments have some variation below 25%, while marginal segments has one segment
that has less than 25% wetlands present .......................................................................... 153

Figure 3.19. Descriptive statistics for stream segments grouped by final rating, with
statistics for reintroduction sites also shown. Horizontal bars represent mean values,
boxes represent the lower and upper quartiles, vertical bars represent the minimum and
maximum scores, and asterisks represent outliers. All mean temperatures fell close to the
suitable range of less than 16°C, while lower ranked segments and reintroduction sites
have high variance ........................................................................................................... 154

Figure 3.20. Descriptive statistics for stream segments grouped by final rating, with
statistics for reintroduction sites also shown. Horizontal bars represent mean values,
boxes represent the lower and upper quartiles, vertical bars represent the minimum and
maximum scores, and asterisks represent outliers. With increasing coarse geology,
average group scores increase. Reintroduction segments have highly variable scores
ranging from poor to suitable. Stream segments in the low rated group had high variation
in coarse geology as well ................................................................................................. 155

Figure 3.21. Descriptive statistics for stream segments grouped by final rating, with
statistics for reintroduction sites also shown. Horizontal bars represent mean values,
boxes represent the lower and upper quartiles, vertical bars represent the minimum and
maximum scores, and asterisks represent outliers. Mean values were similar among
ranked stream segments, but as rankings decreased groups varied more. Reintroduction
sites had a higher mean gradient than what is considered the upper limit for marginal
(0.010) with rankings as high as 0.0170 .......................................................................... 156

Figure 3.22. Descriptive statistics for stream segments grouped by final rating, with
statistics for reintroduction sites also shown. Horizontal bars represent mean values,
boxes represent the lower and upper quartiles, vertical bars represent the minimum and
maximum scores, and asterisks represent outliers. A general trend of decreasing mean
values and variance in 10/90 ratio occurs as stream segments increase in rank.
Reintroduction sites had a mean 10/90 ratio that was suitable (<10), but had segments
that were poor and marginal as well ................................................................................ 157

Figure 3.23. Descriptive statistics for stream segments grouped by final rating, with
statistics for reintroduction sites also shown. Horizontal bars represent mean values,
boxes represent the lower and upper quartiles, vertical bars represent the minimum and
maximum scores, while asterisks represent outliers. A general trend of increasing mean
values and variance of 90/50 ratio occurs as stream segments increase in rank.
Reintroduction sites had a mean 90/50 ratio well below the good score of 55% ............ 158

Figure 3.24. Descriptive statistics for stream segments grouped by final rating, with

statistics for reintroduction sites also shown. Horizontal bars represent mean values,
boxes represent the lower and upper quartiles, vertical bars represent the minimum and

xiii

maximum scores, while asterisks represent outliers. Top and high rated segments have
higher mean Procedure 51 total score values than those with below high and
reintroduction sites, which all have means close to 100. Cutoff for “good” total Procedure
51 total scores is 105 ........................................................................................................ 159

xiv

Introduction

Through wide-spread alteration of the landscape, stream habitat has been
degraded, changing fish assemblages and causing extinctions as well as reducing the
range of many species globally (Helfman 2007). Since 1989, the number of imperiled
and extinct freshwater and diadromous fish taxa identified by the American Fisheries
Society (AFS) in North America, including distinct populations and seasonal runs, has
nearly doubled to 700 (Jelks et al. 2008). Two subpopulations of the Arctic grayling,
Thymallus arcticus, one extinct and one imperiled, are listed within the conterminous
United States.

The Michigan subpopulation of Arctic grayling was extirpated in 193 6. Factors
theorized to contribute to the loss of the grayling include overfishing and competition
from non-native species as well as habitat degradation due to logging and the creation of
dams and their subsequent effects on rivers and streams (Vincent 1962). Since this time,
several reintroductions have been attempted but none have been successful. The last
reintroduction attempts occurred from 1987 to 1991 and were conducted by the Michigan
Department of Natural Resources (MDNR). Sites were selected for reintroduction based
on remoteness of location, ability to support other trout species including brook trout
(Salvelinusfontinalis) and rainbow trout (Oncorhynchus mykiss), and lack of other
species, but not based on habitat requirements specific to the Arctic grayling.

The imperiled subpopulation within the upper Missouri River, Montana, may be
facing a fate similar to the Arctic grayling in Michigan. Currently, the subpopulation has
been restricted to one catchment, the Big Hole River basin, and is thought to occupy

approximately 5% of its historic range (Kaya 1992). The decline in abundance is

attributed to wide-spread habitat degradation, decreases in flow due to water withdrawals
for agriculture, and the expansion of non-native salmonids such as the brown trout, Salmo
trutta, and the rainbow trout (Kaya 1992). In this region, efforts are underway to save the
dwindling subpopulation, and introductions within the state are being attempted to extend
its current range. Increased interest in conservation of the Montana subpopulation has led
to new research and a greater understanding of the species since the last Michigan
reintroductions. Further, it has helped raise the question of whether range expansion to
historically occupied northern Michigan would be possible.

With the use of GIS and with new information on habitat characteristics important
to Arctic grayling, identifying potentially suitable locations for reintroductions is now
possible. Through the landscape approach, we know that landscape factors including
vegetative land covers, anthropogenic land uses, and physical characteristics like geology
and topography affect physical and biological characteristics of streams (Wang et al.
2003, Allan 2004). If effects on in-stream habitat are understood, landscape conditions
can be used as a surrogate or predictor of stream habitat. However, relationships between
landscape and in-stream habitat features are complex. Landscape conditions affect the
stream in a hierarchical fashion, predicting species assemblages through complex
relationships with intermediate mechanisms that alter habitat, such as flow variability
(Infante and Allen in press). Between regions, the effects of land cover on habitat and
fish assemblage can vary (Utz et al. 2010). The scale at which a landscape variable is
summarized also matters, with landscape condition summarized at multiple scales having
different effects on stream habitat (Wiens 2002). Once such relationships are better

understood, landscape characteristics can be used to help predict where habitat conditions

that are specifically needed for a species such as the Arctic grayling may exist within a
region, helping to pinpoint locations most suitable for reintroductions and range
expansion.

In this thesis, I attempt to identify streams with the highest potential for
supporting Arctic grayling habitat within Michigan. I first conduct a literature review of
historical information on the species within Michigan as well as new research in other
areas. Next, I explore relationships between landscape predictors and in-stream habitat in
the potential range for grayling in Michigan to assess how the landscape affects stream
habitat in our study region. Finally, I develop an assessment tool capable of identifying
streams suitable for Arctic grayling habitat using a hierarchical filter approach
encompassing both landscape and modeled reach scale variables.

With this assessment, I intend to answer the question of whether fluvial habitat
capable of supporting the Arctic grayling exists in Michigan, while creating a tool that
can be applied to any area where landscape data exists. My hope is that the information
generated through this research will be useful for conserving a subpopulation of a very

unique and imperiled fish species.

Literature Cited

Allan, J. D. 2004. Landscapes and riverscapes: the influence of land use on stream
ecosystems. Annual Review of Ecology, Evolution, and Systematics 35:257-284.

Helfman, G. S. 2007. Fish conservation: A guide to understanding and restoring global
aquatic biodiversity and fishery resources. Island Press, Washington, DC.

Jelks, H. L., S. J. Walsh, N. M. Burkhead, S. Contreras-Balderas, E. Diaz-Pardo, D. A.
Hendrickson, J. Lyons, N. E. Mandrack, F. McCormick, J. S. Nelson, S. P.
Platania, B. A. Porter, C. B. Renaud, J. J. Schmitter-Soto, E. B. Taylor, and M. L.

Warren Jr. 2008. Conservation status of imperiled North American freshwater and
diadromous fishes. Fisheries 33:372-407.

Kaya, C. M. 1992. Review of the decline and status of fluvial Arctic grayling, T hymallus
arcticus, in Montana. Proceedings to the Montana Academy of Sciences 52:43-70

Utz, R. M., R. H. Hilderbrand, and R.L.Raesly. 2010. Regional differences in patterns of
fish species loss with changing land use. Biological Conservation 143(3): 688-
699.

Vincent, R. E. 1962. Biogeographical and ecological factors contributing to the decline of
Arctic grayling, Thymallus arcticus (Pallas), in Michigan and Montana. Doctoral
dissertation. University Microfilms, Inc., Ann Arbor, Michigan.

Wang, L., P. Rasmussen, P. Seelbach, T. Simon, M. Wiley, P. Kanehl, E. Baker, S.
Niemela, and P. M. Stewart. 2003. Watershed, reach, and riparian influences on
stream fish assemblages in the Northern Lakes and Forest ecoregion, U.S.A.
Canadian Journal of Fisheries and Aquatic Sciences 60:491-505.

Wiens, J. A. 2002. Riverine landscapes: taking landscape ecology into the water.
Freshwater biology 47:501-515.

Chapter 1

Introduction

In recent years, concern over the conservation and restoration of native fish
species has increased due in part to noted declines in numbers and ranges of many
species globally (Helfman 2007). The most recent compilation of imperiled plus extinct
freshwater and diadromous fishes in North America clearly supports the need for action,
as the list has increased from 363 to 700 taxa since 1989 (Jelks et al 2008). Salmonids
represent 11% of listed taxa, with 46% of these being distinct populations or seasonal
runs (Jelks et al 2008). Two of the distinct populations included are the fluvial
population of Thymallus arcticus, the Arctic grayling, found in Montana streams, and the
extinct Arctic grayling of Michigan (Jelks et al 2008). With growing concern and
attention to the species’ decline in both Montana and parts of Canada (Northcote 1993,
Clarke et al. 2007, Lamothe and Peterson 2007), potential consideration for the expansion
of its range to areas that historically or could potentially support the species may
increase.

Because the loss of the Michigan Arctic grayling occurred over 100 years ago in
the Lower Peninsula (Mershon 1923), information on life history and habitat
characteristics are lacking. However, a historical account of the species’ range and
decline (Mershon 1923) and a review of historical information relating to its decline as
well as its habitat requirements (Vincent 1962) have provided a basis for understanding
the species in Michigan. Literature reviews from Alberta (Northcote 1993) and Alaska
(Armstrong 1986) as well as information on life history characteristics in British

Columbia (Clarke et al. 2007) were extremely useful in identifying sources of

information and summarizing habitat needs of the species. New research in the Upper
Missouri River, Montana, helped in identifying habitat characteristics suitable for the
grayling as well as identifying literature sources (Lamothe and Magee 2003, Lamothe
and Magee 2004, Lamothe and Peterson 2007).

The goal of the proceeding literature review is to provide information needed to
identify habitat for the Arctic grayling in Michigan, which can then be used for
management decisions about possible reintroduction attempts in the future. Objectives
include synthesizing information imperative for understanding the extinction of the
Arctic grayling in Michigan, to review new research describing habitat characteristics
that are important to their survival, to supply the reader with an understanding of
previous reintroduction attempts in Michigan, and to describe recovery efforts in other

parts of the species’ range.
The Arctic Grayling

Classification

The Arctic grayling, T hymallus arcticus, is in the subfamily Thymallinae and the
family Salmonidae, which include salmon, trout, Whitefish, and grayling. Thymallus is
the lone genus within Thymallinae, with five species existing throughout the world: T.
arcticus, T. thymallus, T brevirostris, T. grubii and T. nigrencens (Department of
Commerce et al. 2008). The name Thymallus stems from historical reports describing the
odor of the herb “thyme” left on the hands after handling a specimen (McAllister and
Crossman 1973). Arcticus is the only Thymallus species known to exist within North
America. T. thymallus, the European grayling, is found throughout Europe and parts of

Asia. The Arctic grayling and the European grayling are the two most widespread

species of the genus. Populations of T. brevirostris, T. grubii and T. nigrenscens are
found in Russia and Mongolia. Due to the disjunct nature of Arctic grayling within North
America, several populations were originally considered to be species and later
subspecies. The Michigan populations were named T. tricolor, those in Montana were
named T. montanus, and those in Canada, Alaska, and Asia were called T. signifier
(Vincent 1962). Vernacular names for the Arctic grayling include: grayling, American
grayling, bluefish, Back's grayling, sailfin Arctic grayling and Arctic trout (Scott and
Crossman 1973). Arctic grayling populations that reproduce in lakes are described as
lacustrine populations of the species, and populations within rivers are considered fluvial.
Historical and current range

The Arctic grayling is a holarctic species present in North America and the
northeastern portions of Asia (Scott and Crossman 1973). In North America, the species .
was historically found in three different regions: 1) northwestern Canada and Alaska,
extending into Alberta to the south, Nunavut to the east and Alaska to the northwest, 2)
Montana, specifically in the upper Missouri River Basin, and 3) the northern two thirds of
the Lower Peninsula of Michigan and a one river basin in the Upper Peninsula (Figure
1.1) (Vincent 1962, McAllister and Crossman 1973, Northcote 1993). The two
southernmost regions, Montana and Michigan, are considered glacial refuges for the
species that occurred following the Pleistocene glaciations (Vincent 1962, Northcote
1993). Within Michigan, the Arctic grayling is believed to have been found in most
major rivers in the Lower Peninsula north of and including the White River draining to
Lake Michigan in the west, and north of and including the Rifle River draining to Lake

Huron in the east. In the Upper Peninsula, records of Arctic grayling occur only in the

Otter and Sturgeon Rivers (Figure 1.2) (Bissel 1890, Vincent 1962). The Arctic grayling
has been extinct in Michigan since 1936.

During the 20th century, the range of the Arctic grayling diminished in areas of
North America. In northwestern Canada and Alaska, the species remains widespread, yet
human-induced alteration of habitat has caused declines in some rivers (Clark et al.
2007). Native fluvial Montana grayling have been reduced in range and now are
contained entirely within the Big Hole River system. This decline was attributed to
habitat degradation and changes in flow conditions (Kaya 1992, Lamothe and Peterson
2007). Low stream flows due to water withdrawals, diversions for agriculture, and
drought are detrimental to Arctic grayling, which depend largely on stable, continuous
flow for feeding and migrating to spawning and overwintering habitat (Kaya 1992,
Lamothe and Peterson 2007). Introductions are being attempted in several streams, and
the species is present in approximately 30 Montana lakes (Rens and Magee 2007). Red
Rock Lake and Elk Lake are the only known native lacustrine population of the Arctic
grayling in Montana (Campton 2006, Rens and Magee 2007). California, Utah,
Wyoming, Colorado, Idaho and Washington support introduced populations of the
species in high mountain lakes with limited fishing, with many being supported though
heavy stocking (Morrow 1980, McGinnis 1984, Wydoski 2003).
Morphology

The most notable characteristic of the Arctic grayling is its large, sail-like dorsal
fin. The fin contains between 17 and 25 rays, while most salmonids have less than 15
(Morrow 1980, Page and Burr 1991). The dorsal fin tends to be larger in males and

larger at the posterior end of the fish. The female tends to have the opposite pattern, with

the fin higher in the front and lower in the back (McClane 1978). The caudal fin is
forked, with the lower lobe extended slightly farther back than the upper. Like all
salmonids, grayling have an adipose fin. The body is compressed and slender and is
sometimes described as cigar-shaped. Length is approximately 5.5 times its deepest
point, which is directly below the dorsal fin (Oosten 1940-41, Page and Burr 1991).
Average adult length is about 25 cm in Montana populations while in Alaska and Canada
average length is closer to 35 cm (Schrenkeisen 1963, Morrow 1980). In Alaska and
Canada, individuals are known to grow larger than those in the southern subpopulations,
with fish of over 2 kg taken annually (McClane 1978). The largest known individual was
taken from the Katseyedie River in the Northwest Territories, and was 75.9 cm in length
and 2.7 kg (McAllister and Crossman 1973, Morrow 1980). Coloration is generally
described as grayish purple to olive green on the back, silvery light purple on the sides,
and bluish white on the belly. Red, pink, and green spots can appear on the dorsal fin,
while darker purple to black spots appear on the body. A long black slash is sometimes
notable along the chin (Sigler and Miller 1963). The head is described as moderately
small, with a terminal mouth and a large eye (Morrow 1980). In the continental United
States maximum age is 10, with most living only to the age of 6 years, while in Canada
the oldest individual was 22 years old (DeBruyn and McCart 1974, Wydoski 2003).
Habitat requirements

Arctic grayling prefer intermediate to low gradient streams with a slow, steady
flow (Vincent 1962, Kreuger 1981). Adult grayling in Alaska have been recorded
spending summer months in areas with a velocity of 0.26 m/s (Kreuger 1981). In

Michigan, velocities between 0.30 and 0.61 m/s and a gradient of 0.09-0.28% were

considered optimal for summer habitat (Vincent 1962). In Montana, areas with a velocity
of 0.21 m/s and a gradient of 0.29% were found to be ideal summer habitat (Liknes and
Gould 1981). Overwintering habitat, usually in large pools, had recorded velocities of
less than 0.15 m/s in Alaska (Kreuger 1981). In Montana, high grayling densities have
been found in pools with overhanging vegetation (Lamothe and Magee 2004). These
pools tend to have stable banks, as well as being “high quality pools,” a metric defined by
maximum depths over two feet, increases in percentage of in-stream cover, and total area
(Lamothe and Magee 2004). Arctic grayling in Alaska spawn in fast moving water,
usually riffles, with velocities ranging from 0.34 to 1.46 m/s (Kreuger 1981). After
hatching, embryos spend three to four days hiding in gravel for cover (Kratt and Smith
1977). Flow stability is important, especially at this life stage, as fry are extremely
susceptible to flooding and are easily flushed downstream or out of refuge areas (Nelson
1954). Young of the year move to the side margins of stream channels, with optimal
velocities at this time ranging from 0.07 m/s to 0.16 m/s (Elliot 1980). Young of the year
in Alberta, the Northwest Territories, and Montana streams have been shown to only use
stream margins with little or no silt as nursery areas (Lucko 1992, Barndt and Kaya 2000,
Jones et al. 2004).

The Arctic grayling is a cold water species, requiring low temperatures
throughout the year. Optimal temperatures for growth and maximum critical
temperatures vary depending on the subpopulation studied. In a 1985 study of Arctic
grayling in Alaska, estimates of high rates of growth occurred between 7.5 and 16.0 °C
(Hubert et al.1985). The lethal temperature for these populations was determined to be

20.0 °C (Hubert et al. 1985), while recent work in Montana identified a lethal temperature

10

of 25.0 °C and a stress temperature of 20.0 0C for juveniles and adults (Lamothe and
Peterson 2007). This geographic difference in lethal temperature may be explained by
the differences in stream temperature regimes. In Alaska and Canada, the species may be
acclimated to lower temperatures throughout the year than those of Montana, lowering
both the optimal growth and lethal temperature. For example, in the laboratory, Arctic
grayling acclimated at 8.4 °C have a lethal temperature of 23.0 °C, while those acclimated
at 16.0 and 20.0 °C had a lethal temperature of about 25.0 °C (Lohr et al. 1996). The
latter two acclimation temperatures were used to mimic stream conditions in Montana.

Arctic grayling are not limited by dissolved oxygen (DO) levels to the same
extent as other salmonids. In overwintering areas, DO levels have been recorded as low
as 0.6 mg/l (Bendock 1980). Modeled optimal DO levels during low flow summer
conditions are greater than 4 mg/l, with a critical level of 2 mg/l. An optimal summer
DO level is estimated to be 6 mg/l (Hubert et al. 1985). In comparison, brook trout have
been shown to have a critical value of 4 mg/l in overwintering conditions and 5 mg/l in
summer conditions while optimal DO for summer conditions was estimated to be about 9
mg/l (Raleigh 1982).

Adults are known to inhabit areas within the main flow of the stream, using depth
as cover (Byroth and Magee 1998). However, in Alaska, juvenile grayling in both the
tributaries and the main channel use large woody debris (LWD) and large boulders for
shelter from predators (Kreuger 1981). This suggests that in-stream cover has significant
implications for Arctic grayling at early life stages, but may become less important with

age.

11

Substrate composition in streams supporting Arctic grayling varies by region, as
well as between spawning, feeding, and overwintering habitat. Grayling streams contain
spawning areas characterized by a gravel bottom (Tack 1971, Shepard and Oswald 1989).
In Alaska, spawning riffles are dominated by pea-sized gravel (0.08 to 38.10 mm; Tack
1971, Tack 1973). Substrates in British Columbia were composed of 10-20% fine
sediment, 30-80% gravel and 10-50% boulder (Butcher et al. 1981). In the Big Hole
River, Montana, substrate composition consisted of 20% fine sediments, 50% fine gravel
and 30% large gravel (Shepard and Oswald 1989). The US. Fish and Wildlife Service
(USFWS) habitat suitability model developed for Alaskan Arctic grayling estimated
optimal spawning sites as having greater than 20% fine gravel, and less than 10% fine
sediments (Hubert et al. 1985). Earlier work by Liknes (1890) estimated that 85% of
sites with Arctic grayling in the Big Hole River were either cobble (6.4-26.0 cm) or
gravel (2.0-6.3 cm). Although the specific substrate composition of spawning sites for
Arctic grayling in Michigan is unknown, rivers that they inhabited tended to be sandy,
but had fine gravel present in riffles and bars throughout the stream (Vincent 1962).
Spawning in Michigan was not recorded in areas of mud, silt or clay. Eggs found in
these substrate types were thought to have drifted and become attached to local
vegetation (Nelson 1954, Bishop 1971).

Little is known about effect of pH on Arctic grayling. The USF WS habitat
suitability index indicates that grayling occupy streams that have a pH of 6.4 to 7.8
(Hubert et al. 1985). Low pH has been proposed as a contributing factor to the failure of
lake introductions in the Upper Peninsula of Michigan (Nuhfer 1992).

Diet and feeding

12

 

Grayling begin to feed about four days after hatching (Brown and Buc 1939). Fry
feed mostly on zooplankton, and as they proceed through their first year of life, they shift
towards a diet consisting of small macroinvertebrates (Jones et al. 2003). Juvenile adults
tend to focus mainly on stream drift, including terrestrial and benthic invertebrates
(Vascatto 1970). Adults are largely insectivorous, though they also consume small
mammals (DeBruyn and McCart 1974), the eggs and fry of other fish species, and
sometimes their own eggs (Alt 1980). Although the diet of Arctic grayling varies among
regions, it is generally assumed that they are opportunistic and aggressive feeders
(Vincent 1962, Hubert et al. 1985, Northcote 1993).

Adult fish hold in upper reaches of the main stem, while juvenile fish are located
in the lower reaches (Hughes 1998). It is believed that this occurs because of the
favorable conditions upstream, including increased stream drift and lower temperatures
(Hughes 1998). In individual pools, larger fish tend to be found on the upstream margins,
while smaller ones tend to be on downstream margins (Hughes 1998).

O’Brien et al. (2001) found that Arctic grayling feeding efficiency was unaffected
by current velocity in artificial conditions, but a relationship between availability of more
stream drift was balanced by a decrease in the species’ ability to accurately target prey in
higher flow conditions. However, in natural systems, increased turbidity can lead to a
decrease in feeding success (McLeay et a1. 1987).

Migration

The fluvial Arctic grayling is a migratory species that generally completes its

entire life cycle within one or more streams, although historical records in the Midwest

indicate that Arctic grayling were caught in Lake Michigan (Vincent 1962). Migrations

13

take place during the spring and fall, allowing the species access to habitats for
overwintering, spawning, and summer feeding (Armstrong 1986, Northcote 1993).
Much of the following description of Arctic grayling migration patterns was
interpreted from Armstrong (1986) and Northcote (1993), with emphasis on Alaska and
British Columbia populations. In these areas, Arctic grayling begin their life in spring
spawning areas. This habitat tends to be in tributary streams consisting of fine gravel or
the upper reaches of large, unsilted rivers. Age-0 grayling will spend the majority of the
spring and fall near the spawning site in backwaters, side channels and stream margins.
In the winter months, grayling will migrate to overwintering sites. Occasionally, age-0
grayling will not leave spawning sites, indicating that in some cases these sites can also
serve as overwintering habitat for young of year. In colder regions, overwintering
locations enable fish to survive when feeding and spawning areas are frozen (Tack 1980,
Armstrong 1986). Arctic grayling will remain in these areas until spring, at which time
they will travel to a feeding area or a spawning area if they are sexually mature. Spring
migrations tend to occur anywhere between mid-April and late July (Tack 1980, Shepard
and Oswald 1989, Northcote 1993), and are triggered by increased temperature and flow
(Shepard and Oswald 1989). While mature grayling return to spawning sites, juveniles
migrate to summer feeding sites. Summer feeding sites are usually large, unsilted rivers
with long deep runs and in some cases, clear tributaries. In autumn, the migration cycle
resumes, with all grayling returning to overwintering areas. Buzby and Deegan (2000)
found that in the Kuparuk River, Alaska, the majority of Arctic grayling return to the

same habitat each year.

14

Arctic grayling sometimes migrate long distances. Lamothe and Magee (2003)
found that Arctic grayling in the Big Hole River (Montana) move on average about 40
km annually. Alaskan populations were shown to have migrations up to 100 km, with
most fish migrating between 25 and 75 km a year (West et a1. 1992). In Michigan,
limited connectivity of essential habitat throughout stream reaches was posed as one of
the reasons why stream introductions may have failed (N uhfer 1992), although the
historical presence of beaver in the region call to question the ability of the species to
migrate long distance (Vincent 1962).

An example of short migration routes due to close proximity of specific habitat is
the Sunnyslope Canal population in Teton County, Montana. Sunnyslope is a small
irrigation canal that fl0ws from the Pishkun Reservoir. The Arctic grayling has existed
within the canal since an introduction in the 1940’s (Barndt and Kaya 2000). Grayling
exist only in the canal and within 6 km of the reservoir because during the summer, the
rest of the 55 km canal goes dry. Arctic grayling overwinter in small residual pools in
this upper section and then spawn in small riffles created by release from the dam in
spring. Outflow from the reservoir is through releases covered by 2.5 cm grates, therefore
limiting the size and age of potential predators from exiting the reservoir and entering the
canal (Barndt and Kaya 2000). For the 60 years since its origin, Sunnyslope Canal has
supported a self-sustaining population of Arctic grayling. Due to increased interest in
protecting the fluvial Arctic grayling within Montana, this population is now monitored
and managed (Barndt and Kaya 2000). The lack of connectivity to numerous spawning
riffles with the lack of flow to pools during winter months suggest that habitat conditions

within the canal would not be conducive to a sustainable population. However,

15

Sunnyslope canal has the conditions of larger river systems that Arctic grayling
historically inhabited: low gradient, stable flow, limited predation and competition from
invasive species, and shallow gravel spawning areas (Barndt and Kaya 2000).
Spawning

Spawning occurs after spring migrations over riffle areas of fine gravel, usually in
upper tributaries or areas of the main channel with higher velocities (Vincent 1962).

Unlike most salmonids, grayling do not build nests or redds and instead are broadcast
spawners (Tack 1971). Males are territorial, establishing territories about 60 m2 (Tack

1971). Females remain in surrounding pools or deeper areas, entering male territories
only for short periods to spawn (Tack 1971). The dorsal fin plays a role in the mating
ritual, as the male will fold it over the back of the female, pressing her slightly into the
gravel substrate (Tack 1971). Eggs are then deposited by the female, some being buried,
but most adhering to gravel at other points within the riffle. In lakes, spawning can take
place in inlets and outlets as well as within the lake itself (Bendock 1980, Armstrong
1986), with young of year potentially holding with streams until moving to lakes to
overwinter (DeBruyn and McCart 1974).
The decline of the Arctic grayling in Michigan

The Arctic grayling decline in Michigan was first widely recognized in the 1880’s
although historical accounts suggest an even earlier period (Vincent 1962). Extirpation
of the species from Michigan waters is believed to be related to three main factors:
overexploitation, competition with other salmonids, and habitat loss and degradation
(Vincent 1962). The background and evidence for a role of these factors is given below.

Overexploitation

16

Its desirability as a food source coupled with ease of catch made grayling a target
to anglers of all skill levels and exploitation of the species was common. Perhaps due to
this ease in angling, sport fishing for grayling was extremely popular to the people of
Michigan. An unknown author describes his experiences in fishing for the Arctic grayling
in the late 1800’s:

“The following day we fished along leisurely until we had our live-boxes,

containing each sixty pounds, so full that the fish began to die. Then we passed

over splendid pools in which we could see large schools of grayling on the

bottom without casting a fly; for we would not destroy them in mere

wantonness."

- From Michigan Grayling, Scribers bound magazines, 1879.

