mamwwfim
L.» 9,... :.
Wan»...

‘-
.1.
.11sz

. n 3.
1314 It} 9.;
4.5....
.. .. ‘
t . . .
Jaw: .u. i
a

 

3...».mm,

:an a

: :7
an. .52
.r P
’35....5':
‘ :2:

£311.04
.3. A \a
I1.

11 l
.- Aft-4:.nn
1 1 $1.]

2. 1‘4 It}; I : . y. “.31 ._”I v » a: . .....v. :1 A0 ‘. . p- . ...~ Kaljfld.) -»‘=\. , . , . .55.“
q. aaéwiié , . ‘ A vanafifiéwgn 1,, _ n , . , ;

.'II>

 

  

“77ft:

/ 3517' O5§3

This is to certify that the
dissertation entitled

VARIABILITY IN FISH GROWTH RATES
IN RELATION TO PHYSICAL AND LANDSCAPE
CHARACTERISTICS OF MICHIGAN AND

WISCONSIN INLAND LAKES

presented by
NANCY ANN NATE

has been accepted towards fulfillment
of the requirements for the

Ph. degree in Fisheries and Wildlife

MAN/£7 gnu/mam

0 Major Professor” s Siqplature
I2 / I 5/0 4

Date

MSU is an Affirmative Action/Equal Opportunity Institution

 

 

LIBRARIES
MICHIGAN STATE UNIVERSITY
EAST LANSING, MICH 48824-1048

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

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

2/05 C'Elfiwateouejndd-ms

 

VARIABILITY IN FISH GROWTH RATES
IN RELATION To PHYSICAL AND LANDSCAPE CHARACTERISTICS OF
MICHIGAN AND WISCONSIN INLAND LAKES
By

Nancy Ann Nate

A DISSERTATION

Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of

DOCTOR OF PHILOSOPHY
Department of Fisheries and Wildlife

2004

ABSTRACT
VARIABILITY IN FISH GROWTH RATES
IN RELATION TO PHYSICAL AND LANDSCAPE CHARACTERISTICS OF
MICHIGAN AND WISCONSIN INLAND LAKES
By
Nancy Ann Nate
The goal of this research was to explore the use of historic fisheries survey
information to explain variability in fish growth rates in relation to physical and
landscape characteristics of Michigan and Wisconsin inland lakes. First, estimates of fish
mean length at age, and of von Bertalanffy growth parameters estimated from mean
length at age, can be biased if based only on age information collected from length-
stratified sub-samples. To evaluate the potential for using historic length-stratified sub-
samples to compare the average growth of fishes across Michigan lakes, I quantified the
sub-sample bias in estimates of mean lengths at ages 2—4 and von Bertalanffy growth
parameters for bluegill, largemouth bass, and yellow perch. In regression analysis of sub-
sample based estimates versus sample based estimates, only 4 of 15 possible tests
indicated significant bias in sub-samples relative to samples, however the magnitude of
the variability in bias around von Bertalanffy estimates across lakes was large for all
species tested. Therefore, historic growth data sets that lack complete sampling
information (e. g., sample length frequencies) can still provide meaningfitl comparisons of
growth among lakes or over time if comparisons are limited to mean lengths at ages.
Next, I evaluated a lake classification scheme consisting of 6 lake classes
differing in landscape position, connectivity, and maximum depth for use in managing

angling fisheries in inland lakes. I used analysis of variance to determine if lake classes

accounted for variability in fish species richness in 256 Wisconsin lakes and in mean
lengths at age 4 of bluegill, largemouth bass, walleye, and yellow perch in 358 Michigan
and 410 Wisconsin lakes. Fish species richness and mean lengths of age 4 bluegill,
largemouth bass, and walleye varied among lake classes, but not all lake class pairs
differed. Map-based variables related to temperature and habitat area were then
evaluated using regression tree analyses to refine the classification for each response
variable. Lake surface area was an important predictor of fish species richness, and mean
lengths of age 4 bluegill and largemouth bass, whereas latitude was an important
predictor of mean lengths of age 4 yellow perch and walleye. Regression tree results
revealed species-specific separation points for identifying categories of lake size (large
and small lakes), and for stratifying the region into northern and southern zones.