Outside the state, the fish gained popularity as well. A small commercial fishery that
shipped grayling to the Chicago area was believed to exist in some parts of Michigan
(Vincent 1962), while angling for the species made Michigan a popular tourist location in
the Midwest (Mershon 1923). While some anglers like those in the previous quote
caught only what they intended to keep, the intense waste of a seemingly endless
resource by an expanding group of anglers was apparent in the Au Sable River:

“Yet they are not out of reach of slaughter, for while I was in the river in August

last two large camps, all non-residents and strangers killed five thousand fish, not

going beyond five miles of the mouth of the North Branch. They salted and
carried away at least half of them. Many were eaten, more were wasted. For two

miles below from their camps decaying fish whitened the stream, and the offal

l7

and fish entrails lefi unburied in camp tainted the air, as the dead fish poisoned the
water”
- From an 1878 article in Recollections of My Fifty Years in
Hunting and Fishing”, W.B. Mershon, 1923.

Because of the intense fishing pressure described in the accounts above, overexploitation
is believed to have caused initial declines in Arctic grayling populations throughout the
Lower Peninsula (Vincent 1962).
Interspecific competition

Exacerbating impacts of overfishing, competition was a second factor that may
have contributed to decline and extinction of the Arctic grayling. Historically, the brook
trout was known to inhabit streams flowing north while those flowing east and west
contained grayling (Mershon 1923). In 1870, brook trout were first stocked into other
streams within the Lower Peninsula, and by 1883, all streams that contained grayling also
contained brook trout (Mershon 1923). The following description of trends in species
abundance was described by Vincent (1962). After their colonization of a stream, brook
trout numbers would increase and grayling would decline over the next 20 years. Some
believed at the time that predation by the brook trout was the cause of grayling decline
(Mershon 1923). As the populations of Arctic grayling diminished in the stream, their
individual size increased, and few fry or yearlings were seen. This led to the belief that
brook trout and perhaps introduced rainbow trout, Oncorhynchus mykiss, were preying on
Arctic grayling eggs and fry (Mershon 1923). However, at least three rivers in the Lower
Peninsula were known to support both grayling and brook trout before 1850 and before

declines were described in the 1870’s: the Jordan, Boardman, and the Boyne, suggesting

18

that these species may have coexisted (Unknown 1879, Vincent 1962). Further, in
Michigan’s Upper Peninsula, the Arctic grayling was only known to inhabit the Otter and
Sturgeon Rivers within the Sturgeon River basin on the Keweenaw Peninsula, where
brook trout were also present (Vincent 1962).

There is further recent evidence that brook trout and the Arctic grayling can
coexist. Brook trout and Arctic grayling occupy different stream microhabitats, and
intraspecifrc competition is expected to be greater than interspecific competition where
gradients are low and species coexist (Byroth and Magee 1998). Vincent (1962) notes
that although grayling were last recorded in Michigan in the upper tributaries and higher
reaches of large streams, they probably were historically more prominent in lower
reaches with steadier flow and lower gradient. It is suspected that populations were
driven to upper reaches of stream systems in search of suitable habitat as logging and
agriculture increased in lower reaches. In these higher gradient areas, the grayling may
have been more likely to be outcompeted by species that prefer such conditions, such as
the brook trout (Vincent 1962). In summary, the proposed competitive displacement of
grayling by brook trout in Michigan does not seem to be supported by current research.
While predation and competition by other species of trout is not well documented, it is
believed that the presence of brown trout, Salmo trutta, and rainbow trout in the Upper
Missouri river may have caused the initial restriction of species range in Montana (Kaya
1992, Campton 2006)

Habitat loss and degradation
Intense deforestation of northern Michigan is believed to be a main cause of

habitat degradation in streams formerly occupied by Arctic grayling. The removal of

19

Michigan’s white pines (Pinus strobus) occurred throughout the mid and late 1800’s, but
largely between 1875 and 1895 (Maybee 1960, Dickmann and Leefers 2003). This rapid
loss of forest was a result of efforts to rebuild Chicago after the Chicago fire of 1871
(October 18) when over 17,500 buildings were destroyed (Dickmann and Leefers 2003).

The streams were the main artery for transport of logs, an open source for logging
companies where logs were piled on rollways or cleared riparian areas until spring when
they were pushed down river during high seasonal flows (Dickmann and Leefers
2003)(Figure 1.3). Loss of riparian woodland could have caused the decline of terrestrial
input of stream drift and the warming of river waters. Large woody debris (LWD) was
removed to allow for more efficient transportation of logs downstream (Vincent 1962)
(Figure 1.4). The loss of LWD caused both the loss of refuge areas important to juvenile
development and a decrease in the abundance of macroinvertebrates and stream drift
(US. Forest Service 1984, Benke et al. 1985).

Foraging and spawning habitats were greatly altered when logs would scrape
banks and bottoms in shallow areas. Stretches of fine gravel riffles once used for
spawning could have been covered by fine sediments and sand (Maybee 1960, Vincent
1962, Miller 1966). The disappearance of young of year and fry that were thought to be
associated with brook trout and rainbow trout predation may in fact be linked to a loss of
spawning habitat induced by logging’s large scale alterations (Vincent 1962, Miller
1966). Increased sand and silt could have also reduced interstitial spacing important for
macroinvertebrate survival, therefore altering and limiting food supply for all stream
species. Not only did the logging affect habitat conditions, but also took a physical toll on

individuals. As Miller (1966) describes

20

“When the logs came down the rivers they raked the spawning beds, destroyed the

eggs or the young fish. In the jam the bark was ground off the Norway pines,

filling the water with fine particles that sifted into the graylings gills. I found

innumerable dead with festered gills, and in every case the fine particles of bark

were the cause.”

- From “The Old Au Sable” by H. B. Miller 1966

Dams were built and managed by private companies and greatly altered flow
regimes in the region (Vincent 1962, Dickman and Leefers 2003).The build-up of water
behind these dams as well as their quick release led to bursts of high flows in systems
historically know for being stable and slow (Vincent 1962). The dams themselves
fragmented connectivity within stream systems, most likely destroying grayling
migration routes. Due to the intense and multi-leveled impacts of logging on stream
health, it would seem that habitat modification is a likely factor contributing to the local
extinction of the Michigan subpopulation.
Previous recovery eflorts in Michigan

Between 1906 and 1991, over 3,000,000 grayling were stocked in Michigan
waters, with two-thirds of all individuals being stocked between 1926 and 1936 (Nuhfer
1992). Arctic grayling were never successfully re-established in Michigan through these
efforts (Leonard 1949, Nuhfer 1992). The most recent stocking attempts occurred from
1987 to 1991 by the Michigan Department of Natural Resources (Nuhfer 1992).
Stocking locations were selected based on their ability to support trout species,
remoteness of location, and scarcity of other fish species (Nuhfer 1992). Yearlings were

stocked from 1987 to 1991 in a total of 13 inland lakes and 7 streams (Table 1.1). Only

21

two of the seven reintroduction streams were known to historically support the species,
the Au Sable below Mio Dam and the Upper Manistee (Vincent 1962, Nuhfer 1992). Fry
were stocked in East Fish Lake in Montrnorency County in 1991 (Nuhfer 1992). Eggs
were taken from a Wyoming lake for 1987, 1989, 1990 and 1991 introductions, and this
source was presumably a result of previous introductions as there are no historical
records of Arctic grayling in Wyoming. The 1988 introduction used eggs from a creek in
Canada’s Northwest Territories where Arctic grayling are native (N uhfer 1992).
Although Arctic grayling persisted in some of the stocked Michigan lakes up to 5
years after introductions, no natural reproduction was detected, and this was proposed to
result from a lack of movement of grayling to spawning inlets and outlets (Nuhfer 1992).
Nuhfer (1992) noted that spawning might have been limited by the absence of imprinting
to spawning areas. Overall survival was hypothesized to be limited by competition for
food, predation by other species, low pH, hooking mortality, illegal harvest, and fungal
infections (Nuhfer 1992). In streams that were stocked with grayling, fish disappeared
after six months, and no reproduction was detected (N uhfer 1992). Stream fish were
believed to have immediately traveled downstream afier being introduced, taking them
into large reservoirs or the Great Lakes, reducing their chance of survival and
reproduction. Reasons for failed restoration included poor habitat quality, fragmentation
of streams by dams, high water temperatures, overfishing and illegal taking in open
fishing areas, and that lacustrine populations were ill adapted to survive in fluvial systems
(Nuhfer 1992). Nuhfer (1992) concluded that Michigan did not have the kinds of large,
high quality streams with few competing species that grayling require. He noted that

future dam removal might lead to increased availability of grayling habitat. This

22

conclusion was based on the inability of Arctic grayling to establish a population in any
of the seven rivers where introductions took place.
Montana recovery eflorts: past and present

In recent years, increased awareness of the declining Arctic grayling
subpopulations in other regions has prompted new approaches to conservation (N orthcote
2003, Lamothe and Peterson 2007). In Montana, the last native fluvial population (Big
Hole River) experienced a sharp decline in the mid-1980’s. This led to the formation of
the Arctic Grayling Recovery Program (AGRP), a consortium headed and largely funded
by the Montana Fish, Wildlife and Parks. The program contains representatives from the
US. Fisheries and Wildlife Service (USFWS), the US. Forest Service (USFS), the
Montana Bureau of Land Management, the Montana Natural Heritage Program, Montana
State University, the University of Montana, the National Park Service, the Montana
Chapter of the American Fisheries Society, and Montana Trout Unlimited (Rens and
Magee 2007). The AGRP’s goals are to 1) understand the factors contributing to the
decline of the Montana Arctic grayling population, 2) monitor and research current
populations, 3) restore habitat, and 4) keep the public informed of all activities and
progress (Rens and Magee 2007). Since 1991, reports on current populations of the
species, recovery efforts and techniques used, and new research have been issued
annually (Rens and Magee 2007).

In 2006, the AGRP developed and implemented a Candidate Conservation
Agreement with Assurances (CCAA). This agreement states that any private landowner
who voluntarily improves habitat and works to keep the Arctic grayling off the

Endangered Species List will not be subject to further restrictions if the species becomes

23

listed (Lamothe and Peterson 2007). Some local ranchers and private landowners have
demonstrated support for the project and have restore habitat by enrolling over 73,000
acres by the end of 2006 (Lamothe and Peterson 2007). Techniques such as increasing
and stabilizing flows, restoring riparian vegetation, and creating fish ladders have been
used. Although existing populations have not had sharp increases, the efforts have been
successful in preventing extinction of the subpopulation. Landowners along the Big Hole
River reduced divergent flows and malfunctioning headgates which control water flow
into a canal, and this led to an increase in flow of over 200 cfs during the spawning
season from 2005 to 2006 (Rens and Magee 2007). This improved conditions for
spawning, and also improved habitat by flushing fine sediment from the system (Rens
and Magee 2007). Severe drought has occurred in Montana in recent years, leading many
to believe that the Big Hole River population would have become extinct without these
conservation efforts (Rens and Magee 2007).

Along with habitat restoration, reintroductions in several fluvial Montana systems
are being attempted as part of the Montana AGRP. Historically, introducing Arctic
grayling to fluvial systems in Montana has been unsuccessful (Rens and Magee 2007).
However, in 2004 remote site incubators (RSIs) were used to introduce grayling eggs
instead of stocking hatchery fish (Rens and Magee 2007). This was in response to
success in grayling introductions in Red Rock Lake, where RSIs were used to expose
grayling eggs to natural physical conditions before hatching (Kaeding and Boltz 2004).
Success in maintaining fluvial populations in the Ruby River has since occurred with the
use of R815, now the only method of reintroduction used by the Montana AGRP.

Canadian recovery efforts

24

Arctic grayling in some areas of Canada are declining due to habitat alteration and
overfishing (Northcote 1993). In Alberta, it is estimated that the Arctic grayling inhabits
only 40 percent of its historical range, and the species is now listed as “sensitive” in this
province (Alberta Sustainable Resource Development 2005). As of 2010, the Arctic
grayling is considered a high priority candidate for listing by the Committee on the Status
of Endangered Wildlife in Canada (COSEWIC 2010). In British Columbia, the Peace
River contained a sustainable population of grayling before construction of the WAC
Bennett Dam, creating the Williston Reservoir in 1967 (Clarke et al. 2007). Since then,
the Peace River population of grayling has diminished and has been classified as
critically imperiled (Northcote 1993, Clarke et al. 2007). This has led to a call for further
research and for increased attention to potential impacts from human activities (N orthcote
1993). Recently, it was discovered that grayling were no longer entering the main stem
of the Peace River (now the Williston Reservoir) and that the populations has become
spatially segregated into small subpopulations within tributaries (Clarke et al. 2007).

This demonstrates how alteration of natural migration routes and the creation of
impoundments, as occurred in Michigan, can negatively impact grayling.
United States legal status

Legal status of the Arctic grayling depends on whether fluvial and lacustrine
populations are genetically distinct. Possible listing as a threatened/endangered species
for grayling began in 1991, when a petition was filed with USFWS to list the fluvial
grayling as an endangered species (Rens and Magee 2007). In 1994, the Montana fluvial
grayling was classified as a candidate species and as a Distinct Population Segment

(DPS) at varying levels. It was believed that differing ability of lacustrine and fluvial

25

populations to maintain position within the stream current was evidence for genetic
diversity (Kaya and Jeanes 1995). Lacustrine grayling would travel downstream, while
fluvial grayling would hold their position within the stream channel, even after moderate
(1 day) acclimation to conditions (Kaya and Jeanes 1995). In 2004, a petition was filed
to emergency list the fluvial grayling due to drought conditions in Montana (Rens and
Magee 2007). In order to be listed as a DPS, the fluvial grayling had to be considered
discrete from other populations of the species and be significant to the future genetic
viability of the species (Campton 2006). To meet these criteria, the species had to either
be separated from other populations by geographic barriers or by international borders.
For the populations in a region to be considered biologically “significant,” they have to
inhabit an ecological setting that is unusual or unique for the taxa and show evidence that
losing the population would create a significant gap in the species’ distribution.

On April 24, 2007, the species was removed as a listing candidate for two reasons
(Wilson and Katzenberger 2007). First, there was not enough information to conclude
that the fluvial Arctic grayling could be listed simply for its genetic differences from that
of lacustrine populations. Findings from previous studies were thought to be a result of
life history differences. This conclusion was based on a review of genetic information on
the species at that time (Campton 2006). However, while Campton (2006) determined
that lacustrine and fluvial populations were not genetically discrete, the Big Hole River
drainage population in both lakes and rivers was discrete from the Red Rock and Elk
Lakes populations in the upper Red Rock River. Therefore, these two groups met the
criteria based on genetic discreteness, but little scientific data existed at the time that

would justify significance (Campton 2006). Second, it was determined that the loss of

26

the Montana fluvial Arctic grayling would not leave a significant gap in the range of the
species (USFWS 2007). Therefore, the Montana population was no longer considered a
DPS and would no longer be eligible for candidacy.

After the 2007 ruling by the USFWS, a coalition of groups and individuals
including the Center for Biological Diversity, the Federation of Fly Fishers, the Western
Watersheds Project, and some individuals, filed a lawsuit over the loss of candidacy
(Backus 2007). In May 2009, the USFWS initiated a voluntary remand of their finding to
consider the fluvial and/or adfluvial populations a DPS under federal law (USFWS
2009). A new study of population genetic structure demonstrated that grayling from
Wyoming, Saskatchewan, and Montana and surrounding lacustrine Arctic grayling were
genetically distinct from Canadian populations (Peterson and Arden 2009). Additionally,
it found that the Big Hole River subpopulation was the most diverse from Canadian
Arctic grayling (Peterson and Arden 2009). This result may play an important role in the
final ruling, to be made in August 2010.

Research needs

Given the declining or unsustainable populations of Arctic grayling in the
conterrninous United States and some areas of Canada, restoration of the species to its
native range should be considered. Recent research in Montana has shown that different
stocking methods can enhance success of Arctic grayling introductions. The finding that
early acclimation to stream flow conditions is important for holding within the current
has shown that stocking of later developmental stages such as fingerlings, yearlings or

even fry may be reduce survival and establishment (Rens and Magee 2007).

27

Degraded habitat contributes to Arctic grayling population declines. Recent
recovery efforts in areas where the species still exists have shown some success. For
example, introduction attempts in the Ruby River, Montana have shown successful use of
RSI to produce Arctic grayling that remain within the stream and potentially have
reproduced naturally (Rens and Magee 2007). Although populations in the Big Hole
River remain low, it is believed that habitat restorations specific to the Arctic grayling’s
needs, such as increases in flow and flow stability, have saved the population from
extinction during severe droughts (Rens and Magee 2007).

While Arctic grayling may be similar to brook trout in some respects, such as the
need for stable flows and cold water temperatures, there are important differences in their
habitat requirements. Arctic grayling prefer low to moderate gradient streams, while
brook trout prefer higher gradient streams. Brook trout use large woody debris and other
structure as cover, while Arctic grayling primarily use depth. Brook trout are known to do
best in streams with heavy canopy cover (50 —— 75%), while Arctic grayling have been
shown to thrive in streams with much less (<5%) (Raleigh 1982, Lamothe and Magee
2004). If Arctic grayling were introduced to higher gradient streams, they would likely
move to downstream reservoirs if present where they may be subject to predators and
high temperatures. Many of the Michigan streams in which Arctic grayling were
introduced have a moderately high to high stream gradient and were within 16 km of a
reservoir or Great Lake (N uhfer 1992). This limited connectivity coupled with no
acclimation to stream flow could also be a factor for a species that migrates on average
40 km a year to spawning sites, especially if local stream conditions are not optimal.

Grayling, unlike brook trout, can survive at low dissolved oxygen levels, a factor that

28

may make them more apt to thrive in lower gradient, slower streams. Brook trout and
Arctic grayling spawn in different areas, with the latter largely favoring fine gravel and
broadcast spawning. Brook trout construct redds within the riffle and select spawning
locations based on upwellings as opposed to substrate size (Raleighl982). Arctic grayling
seem most suited for large, deep streams with runs and pools that have steady, cold flows
and minimum human influence.

Efforts to reintroduce grayling should first attempt to identify those locations
most likely to have habitat characteristics that could support the species. Using ArcGIS, I
can examine large regions for suitable Arctic grayling habitat using relationships known
to exist between landscape and habitat conditions. This would allow for more in-depth
analysis of the existence of suitable habitat within the state and help direct managers

towards future decisions on possible reintroductions.

29

Table 1.1: Locations, age, and number stocked during most recent Arctic grayling
reintroduction attempt (modified from Nuhfer 1992). Streams marked with an asterisk
were known to contain the species historically.

 

Total
Kg County Stream/lake Age stocked
1987
Alcona Horseshoe Lake yearlings 1500
Alcona Reid Lake yearlings 1500
Alger Ackerman Lake yearlings 1375
Alger Chapel Creek yearlings 800
Alger Kettlehole Lake yearlings 700
Alger Section 34 Creek yearlings 5400
Alger Spray Creek yearlings 5290
Antrim Cedar Creek yearlings 3000
Crawford Kneff Lake yearlings 1424
Crawford Manistee River yearlings 18000
Grand Traverse Sand Lake yearlings 1700
Houghton Penegore Lake yearlings 1000
Kalkaska Manistee River* yearlings 13139
Luce Deer Lake yearlings 1200
Luce Sid Lake yearlings 1000
Marquette Mulligan Creek yearlings 2000
Montmorency East Fish Lake yearlings 1600
Montmorency Fuller Pond yearlings 1500
Oscoda Au Sable River* yearlings 40320
Schoolcraft Dutch Fred Lake yearlings 1000
1988
Alger Ackerman Lake yearlings 926
Alger Section 34 Creek yearlings 4136
Alger Spray Creek yearlings 4259
Alcona O'Brien Lake yearlings 1334
Antrim Cedar River yearlings 2938
Crawford Manistee River* yearlings 9634
Luce Deer Lake yearlings 840
Luce Sid Lake yearlings 1000
Oscoda Au Sable River* yearlings 13795
1990
Alger Ackerman Lake 4-month fingerlings 1400
Luce Deer Lake 4-month fingerlings 900
Luce Sid Lake 4-month fingerlings 646
1991
Baraga West Branch Huron 4-month fingerlings 25230
Luce Deer Lake 4-month fingerlings 2400
Luce Sid Lake 4-month fingerlings 2200
Marquette Mulligan Creek 4-month fingerlings 10000
Montmorency East Fish Lake fry 62160

 

30

 

Historical range ofAlaska and Canadian subpopulation
- Historical range of Michigan subpopulation
I: Historical range of Montana subpopulation

s
O 1.400 2,80016Iometers

Figure 1.1: Historical ranges of the three major Arctic grayling (Thymallus arcticus)
populations in North America. (Modified from Vincent 1962 and Scott and Crossman
1973).

31

SI]°0‘0'W 85°0‘0'W
I |

45°0‘0'N —

 

l |
90°0'0'w 86°0‘0'W
Streams with vouch er specimens

 

:i Basins with historical records ofArctic grayling ‘4

Luv} Major river basins -.- ._ ' _ C
O 80 160 320 Kilometers ‘ _
|

llllllll 5

Figure 1.2: Catchments historically inhabited by the Arctic grayling in Michigan.
Catchments were based on records from Vincent (1962) and voucher specimen
locations from Bailey et a1. (2004).

32

 

Figure 1.3: Shoreline of the Au Sable River after heavy logging in the late 1800’s. Much
of the riparian zone was cleared, limiting terrestrial input into the stream and eliminating
shading. Photo courtesy Lovells Township Historical Society and Glen Eberly.

33

 

Figure 1.4: A log drive on the Au Sable River during the 1800’s. Rivers were a useful
means of log transportation, however, this had detrimental effects on Michigan streams

including loss of riparian zone vegetation, increased sedimentation, increased flashiness,
breaks in connectivity, and alteration of substrate composition. Photo courtesy Lovells

Township Historical Society and Glen Eberly.

34

Literature Cited

Alberta Sustainable Resource Development. 2005. Status of the Arctic grayling
(Thymallus arcticus) in Alberta. Alberta Sustainable Resource Development, Fish
and Wildlife Division, and Alberta Conservation Association, Wildlife Status
Report 57. Edmonton, Alberta.

Alt, K. 1980. Inventory and cataloging of sport fish and sport fish waters of western
Alaska. Alaska Department of Fish and Game, Juneau, Alaska. 1979-1980,
Project F-9-l2, 21(G-I-P):32-59.

Armstrong, R. H. 1986. A review of Arctic grayling studies in Alaska, 1952-1982.
Institute of Arctic Biology, University of Alaska, Fairbanks, Alaska.

Backus, P. 2007. Groups file lawsuit over Arctic grayling ruling. 27 November 2007. The
Missoulian, newspaper article, Missoula, Montana. 28 December 2008.
<www.missoulian.com/articles/2007/1 1/27/news/local/news05.txt>

Bailey, R.M., W.C. Latta and GR. Smith. 2004. An Atlas of Michigan fishes with keys
and illustrations for their identification. University of Michigan, Ann Arbor,
Michigan

Barndt, S. A. and C. M. Kaya. 2000. Reproduction, growth and winter habitat of Arctic
grayling in an intermittent canal. Northwest Science 74:294-3 05.

Bendock, T. 1980. Inventory and cataloging of Arctic area waters. Alaska Department of
Fish and Game, Federal Aid in Fish Restoration, Report 21.

Benke, A. R. Henry, D. Gillespie, and R. Hunter. 1985. Importance of snag habitat for
animal production in southeastern streams. Fisheries 10:8-13.

Bishop, F. G. 1971. Observations on spawning habits and fecundity of the Arctic
grayling. Progressive Fish Culturist 33:12-19.

Bissel, J. H. 1890. Grayling in Michigan. Transactions of the American Fisheries Society
19: 27-29

Brown, C. J ., and C. Buc, Jr. 1939. When do trout and grayling fry begin to take food?
Journal of Wildlife Management 3: 134-140.

Butcher, G. A., J. R. Ellis and R. B. Davidson. 1981. Aspects of life history of Arctic
grayling (T hymallus arcticus) in the Liard River drainage, British Columbia.
British Columbia Ministry of Environment, Aquatic Studies Branch, Victoria.

35

Buzby, K. M., and L. A. Deegan. 2000. Inter-annual fidelity to summer feeding site in
Arctic grayling. Environmental Biology of Fishes 3:319-327.

Byorth, P. A., and J. P. Magee. 1998. Competitive interactions between Arctic grayling
and brook trout in the Big Hole River drainage, Montana. Transactions of the
American Fisheries Society 127:921-931.

Campton, D. E. 2006. Should Arctic grayling in the Big Hole River, Montana be
considered a Distinct Population Segment (DPS) under the US. Endangered
Species Act? An evaluation of existing genetic information. Report for the US.
Fish and Wildlife Service, Abernathy Fish Technology Center, Longview,
Washington.

Clarke, A. D., K. H. Telmer, and J .M. Shrimpton. 2007. Habitat use and movement
patterns for fluvial species, the Arctic grayling, in a watershed impacted by a
large reservoir; evidence from otolith microchemistry. Journal of Applied
Ecology 44:1156-1165.

Committee on the Status of Endangered Wildlife in Canada (COSEWIC). Candidate
wildlife species. 27 November 2003. Committee on the Status of Endangered
Wildlife in Canada. 11 February 2010.
<http://www.cosewic.gc.ca/eng/sct3/index_e.cfm#p3>

DeBruyn, M., and P. McCart. 1974. Life history of the grayling (T hymallus arcticus) in
Beaufort Sea Drainages in the Yukon Territory. Fisheries research associated
with proposed gas pipeline routes in Alaska, Yukon and Northwest Territories,
Report 15. Canadian Arctic Study, LTD, Calgary, Alberta.

Department of Commerce, Department of Interior, Environmental Protection Agency,
Department of Agriculture, Smithsonian Institute. 1 August 2008. Integrated
Taxonomic Information System. 1 December 2008.
<http://www.itis.gov/servlet/SingleRpt/SingleRpt?search_topic=TSN&search_val
ue=162015>

Dickman, D., and L. A. Leefers. 2003. The Forests of Michigan. The University of
Michigan Press, Ann Arbor, Michigan.

Elliott, G. V. 1980. First interim report on the evaluation of stream crossings and effects
of channel modifications on fishery resources along the route of the trans-Alaska

pipeline. US. Fish and Wildlife Service, Special Studies, Anchorage, Alaska.

Helfman, G. S. 2007. Fish conservation: A guide to understanding and restoring global
aquatic biodiversity and fishery resources. Island Press, Washington, DC.

36

Hubert, W. A., R. S. Helzner, L. A. Lee, and P. C. Nelson. 1985. Habitat suitability index
models and instream flow suitability curves: Arctic grayling riverine populations.
US. Fish and Wildlife Services, Biological Report 82(10.110).

Hughes, N. F. 1998. A model of habitat selection by drift-feeding stream salmonids at
different scales. Ecology 79: 281-294.

Jelks, H. L., S. J. Walsh, N. M. Burkhead, S. Contreras-Balderas, E. Diaz-Pardo, D. A.
Hendrickson, J. Lyons, N. E. Mandrack, F. McCormick, J. S. Nelson, S. P.
Platania, B. A. Porter, C. B. Renaud, J. J. Schmitter-Soto, E. B. Taylor, and M. L.
Warren Jr. 2008. Conservation status of imperiled North American freshwater and

diadromous fishes. Fisheries 33:372-407.

Jones, N. E., W. M. Tom, and G. J. Scrimgeour. 2003. Selective feeding of age-0 Arctic
grayling in lake-outlet streams of the Northwest Territories, Canada.
Environmental Biology of Fishes 67: 169-178.

Kaeding, L. R., and G. D. Boltz. 2004. Use of remote-site incubators to produce Arctic

grayling fry of wild parentage. North American Journal of Fisheries Management
24:1031-1037.

Kaya, C. M. 1992. Review of the decline and status of fluvial Arctic grayling, Thymallus
arcticus, in Montana. Proceedings to the Montana Academy of Sciences 52:43-70

Kaya, C. M and E. D. Jeanes. 1995. Rentention of adaptive rheotactic behavior by F1
fluvial Arctic grayling. Transactions of the American Fisheries Society 1242452-
457

Kratt, L. F., and R. J. F Smith. 1977. A post-hatching sub-gravel stage in the life history
of the Arctic grayling, T hymallus arcticus. Transactions of the American Fisheries
Society 106: 241-243.

Kreuger, S. W. 1981. Freshwater habitat relationships: Arctic grayling (T hymallus
arcticus). Alaska Department of Fish and Game, Habitat Division, Anchorage,
Alaska.

Lamothe, P. J ., and J. P. Magee. 2003. Movement and habitat selection of Arctic
grayling, brook trout, and mountain Whitefish during drought conditions in the
Big Hole River, MT. Report prepared for Montana Fish, Wildlife and Parks,
Dillon, Montana.

Lamothe, P. J ., and J. P. Magee. 2004. Linking Arctic grayling abundance to physical

habitat parameters in the Upper Big Hole River, Montana. Report prepared for
Montana Fish, Wildlife and Parks, Dillon, Montana.

37

Lamothe, P. J ., and A. Peterson. 2007. Candidate Conservation Agreement with
assurances for fluvial Arctic grayling in the upper Big Hole River. Report
prepared for the Montana Fish Wildlife and Parks, Dillon, Montana.

Leonard, J .W. 1949. The Michigan Grayling. Report prepared for Michigan
Conservation. 1212: 1-4.

Liknes, G. A., and W. R. Gould. 1987. The distribution, habitat, and population
characteristics of fluvial grayling (Thymallus arcticus) in Montana. Northwest
Science 61:122—129.

Lohr, S. C., P. A. Byroth, C. M. Kaya, and W. P. Dwyer. 1996. High temperature
tolerances of fluvial Arctic grayling and comparisons with summer river
temperatures of the Big Hole River, Montana. Transactions of the American

Fisheries Society 125:933-939.

Lucko, B. 1992. Distribution of the young of year Arctic grayling sport fishery on the
Kakisa River, Northwest Territories, 1983 and 1984. Canadian Data Report on
Fisheries Aquatic Sciences 612.

Maybee, R. H. 1960. Michigan’s white pine era, 1840-1900. Michigan Historical
Commission, Lansing, Michigan.

McAllister, D. E., and E. J. Crossman. 1973. A guide to the freshwater sport fishes of
Canada. National Museums of Canada, Ottawa, Canada.

McClane, A. J. 1978. McClane’s field guide to freshwater fishes of North America. Holt
Reinhart and Winston, New York, New York.

McGinnis, S. M. 1984. Freshwater fishes of Canada. University of California Press,
Berkley, California.

McLeay, D. J ., I. K. Birtwell, G. F. Hartman, and G. L. Ennis. 1987. Responses of Arctic
grayling flhymallus arcticus) to acute and prolonged exposure to Yukon placer
mining sediment. Canadian Journal of Fisheries and Aquatic Sciences. 44:65 8-

673.

Mershon, W. B. 1923. Recollections of my 50 years hunting and fishing. The Stratford
Company, Boston, Massachusetts.

Miller, H. L. 1966. The old Au Sable. B. Eerdmuns Publishing Company, Grand Rapids,
Michigan.

Morrow, J. E. 1980. The freshwater fishes of Alaska. Alaska Northwest Publishing,
Anchorage, Alaska.

38

Nelson, P. H. 1954. Life history and management of the American grayling (Thymallus
arcticus tricolor) in Montana. Journal of Wildlife Management 18:324-342.

Northcote, T. G. 1993. A review of management and enhancement options for the Arctic
grayling (Thymallus arcticus) with special reference to the Williston Reservoir
watershed in British Columbia. Peace/Williston Fish and Wildlife Compensation
Program Research Report 55, Prince George, British Columbia.

Nuhfer, A. J. 1992. Evaluation of the reintroduction of the Arctic grayling into Michigan
lakes and streams. Michigan Department of Natural Resources Fisheries Research
Report, Ann Arbor, Michigan.

O’Brien, J. W., M. Barfield, and K. Sigler. 2001. The functional response of drift feeding
Arctic grayling: The effects of prey density, water velocity and location
efficiency. Canadian Journal of Fisheries and Aquatic Sciences 58:1957-1963.

Oosten, J. V. 1940-1941. Fishes of Michigan. Newspaper Series, Detroit News, Detroit,
Michigan.

Page, L.M., and BM. Burr. 1991. Peterson field guide to freshwater fishes. Houghton
Mifflin Company, New York, New York.

Peterson, D. P and W. R. Arden. 2009. Ancestry, population structure, and conservation
genetics of Arctic grayling (Thymallus arcticus) in the upper Missouri River,
USA. Canadian Journal of Fisheries and Aquatic Sciences 66: 1758-1774.

Raleigh, R. F. 1982. Habitat suitability index models: Brook trout. US. Department of
the Interior, Fish and Wildlife Service Biological Report 82(10.24).

Rens, E., and J. Magee. 2007. Arctic grayling recovery program: Montana Arctic
grayling monitoring report. Montana Fish, Wildlife and Parks, Dillon, Montana.

Scott, W. B. and E. J. Crossman, 1973. Freshwater fishes of Canada. Bulletin of the
Fisheries Research Board of Canada. Fisheries Research Board of Canada, Ottawa

Schrenkeisen, RM. 1963. Field book of freshwater fishes of North America: North of
Mexico. G.P. Putnam’s Sons, New York, New York.

Shepard, B. B. and R. A. Oswald. 1989. Timing, location and population characteristics
of spawning Montana Arctic grayling (Thymallus arcticus montanus) in the Big
Hole River Drainage,1988. Montana Department of Fish, Wildlife and Parks,
Bozeman, Montana,

Sigler, W.F., and RR. Miller. 1963. Fishes of Utah. Utah Department of Fish and Game
publications, Salt Lake City, Utah.

39

Tack, S. L. 1971. Distribution, abundance, and natural history of the Arctic grayling in
the Tanana River drainage. Project F -9-3, Alaska Dept. Fish and Game, Report
12.

Tack, S.L. 1973. Distribution, abundance, and natural history of the Arctic grayling in the
Tanana River drainage. Project F-9-6, Alaska Department of Fish and Game,
Federal Aid in Fish Restoration, Report 15.

Tack, S.L. 1980. Migrations and distribution of Arctic grayling, T hymallus arcticus
(Pallas), in interior and arctic Alaska. Project F-9-l2, Alaska Department of Fish
and Game, Federal Aid in Fish Restoration, Report 21.

Unknown Author. 1879. Michigan grayling. Scribers Bound Magazine. Magazine series.
November 1879- April 1880.

US. Forest Service. 1984. The seen and unseen world of the fallen tree. General
Technical Report PNW-164.

U.S. Fisheries and Wildlife Service. 2007. Endangered and threatened wildlife and
plants: Status review of Arctic grayling (T hymallus arcticus) in the Upper
Missouri River system. Federal Register 72: 20305-20306

U.S. Fisheries and Wildlife Service. 2009. Endangered and threatened wildlife and
plants: Status review of Arctic grayling (Thymallus arcticus) in the Upper
Missouri River system. Federal Register 74(207):55524-55525.

Vascatto, G. L. 1970, Summer ecology and behavior of the grayling of McManus Creek,
Alaska. Masters thesis. University of Alaska, Fairbanks, Alaska.

Vincent, R. E. 1962. Biogeographical and ecological factors contributing to the decline of
Arctic grayling, Thymallus arcticus (Pallas), in Michigan and Montana. Doctoral
dissertation. University Microfilms, Inc., Ann Arbor, Michigan.

West, E. B., M. W. Smith, W. E. Barber, J. B. Reynolds and H. Hop. 1992. Autumn
migration and overwintering of Arctic grayling in coastal streams of the Arctic

National Wildlife Refuge, Alaska. Transactions of the American Fisheries Society
121:709-715.

Wilson, M., and D. Katzenberger. 2007. Upper Missouri River fluvial Arctic grayling no
longer candidate species. 24 April 2007. United State Fish and Wildlife Service,
Mountain Prairie Region, press release, Lakewood, Colorado. 1 December 2007.
<http://www.fws.gov/mountain%2Dprairie/pressrel/07-29.htm>

Wydowski, R. S. 2003. Inland fishes of Washington, 2Ind edition. American Fisheries
Society in association with University of Washington Press, Seattle, Washington.

40

Chapter 2

Introduction

Understanding landscape effects on factors that control local stream physical
habitat and biological assemblages has been evolving since it was first formally
acknowledged that landscape characteristics of river valleys influence streams (Hynes
1975). A seminal work by Vannote et al. (1980) represents an early attempt to consider
streams in a landscape context by classifying stream segments throughout river networks
by types and sources of energy derived in part from the terrestrial environment. Frissell et
al. (1986) created a hierarchical framework for defining streams on several
spatiotemporal scales that acknowledged the way in which catchment factors predict
conditions within stream systems from the river network down to local microhabitats of
individual sites. Following an approach developed by Tom (1990) that attempted to
account for large-scale influences on fish assemblages of lakes and incorporating ideas of
Frissell et al. (1986), Poff (1997) proposed a filter approach for describing landscape
effects on stream systems. Poff (1997) postulated that the combination of landscape and
biological filters over several scales (from the watershed to the microhabitat within a
reach) could predict stream fish distributions of species based on their functional
relationships. Poff (1997) also noted that in order to make filter approaches applicable,
further investigation of scale-specific landscape to habitat relationships needed to be
conducted. Together, these ideas underscore the importance of landscape effects in
understanding mechanisms controlling habitat and biological assemblages of fluvial

ecosystems and also suggest the potential relevance of considering landscape effects at

different spatial scales (Wiens 2002).

41

Many studies have attempted to understand effects of landscape factors at
different scales, focusing largely on anthropogenic land uses (Allan 2004), and both
urbanization and agriculture have been shown to be strongly associated with degraded
fluvial fish and macroinvertebrate communities. For example, Wang et al. (1997) found
that in Wisconsin streams where the basin was dominated by agriculture, catchment land
use explained the most variance in a fish index of biotic integrity (IBI) as well as metrics
describing habitat condition, with increasing catchment agriculture related to declines in
both biological and physical condition. Roth et al. (1996) found similar results in
southeastern Michigan streams; over three spatial scales of land cover examined — the
catchment, the riparian zone 1500 m upstream of the reach, and local reach riparian zone
— agriculture negatively affected fish IBI and habitat index (HI) scores, with catchment
agriculture showing the strongest relationship to stream conditions. In central Michigan,
macroinvertebrate functional groups and species richness were more strongly related to
physical habitat within a reach than to catchment agricultural land use (Richards et al.
1997). However, it was noted that high correlations between land use and surficial
geology at the catchment scale may have masked effects of land use, as row crop
agriculture was highly correlated with lacustrine clay geology (Richards et al. 1997).
Fitzpatrick et al. (2001) found varying effects of different scales of land cover on
agricultural streams in eastern Wisconsin. Fish IBI was most strongly related to land
cover in the riparian buffer zone upstream of the reach and decreased with increasing
human land use at that scale, while a habitat index was related to variables within the
reach as well as geology at the network catchment, land cover at the reach catchment and

network riparian buffer, and riparian vegetation width of the reach. Finally, streams with

42

catchments dominated by urban land use in the Etowah River basin, Georgia, showed no
significant differences in IBI or HI between reaches with forested and non-forested
buffers, suggesting that urban land use in the watershed predominately influenced stream
conditions (Roy et al 2003).

While relationships between anthropogenic land uses at multiple scales are known
to impact stream conditions, the role of natural landscape features like geology, climate,
and topography are also important. Studies have demonstrated how these factors
influence both biological and physical characteristics of stream systems, and even in
regions heavily impacted by human land use, the control of these factors on stream
characteristics must be considered (Van Sickle and Hughes 2000). Wang et al. (2003)
found that in minimally impacted catchments, species distribution and number of species
are explained by in-stream physical habitat. However, Wang et al. (2003) also found that
species assemblages are indirectly controlled by landscape effects on local conditions.
Townsend et al. (2003) examined the influence of land use and landscape characteristics
on fish assemblage, flying macroinvertebrates, and non-flying macroinvertebrates at three
spatial scales, including the network catchment, the reach riparian zone, and the riffle as
well as its immediate riparian zone along with in-stream conditions at the reach and riffle
scale and geographic location. Fish assemblages were found to be most influenced by in-
stream conditions at the reach and riffle scales, however agricultural land use in the
riparian zone at these scales also had predictive power (Townsend et al. 2003).
Macroinvertebrates were influenced by physical landscape conditions at the catchment

scale, such as gradient and drainage area, and to a lesser extent by pasture land in the

43

reach riparian zone. While geographic location was useful in predicting non-flying
macroinvertebrate assemblages, it had no predictive power on flying macroinvertebrates.

As they show the importance of both natural and anthropogenic landscape factors
summarized at multiple spatial scales in controlling stream conditions, the studies
discussed above suggest that mechanisms by which landscape factors influence streams
are complex. Further, they demonstrate the importance of taking a hierarchical approach
when exploring landscape effects on stream conditions. One study that demonstrated the
effectiveness of the hierarchical approach was Infante and Allan (in press), who
quantified effects of natural and anthropogenic landscape factors summarized at the
catchment scale on fish assemblage descriptors through various reach-level habitat
variables to test the notion that landscape affects fish through habitat. Infante and Allan
(in press) demonstrated that in their study region, physical landscape factors (e. g.
drainage area and slope) did indeed have mechanistic effects on fish assemblages when
acting through habitat variables. Riseng et al. (2004) also used a hierarchically-structured
approach to understand the complex role of hydrologic controls and nutrient availability
on macroinvertebrate functional groups through algal abundance and low and high flow
disturbance metrics. Such studies performed in a hierarchical fashion allow us to better
understand specifically how landscape characteristics influence streams and also develop
hypotheses about the hierarchical controls on stream ecosystems as a whole.

The goal of this study is to explore how landscape factors summarized at multiple
spatial scales control stream habitat conditions in northern Michigan streams. My first
objective is to determine at what scale land cover variables account for the most variance

in stream condition, describing a range of habitat characteristics such as habitat

44

heterogeneity through stream reaches, reach substrate variables, descriptors of fish cover,
and estimates of stream bank condition. My second objective is to evaluate how specific
landscape and land cover variables affect stream habitat. To meet this objective, I used a
hierarchical approach to test two specific hypotheses, 1) that land cover and land use
summarized at different spatial scales should affect different characteristics of physical
stream habitat and 2) that land cover variables at multiple scales have indirect effects on
in-stream physical conditions through flow variability. In addressing these hypotheses,
this work will provide insights into landscape and habitat interactions in my study region,
and this understanding can be used to further the use of landscape ecology in stream

management.
Methods

Study area

The study area includes Michigan’s Upper Peninsula and the Lower Peninsula
north of and including the Muskegon and Rifle River basins (Figure 2.1). Rivers within
this region drain into Lakes Huron, Michigan and Superior. All sampling sites were i
selected from within one ecoregion as defined by the World Wildlife Fund, the
Laurentian Great Lakes ecoregion (The Nature Conservancy and World Wildlife Fund
2008). The landscape of this area is dominated by managed forests, and agriculture is
also present in intermittently throughout the study area (Wang et al. 2003, Danz et al.
2007)

Sixty-five study sites were selected; twenty sites in the Upper Peninsula and forty-
five in the Lower (Figure 2.1). Study sites were chosen based on variation in landscape

characteristics including percent wetlands, coarse geology, drainage area, and gradient,

45

but I excluded areas with high urban development. Sites were also selected based on
their accessibility, near road crossings or parks, and were located at least 1 km
downstream from the first upstream tributary inflow.
Stream habitat data collection and summary

Physical habitat data were collected during the summer months of 2009 and
included measures of water depth, channel shape, bank stability, riparian vegetation type,
percent overhanging vegetation, substrate type and embeddedness, types and percentages
of fish cover, and percentages of riffle, run, and pool habitat through study reaches.

Habitat sampling methods followed Simonson et al. (1994). Sample sites were
selected upstream of the stream access point at least 10 m from stream entry, and length
of the sampled area equaled 40 times average stream width (Simonson et al.1994). Each
site was equally divided into 20 transects. Five point measurements of dominant
substrate, depth and embeddedness were taken at regular intervals across each transect.
Percent substrate including silt, sand, fine gravel, coarse gravel, cobble, boulder, bedrock
and clay were visually estimated in the area occurring within 3 m of either side of the
transect. Percent fish cover type measurements including macrophytes, bedrock and
boulder, large woody debris (LWD), and bank undercut were estimated within the same
area. Percent detritus and silt were combined into one category, “percent fines.” The
presence of overhanging vegetation acting as cover was recorded for both banks at each
transect, as was the type and percentage of riparian vegetation or land cover within 5 m
of the stream on each bank. Channel dimensions including wetted width, bankfull width,
and bankfull height were also measured. Wetted and bankfull width were measured

using a rangefinder accurate to 0.5 m, while bankfull height was measured using 2.5-

46

meter segment of PVC piping marked at 0.1 m intervals. Bankfull height, or vertical
channel incision, was the measured as the difference between the elevations of the stream
flow surface at the time of sampling and the point at which the stream would enter the
floodplain (Infante et al. 2006). Depth coefficient of variation of pooled transect points
within a site was calculated to characterize heterogeneity of depth throughout the study
reach. Bank stability was visually scored for each stream bank on each transect with a
score of 4 indicating a stream bank with no visible evidence of erosion, 3 indicating some
evidence of erosion, 2 showing moderate erosion causing unstable soil conditions, 1
showing heavy erosion and a deteriorating bank, and 0 indicating no structural support
and maximal erosion. Overall habitat conditions of the sample reach were also visually
assessed using the Michigan Department of Environmental Quality (MDEQ) Procedure
51 (MDEQ 2002). This method is based on visual scores for a variety of physical
characteristics used to determine stream condition and habitat quality, including
embeddedness, variability in depth/velocity, flow stability, bottom deposition and
sedimentation, diversity of pools, riffles and runs, bank stability, and streamside cover.

Stream flow regimes were characterized using a model of groundwater delivery to
stream channels, precipitation data, surficial geology and land use (Seelbach et al. 2007).
The 10/90 ratio, the annual flow that is exceeded 10% of the time divided by the annual
flow exceeded 90% of the time, was calculated and used as a metric for estimating stream
flow stability. A 90/50 ratio, the annual flow exceeded 90% of the time divided by the
annual flow exceeded 50% of the time, was calculated to indicate areas with high
baseflow (Raleigh 1982).

Landscape data

47

The stream coverage line data was from the 1:100,000 scale National
Hydrography dataset (USGS 2004a). The basic stream unit was the confluence to
confluence stream reach defined as a continuous section of surface water with similar
hydrologic characteristics (USGS 2004a). Drainage area was calculated using reach
catchments delineated from the NHD and the National Elevation Dataset (NED, USGS
2004b, Brenden et al. 2006). Landscape information was organized at four different
spatial scales for each stream reach: local riparian zone, local watershed, network riparian
zone, and network watershed (Brenden et al. 2006). Local refers to the catchment
immediately surrounding and directly draining to the reach, while network refers to the
area upstream of the reach through the watershed or riparian zone. The riparian zone is
classified as a 60 meter buffer on either side of the stream.

Data on land cover (MNDR 2001) and surficial geology (Farrand and Bell 1982)
were assembled in a previous study (Brenden et al. 2006). Stream reach gradient was the
drop in meters per meters of stream reach as depicted by the NED and NHD coverages
(Brenden et al 2006). Land cover types including forest, open and wooded wetlands,
urban land use, and agriculture land use were then summarized as percentages at all four
spatial scales.

Data analysis

I selected 22 habitat variables for analysis from an original set of 110 using both
empirical and analytical procedures. Variables were transformed to achieve normality of
the residuals, and variables with highly skewed residual values were excluded from
further analysis. Variables with limited presence within the study region were also

removed or pooled into one variable indicative of a general habitat characteristic.

48

Percentage variables were transformed using arcsine square root, while continuous
variables were transformed using natural log. MDEQ Procedure 51 metrics did not need
transformations. Pearson pairwise correlations were examined among remaining
variables to help eliminate highly redundant measures determined by an r value greater
than 0.6, resulting in the 22 variables broadly classified into five groups: channel flow
and habitat heterogeneity, reach substrate, channel size, fish cover, and stream bank
condition (Table 2.1).

The 11 landscape variables included wetlands and agriculture percent land cover
at all four spatial scales and physical landscape variables including coarse surficial
geology, drainage area and stream gradient (Table 2.2). Percent coarse geology was
included in analysis because of its known effects on stabilizing stream flow and lowering
stream temperatures (Shepherd 1989, Baker et al. 2003). This variable is the sum of
geologic types with high hydraulic conductivity (k greater than 5.0 m/d), which describes
the relative velocity at which water moves through porous spaces in soil (Shepherd
1989). Percent forest showed a high correlation with total wetlands at several scales
(Appendix A), and because of the high correlation in these landscape factors across
multiple spatial scales, I chose to eliminate forests from analysis because of the potential
effect of wetlands on stabilizing stream flow regimes, sediment loading, and substrate
composition (Wang et al. 2008). Percent urban land use was removed from the dataset
because of its rarity throughout the study region. Box-Cox analysis was used to
determine transformations for individual landscape variables to maximize normality
(Legendre and Legendre 1998). Coarse geology was squared, the arcsine square root

transformation was applied to agriculture at all spatial scales, and gradient was raised to

49

the power of 0.25. Local and network riparian wetlands were normal without
transformation.

My approach to understanding the effects of landscape on habitat consisted of
three steps. First, a principal components analysis (PCA) was conducted on habitat
variables to create principal components (PCs) that portray patterns in habitat. PCA can
be useful in simplifying large multivariate datasets into a smaller number of interpretable
variables (Norman and Streiner 1986, Legendre and Legendre 1998). PCA indentifies
axes of variance within multidimensional space and assigns weights to individual
variables to characterize the various axes. PCs can then be defined based on the variables
that weight heaviest on a given axis. PCs with eigenvalues less than 1 were removed
from future analysis (Norman and Streiner 1986, Legendre and Legendre 1998). PC
scores for individual variables were plotted against the first four axes generated by PCA
to examine potential differences in Upper vs. Lower Peninsula sites. The PCA was used
only to explore possible regional difference within our study region and PC scores were
not used in other multivariate analyses.

The second step was to conduct a redundancy analysis (RDA) using the program
CANOCO to examine the amount of variance in habitat variables explained by each
group of landscape variables. The five groups of landscape factors examined included
physical landscape factors summarized in network catchments (gradient, percent coarse
geology, and drainage area) and land cover variables (percent wetlands and agriculture)
summarized at four spatial scales: local catchments, local riparian buffers, network
catchments, and network riparian buffers of stream reaches. RDA is a direct gradient

analysis that can be used to quantify variation explained in a matrix of response variables

50

(in this case, habitat variables) by constraining their ordination to linear combinations of
predictor variables (in this case, landscape factors) (Legendre and Legendre 1998, ter
Braak and Prentice 1998). RDA is similar to Canonical Correspondence Analysis (CCA)
but is more suitable for data that are linearly vs. unimodally distributed (ter Braak and
Prentice 1998). All 22 habitat variables were included within the RDA to determine total
variance in the group explained by landscape predictors. The amount of variance
explained by each of the five groups of landscape predictors was determined by treating
each individual group of landscape variables as predictors and treating all others as
covariables (Wang et al. 2003). A biplot of the first two axes generated by the RDA was
created to examine relationships among the habitat variables and landscape predictors.

For my third step, covariance structure analysis (CSA) was performed using
AMOS 18.0. CSA is a multivariate technique that quantifies sources of variance in
multivariate data sets (Bollen 1989, Wootton 1994, Maruyarna 1998, Shumacker and
Lomax 2004, Hancock and Mueller 2006). Direct effects, the influence of a predictor
directly on a response variable, and indirect effects, the influences of a predictor through
its relationship with a second predictor variable, can be quantified using CSA
(Shumacker and Lomax 2004). The CSA model tests for the fit of the model to structure
within the data, and can be used to support or refute hypotheses generated about
interrelationships (Bollen 1989).

I used CSA to test the hypothesis that land cover at different spatial scales will
affect physical stream habitat variables differently, both directly and indirectly by
affecting stream flow regimes which I characterized by my estimate of stream flow

variability (i.e., 10/90 flow ratio) (Figure 2.2). This is a technique that can be used in a

51

hierarchical fashion to quantify relationships between a set of predictor and observed
variables while also accounting for correlations between predictor variables (Bollen
1989). As it incorporates a hierarchical approach, CSA can model both direct and
indirect effects of a predictor variable on a response variable (Bollen 1989). I used my
model to predict all habitat variables individually and included land cover variables at
scales shown to explain the most variance in habitat as supported by the RDA.
Landscape variables as well as “Peninsula,” an indicator variable included to broadly
capture variation attributed to the location of a site in either the Lower or Upper
Peninsula, were included as exogenous variables (strictly predictor variables). Ten-
ninety ratio was incorporated as an endogenous variable, used to model the indirect
effects of landscape variables and land cover through its effects on habitat variables
(Bollen 1989). Correlation coefficients (r>0.2) were used to determine whether
correlations should be modeled between exogenous predictor variables to account for
collinearity.

Because CSA requires multivariate normality to generate reliable estimates of fit
(Bollen 1989), bootstrapping was used to help achieve a higher level of multivariate
normality within my dataset (Bollen 1989, Murayama 1998, Hancock and Mueller 2006).
Bootstrapping is considered a reliable technique for correction of multivariate non-
norrnality (Bollen 1989, Hancock and Mueller 2006). Maximum likelihood was the
estimation technique used in my model, as it tends to be robust to deviations from
normality (Bollen 1989, Hancock and Mueller 2006).

Twenty-two separate runs of the model were conducted for each habitat variable,

and indices of overall fit and parsimony of the model were selected following previous

52

studies (e. g., Riseng et al. 2004, Infante and Allan in press). Indices of fit were Chi-
square (P > 0.05), root mean square error of approximation (RMSEA P< 0.05), Tucker-
Lewis Index (TLI>0.9) and Normed Fit Index (N FI>0.9). If the model was determined to
be significant, percent observed variance explained (r2) as well as direct and total effects

and their significance (P<0.05) were evaluated for each habitat variable.

Results

Landscape and in-stream habitat conditions
Study site drainage area ranged from 7 to 716 km2 (Table 2.2). Reach gradient

varied by two orders of magnitude, from 0.0001 to 0.0178. Catchments had relatively
low levels of human land disturbance compared to southern Michigan; the highest mean
value over all scales of agriculture was 5.18%. Wetlands were present at all scales and
reached values of over 80% in the network and local riparian zone. Coarse geology was
dominant throughout study catchments, with a mean value of 84.83%.

Habitat characteristics varied throughout the study sites (Table 2.1). While runs
were the dominant habitat type, riffles had a mean of 13% and a maximum value of 70%.
Pools were less common with a maximum of only 17%. Stream flow regimes varied
across the study region as indicated by multiple variables. Stream flow stability, a visual
estimate, ranged from 2 to 10 (maximum possible stability score), and the 10/90 ratio
ranged from 1.13 to 47.92. Sand was the most dominant substrate type with a mean of
45% across all sites. Fine gravel was less common in sites with a mean of 9% and
maximum of 46%. Wetted width varied greatly, ranging from 2.90 to 38.38 m.
Overhanging vegetation had a mean of 43% and a maximum value of 100%. Overall,

stream bank condition ranged from moderately poor to very good, indicated by visual

53

measures of bank stability. Bank stability averaged across all sites was 3.19, suggesting
that erosion and scouring are uncommon in my study streams.
Patterns of in-stream habitat conditions

PCA of twenty-two habitat variables generated four principle components
explaining 65% of the variance in habitat data (Table 2.3). Axis I explained 27% and
was positively weighted by the presence of riffle habitat; coarse substrate, rock cover —
which includes cobble, boulder, and bedrock — and high interstitial spacing scores, while
runs weighted negatively on the axis. For these reasons, I named this axis “riffle system
habitat and coarse substrate,” although it indicates a range in substrate and habitat types.
Axis 2 explained 15% of the variance in habitat variables, and indicated a range in the
variability of stream flows. Stream flow stability, a visually estimated-variable, weighed
heavily on the axis, as did bank stability. Bankfull minimum height weighted negatively.
Axis 3 was positively weighted by 90/50 ratio, average depth, and wetted width,
indicating that the axis was representative of larger streams. This axis explained 12% of
the variation. The fourth axis was only heavily weighted by pools, again, relatively
uncommon in my study sites, and explained 9% of the variance in habitat data.

Sites scores plotted against axes generated by PCA showed that some
characteristics differed across peninsulas. While no difference was detected in amounts
of riffle habitat and coarse substrate (Figure 2.3B), streams in the Lower Peninsula
tended to be larger than those in the Upper Peninsula (Figure 2.3A). Also, site scores
plotted against axes 1 and 2 indicate that stream stability differs by peninsula (Figure
2.3A). Seventeen of twenty Upper Peninsula sites (85%) fell below the x axis, indicating

that these streams were less stable than many Lower Peninsula streams. Stream stability

54

in the Lower Peninsula varied, but tended to be more stable overall. Because of the
differences in stability and stream size across peninsulas, an indicator variable was
included in the CSA model to account for those regional differences, with 1 indicating
the Lower Peninsula and 0 indicating the Upper Peninsula.