Finally, to evaluate the lake classification scheme developed in Chapter 2 for use
in management, I determined expected ranges of bluegill grth and mortality for
different lake classes in Wisconsin from existing survey data to determine the best
combination of minimum length limits (152, 178, 203, and 229 mm) and daily creel
limits (25, 10, 5) to maximize yield per recruit and number of bluegill harvested per 100
individuals for each lake class. For each combination of minimum length limit and daily
creel limit, yield per recruit increased with lake size, from the smallest lake class to the
largest lake class, and ranged 4—30 grams per recruit among regulation scenarios. My
results provide ranges of bluegill growth, natural mortality, and fishing mortality for lake
classes that may be used to evaluate future regulation proposals and to identify candidate

lakes for experimental regulations.

DEDICATION

For my husband Michael J. Hansen.

iv

ACKNOWLEDGMENTS

I am indebted to my major advisor Mary Bremigan and the members of my
graduate committeee Mike Jones, Pat Soranno, Steve Hamilton, and Jim Breck for their
help and encouragement over the past 5.5 years. I am grateful to my Wisconsin
Department of Natural Resources supervisor Mike Talbot for supporting my degree
completion and allowing a flexible work schedule for the past 4 years. Jim Myrant, and
Roger Parsons of the Michigan Department of Natural Resources, Fisheries Division
were instrumental in providing access to Department databases, and Jim Scheider laid the
foundation for much of the ongoing research on Michigan lakes. Abby Mahan lead the
photocopying and data entry efforts for the historic mean length at age summaries. I am
indebted to Abby and the many students in the Bremigan and Soranno labs for their work
on various aspects of this project. I am also indepted to the many Department of Natural
Resources biologists and technicians in Michigan and Wisconsin for their field work over
the past 60 years, and for the data used in these analyses. Collection of the data used was
supported in part by funds from the United States Federal Aid in Sport Fish Restoration
Act. My participation in this project was supported in part by a Michigan Department of
Natural Resources, Fisheries Division Research Grant, and the Wisconsin Department of
Natural Resources, Bureau of Fisheries Management and Habitat Protection. Doug
Beard was an important early mentor in fisheries. Finally, completion of my dissertation
would not have been possible without the love and support of my husband, Mike Hansen.

Words aren’t enough to express my gratitude.

TABLE OF CONTENTS

LIST OF TABLES ........................................................................................................... viii
LIST OF FIGURES ........................................................................................................... xi
INTRODUCTION ............................................................................................................... 1
CHAPTER 1
COMPARISON OF GROWTH PARAMETERS FROM LENGTH STRATIFIED
SUB-SAMPLES AND SAMPLES IN MICHIGAN LAKES ............................................. 5
Introduction .............................................................................................................. 6
Methods .................................................................................................................... 7
Lake Selection Criteria ................................................................................ 7
Statistical Analysis ....................................................................................... 8
Results .................................................................................................................... 10
Discussion .............................................................................................................. 11
CHAPTER 2
A LAKE CLASSIFICATION SCHEME FOR FISH SPECIES RICHNESS AND FISH
GROWTH IN MICHIGAN AND WISCONSIN LAKES ................................................. 23
Introduction ............................................................................................................ 24
Methods .................................................................................................................. 29
Study Area ................................................................................................. 29
Lake Morphometry and Landscape Features ............................................. 30
Fish Species Richness ................................................................................ 30
Fish Growth ............................................................................................... 31
Statistical Analysis ..................................................................................... 32
Results .................................................................................................................... 34
Fish Species Richness ................................................................................ 34
Fish Growth ............................................................................................... 35
New Lake Classes ...................................................................................... 38
Discussion .............................................................................................................. 38
Fish Species Richness ................................................................................ 38
Fish Growth ............................................................................................... 41
New Lake Classes ...................................................................................... 42
Management Implications ...................................................................................... 46
CHAPTER 3
EVALUATING AN GLING REGULATIONS FOR BLUEGILL IN WISCONSIN ........ 63
Introduction ............................................................................................................ 64
Methods .................................................................................................................. 67
Study Area and Data Sources .................................................................... 67
Growth ....................................................................................................... 69

vi

Mortality .................................................................................................... 70

Exploitation ................................................................................................ 7 1

Yield per Recruit and Number Harvested .................................................. 72

Results .................................................................................................................... 74
Growth ....................................................................................................... 74

Mortality .................................................................................................... 74

Exploitation ................................................................................................ 75

Yield per Recruit and Number Harvested .................................................. 75

Discussion .............................................................................................................. 78
Growth ....................................................................................................... 78

Mortality .................................................................................................... 79

Exploitation ................................................................................................ 80

Yield per Recruit and Number Harvested .................................................. 80

Management Implications ...................................................................................... 83
SUMMARY AND CONCLUSIONS ................................................................................ 95
LIST OF REFERENCES ................................................................................................... 98

vii

LIST OF TABLES

Table 1. Regression parameters for bluegill, largemouth bass, and yellow perch mean
lengths at ages 2—4 from sub-samples (dependent) and samples (independent) in
Michigan lakes during 1993—2002. N represents the number of surveys.