Relationships between landscape and habitat variables

The RDA indicated that site scale habitat conditions were associated with
landscape variables at multiple scales. Together, the fifteen predictor variables explained
41% of total variance in habitat variables (Figure 2.4). The physical landscape grouping
accounted for most of the explained variance (57%), and land cover in the network
riparian zone accounted for 13%. Local riparian zone land cover explained just over 5%,
slightly more than the network watershed. The local watershed land cover explained the
least variance, with only 1% accounted for.

A biplot of habitat and land cover variables in multidimensional space depicted
across two generated RDA axes indicates differences in landscape to habitat relationships
that may be related to the scale at which a land cover variable was quantified (Figure
2.5). The lengths of vectors, representative of landscape variables, indicate the weight of
the individual variable on the habitat dataset. Drainage area, gradient, network riparian
wetlands, local riparian wetlands, and percent coarse geology have the most weight on
habitat variables. Habitat variables in close proximity to landscape variables are highly
positively correlated with said landscape variables, those at ninety degrees angles are
maximally uncorrelated, and those opposite landscape variables are maximally negatively
correlated. Drainage area and gradient are highly correlated with habitat variables

indicative of riffle systems and larger streams; habitat heterogeneity, coarse substrate,

55

wetted width, coarse gravel, and bedrock and boulder that can act as stream cover.
Drainage area and gradient are highly negatively correlated with runs, fine substrate,
macrophytes and overhanging vegetation, while these habitat variables are positively
correlated with local riparian wetlands.

Coarse geology and network riparian wetlands are positively correlated with
indicators of flow stability (stream flow stability, bank stability) and increased baseflow
(90/50 flow ratio), while negatively correlated with indicators of instable flows (10/90
ratio, minimum bankfull height). Local and network catchment wetlands weight weakly
on the dataset, while wetlands at different scales correlate differently with habitat
variables.

The CSA model

My approach in developing the CSA model was to compare the influence of
different landscape factors on stream habitat, including the influence of landscape factors
summarized at multiple scales. I used results of the PCA and RDA as well as
correlations among landscape factors to aid in model development. Given the relatively
large amount of variance (57.8%) that they accounted for out of the total variance
explained (41.2%), all physical landscape variables including coarse geology, catchment
area, and gradient were incorporated into my model as exogenous variables (Figure 2.2).
However, landscape variables summarized in the network watershed and local
watersheds of my study sites were excluded because each grouping was found to account
for less than 5% of the total explained variance with the RDA. In contrast, landscape
variables summarized at the scales of the local and network riparian zones explained

13.1% and 5.3% respectively and were included in the models. Besides these natural and

56

anthropogenic landscape factors, I also included the indicator variable “Peninsula” to
broadly capture regional differences in controls on stream systems.

In assigning relationships within my model, agriculture and wetlands in the local
riparian zone, wetlands in the network riparian zone, and drainage area were modeled to
have both direct and indirect effects on habitat variables. Both the indicator variable
“Peninsula” and gradient were modeled as having direct effects on habitat variables.
Percent coarse geology was modeled as having only indirect effects on habitat due to the
strong influence of geology on Michigan stream flow regimes (Seelbach et al. 2007).
CSA results

All models predicting individual habitat variables were significant based on my
set of fit statistics except for the model predicting 90/ 50 flow ratio (Table 2.5). Therefore,
all models except for that predicting the 90/50 flow ratio were considered. Two out of
the remaining twenty-two models predicting fine gravel and percent pools showed no

significant relationships between predictors and habitat data and explained low amounts
of variation (r2 < 0.10).
Effects of landscape on flow stability

Two of the five variables modeled to influence 10/90 ratio were found to have

significant direct effects (Table 2.6). Both were physical habitat variables, drainage area

and percent coarse geology, and the 10/90 flow ratio decreased as these variables
increased indicating more stable stream flow regimes. The r2 value for the 10/90 flow

ratio was 0.72, indicating that a majority of variance in this factor was explained by the
model. Neither agriculture nor wetlands in local or network riparian zones had a

significant effect on the 10/90 ratio.

57

Effects of landscape variables on in-stream habitat

Natural landscape variables significantly affected many habitat variables (Table
2.7), with direct effects very similar to total effects (Table 2.8). Local riparian agriculture
had no significant effects on any stream habitat variable. Wetlands in both the local and
network riparian zone were found to have different effects on habitat variables indicative

of stream stability, substrate type, and habitat heterogeneity.
Variables describing channel flow and habitat heterogeneity generally had r2

values ranging from 0.36 to 0.48, with the exception of percent pools which was poorly
predicted by the model. Both runs and riffles decreased with increasing drainage area
and gradient, while the 10/90 flow ratio was negatively associated with stream flow
stability. Coarse geology was positively associated with the visual estimate of stream
flow stability, and study sites in the Lower Peninsula were shown to be more stable.
Local riparian zone wetlands had no significant effects on any habitat variables within
this group. Network riparian zone wetlands had negative effects on runs and depth
coefficient of variation and were positively related to stream flow stability and riffle

habitat.
Overall amounts of variation explained in substrate variables were fairly high (r2

= 0.37-0.47) with the exception of fine gravel. Reach substrate generally increased in
size with increasing drainage area and gradient, while coarse substrate, which included
boulder and bedrock, decreased as coarse surficial geology increased in catchments of the
study region. Sand was less common in systems with variable stream flow regimes, while
coarse substrate increased as flow stability decreased. As wetlands in the network

riparian zone increased, substrate size also generally increased, but this variable was not

58

shown to have a significant effect on fine substrate. Local riparian zone wetlands were
negatively correlated with the presence of coarse gravel.

Average depth and wetted width were well predicted by my model, and over 68%
of the variance was explained for depth. Drainage area had a large direct effect on wetted
width (0.81), while network riparian wetlands and 10/90 ratio had a weaker positive
effect.

Fish cover variables were only significantly predicted by drainage area, gradient
and the 10/90 flow ratio with explained variance ranging from 21% to 35%; this was the
lowest of all habitat variable groupings. Larger streams were associated with decreases
in cover variables except rock cover (BED) which increased significantly with drainage
area (total effect of 0.29). Increases in boulders and bedrock that could supply fish cover
were also significantly associated with increased gradient and 10/90 ratio. Macrophytes
decreased with increasing gradient and drainage areas. Local riparian wetlands had no
significant effects on macrophytes.

Variables describing stream bank condition were well predicted by the models.
Bank stability and minimum bankfull height had 53% and 60% of their variation
explained, respectively. The visual metric describing riparian vegetation type and
condition decreased with larger drainage areas and was generally of higher quality in the
Lower Peninsula. Bank stability increased as 10/90 ratio decreased and as coarse
geology increased, and it has a positive but insignificant relationship with network
riparian wetlands. Minimum bankfull height significantly decreased in stable streams
with network riparian zone wetlands, but showed no relationship with local riparian zone

wetlands.

59

Discussion

Overview

My analyses demonstrated ways in which landscape factors relate to stream
habitat, including how factors summarized at different spatial scales predict different
habitat characteristics. By using ecological understanding of landscape to habitat
relationships along with results of a multivariate analysis (RDA), I developed a CSA
model to test hypotheses about the response of habitat to landscape predictors at multiple
spatial scales in my study region. My first hypothesis, that land cover and land use
summarized at different spatial scales should affect different characteristics of physical
habitat, was supported. I found that wetlands in the network riparian zone were
associated with increases in stream flow stability and substrate size, while wetlands in the
local riparian zone were negatively associated with increasing amounts of coarse gravel
and not significantly associated with any of the other habitat characteristics that I
considered. Agriculture in the local riparian zone had no significant effects on habitat
conditions; however, natural landscape variables including percent coarse surficial
geology, drainage area, and gradient significantly affected many of the habitat variables.
Despite being strongly affected by coarse geology and drainage area, the 10/90 flow ratio
was not significantly affected by any of my land cover variables, and this finding failed to
support my second hypotheses. I conclude that in my study streams, natural landscape
variables are closely associated with stream habitat; wetlands at the network riparian zone
scale are secondary predictors of habitat heterogeneity, substrate size and stream flow
stability; and local riparian wetlands have minimal effects. By understanding specific

effects of land cover at multiple scales in natural systems, we can more appropriately

60

incorporate landscape analyses into management and conservation practices.
Physical landscape controls on stream habitat

My results indicate that drainage area, gradient, and geology have the strongest
effects on physical conditions of my study streams. Drainage area had a significant
positive effect on habitat heterogeneity, larger substrate, and riffle habitat, a result that
may be due to an increase in stream power with greater catchment area. Stream power
increases with greater stream flows which can increase with drainage area. More stream
power provides streams more opportunity to do work, potentially leading to greater
variability in depth, greater diversity in habitat types present, and increased movement of
fine substrate compared to smaller streams (Hauer and Larnberti 1996, Knighton 1998).
Drainage area’s significant negative relationship with LWD may be caused by the effect
of stream power, increasing movement of wood through the stream. Drainage area also

had significant negative effects on the presence of overhanging vegetation, and the
presence of overhanging vegetation and LWD were positively correlated (r2: 0.27).

Less overhanging vegetation in a reach may indicate a decreased presence of woody
plants within the riparian zone, potentially lessening LWD input into the stream. Another
reason for drainage area’s significantly negative relationship with LWD may be that I did
not account for stream size differences when calculating my LWD measurements within
the transect. It is possibly that the trend of decreasing LWD with increasing drainage area
may be caused by the relatively larger wettedeidth in larger drainage areas, meaning
that if a small stream has the same amount of LWD as a large one, the percentage of the

stream containing LWD may be greater in the stream with a smaller wetted width.

61

Controlling for stream size may better demonstrate how LWD may be affected by
drainage area.

Stream gradient had similar relationships to habitat conditions as drainage area,
and was significantly related to increases in habitat heterogeneity, substrate size, and
riffle habitat. Increased gradient is related to increased stream power within a system

(Knighton 1998). However, in my study region, stream gradient and drainage area are
strongly negatively correlated (r2: -0.59). A study conducted in southeast Michigan

found similar effects of drainage area and gradient on habitat heterogeneity, riffle
presence and substrate, postulating that maximum stream power was achieved in
meditun-sized streams of the study region where drainage area and gradient were
maximized (Infante and Allan in press). Southeast Michigan is an area of comparatively
heavy human disturbance while surficial geology varies from regions dominated by
coarse materials to areas dominated by clay and sand lakeplain (Infante and Allan in
press). Similar findings on the role of drainage area and gradient on stream habitat
despite differences in land cover and surficial geology across northern vs. southern
Michigan demonstrates that gradient and drainage area are important drivers of substrate
type and channel habitat heterogeneity in multiple systems.

My results are consistent with other recent work in the region by Wang et al.
(2003) who found that natural landscape factors including drainage area and surficial
geology at the network catchment scale explained much of the variance in stream habitat
variables including channel morphology, substrate type, and assessments of dissolved
oxygen, conductivity, pH, alkalinity, hardness and turbidity. The role of natural

landscape variables in the network catchment in an overlapping study region emphasizes

62

the importance of considering these variables when attempting to predict stream
conditions.

Increasing coarse surficial geology was associated with lower 10/90 flow ratios
within my study region. In Michigan, surficial geology has been shown to have major
effects on stream flow and temperature regimes through effects on groundwater delivery
(Seelbach and Wiley 1997). Surficial geology was found to have limited effects on other
factors examined in this study, but its significant negative relationship with coarse
substrate may be indicative of its control on substrate type within a reach. Areas with
small amounts of coarse geology are more prone to high flow events and therefore
increased stream power, which would move fine sediments and smaller substrate to lower
reaches or out of a system, potentially increasing the percentage of coarse substrate
within the stream.

The lack of a significant effect of land cover on 10/90 ratio may be due to the fact
that land cover, specifically wetlands, may be influencing different characteristics of
catchment hydrology than that captured by the 10/90 flow ratio, a ratio quantifying
annual high over annual low flows. Perhaps a more seasonal measure, such as spring or
fall high flow events controlled for by drainage area, may indicate wetlands having a
more direct effect on attenuation of flows. Similarly, my failure to detect effects of
agriculture on the 10/90 flow ratio may be due to the fact that agriculture was relatively
uncommon throughout my study region and its influence was not readily detectable.
General eflects of land cover at difierent spatial scales: watershed versus riparian zone

The RDA results indicate that land cover in the riparian zone explained some

variance in my group of stream habitat characteristics, while watershed land cover

63

explained comparatively little variance within the dataset. This result suggests that
riparian zone land cover is more useful in predicting habitat conditions than catchment
area land cover in my study region, and the CSA confirms that wetlands in the riparian
zone do explain significant amounts of variance in habitat heterogeneity, flow stability,
and substrate percentage.

Network catchment land cover has been shown to be the strongest predictor of
stream habitat condition in several studies of systems dominated by human land use
including urban, agriculture, or a combination of both (Roth et al. 1996, Wang 1997, Roy
et al. 2003, Stephenson and Morin 2009). While these studies examine different regions
and a variety of landscape conditions, they all focus on areas of heavy anthropogenic
disturbance. My study region is different in that it focuses on areas with minimal current
human impact. My results indicate that land cover in network riparian zone explains a
larger amount of variance within stream habitat conditions in systems dominated by
natural conditions. Wang et al. (2003) found similar results in a study region
encompassing much of northern Wisconsin and a portion of my study region, an area
dominated by natural conditions, stating that the local riparian zone land cover explained
more variance than network catchment land cover overall. The results of Wang et al.
(2003) and my study compared to studies with largely disturbed catchments suggests that
as a watershed is increasingly disturbed through human land use, the effects of the
modified catchment on stream stability, habitat heterogeneity and substrate type may
outweigh the effects of the network riparian zone.

While wetlands at the catchment scale were positively associated with RDA Axis

2, suggesting a relationship with stream instability, they were not included in my CSA

64

model because of the relatively low amount of variance explained in my set of habitat
variables by watershed land cover in general. In contrast, riparian zone wetlands were
negatively associated with RDA Axis 2, suggesting a relationship with increased stream
flow stability. In the Tittabawasee River system, Michigan, catchments with greater than
60% wetlands, including forested and open water wetlands, were shown to have lower
flow stability and greater total water yield compared to those with less wetlands
(Tompkins et al. 1997). However, Tompkins et al. (1997) did not examine the role of
wetlands in the riparian zone. While my results did not contradict Tompkins et al.
(1997), I saw a different effect of wetlands when summarized at the network riparian
zone versus the catchment scale. The results of my study emphasize that the scale at
which landscape is quantified in a study can lead to potentially differing conclusions
about controls on habitat condition.
Specific effects of network and local riparian wetlands on habitat

Wetlands summarized at the network vs. local riparian zones had significantly
different effects on reach substrate, responses to flow stability, and habitat heterogeneity.
Network riparian zone wetlands had a significant negative, effect on bankfull minimum
height and a significant positive effect on the visual metric of stream flow stability.
Together, these results indicate that at the network riparian zone, wetlands may be acting
to mitigate high, channel shaping stream flows, absorbing surface runoff and rainwater
and pooling it for increased periods of time (Cohen and Brown 2007, Wang et al. 2008).
Also, wetlands in the network riparian zone have greater control over catchment
hydrology influencing a particular reach than wetlands at a local scale. A recent Florida

study focused on modeling the effectiveness of wetlands at multiple scales at removing

65

sediment and attenuating flow for storm water removal (Cohen and Brown 2007). While
local headwater wetlands had small effects on decreasing flood levels, in a system
containing the same percentage of wetlands located throughout the stream network, the
intensity of storm flows was lowered to an even greater degree (Cohen and Brown 2007).
My results also suggest that network riparian wetlands may be affecting reach
substrate. Wetlands are known to trap sediments and fine particles delivered from
upstream catchment areas, therefore reducing fine sediment presence within the stream
(Johnston 1991, Cohen and Brown 2007, Wang et al. 2008). In my study area the
network riparian zone wetlands may be affecting stream substrate in this way, a result
that is supported by their significantly positive relationship with riffle habitat. Because
network riparian zone wetlands were not significantly correlated with drainage area or
gradient, other variables significantly affecting hydrology, its control on substrate occurs
independently. I failed to see an effect of wetlands on fine substrate, however this may
be because I grouped sediment and detritus in my analysis. Detritus, decaying masses of
organic material, is very common in areas dominated by wetlands, while fine sediment
has been shown to be removed by wetlands in the riparian zone (Tompkins et a1 1997,
Wang et al. 2003, Wang et al. 2008). By pooling these two substrates types into one
category, I may have masked the influence of wetlands on fine sediments. In contrast,
local riparian zone wetlands have a significantly negative effect on the presence of coarse
gravel, leading us to the conclusion that wetlands at these two scales impact stream
substrate composition in a different way. In minimally impacted streams of Wisconsin
and some of northern Michigan, the effects of local riparian wetlands was the best single

predictor of substrate compared to watershed landscape and land cover conditions as well

66

as other local riparian land cover, but the network riparian zone was not assessed (Wang
et al 2003). The Florida study on the effects of multiple scales of wetlands that predicted
decreases in flood intensity with network versus local wetlands also predicted that
wetlands throughout the network were more effective than local riparian wetlands at
removing sediments within the stream (Cohen and Brown 2007).
Utility of CSA and RDA

By combining these two multivariate analytical technique (RDA and CSA), 1 was
able to develop and test hypotheses suggested by actual relationships within my data to
account for variance explained at multiple scales. RDA biplots allowed for a visualization
of how patterns in habitat conditions may be related to patterns in landscape, including
landscape factors expressed at multiple scales. RDA was also a useful tool in creating the
CSA model by allowing me to assess which landscape scale groupings explained the
most variance in habitat factors. While RDA allowed for the examination of broad
relationships between sets of landscape factors and habitat variables, it does not allow for
examining specific relationships among pairs of variables or tests of significance. Due to
the nature of the biplots that 1 generated, which represent a compression of
multidimensional space down to two axes, interpretations of results must be made
cautiously. While a habitat variable may be associated with a certain landscape variable
in the biplot, it does not mean that a causal relationship exists. The RDA biplot (Figure
2.4) shows that local riparian zone wetlands are associated with habitat variables like
riparian vegetation condition, sand, fine substrate and macrophytes. However, when each

habitat variable was independently examined within the CSA model, my results showed

67

that drainage area and gradient are actually the strongest predictors within the dataset,
located at the opposite end of Axis I.

The value of CSA is that it allows for the examination of specific effects of
landscape predictors on habitat variables while accounting for correlations among the
predictors (Bollen 1989). CSA also allows for the assessment of significance of effects
on a particular habitat variable, so one can then understand which predictors are strongly
acting upon the predicted variable, which is not available through multivariate
approaches like RDA.

Implications

My study area has a unique history associated with its landscape. While today the
region is largely forests and wetlands, in the late 1800’s and early 1900’s much of the
land was altered through logging (Mershon 1923, Maybee 1960, Vincent 1962,
Dickmann and Leefers 2003). The streams were the main artery for the transport of logs,
which were piled on cleared riparian areas until spring at which time they were pushed
down river during high seasonal flows (Dickmann and Leefers 2003). Stream systems
were altered significantly in terms of substrate, channel morphology and hydrologic
conditions through these intense disturbances (Vincent 1962, Miller 1966, Dickmann and
Leefers 2003, Johnson et al. 2003). Gravel riffles were scoured by the driving of logs
and covered with sand deposited from the surrounding watershed, while channels were
deepened and widened and woody debris was lost, reducing overall channel complexity
and permanently altering the streams (Maybee 1960, Vincent 1962). Harding et al.
(1998) found that catchments dominated by agriculture in the 1950’s now dominated by

second growth forests still showed macroinvertebrate assemblages similar to those

68

catchments currently dominated by agriculture. In my study region, it is possible that
some of the unexplained variance in my dataset could be attributed to the historical
alterations of catchments. For instance, while network riparian wetlands are correlated

with most substrate conditions, fine gravel is not significantly related to this variable.
Fine gravel’s very low r2 value (0.09), lack of any significant predictors, and limited

presence within the study area suggests that it is not well predicted by variables within
my model. It is possible that the increases in substrate deposition in the form of sand
from former periods of high erosion and runoff may have altered stream systems so that
areas historically dominated by fine gravel are now dominated by sand, therefore causing
the variable to be poorly predicted by landscape predictors.

While the unique characteristics of my study region may limit the application of
my results beyond this ecoregion, these results supply a base for considering similar
questions in other regions. For instance, Utz et al. (2010) showed that even within a
relatively small area of coastal Maryland, variations in physical landscape attributes
across regions within the same stream systems can cause a change in the effect of urban
land cover on fish assemblages. By comparing studies and results across regions, we can
develop a deeper understanding of the complexities of landscape control on habitat
conditions while looking for broad, overarching relationships that may exist despite
regional differences.

My study supports the theory that the landscape affects stream habitat at multiple
scales (Poff 1997, Allan 2004, Wiens 2002). By understanding the specific relationships
between landscape at different scales and habitat conditions, I can apply hierarchically-

structured approaches to identify suitable habitat for species and to better understand

69

stream community assemblage. To apply methods like the filter approach within the
stream as described by Poff (1997) in largely natural systems, a greater understanding of
controlling landscape factors is needed in these regions. Identifying which scale of
natural land cover affects habitat conditions like substrate composition, flow stability,
and habitat type can be useful in predicting stream reaches that may have conditions that
are suitable for spawning, feeding and refuge sites for fish species (F ausch et al. 2002,
Durance et al 2006). With greater understanding of controlling factors in a region, we
can improve the results of such an analysis and conduct studies quickly and efficiently
using the growing amount of landscape and land cover data available across the United
States.

The identification of landscape and land cover effects at multiple spatial scales
also has direct benefits for management. For example, my study demonstrated that at the
network riparian zone scale, wetlands have a more significant effect on the limitation of
fine sediments within the stream than at any other scale. Managers considering the use of
wetlands for sediment control could benefit from this information when deciding how
and where wetlands could be constructed within a stream system. Similarly, I concluded
that network riparian zone wetlands play a role in stabilization of flows within the stream,
while local riparian wetlands have a limited effect on attenuating flow and catchment
wetlands may contribute to flow instability. Managers hoping to use wetlands to
attenuate flow can use studies such as this as a guide, and avoid adversely affecting
stream stability by constructing wetlands at other scales. Overall, my work accentuates
the point that when attempting to manage stream systems through the alteration of land

cover or when trying to identify stream conditions based on the landscape approach, the

70

scale at which land cover is summarized must be considered, even in largely natural

systems.

71

Table 2.1: Mean, range and standard deviation of habitat variables used in analyses.

 

 

Standard
Category Variable name Code Mean Range deviation
Channel flow and habitat heterogeneity
Pools (%) Pool 3.00 0.00-17.00 4.00
Riffles (%) Rif 13.00 0.00-70.00 19.00
Run (%) Run 84.00 3000-100.00 19.00
Depth coefficient of variation DCV 0.46 0.22-0.98 0.15
Habitat heterogeneity HH 12.45 GOO-20.00 4.17
Stream flow stability SFS 7.12 2.00-10.00 2.19
10/90 flow ratio 10/90 5.18 1.13-47.92 6.09
Reach substrate
Fine substrate (%) FS 11.00 0.00-54.00 12.00
Sand (%) SD 45.00 0.00-94.00 31.00
Fine gravel (%) FG 9.00 0.00-46.00 11.00
Coarse gravel (%) CG 22.00 0.00-74.00 24.00
Coarse substrate (%) CS 13.00 0.00-78.00 20.00
Channel size
90/50 flow ratio 9050 0.55 0.19-0.97 0.13
Average depth (cm) DEP 50.63 1106-13275 23.82
Wetted width (m) W 12.78 2.90-38.38 7.17
Fish cover
Undercut (%) UNC 1.00 0.00-7.00 1.00
Boulder or bedrock (%) BED 1.00 0.00-14.00 3.00
Large woody debris (%) LWD 10.00 0.00-34.00 7.00
Macrophytes (%) MAC 5.00 0.00-27.00 7.00
Overhanging vegetation (%) OV 43.00 2.50-10000 21.00
Stream bank condition
Riparian vegetation RV 8.16 3.00-20.00 1.67
Bank stability BS 3.19 2.18-4.00 3.70
Minimum bankfull height (cm) HM 59.29 450-27250 47.12

 

72

Table 2.2: Mean, range and standard deviation of landscape variables used in analyses

 

 

Standard

Category Variable name Code Mean Range deviation
Physical landscape

Coarse geology in network

watershed (%) Coarse 84.83 0.00-100.00 27.01

Gradient of stream reach Gradient 0.0022 0.0001-0.0178 0.00

Drainage area (km’) DA 127.17 669-71555 134.75
Network riparian

Agriculture in the network

riparian (%) NRAg 2.93 0.00-30.28 5.11

Wetlands in the network

riparian (%) NRWet 44.81 460-8290 18.04
Network watershed

Agriculture in the network

watershed (%) NWAg 5.17 0.00-23.64 6.13

Wetlands in the network

watershed (%) NWWet 18.90 1.91 -73.05 13.60
Local riparian

Agriculture in the local

riparian (%) ‘ LRAg 1.81 0.00-17.78 3.81

Wetlands in the local

riparian (%) LRWet 50.03 703-829 20.64
Local watershed

Agriculture in the local

watershed (%) LWAg 4.12 0.00-27.42 6.48

Wetland in the local

watershed (%) LWWet 17.38 1.91-65.74 13.29

 

73

Table 2.3: PCA results of 22 habitat variables. Four axes explained 65.05% of the
variation in habitat data. Bold values indicate variable weights with an absolute value
greater than 0.6 that were used to interpret individual axes.

 

 

Axis 1 Axis 2 Axis 3 Axis 4

Riffle systems Pools and

and coarse Responses to Stream overhanging

substrate flow stability size vegetation
% Variance
explained 27.15 15.70 12.23 9.25
Code
Rif 0.90 -0.10 -0.15 0.00
CS 0.80 -0.16 -0.14 -0.08
CG 0.77 0.00 0.20 012
BED 0.62 -0.21 -0.23 -0.09
SD -0.88 -0.06 0.09 0.31
Run -0.89 0.08 0.19 -0.15
SFS -0.06 0.87 0.12 0.01
BS 0.09 0.82 0.05 0.03
DCV 0.26 -0.62 -0.35 -0.22
HM 0.09 -0.90 0.00 -0.03
W 0.08 -0.06 0.81 -0.37
90/50 -0.1 1 0.40 0.74 -O.13
DEP -0.25 0.15 0.81 -0.01
Pool 0.20 -0.12 -0.09 0.61
0V -0.22 0.21 0.00 0.73
HH 0.50 -0.28 0.51 0.18
UNC 0.08 0.08 -0.23 -0.02
FG -0.07 -0.01 0.45 -0.12
RV -0.16 0.42 -0.21 0.52
LWD -0.33 -0.27 -0.16 0.45
MAC -0.35 0.53 -0.20 -0.28
FS -0.50 0.17 -0.52 -0.24

 

74

Table 2.4: Pearson pairwise correlations for landscape variables used in CSA. Variables
with correlations higher than an absolute value of 0.2 are shown in bold and were
included in the model.

DA LRWet NRWet Gradient Coarse LR:Ag 10/90 Peninsula
DA 1.00

 

 

LRWet 0.18 1 .00

NRWet 0.05 0.46 1.00

Gradient -0.59 -0.49 -0.23 1 .00

Coarse 0.11 -0.06 -0.18 -0.02 1.00

LRAg -0.11 -0.23 -0.23 -0.01 -0.05 1.00

10/90 -0.59 -0.22 -0.05 0.36 -0.65 0.18 1.00

peninsula 0.23 0.21 0.1 1 -0.29 0.46 0.01 -0.41 1.00

 

75

Table 2.5: Fit statistics for CSA models predicting habitat variables from landscape

factors. Statistics include Chi-square goodness of fit (X2), root mean square error of
approximation (RMSEA), Tucker-Lewis Index (TLI) and Normed Fit Index (N F 1). Fit of
each model is indicated by “yes” or “no.”

 

Category Variable

 

X2 X29 RMSEA TLI NFl Fit?
Channel flow and habitat heterogeneity
Pool 10.89 0.62 0.00 1.04 0.94 Yes
Rif 11.4 0.59 0.00 1.02 0.95 Yes
Run 11.21 0.59 0.00 1.02 0.95 Yes
DCV 12.37 0.5 0.00 1.01 0.95 Yes
HH 12.65 0.48 0.00 1.01 0.94 Yes
SFS 11.41 0.58 0.00. 1.02 0.95 Yes
Reach substrate
FS 15.07 0.3 0.05 0.97 0.93 Yes
so 11.7 0.55 0.00 1.02 0.95 Yes
FG 12.04 0.52 0.00 1.02 0.94 Yes
CG 10.9 0.62 0.00 1.03 0.95 Yes
cs 10.9 0.62 0.00 1.03 0.95 Yes
Channel size
9050 1.9 0.03 0.12 0.91 0.94 No
DEP 11.82 0.54 0.00 1.01 0.96 Yes
W 13.27 0.43 0.02 1.00 0.96 Yes
Fish cover
UNC 10.95 0.62 0.00 1.03 0.95 Yes
BED 12.07 0.52 0.00 1.01 0.94 Yes
LWD 10.94 0.62 0.00 1.03 0.95 Yes
MAC 11.21 0.59 0.00 1.03 0.95 Yes
0v 15.38 0.29 0.05 0.96 0.92 Yes
Stream bank condition
RV 11.16 0.6 0.00 1.03 0.95 Yes
BS 10.93 0.62 0.00 1.03 0.95 Yes
HM 10.98 0.61 0.00 1.03 0.96 Yes

 

76

Table 2.6: Standardized total effects generated through CSA of landscape factors on
10/90 flow ratio. Significant effects are shown in bold (P<0.05).

 

Landsge factors 10/90 flow ratio
Drainage area -0.49
Coarse geology -0.62
LRAg 0.06
LRWet -0.14
NRWet 0.06
r2 0.72

 

77

Table 2.7: Standardized total effects generated through CSA of landscape variables on
individual habitat variables. Effects that are statistically significant (P<0.05) are shown

 

 

in bold.
Standardized total
Category Variable effects
Drainage
Area Gradient Coarse Peninsula 10/90
Channel flow and habitat heterogeneity
P 0.07 -0.24 -0.06 -0.03 -0.03 0.04
Rif 0.41 0.40 0.77 -0.11 0.01 0.18
Run 0.36 -0.30 -0.68 0.11 0.01 -0.18
DCV 0.47 0.08 0.25 -0.02 -0.52 0.05
HH 0.35 0.61 0.40 0.11 -0. 14 -0.17
SFS 0.48 -0.14 0.00 0.27 0.34 -0.44
Reach substrate
FS 0.37 -0.69 -0.62 -0.04 0.08 0.07
SD 0.44 -0.49 -0.65 0.17 -0.07 -0.28
FG 0.09 0.18 -0.04 0.08 0.08 -0.13
CG 0.43 0.56 0.58 -0.03 0.07 0.05
CS 0.47 0.47 0.58 -0.27 -0.08 0.43
Channel size
DEP 0.68 0.50 -0.34 0.14 0.16 -0.23
VWV 0.84 0.90 -0.01 0.11 -0.10 -0.18
Fish cover
UNC 0.21 -0.25 0.17 -0.15 0.19 0.23
BED 0.35 0.29 0.41 -0.18 -0.22 0.30
LWD 0.24 -0.39 -0.10 0.10 -0.23 -0.17
MAC 0.24 -0.44 -0.57 0.04 0.02 -0.07
0V 0.17 -0.43 -0.28 0.11 0.00 -0.18
Stream bank condition
RV 0.27 -0.44 -0.11 0.05 0.30 -0.08
BS 0.53 -0.15 0.13 0.32 0.45 -0.52
HM 0.60 0.34 0.10 -0.18 -0.57 0.29

 

78

Table 2.7: Continued

 

Category Variable

 

LRAg LRWet NRWet
Channel flow and habitat heterogeneity
P -0.05 -0.16 0.10
Rif 0.09 0.03 0.23
Run -0.05 0.04 -0.25
DCV 0.01 -0.04 -0.24
HH 0.15 -0.10 0.06
SFS 0.02 0.10 0.34
Reach substrate
FS 003 -0.07 -0.16
SD 0.02 0.18 -0.33
FG 001 -0.06 -0.11
CG 0.01 -0.26 0.21
CS 0.10 -0.09 0.36
Channel size
DEP 0.02 -0.10 0.08
W 0.01 -0.08 0.14
Fish cover
UNC -0.21 0.05 013
BED 0.06 -0.16 0.14
LWD 0.22 0.10 -0.06
MAC 015 -0.08 0.05
0V 0.03 0.01 0.09
Stream bank condition
RV -0.19 0.07 -0.16
BS 005 -0.04 0.15
HM 0.10 0.02 -0.27

 

79

Table 2.8: Standardized direct effects of landscape variables generated through CSA
on individual habitat variables. Effects that are statistically significant (P<0.05) are

shown in bold.

 

 

Category Variable Standardized direct effects
Drainage Grad
Area -ient Peninsula LRAg 10/90 LRWet N RWet

Channel flow and habitat heterogeneity

P -0.22 -0.06 -0.03 0.00 0.04 -0.16 0.10

Rif 0.49 0.77 -0.02 0.09 0.18 0.05 0.24

Run -0.39 -0.68 0.01 -0.05 -0.18 0.01 -0.26

DCV 0.10 0.25 -0.52 0.01 0.03 -0.03 -0.24

HH 0.52 0.40 -0.14 0.15 -0.17 -0.12 0.05

SFS -0.36 0.00 0.34 0.02 -0.44 0.04 0.31
Reach substrate

FS -0.65 -0.62 0.07 -0.03 0.07 -0.06 -0.15

SD -0.63 -0.65 -0.07 0.02 -0.28 0.14 -0.35

FG 0.12 -0.04 0.08 -0.01 -0.13 -0.08 -0.12

CG 0.58 0.58 0.07 0.01 0.05 -0.25 0.21

CS 0.69 0.58 -0.08 0.10 0.43 -0.02 0.39
Channel size

DEP 0.38 -0.34 0.16 0.02 -0.23 -0.14 0.06

VWV 0.81 -0.01 -0.10 0.01 -0.18 -0.11 0.13
Fish cover

UNC -0.14 0.17 0.19 -0.21 0.23 0.08 -0.11

BED 0.44 0.41 -0.22 0.06 0.30 -0.11 0.16

LWD -0.48 -0.10 -0.23 0.22 -0.17 0.07 -0.07

MAC -0.48 -0.57 0.02 -0.15 -0.07 -0.09 0.04

0V -0.53 -0.28 0.00 0.03 -0.18 -0.01 0.08
Stream bank condition

RV -0.48 -0.11 0.30 -0.19 -0.08 0.06 -0. 16

BS -0.42 0.13 0.45 -0.05 -0.52 -0.12 0.11

HM 0.49 0.10 -0.57 0.10 0.29 0.06 -0.25

 

80

90°O'O"W 85°O'O"W
1 I

 

 

 

 

. 3.1.7 ,2.
~ " 0 r.’ .0
a}
45°0'0"N- . -‘ -45°0'0"N
{ _.-".'~:
09 ~ " .
‘3‘: 1' ~ O
--c-- '- *
I ‘,.
90°o"o"w 85°0|'0"W
0 Sampling sites
. N
‘ Major basins
W E
0 80 160 320 Kilometers

l 1 1 1 l 1 1 1 J S

 

Figure 2.1: Study sites locations and major river basins in Michigan. Forty five sites
were located in the Lower Peninsula and twenty were located in the Upper Peninsula.

81

 

accuse—oboe 8865 £88.“ 383:3 wag: @505 pogo USN 839.8.» goaowocno com 39:65 0.8 EB “ohm .Bom gum
nmsefi floofio 8865 use “832 Emobm no E88.“ camomv§_ mo Scuba 82% 8.335 8 com: <mo new Baggage ESE "ad 9.5»:—

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

>m2oom a reigns
0230 2269.0 awed mucosa; accuse? 23.3: < Canaan 5
o «529 m2 m4 m.— «Ems—Sm

0:2

26:

cm}:

A
. . . we

82

 

A- 0 Lower Peninsula sites<> Upper Peninsula sites
a ‘1
a . 3
2
g <> ° 0 o
g o
8’ 0 o ‘00 o $
> 0 <2) "”0 '* ' T "‘ “'fl*6' 7
g -3 ~2 -10 0 .0 i y 1.0 2 3
E’ o -1 4° 0 0 ' I .
e ‘ °'
8 -31
Stream size
B. 0 Lower Peninsula sites<> Upper Peninsula sites
3i
. .
g . 2.
8 ' 9 .w ‘ .
tn 0 . 1i .0 ~ .
E 0 O 0- ° 0 o
0 o
1:: 00
s r : ’**‘*°° 5 ° '* 3'. ““““ X“ <>
§-2 0-10 .°i%’ .01. 02
g. C 0 -1 I 0 o
o 0 '
a: o -2- o
o I
o l
-33

Riffle and coarse substrate

Figure 2.3: Sampling sites plotted over A. the first two axes generated by PCA of habitat
variables named “Riffle and coarse substrate” and “Stream stability” respectively and B.
The third and fourth axes named “Stream stability” and “Percent pools and overhanging
vegetation.” Note that two of the axes, responses to flow stability and stream size, show
separation of study sites by peninsula.

83

70.0

 

60.0 57-8
E
8 50.0
E
': 40.0
a
>
'3 30.0
E
g 20.0 18.5
x
ui
10.0 53 4'4 1 0
0.0 . l:l - _‘
Physical Network Local Network Local Interaction
landscape riparian riparian watershed watershed
variables zone zone

Figure 2.4: Percent of total explained variance by RDA predicting habitat variables from
groups of landscape variables. A total of 41 .2% of the variance within habitat variables
was explained by all groups of landscape variables

84

.0
N

RDA Axis II
o

.b
N

.b
is

4_A_

.5
o:

-0.8 " —“

-0.8

   
 

 

.HM
,ocv
.1090
BED "RAQ LWWet
0
HH DA
o .08 Run SD
06 ° ° 3
. WW NRAg FS
Rif ° .OV.MAC
Gradient NW A9
RV LRWet
O
9050 NRWet
C
as
.SFS
-0.6 -0.4 0.2 0 0.2 0.4 0.6 0.8
RDA Axisl

Figure 2.5: Plot of the first two axes from redundancy analysis (RDA) with vectors
representing landscape features and points representing physical habitat variables. Gray
text represents landscape variables while habitat variables are shown in black.

85

Literature Cited

Allan, J. D. 2004. Landscapes and riverscapes: The influence of land use on stream
ecosystems. Annual Review of Ecology, Evolution, and Systematics 35:257-284.

Baker, M. E., M. J. Wiley, and P. W. Seelbach. 2003. GIS-based models of potential
groundwater loading in glaciated landscapes: Considerations and development in

Lower Michigan. Michigan Department of Natural Resources, Fisheries Research
Report 2064.

Bollen, K. A. 1989. Structural equations with latent variables. John Wiley and Sons, Inc.,
New York.

Brenden, T. O., R. D. Clark, A. R. Cooper, P. W. Seelbach, and L. Wang. 2006. A GIS
framework for collecting, managing, and analyzing multiscale landscape variables
across large regions for river conservation and management. American Fisheries
Society Symposium 48: 49-74.

Cohen M. J. and M. T. Brown. 2007. A model examining hierarchical wetland networks
for watershed storrnwater management. Ecological Modeling 201 :179-193.

Danz, N. P., G. J. Nierrri, R. R. Regal, T. Hollenhorst, L. B. Johnson, J. M. Hanowski, R.
P. Axler, J. J. H. Ciborowski, T. Hrabik, V. J. Brady, J. R. Kelly, J. A. Morrice, J.
C. Brazner, R. W. Howe, C. A. Johnston, and G. E. Host. 2007. Integrated

measures of anthropogenic stress in the US. Great Lakes basin. Environmental
Management 39: 361-647.

Dickmann, D. I. and L. A. Leefers. 2003. The forests of Michigan. University of
Michigan Press, Ann Arbor, Michigan.

Durance I., C. Lepichon, and S. J. Ormerod. 2006. Recognizing the importance of scale
in the ecology and management of riverine fish. River Research and Applications
22: 1143-1152.

Farrand, W. R., and D. L. Bell. 1982. Quaternary geology of southern Michigan (map).
Department of Geological Sciences, University of Michigan, Ann Arbor,
Michigan.

Fausch, K. D., C. E. Torgersen, C. V. Baxter, and H. W. Li. 2002. Landscape to

riverscapes: Bridging the gap between research and conservation of stream fishes.
BioScience 52: 483-498.

86

Fitzpatrick, F. A., B. C. Scudder, B. N. Lenz, and D. J. Sullivan. 2001. Effects of multi-
scale environmental characteristics on agricultural stream biota in eastern

Wisconsin. Journal of the American Water Resources Association 37: 1489-1507.

Frissell, C. A., W. J. Liss, C. E. Warren, and M. D. Hurley. 1986. A hierarchical
framework for stream habitat classification: Viewing streams in a watershed
context. Environmental Management 10: 199-214.

Hancock, G. R. and R. O. Mueller. 2006. Structural equation modeling: A second course.
Information Age Publishing, Greenwich, Connecticut.

Harding, J. S., E. F. Benfield, P. V. Bolstad, G. S. Helfman, E. B. D Jones. 1998. Stream
biodiversity: The ghost of land use past. Proceedings to the National Academy of
Sciences, USA 95:14843-14847.

Hauer, R. H. and G. A. Larnberti. 1996. Methods in stream ecology. Academic Press Inc,
San Diego, California.

Hynes, H. B. N. 1975. The stream and its valley. Verhandlungen—Intemationalen der
Vereinigung fiir theoretische und angewandte Limnologie 19:1-15.

Infante, D. M., M. J. Wiley, and P. W. Seelbach. 2006. Relationships between channel
shape, watershed characteristics, and fish assemblage structure in lower Michigan.
Pages 339-358 in R. M. Hughes, L. Wang and P. W. Seelbach, editors. Influences
of landscapes on stream habitats and biological assemblages. American Fisheries
Society, Bethesda, Maryland.

Infante, D. M. and J. D. Allan. In press. Response of stream fish assemblage to local-
scale habitat as influenced by landscape: A mechanistic investigation of stream
fish assemblage in K. Gido and D. Jackson, editors. Community ecology of
stream fishes. American Fisheries Society, Bethesda, Maryland.

Johnson, L. B, D. H. Breneman, and C. Richards. 2003. Macroinvertebrate community
structure and fimction associated with large wood in low gradient streams. River
Research and Applications 19: 199-218.

Johnston, C. A, N. E. Detenbeck, and G. J. Niemi. 1990. The cumulative effect of
wetlands on stream water quality and quantity: A landscape approach.

Biogeochemistry 10: 105—141.

Knighton, D. 1998. Fluvial forms and processes: A new perspective. John Wiley and
Sons, Inc., New York, New York.

Legendre, P., and L. Legendre. 1998. Numerical ecology, second edition. Elsevier,
Amsterdam.

87

Maruyama, G. M. 1998. Basics of structural equation modeling. Sage Publications, Inc.,
Thousand Oaks, California.

Maybee, R. H. 1960. Michigan’s white pine era, 1840-1900. Michigan Historical
Commission, Lansing, Michigan.

Mershon, W. B. 1923. Recollections of my 50 years hunting and fishing. The Stratford
Company, Boston, Massachusetts.

Michigan Department of Environmental Quality. 2002. Qualitative biological and habitat
survey protocols for wadeable streams and rivers. MI/DEQ/SWQ-96/068. Surface
Water Quality Division. Great Lakes and Environmental Assessment Section.
Staff Report. Lansing, Michigan.

Michigan Department of Natural Resources, Forest, Mineral, and Fire Management
Division. 2003. Integrated Forest Monitoring, Assessment, and Prescription
(IFMAP).

Miller, H. L. 1966. The old Au Sable. B. Eerdmuns Publishing Company, Grand Rapids,
Michigan.

Norman, G. R. and D. L. Streiner. 1986. PDQ Statistics. B.C. Decker, Toronto, Canada.

Poff, N. L. 1997. Landscape filters and species traits: Towards mechanistic
understanding and prediction in stream ecology. Journal of the North American
Benthological Society 16:391-409.

Raleigh, FR. 1982. Habitat suitability index models: Brook trout. United States
Department of the Interior, Fish and Wildlife Service. FW/OBS-82/10.24.

Richards, C., L. B. Johnson, and G. E. Host. 1996. Landscape-scale influences on stream
habitats and biota. Canadian Journal of Fisheries and Aquatic Sciences 53:295-
311.

Riseng, C. M., M. J. Wiley, and R. J. Stevenson. 2004. Hydrologic disturbance and
nutrient effects on benthic community structure in midwestem US streams: A

covariance structure analysis. Journal of the North American Benthological
Society 23:209-326.

Roth, N. E., J. D. Allan, and D. L. Erickson. 1996. Landscape influences on stream biotic
integrity assessed at multiple spatial scales. Landscape Ecology 11:141-156.

Roy, A. H., A. D. Rosemond, M. J. Paul, D. S. Leigh, and J. B. Wallace. 2003. Stream
macroinvertebrate response to catchment urbanization (Georgia, USA).
Freshwater Biology 48:329-346.

88

Seelbach, P. W., L. C. Hinz, M. J. Wiley, and A. R. Cooper. 2007. Using multiple linear
regression to estimate flow regimes for all rivers across, Illinois, Michigan, and
Wisconsin. Michigan Department of Natural Resources, Fisheries Technical
Report, Ann Arbor, Michigan.

Schumacker, R. E. and R. G. Lomax. 2004. A beginner’s guide to structural equation
modeling. Lawrence Erlbaum Associates, Mahwah, New Jersey.

Shepherd, R. G. 1989. Correlations of permeability and grain size. Ground Water 27:633-
638.

Simonson, T. D., J. Lyons, and P. D. Kanehl. 1994. Guidelines for evaluating fish habitat
in Wisconsin streams. General Technical Report NC-164, US. Forest Service, St.
Paul., Minnesota.

Stephenson, J. A. and A. Morin. 2009. Covariation of stream community structure and
biomass of algae, invertebrates and fish with forest cover at multiple spatial
scales. Freshwater Biology 54: 2139- 2154.

Ter Braak C. F. and I. C. Prentice. 2004. A theory of gradient analysis. Advances in
Ecological Researchz34: 235-282.

The Nature Conservancy and World Wildlife Fund, Inc. 2008. Freshwater Ecoregions of
the world. < www.feow.org>.

Tompkins, T. M., W. W. Whipps, L. J. Manor, M. J. Wiley, C. W. Radcliffe, and D. M.
Majewski. 1997. Wetland effects on hydrological and water quality characteristics
of a Mid-Michigan system pages 273-285 in C. C. Trettin, M. F. Jurgensen, D. F.
Grigal, M. R. Gale, and J. K. Jeglum, editors. Northern Forested Wetlands. CRC
Press, Boca Raton, Florida.

Tonn, W. M., J. J. Magnuson, M. Rask, and J. Toivonen. 1990. Intercontinental
comparison of small-lake fish assemblages: The balance between local and
regional processes. The American Naturalist 136:345-375.

Townsend, C. R., S. Doledec, R. H. Norris, K. Peacock, and C. J. Arbuckle. 2003. The
influence of scale and geography on relationships between stream community

composition and landscape variables: Description and prediction. Freshwater
Biology 48:768-785.

United States Geological Survey. 2004a. National hydrography dataset.
<www.nhd.usgs.gov> February 2010.

United States Geological Survery. 2004b. National elevation dataset.
<www.ned.usgs.gov> February 2010.

89

Utz, R. M., R. H. Hilderbrand, and R .L. Raesly. 2010. Regional differences in patterns of
fish species loss with changing land use. Biological Conservation 143: 688-699

Vannote, R. L., G. W. Minshall, K. W. Cummins, J. R. Sedell, and C. E. Cushing. 1980.
The river continuum concept. Canadian Journal of Fisheries and Aquatic Sciences
37:130-137.

Van Sickle, J ., and R. M. Hughes. 2000. Classification strengths of ecoregions,
catchments, and geographic clusters for aquatic vertebrates in Oregon. Journal of
the North American Benthological Society 19:370-384.

Vincent, R. E. 1962. Biogeographical and ecological factors contributing to the decline
of Arctic grayling, Thymallus arcticus (Pallas), in Michigan and Montana.
Doctoral Dissertation. University Microfilms, Inc., Ann Arbor, Michigan.

Wang, L., J. Lyons, P. Kanehl, and R. Gatti. 1997. Influences of watershed land use on
habitat quality and biotic integrity in Wisconsin streams. Fisheries 22:6-12.

Wang L., P. Rasmussen, P. Seelbach, T. Simon, M. Wiley, P. Kanehl, E. Baker, S.
Niemela, and RM. Stewart. 2003. Watershed, reach, and riparian influences on
stream fish assemblages in the Northern Lakes and Forest ecoregion, U.S.A.
Canadian Journal of Fisheries and Aquatic Sciences 60:491-505

Wang, Y.,W. Hong, C. Wu, D. He, S. Lin, H. Fan. 2008 Application of landscape
ecology to the research on wetlands. Journal of Forestry Research 19:164-170.

Wiens, J. A. 2002. Riverine landscapes: Taking landscape ecology into the water.
Freshwater biology 472501-515.

Wootton, J. T. 1994. Predicting direct and indirect effects - an integrated approach using
experiments and path analysis. Ecology 75:151-165. '

90

Chapter 3

Introduction

The decline in the biodiversity of freshwater organisms globally has been
described in recent assessments of imperiled organisms. Over 47% of known crayfish
species within North America are now considered imperiled (Taylor et al. 1996). Mussel
taxa are facing a bleaker picture within the United States and Canada, with 72%
threatened (Abell et al. 2000). The number of imperiled and extinct freshwater and
diadromous fish taxa in North America has increased from 363 to 700 since 1989 (Jelks
et al. 2008). Salmonids alone represent 11% of listed taxa, with 46% of these being
distinct populations or seasonal runs (Jelks et al. 2008).

While various factors have been cited as leading to imperilment including
overfishing and the spread of invasive organisms, the degradation of habitat is considered
to be a main cause of the decline of freshwater taxa, with over 71% of freshwater fish
extinctions world-wide directly relatable to flow alterations, fragmentation by dams,
removal of woody debris, sediment deposition, increased pollution, and sediment load
(Helfinan 2007). In stream systems, degraded conditions in localized reaches can further
fragment suitable habitat along the longitudinal stream gradient creating unsuitable
patches that constrain the movement of species and that may increase risk of species loss
(Fausch et al. 2002). Habitat restoration can be used to increase the suitability of a
stream system, but in some cases habitat has been altered in ways that are irreversible. In
such instances, the identification of stream networks that contain ample amounts of

suitable habitat may be imperative to protecting imperiled species.

91

The use of a Geographic Information System (GIS) is one approach that allows
for the identification of potential habitat over large geographic areas. Large datasets of
information on land cover, gradient, and surficial geology are available for many regions
and several studies over large areas have used similar datasets to assess stream condition.
At the national scale, Essleman et al. (in press) created a disturbance index based on 15
land use disturbance metrics summarized at two scales, the local catchment draining into
a stream reach and the entire catchment upstream of the reach, capable of quantifying
anthropogenic stress on individual stream reaches across the United States. Danz et al.
(2007) summarized data on atmospheric deposition, human population, land cover, and
point source pollution in order to create a regional stress index for river basins of the
Great Lakes and then examined the role of stress levels on determining fish community
characteristics. Such studies performed across large regions illustrate how landscape
factors can be used as surrogates for stream condition, and are based on numerous studies
showing specific relationships between landscape factors such as gradient, surficial
geology, and land use on stream habitat conditions and quality (Roth et al. 1996,
Richards et al. 1997, Wang et al. 1997, Wang et al. 2003, Infante and Allan in press).

Together, the use of landscape and in-strearn habitat characteristics can be used to
understand multiple scale controls on fish community composition (Tonn et al. 1990,
Poff 1997, Quist et al. 2005). Tonn et al. (1990) attempted to predict community
composition in lakes in two distinct geographic regions, creating a filter approach that
organized different controls on fish presence at multiple spatial scales. This approach
used known relationships between large scale variables and local lake conditions to

predict habitat within the lake. Poff (1997) made this approach applicable to streams by

92

suggesting the use of a hierarchical framework based in part on the spatiotemporal scale
presented in Frissell et al. (1986) in which geographic, landscape, and biological filters at
several spatial scales could be used to predict community composition based on
characteristics needed to support specific fish species. Quist et al. (2005) used field
collected data on in-stream habitat conditions to predict individual species presence or
absence within a study reach. While these studies incorporate a filter approach for
understanding community composition, utilizing landscape and reach scale information
to predict habitat characteristics known to control a species distribution can be used to
identify stream systems that a species could potentially occupy.

The Arctic grayling (T hymallus arcticus) is a holarctic species and in North
America exists through Alaska and into Canada, extending south to Alberta and British
Columbia and cast into Nunavut (McAllister and Crossman 1973, Scott and Crossman
1973). The Michigan subpopulation was one of two glacial refugia subpopulations of the
species within the conterminous United States, while the other is located in the upper
Missouri River, Montana (Vincent 1962, Scott and Crossman 1973, Northcote 1993). The
Arctic grayling was extirpated from Michigan in 1936 (Vincent 1962). A combination of
habitat degradation, overfishing and competition from non-native species was believed to
cause the grayling’s decline and eventual disappearance (Mershon 1923, Vincent 1962,
Nuhfer 1992). While several reintroductions have been attempted, none have been
successful (Nuflier 1992). The last reintroductions occurred from 1987 through 1991,
and were attempted in a small number of lakes and rivers selected based on their
remoteness of location, lack of competitor or predator species, and ability to support

other trout species (Nuhfer 1992). Some individuals survived for several years in lakes

93

but no spawning was observed, while fish stocked in rivers disappeared soon after
reintroduction (within 6 months) (N uhfer 1992). Failed reintroductions were attributed to
a number of factors including lack of suitable grayling habitat, fragmentation of streams
by dams, high water temperatures in some reintroduction sites, overfishing and illegal
taking in open fishing areas, fungal infections, and the inability of fish to remain in the
area in which they were stocked (Nuhfer 1992).

Recently, concern has been growing over dwindling subpopulations of grayling in
both Alberta and British Columbia and the Big Hole River (Kaya 1992, Northcote 1993,
Clarke et al. 2007, Backus 2007, Lamothe and Peterson 2007). Because of the increasing
awareness of threats facing the species as well as strong local support, new research since
the last Michigan reintroduction attempts has become available. Given new knowledge
on habitat requirements and preferences of grayling and advances in the ability of
identifying stream conditions over large areas using landscape data, the question of
whether suitable Arctic grayling habitat in Michigan exists can be revisited.

My research addresses the question: does suitable grayling habitat currently exist
in Michigan? To answer this question, I implemented a three-step approach to
characterize stream habitat. First, for all stream reaches in my study area, I develop a
hierarchical assessment tool capable of evaluating habitat suitability. Next, I rate stream
segments, or connected stream reaches free of barriers to migration, based on total
connectivity in the context of reach habitat suitability scores. Finally, I use in-stream
data collected in the field to determine if habitat characteristics important to Arctic
grayling that cannot be incorporated into my assessment tool exist in high rated stream

segments. By using new data and tools to attempt to locate suitable habitat specific to a

94

species, 1 plan to develop a novel approach that can be used to determine if Arctic
grayling can exist within Michigan streams. My hope is that this approach will be
applicable to management projects in which the goal is to determine where suitable

habitat may exist for a range restricted or locally extinct species.
Methods

Study area

Historically, Arctic grayling were found in all catchments in the Lower Peninsula
of Michigan north of and including the White River draining to Lake Michigan in the
west and north of and including the Rifle River draining into Lake Huron in the east
(Vincent 1962, Figure 3.1). In the Upper Peninsula, the species was only found in the
Sturgeon River basin, in both the Otter and Sturgeon Rivers (Vincent 1962). My study
area included the historic species’ range in the Lower Peninsula as well as the entire
Upper Peninsula. The Upper Peninsula was included because of its relative lack of
human impact and because reintroductions were attempted in streams throughout this
region. In general, the landscape of my study region is dominated by forests and
wetlands, with small areas of agricultural and urban land uses (Wang et al. 2003, Danz et
al. 2007). l

A total of 69 reaches were sampled during the summer months of 2009. A reach
is defined as a continuous section of surface water with similar hydrologic characteristics
separated by the confluence with another reach above and below it (U SGS 2004). Study
reaches were selected to capture a range of landscape and land cover characteristics, to
include locations with historical voucher specimens of Arctic grayling, and to include

locations where recent reintroductions were attempted from 1987 to 1991 (Nuhfer 1992).

95

Study sites were selected within reaches based on accessibility and were located at least 1
km downstream from the first upstream confluence and at least 10 m upstream from
stream entry. Site length equaled 40 times average stream width following a stream
sampling protocol developed for Wisconsin by Simonson et al. (1994).
Habitat data collection and summary

Habitat data collected included measures of water depth, channel shape, bank
stability, riparian vegetation type, percent overhanging vegetation, substrate type and
embeddedness, percent and type of fish cover, percent of geomorphic units (run, riffle,
pool), and summer water temperatures. Habitat sampling methods were based on
Simonson et al. (1994). At 20 transects equally spaced throughout a sample site, five
measurements were taken at regular points across each transect to characterize stream
depth, dominant substrate (clay, silt, sand, fine gravel, coarse gravel, cobble, boulder, or
bedrock), and embeddedness of gravel or cobble. Embeddedness was ranked in quartiles
from 0-100%, with 0% indicating no or very little embeddedness, 25% indicating that
individual particles of substrate are embedded to half their height, 50% indicating that
substrate is heavily embedded on its sides but lacks surface cover, 75% indicating that
substrate is partially covered by sediment or sand, and 100% indicating that substrate is
completely covered in sediment or sand. Percent substrate types and percent fish cover
type measurements including macrophytes, bedrock, boulder, and large woody debris
(LWD) were visually estimated over an area 3 m on either side of the transect line.
Occurrence of bank undercuts and percent of overhanging vegetation acting as cover
were noted for both banks at each transect, as was the percent of riparian vegetation or

land cover type within 5 m of the stream on each bank. Channel dimensions including

96

wetted and bankfull width were measured using a rangefinder accurate to 0.5 m, while
bankfull height was estimated to the nearest 0.1 m using a marked 2.5-meter segment of
PVC piping. Bankfull height was measured as the difference in height between the
stream flow surface at the time of sampling and the point at which water would enter the
floodplain. Bank stability was visually scored for each stream bank on each transect with
4 indicating a stream bank with no visible evidence of erosion, 3 indicating some
evidence of erosion and recent stream overflow, 2 showing moderate erosion causing
unstable soil conditions, 1 showing heavy erosion and a deteriorating bank, and 0
indicating no structural support and maximal erosion.

To characterize habitat conditions that I was unable to measure quantitatively, I
used the Michigan Department of Environmental Quality (MDEQ) Procedure 51 habitat
assessment (MDEQ 2002). This method visually scores a variety of physical
characteristics used to determine stream condition and habitat quality including
embeddedness, variability in depth/velocity, flow stability, bottom deposition and
sedimentation, geomorphic unit diversity, bank stability, bank vegetation stability, and
streamside cover. July temperatures were recorded at 27 of the 69 sites using Hoboware
Pro temperature loggers. Temperature loggers were placed in streams throughout the
month of June and collected in August. Mean daily temperature was calculated for each
reach for the month of July.

Landscape and modeled reach data

The stream coverage used in this analysis was the National Hydrography Dataset

(NHD) defined at a 1:100,000 scale (USGS 2004a). Drainage area was calculated using

reach catchments delineated from the NHD and National Elevation Dataset (NED, USGS

97

2004b). Landscape data used in this study were organized at two different spatial scales
for each stream reach: the network catchment and the network riparian zone. Network
refers to the area upstream of the reach throughout the watershed or riparian zone. The
riparian zone is classified as a 60 meter region on either side of the stream. Data on land
cover (IFMAP 2001) and surficial geology (F arrand and Bell 1982) were summarized,
with percent forest and percent coarse geology summarized at the network catchment
scale, while percent urban, agriculture, and open and forested wetlands summarized at the
network riparian zone scale. Coarse geology is known to influence stream stability and
stream temperature (Seelbach and Wiley 1997, Seelbach et al. 2007) and is the percent of
geologic types with high hydraulic conductivity (values greater than 5.0 m/d), which
describes the relative velocity at which water moves through porous spaces in soil
(Shepherd 1989). Stream gradient summarized at the reach scale was the drop in meters
per meters of stream reach as depicted by the NHD coverage. See Brenden et al. (2006)
for further explanation of catchment delineation.

Stream flow regimes were characterized for each stream reach using a model of
groundwater delivery to stream channels, precipitation data, surficial geology and land
use (Seelbach et al 2007). The 10/90 ratio, the annual flow that is exceeded 10% of the
time divided by the annual flow exceeded 90% of the time, was calculated and used as a
metric for estimating stream flow stability. A 90/50 ratio, the annual flow exceeded 90%
of the time divided by the annual flow exceeded 50% of the time, was calculated to
indicate areas with high baseflow (Raleigh 1982). Modeled July mean stream
temperatures were predicted using statistical models based on relationships of sample

temperature measurements and landscape data from throughout Michigan (Werhly et al.

98

2009). The majority of observed July mean temperatures (89%) were lower than the
modeled temperatures used in my assessment tool and close to observed values, with a
4.6° C maximum deviation from the observed value (Figure 3.2). Only three of twenty—
seven sites had higher temperatures than predicted, all within one degree of their modeled
temperature. Given these results, I felt that using modeled temperatures within my
assessment was justifiable.
Fish data and index of biotic integrity

Brook trout abundance and fish assemblage data were from the Michigan
Department of Natural Resources (MDNR) Fish Collection System and Michigan Rivers
Inventory databases (Seelbach and Wiley 1997). These collections were made largely
through backpack or barge electrofishing, and samples that were collected using rotenone
were corrected using information from sites that contained both electrofishing and
rotenone samples (Seelbach et al. 1994). Fish were identified and counted in the field.

Wang et al. (in press) calculated index of biotic integrity (IBI) scores for coldwater
stream sites using the Wisconsin coldwater IBI procedure (Lyons et al. 1996), while
wadeable warmwater IBI scores were calculated using methods outlined in Procedure 51
(MDEQ 2002) then adjusted for ecoregion differences (Wang et al. 2008). Both
coldwater and warmwater IBI scores were calculated for marginal (coolwater) trout
streams and the high of the two values was used to represent a stream reach, because no
coolwater IBI for the region currently exists (Wang et al. in press). IBI scores scaled
from 0 to 100, with 90-100 being deemed excellent, 60-80 good, 30-50 fair, 10-20 poor,
and 0-10 being very poor (Lyons et al.1996).

Overview of the assessment tool

99

The tool developed to assess habitat suitability incorporated a series of
hierarchical filters that accounted for stream habitat conditions important to Arctic
grayling. Filters included natural catchment landscape variables, land cover at the
network catchment and network riparian zone scales, and modeled variables specifically
describing conditions in stream reaches (stream flow and temperature). The method
follows approaches and ideas of Tonn (1990), Poff (1997) and Quist et al. (2005), who
use hierarchical filters to predict fish assemblages in lake and river systems using
geographic (historical species distribution and separation due to physical barriers),
landscape, and biotic variables. My use of the filter approach did not focus on predicting
fish distributions directly, but instead was used to rate streams based on their suitability
of habitat characteristics for Arctic grayling. These filters were organized over three
spatial scales; the network catchment, the network riparian zone, and the reach. My
framework contained five hierarchical filters: drainage area; network catchment
conditions, network riparian conditions, reach conditions; and stream temperature (Figure
3.3). Once an individual reach score was determined, the length of a stream between
barriers to migration, i.e., the total connectivity of a stream segment, was calculated. A
“habitat-weighted connectivity” score was then applied to each reach, a measurement
created to address the concept that fish migration length is dependent on the quantity and
quality of habitat within a connected stream segment (Benke 1992, Fausch et al. 2002).
Habitat-weighted connectivity scores were then summarized and weighted by total length
within a stream segment to create a final stream segment rating.

In order to determine breaks in suitability of landscape drivers, I graphed

relationships between landscape conditions and potential response variables. The

100

frequently non-linear relationships of habitat and biological characteristics with
landscape features have been used to mark theoretical thresholds indicative of ecosystem
condition or health (Huggett 2005). In this study, I defined a threshold as a zone rather
than a point, indicative of a gradual shift from one ecological state to another (Marudian
2001, Huggett 2005). The zones above and below this range would then represent
differing ecological states. Monitoring disturbance variables in an attempt to keep
conditions from reaching a point where degradation may be irreversible is one application
of thresholds (Wiens et al. 2002). However, I used threshold zones in a different way, to
examine changes in relationships between landscape factors and variables indicative of
stream quality and condition potentially important to Arctic grayling. Specifically, I used
fish IBI scores, 10/90 ratio, and a visual metric of stream flow stability as response
variables. I believed that for values within the suitable range determined for a predictor
variable, it was more likely that the specific reach would have characteristics suitable for
grayling. The marginal range would be less likely to have suitable habitat, while the poor
range would be the most likely to have conditions unsuitable for grayling.
Scoring stream reaches

Predictor variables were assigned a suitability score ranging from 0 to 2 (Table
3.1), with breaks defined by conditions known to support Arctic grayling, habitat
responses to landscape characteristics, or thresholds identified to depict variable
responses to large-scale controls (Table 3.2). Depending on the cumulative value of
variables at each filter, a stream reach was then assigned a score ranging from 1 to 2 or 0
to 2, and the sum of all filter scores was considered the individual reach score. I chose to

include a score of “0” for reach scale and temperature filters because of the limiting

101

effects of variables relating to flow stability, depth for cover, gradient, and temperature
on Arctic grayling survival. At the network catchment and network riparian zone, high
human disturbance and low coarse geology are indicative of conditions unsuitable for
Arctic grayling, but the response of stream health metrics to these large-scale variables
can be affected by other factors and may vary in their response across reaches despite
general trends. Therefore, I felt it would be inappropriate to apply a score of zero or
consider a stream unsuitable because of a single landscape feature. The reach scale
variables had a more direct effect on the physical habitat available to the Arctic grayling,
and therefore the cumulative total score of 10/90 ratio, 90/ 50 ratio, and gradient was re-
scored 0 to 2, with 0 indicating that the reach was most likely unsuitable for sustaining a
population. Temperature was treated as its own filter due to the importance of stream
temperatures to salmonids (Brett 1979, Richter and Kolmes 2005) and zero values were
given to sites above 20°C. I believe that of all the variables I examined temperature was
the only one that could reasonably exclude grayling existence within a stream reach, and
therefore treated an unsuitable temperature score (0) as multiplicative.
Rating of stream segments

Due to the Arctic grayling’s tendency to migrate long distance between spawning,
overwintering and feeding sites (West 1992, Lamothe and Mage 2003), I attempted to
account for habitat-weighted connectivity among stream reaches of various rankings
within my assessment tool. Locations of dams were determined using a dam layer
created by the Michigan Department of Environmental Quality in 2000, with locations
corrected by the Institute for Fisheries Research at Ann Arbor in 2004. I used a Michigan

atlas, aerial photography, information on individual dams within the dataset, and Google

102

Earth to estimate the position of each dam on the NHD stream layer. Dams that occurred
on reaches not included in the NHD stream layer were deleted, while all others were
manually linked to the stream layer. A data layer consisting of points representing
inflows into the Great Lakes was joined with the dam layer. The NHD stream layer was
transformed into a network dataset, and the network analyst tool within GI S was used to
estimate the distances between barriers in stream systems. Barriers could be dams, Great
Lakes, or large natural inland lakes in lower reaches of a system.

Reaches were then re-rated based on the total length of the stream segment in
which it occurred (Figure 3.4). Because Arctic grayling are known to migrate long
distances (40-50 km) but are not likely to travel to each reach summarized in the stream
segment (West 1992, Lamothe and Magee 2003), I used a break value double that of
expected migration length to define streams with ample connectivity (100 km). The
shortest known area to contain a fluvial Arctic grayling population is just less than 6 km,
but exists in a unique and highly modified canal (Barndt and Kaya 2000). Therefore, I
used 10 km as a lower break for scoring between marginal and unsuitable connectivity.
Reaches located within stream segments with a length of greater than 100 km kept their
previous score of 0 to 2 (Table 3.3). Reaches with segment lengths between 10 and 100
km were dropped one score (2 dropped to 1 and I dropped 0), while all reaches with a
segment length less than 10 were given a scored a 0. Reach scores were then weighted
by their individual length, which were then summarized as a total value across the stream
segment divided by the total length of the stream segment in order to create a final stream
segment rating between 0 and 2.

Identification of in-stream habitat and validation of assessment tool

103

Segments were summarized by their final rating into top (1.50-1.70), high (1.30-
1.49), marginal (1.10-1.29), low (0.90-1.09), and lowest (0.00-0.89) groupings.
Individual variables included within filters were then compared across segment groups to
examine trends in variable scores as segment rating increased. Average total scores for
the visual habitat assessment protocol were also compared to final stream segments
groupings.

Additional in-stream data collected from study sites were summarized to identify
if habitat characteristics important to Arctic grayling survival but not included in my
assessment tool were present in high rated stream segments. Characteristics included
percent and type of gravel in riffles for spawning and presence of deep, stable pools and
runs for summer feeding habitat. Presence of various game fish species were also
summarized. For high rated stream segments within my study region I also considered

the occurrence of potential predators and competitors of Arctic grayling in these streams.
Results

Landscape conditions
Drainage area of my study reaches ranged from less than 1 to 10,373 km2 (Table

3.4). Forest land cover in reach catchments had a mean value of 60%, and coarse
geology was dominant, with a mean value of 71%. Reach network riparian zones had
relatively low levels of human land uses; percent agriculture’s mean value was 6% while
urban land use was low with a mean of 2%. Wetlands were common in network riparian
zones, with a mean value of 37%. Reach gradient varied by two orders of magnitude,

from 0. 01% to 1.78% across study reaches, and 10/90 flow ratio varied from as low as

104

1.1 to 640.9, with a mean of 16.3. July mean temperatures were low overall, with an
average of 164°C.

Break points of variables included in the habitat scoring system
Reaches with a drainage area between 40 km2 and 620 km2 were deemed most

suitable for the Arctic grayling and given a score of 2, while all others were given a score
of 1. The break in suitability for small drainages was determined by multiple factors.
First, grayling were known to historically inhabit medium to large streams in Michigan
(Vincent 1962) and therefore smaller drainage areas may be indicative of less suitable
habitat. Also, when brook trout abundance was plotted against drainage area (Figure
3.4), high abundances of brook trout occurred in small drainage areas. Given this
information, I concluded that grayling are more suited for larger streams, and in smaller
streams, brook trout are a potential competitor. The break between the suitable zone and
the marginal threshold zone was set at the point at which brook trout abundance
decreased, 40 krnz, which was also the upper end of “small” sites as defined by the
Michigan Department of Natural Resources (MDNR) Status and Trends Sampling
protocol. The break in suitability for larger drainages was determined to be 620 kmz, the
“very large” distinction in the Status and Trends Sampling protocol (Wills et al. 2008),
and was chosen because grayling may be limited by predation in these bodies of water
(Vincent 1962).

Percent coarse geology and forest land cover were both included in the network
catchment condition filter. Reaches with less than 50% coarse geology scored a 0,
reaches with 50% to 80% scored a 1, and reaches with greater than 80% scored a 2.

These breaks were determined by plotting the 10/90 flow ratio against coarse geology

105

(Figure 3.5). Sites with greater than 50% coarse geology have a notable decline in the
lower limit of 10/90 ratios, indicating increasing stability. This trend continues until
approximately 80% where a higher density of low values can be found and the lower
limit stabilizes at about 2. Suitability of forest cover in the network catchment was also
broken into three groups; reaches with greater than 60% forest in the network catchment
scored a 2, 30-60% a 1, and less than 30% a 0. These breaks were determined by plotting
fish IBI scores against percent forest land cover to determine points at which biological
integrity changed (Figure 3.6). IBI scores for fishes are commonly used to indicate the
“health” of a stream and can be useful when considering effects of land use disturbance
(Fausch et al. 1990). Above approximately 30% forest land cover, the number of low IBI
scores decreases and continues to decrease until 60% forest land cover, after which no
IBI values below 40 exist. Total scores for percent forest and coarse geology
cumulatively ranged from 0 to 4, with 0 to 2 being ranked a 1 and 3 to 4 being ranked a 2
for the network catchment scale.

Percent urban, agriculture, and wetlands land cover were assessed within the
network riparian landscape condition filter. Reaches with percent urban land cover less
than 5% were scored a 2, 5-15% a 1 and greater than 15% a 0. Fish IBI scores from
Michigan were plotted against percent urban in the network riparian zone to determine
thresholds in biological integrity (Figure 3.7). No sites with greater than 5% urban land
cover were ranked as “excellent” in terms of their IBI scores; therefore, I determined that
sites with less than 5% urban land use have the most likelihood of exhibiting high
biological integrity. In contrast, sites with greater than 15% urban land use had IBI

scores that were almost entirely ranked “fair” or “poor,” and I used that to determine a

106

point above which sites would be least suitable for grayling. Reaches with agriculture
land use less than 10% were scored a 2, 10-40% a 1, and greater than 40% a 0. Fish IBI
scores were also plotted against percent agricultural land use in network riparian zone to
determine breaks in suitability (Figure 3.8). Above 10% agriculture, a general trend in
decreasing “good” scores was observed, with increasing “fair” and “poor” scores and a
sharp drop-off in scores overall at 40% agriculture in the network riparian zones. The
visual estimate of stream flow stability was plotted against percent wetlands in the
network riparian zone (Figure 3.9). Reaches with greater than 25% wetlands were scored
a 2, while those with less than 25% wetlands were scored a 1. Reaches with more than
25% wetlands in the network riparian zone were found to have a maximum score of 10
(the highest possible) and a minimum of 4, while those with less 25% had a maximum
score of 7 and a minimum score of 2; consequently, I set breaks in suitability scores to
reflect these trends. Cumulative scores for urban, wetlands and agriculture in the
network riparian zone were 0 to 6, with values of 0 to 3 being given an overall network
riparian score of l and values from 4 to 6 given an overall score of 2.

Stream gradient, 10/90 ratio, and 90/50 ratio were assessed within the reach
condition filter. Stream gradient values from 0.09% to 0.30% were given a score of 2,
0.31% to 1% given a score of 1, and values less than 0.09% or greater than 1% were
scored 0. Stream gradient breaks were determined based on historical accounts of Arctic
grayling, which described gradient for optimal environments ranging from 0.09% to
0.29% in Michigan (Vincent 1962) and 0.28% in Montana (Liknes and Gould 1987).
While Vincent (1962) describes a maximum gradient of 0.38% for Arctic grayling, I

chose a higher value of 1% because brook trout and grayling in the Upper Big Hole River

107

showed low interspecific competition in areas with less than 1% gradient (Byroth and
Magee 1998). Ten-ninety ratios above 20 were given a score of 0, 10 to 20 a 1, and
below 10 a 2. Fish IBI scores were plotted against 10/90 ratio across Michigan to assess
thresholds in biological integrity associated with stream stability (Figure 3.10). At a
10/90 ratio of 20, IBI scores show a sharp decrease in general, and “excellent” scores do
not occur after this point. I considered values greater than 20 to be unsuitable in the
rating system. Because of the importance of stable flow to the survival of Arctic grayling
(Vincent 1962, Kreuger 1981, Liknes and Gould 1981, Lamothe and Magee 2004), I
chose to create a second break at a 10/90 ratio of 10, differentiating between streams
stable enough to support most fish species and streams that could be considered
extremely stable. Reaches with 90/50 ratios greater than 0.55 scored a 2, 0.25 to 0.55 a l,
and less than 0.25 scored a 0. Ninety-fifty ratio was considered a surrogate for depth and
baseflow (Raleigh 1982, Raleigh et al. 1986), and breaks in 90/50 ratio within the habitat
suitability index model for brown trout, Salmo trutta, were used in my assessment
(Raleigh et al. 1986). I believe that given the importance of depth as cover to both brown
trout and grayling and in the absence of information specific to grayling, that 90/50 ratio
breaks used to identify brown trout habitat can be applied to Arctic grayling as well
(Raleigh et al. 1986, Byroth and Magee 1998). Stream gradient, 10/90 ratio and 90/50
ratio have a cumulative possible score of 0 to 6 at the reach scale. This cumulative score
was then rescored for the filter, with 0 to 2 being scored a 0, 3-4 scored a 1, and 5-6
scored a 2.

Suitability of July mean temperature was based on known estimates of critical and

optimal temperatures for growth in different regions for the Arctic grayling. Streams with

108

July mean temperature less than 16 °C were scored a 2, 16-20 °C a l, and greater than 20
°C a 0. Because I was using modeled mean temperatures, I chose a value less than
described critical maximum values for Montana populations (23 -25 °C) (Lohr et al.
1996, Lamothe and Peterson 2007) as the break between marginal and unsuitable river
systems, while high rates of growth were shown to occur between 75° and 16 oC (Hubert
et al. 1985).

Overall, land cover and landscape characteristics were suitable for many the
reaches in the study area, with a majority at each filter, 60.12% at the network catchment
and 85.04% at the network riparian zone, receiving rankings of 2 (Figure 3.11). Drainage
area and variables at the reach scale acted as more discriminating filters. Drainage area
had only 21.55% of reaches score 2, while the reach habitat filter had 44.35% scored as 2,
49.49% as 1, and 33.57% as 0. Nearly 10% of streams scored a 0 for July mean
temperature, while 44.35% of streams were considered suitable and received a score of 2.
Scored stream reaches

Overall individual reach scores ranged from 0 to 10 before connectivity was
applied. High scoring reaches (cumulative score greater than 8) made up 9.3% of the
total reaches in Michigan and were located in the northern portion of the Lower Peninsula
(a total of 13.4% of Lower Peninsula reaches) (Figure 3.12). In the Upper Peninsula,
high rated reaches were less common (6.2% of Upper Peninsula reaches), with the
majority existing within the central portion just west of Munising as well as a small
amount in the west. The majority of reaches scored 5 to 8 in my assessment tool
(78.8%).

Rated stream segments

109

Segments were summarized by their final rating into top (1.50-1.70), high (1.30-
l.49), marginal (1.10-1.29), low (0.90-1.09), and lowest (0.00-0.89) groupings. After
incorporating connectivity across stream reaches, final stream segment rating ranged
from 0.00 to 1.70 (Figure 3.13). The top rated segment was the Little Manistee River
(1.70) in the Manistee watershed (Table 3.5). Five of the six remaining high-rating
segments also occurred in the Lower Peninsula: the Pine, Sturgeon, Pere Marquette,
Upper White and Upper Au Sable Rivers (Figure 3.14, Table 3.5). All of these Lower
Peninsula streams had historical records of grayling being present, supported by voucher
specimens or written accounts (Unknown 1879, Mershon 1923, Bailey et al. 2006). The
Sucker River was the highest rated Upper Peninsula river (1.38), located east of Munising
and flowing into Lake Superior. The majority of marginal stream segments occur in the
Lower Peninsula (63.6%), with the other four occurring in the central and western Upper
Peninsula. Segment scores in which reintroductions were attempted ranged from 0.00
(Section 34, Spray Creek) to 1.21 (Upper Manistee), with an average score of 0.63.
Reaches in Spray and Section 34 Creeks had marginal rankings before connectivity was
applied, but ranked lower after due to their relatively short lengths.
Evaluation of grouped stream segments

Stream segment groups plotted against individual predictor variables had
increasing suitability scores as rating increased. Overall, network watershed and riparian
zone land cover characteristics for high rated and top rated segments fell within the range
considered suitable for Arctic grayling (Figures 3.15-3.18). July mean temperature
scores for all segment groupings had mean values close to the suitable range (Figure

3.19). Percent coarse geology and reach specific variables (gradient, 10/90 ratio, and

110

90/50 ratio) increased across groups of sites with higher ratings (Figures 3.20-3.23).
Segments that rated marginal or low in the assessment also showed greater variability in
individual variables scores than the high or top segment groupings.

Stream segments in which reintroductions were attempted had land use, land
cover, and July mean temperature scores that fell within the suitable range. Average
scores of reach scale variables ranked lower for reintroduction sites than for high and top
stream segment groupings, and were more variable. Reintroduction segments had a wide
range in percent coarse geology, from 0 to 90%, with the mean value falling within the
unsuitable range. Reintroduction segments also had mean gradients within the unsuitable
range (>0.01) and the lowest average 90/50 ratio of all groups.

Total score for the visual habitat assessment protocol showed an increasing trend
with segment score, however marginal, low, and lowest grouped segments were very
similar in mean and distribution of scores (Figure 3.24). Total scores greater than 154 are
considered excellent, 105-154 are considered good, 56-104 are classified as marginal, and
less than 56 are classified as poor (MDEQ 2002). Sites within top ranked segments had a
mean value on the high end of good (138) with some values ranging into excellent.
Visual assessment scores at reintroduction sites (Upper Manistee River, Mulligan Creek,
and the Huron River) ranged from marginal to good, with a mean value of 102, the lowest
of all groups.

Identification of in-stream habitat critical to Arctic grayling life history characteristics

Geomorphic unit composition in the top rated segments was dominated by runs,
with some pools and riffles present (Table 3.6). On average, pools comprised less than

10% of reaches within top rated segments with the Little Manistee (7.4%), the Pere

lll

Marquette (6.0%), and the Pine (5.9%) having the most. Riffle habitat varied from
2.19% throughout reaches in the Pere Marquette to as high as 40.7% in the Sturgeon
River.

Riffle substrate composition varied across high rated streams, ranging from clay
to boulder (Table 3.7). Four out of six streams were dominated by fine or coarse gravel
(cumulative totals greater than 50.0%), while the Sucker River had 40.18% gravel and the
Pine had 19.2%. Fine gravel was present in all streams, ranging from 1.04% for the
Sturgeon River to 30.0% for the Pere Marquette River. Average embeddedness was less
than or equal to 50% in most streams, while fine gravel in the Sucker and Pere Marquette
Rivers had higher embeddedness of 70.8% and 75.0%, respectively.

Pools within stream segments had overhanging vegetation present over 30% of
the time (Table 3.8). Average in-stream pool cover was less than 35% in all pools, with
the highest percent cover in the Sucker River (33%). Banks were largely stable, with
mean value greater than 3. Average pool depth was greater than 0.6 m in all segments
except the Pine and Sturgeon River, while the Sucker River in general had the deepest
sampled pools (mean of 0.80 m).

Runs in top rated segments all had average depths greater than 0.40 m (Table 3.9).
Bank stability was good, with averages close to or above 3 in all segments. Mean in-
stream cover was less than 30% over all segments, with the Sucker River having the most
(28.6%).

Presence/absence of other species
All stream segments within high or top ranked groupings have game species

present based on MDNR sampling records that could be potential predators or

112

competitors of Arctic grayling. Brook trout and brown trout were present in every stream
segment (Table 3.10), while rainbow trout, Oncorhynchus mykiss, were present in every
segment except the Upper White River. Salmon, Salmo spp., are found in the Sucker,
Pere Marquette, and Little Manistee Rivers, while steelhead were also recorded in the
Little Manistee. The Pere Marquette and Pine River also support large non-salmonid
species including northern pike (Esox lucius) and largemouth bass (micropterus

salmoides) that may act as predators.

Discussion

Overview

By creating an assessment tool capable of integrating environmental factors
across multiple spatial scales, 1 was able to score northern Michigan streams based on
their potential ability to support the Arctic grayling. My assessment tool provided a
hierarchical framework for understanding the influence of landscape on habitat suitability
specific to the Arctic grayling. My results showed that characteristics summarized across
large (network catchment) and intermediate (network riparian zone) scales were suitable
over a wide range of my study region, while reach-specific variables (90/50 ratio, 10/90
ratio, stream gradient) and drainage area were limiting in determining a segments
suitability. In addition, a measure combining a habitat score for stream reaches with a
score for unfragmented lengths of stream segments in which reaches are located was
developed to indicate whether long migration paths containing high quality habitats were
available. This allowed us to create a final rating applied to all stream reaches throughout
northern Michigan that described the habitat-weighted connectivity of a stream segment.

My results suggest that there are stream segments within Michigan that have physical

113

habitat conditions suitable of supporting Arctic grayling, including the Little Manistee,
the Upper White and others in the Lower Peninsula and the Sucker River in the Upper
Peninsula. However, biological interactions with other fish species could be a limiting
factor and should be considered.
Habitat assessment

Habitat assessment occurred in two steps: application of a hierarchical filter
approach over multiple spatial scales and an assessment of habitat-weighted connectivity.
The first step followed a landscape approach in which I considered the stream in the
context of surrounding and upstream landscape conditions at multiple spatial scales
(Wiens 2002) as well as incorporating modeled flow and temperature at the reach scale.
Modeled after the works of Tom (1990), Poff (1997) and Quist et al. (2005), I used
hierarchical filters over several spatial scales to determine the suitability for a fish species
based on known and theorized responses to habitat characteristics. Tom (1990) first
proposed that historical factors and geographical location were major controls on fish
assemblage in lakes with individual characteristics acting as filters to constrain the types
of species that a given lake could support, a theory later incorporated by Poff (1997) and
applied by Quist et a1. (2005) in stream systems. While Poff (1997) focused on large-
scale predictors of fish assemblage in stream systems, Quist et al. (2005) used elevation
and reach-scale data such as stream width, depth, presence of LWD and geomorphic unit
composition to predict fish assemblages. While applying the same conceptual filtering
approach developed by Tom et al. (1990), Poff (1997), and Quist et al. (2005), my
approach in assessing stream reaches differed in both the overall goal and methodology.

First, I did not attempt to predict a fish assemblage that already existed (Tonn et al. 1990,

114

Quist et al. 2005), or focus on the presence of functional groups of fishes (Poff 1997).
Instead, I focused on identifying the most suitable habitat for a species. Another
difference was that while the methods above incorporated the removal of stream fish at
each filter to determine which were mostly likely to occur in a particular stream, I
allowed all reaches to be considered at all scales, then implemented a final rating for
each. This was because I believed that reach scale factors could actually limit but not
necessarily exclude the existence of Arctic grayling within a segment. Finally, my reach
assessment also attempted to incorporate both landscape scale data to broadly capture
habitat characteristics and modeled information specific to stream reaches.

The second step in the application of the assessment tool was to consider the
habitat-weighted connectivity of individual stream reaches. F ausch et al. (2002) notes
that understanding the interactions with the landscape and habitat suitability across the
entire river system in which a species is found is important for developing effective
reintroduction strategies. Fausch et al. (2002) defines this spatial unit of the stream
system and the landscape around it the “riverscape,” a unit that I attempt to incorporate
into my investigation. While Arctic grayling are known to migrate long distances, it can
be assumed that the length of annual migration is in part a response to availability of
habitat within the segment through which it migrates (Benke 1992, Fausch et al. 2002).
Because of this, after the application of the habitat suitability rating to stream reaches, 1
re-scored stream reaches individually based on their total connected length within a
segment. By making the score dependent on the previous habitat score and total distance,
I developed the habitat-weighted connectivity score. Then, by rating stream segments

based on the length of each functional connectivity score (0,1 or 2) within it, I attempted

115

to establish that the context of the individual reach within a segment matters, following
from the concept of the riverscape (F ausch et al. 2002) and landscape ecology concept
that patch context matters (Wiens 2002).

High rated stream reaches versus former reintroduction locations

My results show that the Little Manistee, Pine, Sturgeon, Upper White, Upper Au
Sable and Pere Marquette Rivers in the Lower Peninsula and the Sucker River in the
Upper Peninsula may offer the best chance for Arctic grayling survival based on their
habitat characteristics. The Little Manistee was the top-rated system, with both reach
characteristics and connectivity that suggest slow stable flows, cold stable temperatures,
relatively deep water for cover, and high water quality due to limited hmnan impact.
Greater variability in specific variable scores below these high rated segments indicates
that stream segments rating marginal and lower tended to have one or more variables that
were unsuitable. Further, the lower mean value for the total Procedure 51 score for
marginal vs. high rated segments indicates that marginal streams have lower habitat
quality in general.

Segments that included reintroduction sites had a mean rating that fell within the
lowest group of stream segments. Scores for three of the reintroduction sites (Spray
Creek, Section 34, and the Cedar River) were lowered significantly due to low
connectivity. The Upper Manistee River was the only site that historically supported the
Arctic grayling that has not had severe alteration due to impoundments. It was the
highest rated segment with a reintroduction attempt but fell within the segment group that
I classified as marginal. The Middle Au Sable River was also know to historically contain

Arctic grayling, but now has the Mio Dam as its upper boundary and the Alcona Dam

116

impoundment as its lower. Because of their overall low mean rating, I believe that sites
chosen for past reintroductions may not have been the most suitable. Again,
reintroduction locations were largely selected based on remoteness of location, the ability
to support trout, and lack of predator species, and it was concluded that reintroduction
failures may have been due to lack of suitable grayling habitat, fragmentation of streams
by dams, and high water temperatures (Nuhfer 1992). Reach specific variables including
gradient, 10/90 ratio, and 90/50 ratio were the limiting variables — those that lowered the
reach scores — within reintroduction segments. Reintroduction locations also had wide
variation in amounts of coarse geology in their catchments, which can lead to stable
flows and higher baseflows in Michigan river systems (Seelbach and Wiley 1997,
Seelbach et a1 2007). Large amounts of forest in their network catchments as well as low
urban and agricultural land use in the network riparian zone of reintroduction streams
indicate that disturbances resulting from anthropogenic land uses probably did not play a
role in limiting reintroduction success.
In-stream habitat of high rated segments

My assessment tool focuses on identifying stream segments most likely to have
habitat conditions capable of supporting Arctic grayling based on landscape factors and
modeled estimates of stream habitat data. However, in-stream habitat data collected from
a subset of sites allows us to validate the results of the assessment tool and also consider
specific characteristics that I was unable to account for but are known to be important to
grayling. These characteristics include presence of stable pools important for
overwintering, deep, stable runs important for feeding, and the presence of riffles and

gravel substrate important for reproduction.

117

Suitable stream geomorphic unit composition for Arctic grayling includes an
abundance of deep, slow runs with pools for overwintering and riffle habitat for spawning
(Vincent 1962, Hubert et a1. 1985, Lamothe and Peterson 2004). Through the Big Hole
River, Montana, reaches were quantified as 66% runs, 20% pools and 14% riffle habitat
(Lamothe and Peterson 2004). The same study showed that pools greater than 0.60 m in
depth having stable banks and high percentages of overhanging vegetation had high
numbers of grayling (Lamothe and Peterson 2004). While little information is given
specific to historical condition in Michigan, grayling were known to occupy runs of most
large rivers in my study region, suggesting that they may have used runs to feed, making
depth and stability important within runs (Mershon 1923, Vincent 1962).

Areas suitable for spawning are riffles with substrate dominated by gravel, with
fine gravel being more present than coarse, and limited amounts of fine sediments (clay,
silt, sand) and coarse substrate (cobble, boulder, bedrock) (Liknes 1890, Vincent 1962,
Tack 1971, Tack 1973, Butcher et al. 1981, Hubert et a1. 1985, Shepard and Oswald
1989). Grayling eggs tend to adhere to fine gravel, making gravel size an important
factor for spawning success (Tack 1971, Tack 1973).

Overall, the Little Manistee River is the most suitable for Arctic grayling based on
the landscape assessment tool and in-stream habitat characteristics selected from a subset
of sites. Its deep, stable runs and pools combined with limited sediment load and gravel-
dominated riffles would theoretically support many of the life history needs of the
species.

Most of the top ranked streams had stable banks, deep runs and pools for cover

and ranked highly using the Procedure 51 visual habitat assessment protocol. The limited

118

amount of cover in most streams, with the Sucker River being an exception, is not a
concern for their ability to support Arctic grayling, given that adults occupy streams with
limited in-stream cover successfirlly, using depth instead of LWD, rock cover, or
macrophytes (Byroth and Magee 1998). The Pere Marquette and Upper Au Sable Rivers,
although scored as suitable, had higher levels of fine sediment and embeddedness than
other systems and are not ideal for spawning. The Sucker River may be limited due to
high levels of embeddedness within spawning sites as well (Raleigh 1982, Byroth and
Magee 1998,). The Pine and Sturgeon Rivers may have riffle substrate that is larger than
ideal for grayling spawning, while the presence of riffles with large substrate and low
percentage of runs in both streams indicates a swifter flow, potentially less suitable for a
species that prefers high velocity locations of the stream (Byroth and Magee 1998).
Consideration of biological factors

While suitable physical habitat seems to be present in several stream segments
within Michigan, the presence of species that could potentially act as competitors or
predators is a concern if reintroductions were considered. In all of the top-rated stream
segments, brook trout, brown trout and rainbow trout are present, and at least one of the
streams, the Little Manistee, supports steelhead. Although competition from brook trout
may not be a factor in larger streams where depth is used as cover, competition with
brown trout could be. Brown trout are known to occupy similar areas of the stream as
grayling, using depth as cover and feeding heavily on stream drift (Raleigh et al. 1986).
Rainbow and brown trout are also know to feed on other fish, and in many of the top-
ranked stream segments both of these species are known to reach large sizes, potentially

feeding on grayling, limiting the success of fry and juveniles. Although I am not aware

119

of any research on the effects of these two species on the Arctic grayling, it is
hypothesized that one factor in the restriction of populations to the extreme upper
Missouri was initially caused by the movement of rainbow and brown trout into lower
reaches, but constrained by limited suitability for these two species in the upper reaches
(Kaya 1992, Carnpton 2006).

The use of thresholds

Thresholds are being incorporated into methods used to manage, conserve, and
understand ecosystems across the globe in different ways (Huggett 2005). Drinnan
(2005) examined effects of fragmentation parameters on the persistence of frog, bird,
fungi, and plant species, finding that remnant areas of habitat had clear threshold values
for all taxa groups. Thresholds have also been used to show responses to disturbances
on stream condition. In Michigan, Wang et al. (2008) used them to identify reference
streams by comparing anthropogenic land use to drops in stream IBI scores and percent
intolerant fish, using the point at which drops occurred to identify when the effects of
land use begin to directly influence stream health. In the central plains, Evans-White et
al. (2009) used them to measure the response of functional groups and assemblage in
marcoinvertbrates to nutrient metrics in a stream, specifically carbon- phosphorus ratio,
total phosphorus, and total nitrogen.

The use of thresholds has been criticized because of the potential for regional
differences in response variables to the same predictor (Huggett 2005, Lindenmayer et al.
2005). To address this concern, I compared threshold breaks to results from other studies
in the Midwest. In Wisconsin, as forest cover in the network catchment increased within

watersheds, IBI scores increased as well (Wang et al. 1997). In reaches where forest

120

cover was greater than 20%, few poor scores existed, while reaches with greater than
70% forest, no IBI scores under 40 existed (Wang et al. 1997), and these results were
similar to the values that I used as breaks in scoring. Another study focused in eastern
Wisconsin and western Michigan found that declines in IBI scores began to occur at 10-
20% agriculture in the network riparian zone (Fitzpatrick et a1. 2001 ), a value at which I
observed the initial shift towards lower fish IBI scores. While Fitzpatrick et al. (2001)
saw no score higher than fair afier this initial decrease, this could be because some
streams in their study region also had high levels of agriculture in their network
catchments which was not a factor in the region. For some variables where thresholds
were used to identify breaks, I did not have comparable studies within the region, but I
believe the use of fish IBI scores and measured habitat variables specific to Michigan
helped eliminate concern over regional bias.
Regional separation

The high scoring of many stream reaches at the network catchment landscape
condition and the network riparian zone landscape condition filters indicate that streams
of the region had habitat conditions suitable for the Arctic grayling. However, the
northern Lower Peninsula, which contained nearly all the Arctic grayling’s native range,
had more than double the number of suitable reaches (13.40%) than the Upper Peninsula
(6.19%), which had only one watershed where the species was described as existing
historically. Given these results, I believe some regional separation based on

geographical and possibly historical conditions occurred within the study region.
Management Implications and Future Needs

Conservation of the Arctic grayling

121

While this work suggests that habitat capable of supporting Arctic grayling exists
within Michigan, biological factors, specifically competition with and predation from
other species, may limit their successful reintroduction to the state. If a reintroduction
were attempted in Michigan, it should be implemented following an adaptive approach.
Goals should be based on ecological constraints as well as conservation objectives and
social values, and reintroductions should be planned carefully following techniques
shown to be successful in other areas (i.e., identification of potential spawning areas and
use of remote site incubators). Further, research projects focused on the greatest research
needs specific to understanding the Arctic grayling, such as the effects of predation and
competition by brown and rainbow trout as well as habitat use in different life stages,
should be conducted concurrently with reintroductions. Following an adaptive
management approach would ensure that if a reintroduction attempt failed, we can still
learn about the life history and biological interactions of the Arctic grayling, improving
future chances for reintroductions and range expansion in Michigan and other areas.

With the potential for a changing climate to alter ecosystem structure and function
across the globe (IPCC 2007) coupled with increasing human population and demands
for natural resources (Helfman 2007), subpopulations like the Montana Arctic grayling
occurring at southern ends of their range are at potentially greater risk for extirpation.
The upper Missouri River is largely fed by snowmelt, controlling both temperature and
flow conditions (Lamothe and Peterson 2007). Western mountain ranges are predicted to
have decreased snowpack in the future (IPCC 2007), which may result in flashier stream
conditions during winter months. In the summer, snowmelt may no longer be available

to stabilize flows and lower temperatures through the entire season. The upper Missouri

122

River, thought to have been a glacial refugia for the grayling because of its stable flows
and cold temperatures, may no longer possess the characteristics needed to support the
Arctic grayling.

The effects of climate change on Michigan streams will vary depending on main
sources of water to the stream. Streams dominated by groundwater input will potentially
have lesser increases in temperature with increasing air temperatures, because
groundwater is stored below the surface and cools before entering the stream (Stefan and
Preud’homme 1993). Unlike snowpack-driven streams which may become less stable
due to changes in timing and amount of input from snowmelt, groundwater driven
systems will continue to maintain stable flows, given precipitation changes are not
dramatic. The Michigan streams we identified as high rating are controlled by
groundwater input and therefore may maintain their cooler temperatures and stable flows
(Stefan and Preud’homme 1993, Steen et al. 2010). Based on the premise that suitable
habitat remains in Michigan, and if effects of climate change are determined to be less
severe in groundwater- vs. snowmelt-driven systems, reintroductions of Arctic grayling
to Michigan could be critical for supporting the conservation of this species in its native
range within the United States.

The assessment method

Overall, the method developed to answer the question of whether fluvial habitat
capable of supporting the Arctic grayling exists in Michigan was effective and holds
promise as an approach that may be applicable to for reintroduction, conservation, and
management of many species with specific habitat requirements. Establishing habitat

needs specific to grayling such as stable flows, high baseflows, and low gradients, helped

123

identify locations that are most suitable, as well as offering insights into why previous
reintroduction attempts in reaches with less suitable habitat conditions may have failed.
Further, determining how landscape factors, specifically drainage area, geology, and
wetlands in the network riparian zone predict habitat conditions, helped improve overall
understanding of landscape effects in my study region while also improving the
assessment tool. By accounting for habitat suitability of a reach in the context of overall
connectivity with other suitable reaches, 1 created a final rating that helped classify
habitat for a species that typically moves long distances within the year, a factor that is
often not considered in fisheries management (F ausch 2002).

The assessment tool itself is adaptable; new research can lead to the alteration and
improvement of habitat identification. This would allow for information acquired
through adaptive management approaches to be incorporated to better predict grayling
habitat. For instance, if brown trout predation is found to be a limiting factor on graying
survival, a biotic filter showing the presence of brown trout could be included at the reach
scale, eliminating those reaches that contain the species. While easily adaptable, my
approach remains both time and cost efficient if landscape datasets are available and
landscape to habitat relationships are understood.

The assessment method that I developed to rank stream segments based on habitat
quality specific to the Arctic grayling may be a beneficial tool in the location of habitat
for other species that may be range-restricted or locally extinct. For instance, the
reintroduction of brook trout range in Illinois is currently being considered by Illinois
Department of Resources (Hinz et al. unpublished). Brook trout were once believed to

exist within several northern stream systems in Illinois, but have been eliminated except

124

for Lake Michigan (Hinz et al. unpublished). Using the available information specific to
brook trout life history, the best possible streams for reintroduction could be indentified
using our multiple scale filter approach (Hinz et al. unpublished).

As GIS landscape data and modeled reach variables become more available,
development and application of management tools like ours represents a step towards
more time and cost efficient approaches. The assessment tool is easily interpretable and
can be related to stakeholders and the public, supplying clear justification for where and
why reintroductions were made. I believe my method of assessing grayling habitat
suitability is useful for identifying if and where a range restricted or locally extinct

species could possibly be sustained within a region.

125

Table 3.1: Scores applied to each stream reach over all five filters. Two indicates
suitable, 1 is marginal, while 0 is considered unsuitable for Arctic grayling.

 

 

Scale Variable SuitabilitLrank
2 1 0
Network watershed
2 40.00 - <40.00 or >
Drainage area (km ) 62000 620.00 --
Coarse geology (%) >80.00 50.00-80.00 <50.00
Forest land cover (%) >60.00 30.00 -60.00 <30.00
Network riparian zone
Wetlands land cover (%) >25.00 <25.00 ----
Urban land use (%) <5.00 5.00-15.00 >15.00
Agriculture land use (%) <10.00 10.00 -40.00 >40.00
Reach
0.0009- 0.0031- > 0.0100 or
Gradient 0.0030 0.0100 <0.0009
10/90 ratio <10.00 10.00-20.00 >20.00
90/50 ratio >055 0.25-0.55 <0.25
July mean temperature (C°) <16.00 16.00-20.00 >20.00

 

126

Table 3.2: Variables included in the stream segment rating tool. The reason for including
each variable is described briefly, and the sources of data and literature used to create

suitability scorings are listed.

Scale

Variable

Purpose

Data/literature source

 

Network watershed

Drainage area
<ka)

Coarse geology
(%)

Forest land cover
(%)

Network riparian zone

Reach

Wetlands land
cover (%)

Urban land use
(%)

Agricultural land
use (%)

Stream gradient

10/90 ratio

90/50 ratio

July mean
temperature(C°)

Greater depth for cover, less
competition with brook trout

Greater potential for stable flows
and cold stable temperatures

Greater potential for high quality
habitat

Greater stream flow stability with
more wetlands in the network
riparian zone, as well as a
decrease in fine sediments within
the stream channel

Greater potential for degraded
habitat

Greater potential for degraded
habitat

Arctic grayling prefer low to
moderate gradients

Arctic grayling prefer streams with
stable flow all season

Indicates greater depth for cover

Arctic grayling Maximum critical
temperatures ranging from 20.0
(Alaska) to 250° C (Montana);
USFWS habitat suitability index
shows optimal growth between 7.5
and 16° C

Krueger (1981 ),
Vincent (1962)

Seelbach et al (2007)

Relationship between
fish IBI and forest,
Wang et al.(1997)

Relationships between
visual stream stability
metric and wetlands,
Tingley (this
manuscript),

Cohen and Brown
(2007)

Relationship between
fish IBI and urban

Relationship between
IBI and agriculture,
Fitzpatrick et al. (2001)

Vincent (1962),

Liknes and Gould
(1981)

Infante (2006).
Seelbach et al. (2007),
Vincent (1962)

Raleigh (1982),
Raleigh (1986),
Vincent (1962)

Hubert et al. (1985),
Lohr et al. (1996),
Lamothe and Peterson
(2007)

 

127

Table 3.3: Length of connected stream segments and the re-ranked score given based on
individual reach habitat score

 

Individual reach Total length of connected Individual reach
habitat score reaches (km) connectivity score
> 8

0 to 10 0

10 to 100 1

>100 2
5 - 8

< 100

> 100 1
< 5

All 0

 

128

Table 3.4: Mean, range and standard deviation (SD) of variables used as filters in the
assessment tool throughout the study region.

 

Scale Variable name Mean Range SD
Network catchment
Drainage Area (kmz) 170.84 0.02-10373.28 718.18
Forest land cover (%) 59.96 0.00-100.00 22.91
Coarse geology (%) 71.30 0.00-100.00 37.98
Network riparian zone
Urban land use (%) 1.71 0.00-60.47 3.08
Agricultural land use (%) 5.76 0.00-95.29 12.66
Wetlands land cover (%) 37.18 0.00-1.00 25.54
Reach
Stream gradient (%) 0.78 0.00-18.06 1.09
10/90 flow ratio 16.30 1.13—640.92 25.82
90/50 flow ratio 0.37 0.04-0.97 0.15
July mean temperature (C°) 16.38 470-2450 0.15

 

129

 

Table 3.5: Final rating of a selection of stream segments including top ranked segments,
segments where reintroductions were performed, and a collection of other segments
where stream field data were available. River segments marked with * are those that
historically contained Arctic grayling and a voucher specimen. Those marked with **
are streams where a historical record or written account of Arctic grayling occurs but a

voucher specimen does not exist.

 

 

Final

River segment Peninsula Watershed rating—
Top rated

Little Manistee River" Lower Manistee 1.70
High rated

Pine River* Lower Manistee 1.42

Sturgeon River * Lower Cheboygan 1.41

Upper White River“ Lower Pere Marquette-White 1.39

Sucker River Upper Betsey-Chocolay 1.38

Pere Marquette

River“ Lower Pere Marquette-White 1.32

Upper Au Sable

River* Lower Au Sable 1.30
Former reintroduction sites

Upper Manistee

River" Lower Manistee 1 .21

Huron River Upper Dead-Kelsey 0.97

Mulligan Creek Upper Dead-Kelsey 0.96

Middle Au Sable

River“ Lower Au Sable 0.76

Cedar River Lower Boardman 0.54

Section 34 Upper Betsey-Chocolay 0.00

Spray Creek Upper Betsey—Chocolay 0.00
Other sites with field data

Black River“ Lower Black 1.28

Jordan River‘ Lower Boardman-Charlevoix 1.15

Otter River* Upper Sturgeon 1.04

Two Hearted River Upper Betsey-Chocolay 0.97

Upper Muskegon

River Lower Muskgeon 0.89

Sturgeon River* Upper Sturgeon 0.78

Rifle River Lower Au Gres-Rifle 0.74

Pigeon River Lower Cheboygan 0.50

Ocqueoc River Lower Lone-lake Ocqueoc 0.00

 

130

 

Table 3.6: Percent of geomorphic units averaged across sites in top-ranked stream

 

 

segments.
River segment
Pool Riffle Run

Little Manistee River 7.36 8.58 84.06
Pine River 5.89 21.48 72.63
Sturgeon River 1.33 40.67 58.00
Sucker River 4.83 4.82 90.35
Pere Marquette River 6.01 2.19 91.80
Upper Au Sable River 2.16 4.14 93.69

 

131

Table 3.7: Riffle substrate composition and embeddedness averaged over all riffles
within each high rated stream segment.

 

 

Embeddedness
River segment Riffle substrate composition (%) 1%)
Fine Coarse Fine Coarse
Sand gravel gravel Cobble Boulder Clay gravel gravel
Little Manistee 40.63 6.25 45.31 7.81 0.00 0.00 31.25 24.60
River
Pine River 31.15 2.50 17.12 42.38 5.71 1.14 50.00 32.12
Sturgeon River 1.02 1.04 56.82 41.11 0.00 0.00 75.00 35.23
Sucker River 39.29 27.68 12.50 14.29 0.00 16.25 70.83 50.00
Pere Marquette 20.00 30.00 30.00 10.00 0.00 10.00 50.00 50.00
River
Upper Au Sable 25.00 12.50 62.50 0.00 0.00 0.00 30.00 15.83
River

 

132

Table 3.8: Mean and maximum depth, average bank stability, average in stream cover
and overhanging vegetation for all pools in high rated stream segments.

 

 

Depth (m) Bank Overhanging Cover
River figment Mean Max stability vegetation (%) (%)
Little Manistee River 0.74 2.50 3.33 33.33 17.27
Pine River 0.57 1.40 3.17 62.50 20.00
Sturgeon River 0.51 1.00 2.67 50.00 10.67
Sucker River 0.80 1.30 3.19 58.33 32.96
Pere Marquette River 0.64 1.70 3.10 37.80 19.69
Upper Au Sable River 0.66 1.30 3.33 41.67 20.33

 

133

Table 3.9: Mean and maximum depth, average bank stability, and average in-stream
cover for all runs in high rated stream segments. The average percent of left and right
transect points (stream margins) in which silt was the dominate substrate type is also
presented.

Depth (m) Bank

 

River segment Mean Max stability Cover (%)
Little Manistee River 0.60 1.40 3.52 17.96
Pine River 0.60 2.10 3.12 12.73
Sturgeon River 0.40 1.20 3.39 17.31
Sucker River 0.44 1.80 2.98 28.59

Pere Marquette River 0.74 2.00 3.35 17.62

Upper Au Sable River 0.67 1.70 3.62 18.58

134

Table 3.10: Presence of fish species that could act as competitors or predators of the
Arctic grayling within high rated stream segments. Those species marked with as "X"
were present in sampling efforts. Fish assemblage data were obtained from the Michigan
Department of Natural Resources (DNR) Fish Collection System and Michigan River
Inventory databases (Seelbach and Wiley, 1997).

 

 

 

Species Stream Segment
Pere
Little Upper White Sucker Marquette Pine
Manistee Au Sable River River River River
Salve/mus

Brook trout fontinalus X X X X X X
Brown trout Salmo trutta X X X X X X :
Rainbow Oncomynchus
trout mykiss X X X X X
Salmon
spp. Salmo spp. X j X X
Northern
pike Esox Iucius X X
Largemouth Micropterus
bass salmoides X

 

135

90°0'0‘W 85° O'O‘W
l l

 

 

45°0'0'N — —45°0'0"N
l I
90°0'0'w 85°0'0'W
Streams with voucher specimens

II Basins with historical records of Arctic grayling ‘1

[“1 Major river basin boundaries . E
0 80 160 320 Kilometers '
L | | | l | | 1 | g,

 

Figure 3.1: Major river basins of Michigan historically inhabited by the Arctic grayling
(Vincent 1962) with streams from which voucher specimens were collected indicated
(Bailey et al. 2004).

136

21

20

19

Modeled July mean temperature (C°)

13:

18 *1
17 ~
16 :

15:

y = 0.2974x + 13.132

14'

- r2: 0.0539 .
,. ° . e
l ’ .0 , o ”
O
O
O O
O
O
O
O
O
13 14 15 16 17 18

Observed July mean temperature (C°)

Figure 3.2: Comparison of modeled July mean temperatures with measured July means
for 26 sites in northern Michigan. Modeled temperatures tended to be overestimated for

2009.

137

 

Network Drainage area 1,2
catchment 1

 

 

Coarse geology (%) 1,2
Forest (%)

_______________________________ l'___________-________

Network Urban (%) 1,2

riparian Agriculture (%)
Wetlands (%)

 

 

 

 

 

 

 

R 10/90 ratio 0,1,2
each 90/50 ratio

Stream gradient

1
July mean temperature 0*,1,2
Zero is multiplicative

l\.

Marginal Suitable
5—8 9-10

conneCtiVity Reach length weighted
scores by habitat score

[Unsuatable] [ Marlglnal ] LSultzable]

Fina| rating Ranked‘stream
Segments
Scores range from 0-2

 

 

 

 

 

 

 
  

      
   
 

Unsuitable
0. 3-4

 

  

 

 

 

 

 

 

Figure 3.3: Habitat assessment tool used to rate stream segments within Michigan as
the most likely to support a population of Arctic grayling. I organized the assessment
over 3 scales, the network catchment, network riparian zone, and reach inclusive of 5
filters including drainage area, catchment landscape condition, riparian zone landscape
condition, reach condition, and temperature. After the cumulative scoring of a reach, I
rescored based on total connectivity and habitat score then rated a stream segment on
total length of unsuitable, marginal, and suitable reaches.

138

250-

200

Brook trout abundance/100 m

 

O 20 40 60 80100120140160180200220240260280300320340

Drainage area (km"2)

Figure 3.4: Brook trout abundance in stream reaches of my study region plotted against
drainage area. As drainage area increases, brook trout abundance decreases, with a

. . 2
marked decrease 1n numbers occurring at 40 km .

139

20

10I90 flow ratio
8

 

 

 

o 10 20 30 4o 50 60 70 80 90 100
Surficial coarse geology in network catchment (%)

Figure 3.5: 10/90 ratio versus network catchment coarse geology (%) for all reaches in
Michigan. Reaches with coarse geology percentages less than 50% generally have 10/90
ratios greater than 8; reaches with coarse geology between 50 and 80% show a decrease
in minimum 10/90 ratio values, and those values greater than 80% have a minimum value
of about 2.

140

 

Fish IBI total score
8

20- i

10‘

 

 

0 _ - L . n, L. ..-___-__L__.-__-_ ___..L.__. F_ ___ -_____ _

0 1o 20 30 4o 50 60 7o 80 90 100
Forest in network catchment (%)

Figure 3.6: Fish Index of Biotic Integrity (IBI) total score versus forest in the network
catchment (%) across Michigan. Less than 30% forest is related to increased numbers of
sites with low scores. At 30 to 60%, the number of low scores decreases, and few sites
with greater than 60% forest have an IBI less than 40.

141

120

100’

Fish IBI total score
8

20.

A “l

 

T _fi‘.*T'—‘ —

‘_' _ _'T'_ —'—T_ ' "T I '7 i i

 

 

o P» e——

05101520253035404550556065707580

 

Urban land use in the network riparian zone (%)
Figure 3.7 Fish Index of Biotic Integrity (IBI) total score versus urban land use in the

network riparian zone (%) across Michigan. At 6%, sites show marked decrease in IBI
score, with few values above 70 present. At 15%, only one value exists above 60.

142

Fish IBI total score

 

 

0 10 20 30 4o 50 60 7o 80 90 100
Agricultural land use in network riparian zone (%)

Figure 3.8: Fish Index of Biotic Integrity (IBI) total score versus agriculture in the
network riparian (%) across Michigan. Sites with agricultural values of 10 to 40% show
a decline in IBI score overall, and at 40%, values of agriculture drop sharply.

143

Visual metric of stream flow stability

Figure 3.9: Visual metric of stream flow stability versus network riparian zone wetlands
(%). Sites scored 8 or higher first appear with more than 25% wetlands, and sites with

 

20

 

30 40 50

Network riparian zone wetlands (%)

less than 4 do not occur at about 25% wetlands.

144

 

100

804

70'

Fish IBI total score
8 8

20 ~

103

 

 

0..- mamas - Time

0 10 20 30 40 50 60 70 80 90 100110120130140150160170180
10I90 ratio

Figure 3.10: Fish Index of Biotic Integrity (IBI) total score plotted against 10/90 flow
ratio in Michigan. IBI scores decrease at a 10/90 ratio of 20.

145

90.00

80.00-
‘ 130‘
1311
.21

70.00 ‘
60.00 1‘
50.00
40.00

30.00

Stream reaches (%)

20.00
10.00

       

0.00 , ~ 7 -

Drainage area Network Network riparian Reach condition Temperature
catchment condition
condition

Filter
Figure 3.11: Percent of stream reaches scoring a 0, 1, or 2 at each hierarchical filter.
Note that only reach condition and temperature could score a 0. Landscape variables

seem to be overall suitable across segments, while drainage area and reach condition
(10/90 ratio, 90/50 ratio, and gradient) seem to be the most limiting filters.

146

 

 

 

 

 

Stream reach ranklngs 0 37.5 75 150 Kilometers
< 5 l_l_l_.I_L_l_l_L_.l

‘ 5-8

 

> 8 304' T I—

E: Major river basln bomdaries

Figure 3.12: Rating of individual stream reaches before application of connectivity. The
majority of high ranked reaches occur in the Lower Peninsula, while in the Upper
Peninsula the number of high scoring reaches is lower.

147

 

 

 

 

 

 

 

 

 

18
1
fl 1
:16 l
9
§14
{Ill
.1-
”12-lgii
g 11H , ,—
210'lllli'7ii ’3
*1 illill; 1
.2 8! -_ ‘
° Minus: 3
a; 6 Vilnidl % r:
'.l.' .- 1 ., 7"..
'2 11?: ll 111112
Iii? i 2 1: {xi
= 4 21;. . ~21: ’iv:
2 . L I: . _: 3‘ ll 1 ‘:‘
.zlt.‘ Z} flfllj‘. ‘.‘:;

2 é:"l;i‘ 'll l 1.“:i: 2
22:12: lL'i‘izile .-

0 1,. :l 1; 11.13.11 ll 1‘ al.-2.1, i..- -221 ._,
00000000000000 0
o‘— wauqcorxmeu—qq Q
OOOOOOOOOOv—Frr N

 

Final segment ranking

Figure 3.13: Frequency distribution of final stream segment rankings grouped by 0.10
changes in segment score. Black lines indicate breaks between groupings. Sites were
grouped by final rating score into top (1.50-1.70), high (1.30-1.49), marginal (1.10-1.29),
low (0.90-1.09), and lowest (0.00-0.89). Stream segments with 0.00 values were not
included within this figure.

148

 

 

 

 

 

Stream segment ranklngs 0 4O 80 160 Kilometers

000-109 | i i I [ 1 i 1 |

110—129

130—149 W . r
150-170

 

 

; , Major river basin boundaries

Figure 3.14: Final rating of stream segments across Michigan. The majority of high
rankings streams occurs in the Lower Peninsula.

149

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

100-
A
g is
v
90..
12'
.1: 80-
ii
a: 70-
8
3 50-
.5
i .0- l
b
O
u.
30- l I l l l
Low Marginal I-igh Top Reintroduction sites
(n=30) (n=11) (n=6) (n=1) _ (n=7)

Stream segments grained by final rabng
Figure 3.15: Descriptive statistics for stream segments grouped by final rating, with

statistics for reintroduction sites also shown. Horizontal bars represent mean values,
boxes represent the lower and upper quartiles, vertical bars represent the minimum and
maximum scores, and asterisks represent outliers. The top ranked group of segments has
the highest mean of forest land cover. Reintroduction sites are shown to have a wide
range in forest land cover, but the mean falls above 60%, the percentage determined to be
suitable for Arctic grayling.

150

 

 

 

0i

 

 

Urban land use in the network riparian zone (%)

Low Marginal High Top Reintroduction sites

(n=30) (n=11) ( =6) (n=1) (n=7)
Stream segments grouped by final ranking

Figure 3.16: Descriptive statistics for stream segments grouped by final rating, with
statistics for reintroduction sites also shown. Horizontal bars represent mean values,
boxes represent the lower and upper quartiles, vertical bars represent the minimum and
maximum scores, and asterisks represent outliers. All values fall in the suitable range of
urban land use (less than 5%) within my study region.

151

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

25°. 251
3 at
3
c 20"
.fl
3
.9- ie ’6
b
x 15-
L
E, it
3 10- *
’ l
a: —
H
.5
| .
3 l
2 l
3
-‘-’ |
:2 °-
de Marginal High po Reintroduction sites
(n=3 0) (n=1 1) (n=6) (n=1) (n=7)

Stream segments grouped by final ranking
Figure 3.17: Descriptive statistics for stream segments grouped by final rating, with
statistics for reintroduction sites also shown. Horizontal bars represent mean values,
boxes represent the lower and upper quartiles, vertical bars represent the minimum and
maximum scores, and asterisks represent outliers. Reintroduction sites have the lowest
mean agricultural percentage, while the top ranked site have the highest average, yet is
below 10%, the break point between good and suitable ranges.

152

sol

70'

 

 

 

 

 

 

 

 

 

30‘

 

 

 

 

 

 

 

20-

10-
*-

 

Wetlands in the network riparian zone (%)
8

Low Marginal High Top Reintroduction sites
(n=30) (n=11) (n=6) (n=1) (n=7)
Stream segments grouped by final ranking

Figure 3.18: Descriptive statistics for stream segments grouped by final rating, with
statistics for reintroduction sites also shown. Horizontal bars represent mean values,
boxes represent the lower and upper quartiles, vertical bars represent the minimum and
maximum scores, and asterisks represent outliers. All mean values are above 25%, but
low segments have some variation below 25%, while marginal segments has one segment
that has less than 25% wetlands present.

153

 

19-

18-

Mean July tenperature (C°)

 

 

14 -
13 ~
12 -
it
1 1 _ l l l I 1
Low Marginal High Top Reintroduction sites
(n=30) (n=1 1) (n=6) (n=l) (n=7)

Stream segments grouped by final rating
Figure 3.19: Descriptive statistics for stream segments grouped by final rating, with
statistics for reintroduction sites also shown. Horizontal bars represent mean values,
boxes represent the lower and upper quartiles, vertical bars represent the minimum and
maximum scores, and asterisks represent outliers. All mean temperatures fell close to the
suitable range of less than 16°C, while lower ranked segments and reintroduction sites
have high variance.

154

8
l

 

 

 

 

 

 

 

 

 

on
O

l
a

 

 

 

 

 

 

 

e

 

N
O
1

geology in the network catchment (%)
8

 

 

 

 

 

 

 

Low Marginal High Top Reintroduction sites
(n=3 0) (n=1 1) (n=6) (n=1) (n=7)
Stream segments grouped by final rating
Figure 3.20: Descriptive statistics for stream segments grouped by final rating, with
statistics for reintroduction sites also shown. Horizontal bars represent mean values,
boxes represent the lower and upper quartiles, vertical bars represent the minimum and
maximum scores, and asterisks represent outliers. With increasing coarse geology,
average group scores increase. Reintroduction segments have highly variable scores
ranging from poor to suitable. Stream segments in the low rated group had high variation
in coarse geology as well.

Coarse

155

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

0.030-
is
0.025-
5 0.020-
'3 i6
:1 0.015- * I
51, 0.010-
o.005- I '
l I
0.000- ' I . I I
Low Marginal High Top Reintroduction sites
(n=3 0) (n=1 1) (n=6) (n=1) (n=7)

Stream segments grouped by final rating

Figure 3.21: Descriptive statistics for stream segments grouped by final rating, with
statistics for reintroduction sites also shown. Horizontal bars represent mean values,
boxes represent the lower and upper quartiles, vertical bars represent the minimum and
maximum scores, and asterisks represent outliers. Mean values were similar among
ranked stream segments, but as rankings decreased groups varied more. Reintroduction
sites had a higher mean gradient than what is considered the upper limit for marginal

(0.010) with rankings as high as 0.0170.

156

35‘

301

25-

20- it

 

 

15'

10I90 ratio
)1:

 

 

 

 

 

10- l

.1 . e:

 

 

 

 

 

 

 

 

 

 

 

I

Low Marginal High po Reintroduction sites
(n=3 0) (n=1 1) (n=6) (n=1) (n=7)
Stream segments grouped by final rating
Figure 3.22: Descriptive statistics for stream segments grouped by final rating, with
statistics for reintroduction sites also shown. Horizontal bars represent mean values,
boxes represent the lower and upper quartiles, vertical bars represent the minimum and
maximum scores, and asterisks represent outliers. A general trend of decreasing mean
values and variance in 10/90 ratio occurs as stream segments increase in rank.
Reintroduction sites had a mean 10/90 ratio that was suitable (< 10), but had segments
that were poor and marginal as well.

157

0.65 7
0.60 - —

0.55 7 l

 

 

0.50-

 

 

 

 

0.45 -

 

 

 

 

0.40 -

 

90/50 ratio

 

 

 

I

 

 

 

 

 

 

0.30 -

 

0.255

 

 

W

Low Marginal _ High Top Reintroduction sites
(n=3 0) (n=] 1) (n=6) (n=1) (n=7)
Segments grouped by rating
Figure 3.23: Descriptive statistics for stream segments grouped by final rating, with
statistics for reintroduction sites also shown. Horizontal bars represent mean values,
boxes represent the lower and upper quartiles, vertical bars represent the minimum and
maximum scores, while asterisks represent outliers. A general trend of increasing mean
values and variance of 90/50 ratio occurs as stream segments increase in rank.
Reintroduction sites had a mean 90/50 ratio well below the good score of 55%.

158

160-

 

150-

 

 

 

 

 

 

1.. I I l I

 

 

130‘

 

 

 

 

 

120‘ ’

110~

 

 

 

 

 

 

1004

 

 

 

 

 

 

 

 

 

 

Total score for Procedure 51

80-

 

 

 

704 !

Lowest de Marginal High po Reintroducton sites
(n=15) (n=15l (n=14l (n=lOi (n=5) (n=11l
Segments grouped by rating

Figure 3.24: Descriptive statistics for stream segments grouped by final rating, with
statistics for reintroduction sites also shown. Horizontal bars represent mean values,
boxes represent the lower and upper quartiles, vertical bars represent the minimum and
maximum scores, while asterisks represent outliers. Top and high rated segments have
higher mean Procedure 51 total score values than those with below high and
reintroduction sites, which all have means close to 100. Cutoff for “good” total Procedure
51 total scores is 105.

 

159

 

Literature Cited

Abell, R. A., D. M. Olson, E. Dinerstein, P. T. Hurley, J. T. Diggs, W. Eichbaum, S.
Walters, W. Wettengel, T. Allnutt, L. C]. and P. Hedao. 2000. Freshwater Ecoregions of
North America: A Conservation Assessment. Island Press, Washington, DC.

Backus, P. 2007. Groups file lawsuit over Arctic grayling ruling. The Missoulian,
newspaper article, Missoula, Montana. 11/27/2007
www.missoulian.com/articles/2007/1 l/27/news/local/newsOS .txt

Bailey, R.M., W.C. Latta and GR. Smith. 2004. An Atlas of Michigan fishes with keys
and illustrations for their identification. University of Michigan, Ann Arbor,
Michigan

Barndt, S. A. and C. M. Kaya. 2000. Reproduction, growth and winter habitat of Arctic
grayling in an intermittent canal. Northwest Science 74:294-305.

Benke, R.J. 1992. Native trout of Western North America. American Fisheries Society,
Bethesda, Maryland. Monograph 6.

Brenden, T.O., R.D. Jr. Clark, A.R. Cooper, P.W. Seelbach, and L. Wang. 2006. A GIS
framework for collecting, managing, and analyzing multiscale landscape variables
across large regions for river conservation and management. American Fisheries
Society Symposium 48: 49-74.

Brett, J .R. 1979. Environmental factors and growth in W. S. Hoar and D. J.
Randall,editors. Bioenergetics and Growth, Fish Physiology,Vol. 8. Academic
Press, New York, New York.

Byorth, P. A., and J. P. Magee. 1998. Competitive interactions between Arctic grayling
and brook trout in the Big Hole River drainage, Montana. Transactions of the
American Fisheries Society 127:921-931. '

Campton, DE. 2006. Should Arctic grayling in the Big Hole River, Montana be
considered a Distinct Population Segment (DPS) under the US. Endangered
Species Act? An evaluation of existing genetic information. Report for the US.
Fish and Wildlife Service, Abernathy Fish Technology Center, Longview,
Washington.

Clarke, A. D., K. H. Telmer, and J. M. Shrimpton. 2007. Habitat use and movement
patterns for fluvial species, the Arctic grayling, in a watershed impacted by a

large reservoir; evidence from otolith microchemistry. Journal of Applied
Ecology 44:1156-1165.

160

 

Danz, N. P., G. J. Niemi, R. R. Regal, T. Hollenhorst, L. B. Johnson, J. M. Hanowski, R.

P. Axler, J. J. H. Ciborowski, T. Hrabik, V. J. Brady, J. R. Kelly, J. A. Morrice, J.

C. Brazner, R. W. Howe, C. A. Johnston and G. E. Host. 2007. Integrated
measures of anthropogenic stress in the US Great Lakes basin. Environmental
Management 39: 631-647.

Drinnan, I. N. 2005. The search for fragmentation thresholds in southern Sydney suburb.
Biological Conservation 124:339-349.

Esselman, P. C., D. M. Infante, L. Wang, D. Wu, A. Cooper, W. W. Taylor. In press. An
index of cumulative disturbance to river fish habitats of the conterrninous United
States from anthropogenic activities in landscapes. Ecological Restoration.

Evans-White, M. A., W. K. Dodds, D. G. Huggins, and D. S. Baker. 2009. Thresholds in
macroinvertebrate biodiversity and stiochiometry across water-quality gradients
in Central Plains (USA) streams. Journal of the North American Benthological
Society 28: 855-868.

Farrand, W. R., and D. L. Bell. 1982. Quaternary geology of southern Michigan (map).
Department of Geological Sciences, University of Michigan, Ann Arbor,
Michigan.

Fausch, K. D., J. Lyons, J. R. Karr, and P. L. Angerrneier. 1990. Fish communities as
indicators of environmental degradation. American Fisheries Society Symposium
81123-144.

Fausch, K. D., C. E. Torgersen, C. V. Baxter and H. W. Li. 2002. Landscapes to
riverscapes: Bridging the gap between research and conservation of stream
fishes. Bioscience 52:483-498.

Fitzpatrick, F. A., B. C. Scudder, B. N. Lenz, and D. J. Sullivan. 2001. Effects of multi-

scale environmental characteristics on agricultural stream biota in eastern
Wisconsin. Journal of the American Water Resources Association 37:1489-1507.

Frissell, C. A., W. J. Liss, C. E. Warren, and M. D. Hurley. 1986. A hierarchical
framework for stream habitat classification: Viewing streams in a watershed
context. Environmental Management 10:199-214.

Helfman, GS. 2007. Fish conservation: A guide to understanding and restoring global
aquatic biodiversity and fishery resources. Island Press, Washingont, D.C.

Hinz, L. C. unpublished. Evaluating Illinois streams for establishing brook trout

populations. Proposal prepared for submission to the Illinois Department of
Natural Resources.

161

j

Hubert, W. A., R. S. Helzner, L. A. Lee, and P. C. Nelson. 1985. Habitat suitability index
models and instream flow suitability curves: Arctic grayling riverine populations.
United States Fish and Wildlife Services, Biological Report 82(10.110).

Huggett, A.J. 2005. The concept and utility of ecological thresholds in biodiversity
conservation. Biological Conservation 124:301 —3 10.

Infante, D. M. and J. D. Allan. In press. Response of stream fish assemblage to local-
scale habitat as influenced by landscape: A mechanistic investigation of stream
fish assemblage in K. Gido and D. Jackson, editors. Community ecology of
stream fishes. AmeriCan Fisheries Society, Bethesda, Maryland.

IPCC, 2007. Climate Change 2007: Synthesis report. Contribution of working groups I, II
and III to the Fourth Assessment Report of the Intergovernmental Panel on
Climate Change. Core Writing Team, Pachauri, R.K and Reisinger, A., editors.
IPCC, Geneva, Switzerland.

Jelks, H. L., S. J. Walsh, N. M. Burkhead, S. Contreras-Balderas, E. Diaz-Pardo, D. A.
Hendrickson, J. Lyons, N. E. Mandrack, F. McCormick, J. S. Nelson, S. P.
Platania, B. A. Porter, C. B. Renaud, J. J. Schmitter-Soto, E. B. Taylor, and M. L.
Warren Jr. 2008. Conservation status of imperiled North American freshwater and
diadromous fishes. Fisheries 33:372-407.

Kaya, CM. 1992. Review of the decline and status of fluvial Arctic grayling, Thymallus
arcticus, in Montana. Proceedings to the Montana Academy of Sciences 52:43-
70.

Lamothe, P. J., and J. P. Magee. 2003. Movement and habitat selection of Arctic
grayling, brook trout, and mountain Whitefish during drought conditions in the
Big Hole River, MT. Report prepared for Montana Fish, Wildlife and Parks,
Dillon, Montana.

Lamothe, P. J. and A. Peterson. 2007. Candidate conservation agreement with assurances
for fluvial Arctic grayling in the upper Big Hole River. Report prepared for the
Montana Fish Wildlife and Parks, Dillon, Montana.

Liknes, G. A., and W. R. Gould. 1987. The distribution, habitat, and population
characteristics of fluvial grayling (Thymallus arcticus) in Montana. Northwest
Science 61:122—129

Lindenmayer, D. B., J. Fischer, R.B. Cunningham. 2005.1Native vegetation cover
thresholds associated with species responses. Biological Conservation 124: 311-
316.

Lohr, S. C., P. A. Byroth, C. M. Kaya, and W. P. Dwyer. 1996. High temperature
tolerances of fluvial Arctic grayling and comparisons with summer river

162

 

temperatures of the Big Hole River, Montana. Transactions of the American
Fisheries Society 125:933-939.

Lyons, J, L. Wang L, and T. D. Simonson. 1996. Development and validation of an index
of biotic integrity for cold-water streams in Wisconsin. North American Journal
of Fisheries Management 16: 241-256.

Muradian, R. 2001. Ecological thresholds: A survey. Ecological Economics 3817-24.

McAllister, D. E., and E. J. Crossman. 1973. A guide to the freshwater sport fishes of
Canada. National Museums of Canada, Ottawa, Canada.

Mershon, W.B. 1923. Recollections of my 50 years hunting and fishing. The Stratford
Company, Boston, Massachusetts.

Michigan Department of Environmental Quality. 2002. Qualitative biological and
habitat survey protocols for wadeable streams and rivers. MI/DEQ/SWQ-96/O68.
Surface Water Quality Division. Great Lakes and Environmental Assessment
Section. Staff Report. Lansing, Michigan.

Northcote, T.G. 1993. A review of management and enhancement options for the Arctic
grayling (Thymallus arcticus) with special reference to the Williston Reservoir
watershed in British Columbia. Peace/Williston Fish and Wildlife Compensation
Program Research Report 55, Prince George, British Columbia.

Nuhfer, A.J. 1992. Evaluation of the reintroduction of the Arctic grayling into Michigan
lakes and streams. Michigan Department of Natural Resources Fisheries Research
Report, Ann Arbor, Michigan.

Poff, N. L. 1997. Landscape filters and species traits: Towards mechanistic
understanding and prediction in stream ecology. Journal of the North American
Benthological Society 16:391-409.

Quist, M. C., F. J. Rahel, W. A. Hubert. 2005. Heirarchical faunal filters: An approach to
assessing effects of habitat and normative species on native fishes. Ecology of
Freshwater Fish 14: 24-39.

Raleigh, RF. 1982. Habitat suitability index models: Brook trout. United States
Department of the Interior, Fish and Wildlife Service Biological Report
82(10.24).

Raleigh, R. F ., L. D. Zuckerman, and P. C. Nelson. 1986. Habitat suitability index

models and instream flow suitability curves: Brown trout, revised. United States
Fisheries and Wildlife Service Biological Report 82(10.124).

163

 

 

Richards, C., L. B. Johnson, and G. E. Host. 1996. Landscape-scale influences on stream
habitats and biota. Canadian Journal of Fisheries and Aquatic Sciences 53:295-
311.

Richter, A and S. A. Kolmes. 2005. Maximum temperature limits for chinook, coho, and

chum salmon, and steelhead trout in the Pacific Northwest. Reviews in Fisheries
Science 13:23-49.

Roth, N. E., J. D. Allan, and D. L. Erickson. 1996. Landscape influences on stream biotic
integrity assessed at multiple spatial scales. Landscape Ecology 11:141-156.

Scott, W. B. and E. J. Crossman, 1973. Freshwater fishes of Canada. Bulletin of the
Fisheries Research Board of Canada. Fisheries Research Board of Canada,
Ottawa.

Seelbach, P. W., R. N. Lockwood, and J. R. Ryckman. 1994. Efficiency of sampling
river fishes with rotenone. Michigan Department of Natural Resources Fisheries
Research Repport 2009, Ann Arbor, Michigan.

‘I

Seelbach, P. W., and M. J. Wiley. 1997. Overview of the Michigan Rivers Inventory
(MRI) project. Michigan Department of Natural Resources Fisheries Technical
Report 97 (3).

Seelbach, P. W., L. C. Hinz, M. J. Wiley, and A. R. Cooper. 2007 Using multiple linear
regression to estimate flow regimes for all rivers across, Illinois, Michigan, and
Wisconsin. Michigan Department of Natural Resources, Fisheries Technical
Report, Ann Arbor, Michigan.

Shepard, B. B. and R. A. Oswald. 1989. Timing, location and population characteristics
of spawning Montana Arctic grayling (Thymallus arcticus montanus) in the Big
Hole River Drainage,1988. Montana Department of Fish, Wildlife and Parks,
Bozeman, Montana, Report.

Shepherd, R.G. 1989. Correlations of permeability and grain size. Ground Water 27:633-
63 8.

Simonson, T. D., J. Lyons, and P. D. Kanehl. 1994. Guidelines for evaluating fish habitat
in Wisconsin streams. General Technical Report NC-l64, US. Forest Service, St.
Paul., Minnesota.

Steen, P.J., M. J. Wiley, and J. S. Schaeffer. 2010. Predicting future change in Muskegon
River watershed game and fish distributions under future land cover alteration and

climate change scenarios. Transactions of the American Fisheries Society
139:396-412.

164

Stefan, H. G. and E. B. Preud’homme. 1993. Stream temperature estimation from air
temperature. Water Resources Bulletin. 29:27-45.

Tack, S. L. 1971. Distribution, abundance, and natural history of the Arctic grayling in
the Tanana River drainage. Project F-9-3, Alaska Dept. Fish and Game, Report
12.

Tack, S.L. 1973. Distribution, abundance, and natural history of the Arctic grayling in the
Tanana River drainage. Project F-9-6, Alaska Department of Fish and Game,
Federal Aid in Fish Restoration, Report 15.

Taylor, C. A., M. L. Warren, J. F. Fitzpatrick, H. H. Hobbs, R. F. Jezerinac, W. L.
Pflieger and H. W. Robinson. 1996. Conservation status of crayfishes of the
United States and Canada. Fisheries 21: 25-38.

Tonn, W. M., J. J. Magnuson, M. Rask, and J. Toivonen. 1990. Intercontinental
comparison of small-lake fish assemblages: The balance between local and
regional processes. The American Naturalist 136:345-375.

United States Geological Survey. 2004. National hydrography dataset.
www.nhd.usgs.gov February 2010.

Unknown Author. 1879. Michigan grayling. Scribers Bound Magazine. Magazine series.
November 1879- April 1880.

Vincent, RE. 1962. Biogeographical and ecological factors contributing to the decline of
Arctic grayling, Thymallus arcticus (Pallas), in Michigan and Montana. Doctoral
dissertation. University Microfilms, Inc., Ann Arbor, Michigan.

Wang, L., J. Lyons, P. Kanehl, and R. Gatti. 1997. Influences of watershed land use on
habitat quality and biotic integrity in Wisconsin streams. Fisheries 22:6-12.

Wang, L., P. Rasmussen, P. Seelbach, T. Simon, M. Wiley, P. Kanehl, E. Baker, S.
Niemela, and PM. Stewart. 2003. Watershed, reach, and riparian influences on
stream fish assemblages in the Northern Lakes and Forest ecoregion, U.S.A.
Canadian Journal of Fisheries and Aquatic Sciences 60:491-505

Wang, L, T. Brenden, P. W. Seelbach, A. Cooper, J. D. Allan, R. Clark Jr., and W.
Wiley. 2008. Landscape based identification of human disturbance gradients and
references for streams in Michigan. Environmental Monitoring and Assessment

141:1-17.
Wang, L., D. Infante, J. Lyons, J. Stewart, and A. Cooper. In press. Effects of dams in

river networks on fish assemblages in non-impoundment sections of rivers in
Michigan and Wisconsin, USA. River Research and Applications.

165

West, E. B., M. W. Smith, W. E. Barber, J. B. Reynolds and H. Hop. 1992. Autumn
migration and overwintering of Arctic grayling in coastal streams of the Arctic
National Wildlife Refuge, Alaska. Transactions of the American Fisheries Society
121:709-715.

Wehrly, K. E., T. O. Brenden, and L. Wang. 2009. A comparison of statistical approaches
for predicting stream temperatures across heterogeneous landscapes. Journal of
the American Water Resources Association 45:986-997.

Wiens, J .A. 2002. Riverine landscapes: Taking landscape ecology to the water.
Freshwater Biology 47:501-515.

Wiens, J. A., B. Van Homer, and B. R. Noon. 2002. Integrating landscape strucrure and
scale into natural resource management in Lui, J. and W. W. Taylor, editors.
Integrating Landscape Ecology into Natural Resource Management. Cambridge
University Press, United Kingdom.

Wills, T. C., T. G. Zorn, A. J. Nuhfer, and D. M. Infante. 2008. Stream Status and
Trends Program sampling protocols. Chapter 26 in Manual of fisheries survey

methods. Michigan Department of Natural Resources, Fisheries internal
document, Ann Arbor, Michigan.

166

 

APPENDIX

167

 

 

3.0

co”.
m:

N00 Nvd
0N0 N00
50.0

.ou eon.
>>Z >3

_.N.0-
0.0-
00.0-
0N0-

m<

>>Z

3.0
2.0
0N0-
00.0-
05.0

9..

m2

00.0
No.0-
0.20-
0N0-

3.0

3.0

m<

m...

00.0-
0.20-
00.0-
00.0-
00.0
00.0
00.0

00.0
_._..0
0V0
00.0
#00-
0 P.0-
00.0-
00.0-

3.0
00.0
0N0
00.0
0N0-
00.0
5.0-
3.0-
No.0-

m< $.80 080

>>-_

00.0-
00.0-
mvd-
3.6-
00.0-
0 _. .0-
00.0-

00.0
00.0-
#00-

~o>>
>>-_

V0.0-
00.0-
0V0-
N00-
no.0-
00.0-
0N0-
0 F .0-
n r .0-
«N0-

0V0

uw>>
m 2

N00-
5N0-
00.0-
0V0-
$0-
2.0-
No.0-

3.0
0V0-
0 _. .0-

n00

00.0

$3
>>Z

52.0-
00.0-
9.0-
0V0-
00.0
vwd-
0N0-
No.0-
00.0-
med-
00.0
0V0
00.0

uo>>
m.-

Nud-
h _. .0-
00.0-
00.0
nwd
00.0-
—. _..0-
5.0
5.0
00.0-
«0.0
00.0
v0.0
0 F .0

<0

5an
an”...
.8052
.82:
232
9E2
9»:
25
09000
new
5253
325.2
$332
$35
<0

 

moEmtg oomom0§_ 0:085 meets—9:00 SJ. «5:.

mo—n—fimha> oflaOm—ufifi_ Macaw mflcmua—Ohhémv n< XHA—ZHA—g

168

  

 

3 1293 03

tiltmilllulluultllfillltjllllll