Lengths were measured in inches .......................................................................... 17

Table 2. Regression parameters for bluegill, largemouth bass, and yellow perch von

Bertalanffy growth parameters Loo (in) and a) (in/year) from sub-samples
(dependent) and samples (independent) in Michigan lakes during 1993—2002. N
represents the number of surveys. .......................................................................... 18

Table 3. Relative sub-sample bias for bluegill, largemouth bass, and yellow perch mean

length at ages 2—4 (in) and von Bertalanffy growth parameters Loo (in) and a)
(in/year) from sub-samples and samples in Michigan lakes during 1993—2002 ...19

Table 4. Wisconsin lake classification scheme based on maximum depth (deep: 2 5.5 m,
shallow <55 m), landscape position (high: watershed area < 1036 ha, low:
watershed area 2 1036 ha), and connectivity (drainage: inlet or outlet present,
seepage: no inlet or outlet present). ....................................................................... 48

Table 5. Summary of lake morphology and latitude used in regression tree analyses to
predict fish species richness in 256 Wisconsin lakes during 1998-2004 and mean
length at age 4 for bluegill, largemouth bass, walleye, and yellow perch in 358
Michigan and 410 Wisconsin lakes during 1947—2003 ......................................... 49

Table 6. Analysis of variance of species richness and mean length at age 4 for bluegill,
largemouth bass, walleye, and yellow perch among six classes of Michigan and
Wisconsin lakes during 1947—2003 ....................................................................... 50

Table 7. Bonferonni matrix of pairwise comparison probabilities for species richness
and mean length at age 4 for bluegill, largemouth bass, walleye, and yellow perch
among the six classes of Michigan and Wisconsin lakes during 1947—2003. Bold
indicates significant difference. ............................................................................. 51

Table 8. Regression tree results using latitude (decimal degrees), surface area (ha) and
maximum depth (m) to predict fish species richness and mean length at age 4 for
bluegill, largemouth bass, walleye, and yellow perch for 358 Michigan and 410
Wisconsin lakes during 1947—2003. PRE = proportional reduction in error,
Values = back-transformed hectares or meters, and latitude in decimal degrees,
Means = the mean value (number of species or length at age 4) for each split
(group), N = the number of lakes in each group, and New Class = the new
response-specific class assignment based on regression tree results ..................... 52

Table 9. Regression tree results using latitude (decimal degrees), Loge surface area (ha)
and Loge maximum depth (m) to predict fish species richness in 256 Wisconsin
lakes and fish mean length at age 4 for Michigan and Wisconsin lakes during

viii

1947—2003. PRE = proportional reduction in error, Values = back-transformed
hectares or meters, and latitude in decimal degrees, Means = the mean value
(number of species or length at age 4) for each split (group), N = the number of
lakes in each group, and New Class = the new response-specific class assignment
based on regression tree results .............................................................................. 53

Table 10. Analysis of variance of species richness and fish mean length at age 4 among
new lake classes developed from regression tree analyses for Michigan and
Wisconsin lakes during 1947—2003. ...................................................................... 55

Table 11. Bonferonni matrix of pairwise comparison probabilities for species richness
and fish mean length at age 4 among new classes developed from regression tree
analyses for Michigan and Wisconsin lakes during 1947—2003. Bold indicates
significant difference ............................................................................................ 56

Table 12. Lake classification criteria for predicting bluegill mean length at age 4
(Chapter 2). ............................................................................................................ 85

Table 13. Minimum, maximum, and average Von Bertalanffy growth parameters
estimated for bluegill in four classes of Wisconsin inland lakes surveyed during

1948—2003, where N = number of lake surveys, L00 = the asymptotic average

length (m), K = the instantaneous rate at which L, approaches Loo (1/year), to =
the hypothetical age at which length is zero (years), and average weight—length
coefficients estimated for bluegill in four classes of Wisconsin inland lakes

surveyed during 1944—2002, where N = number of lake surveys, Loge (a) = a
scaling constant, and ,6 = the allometric shape parameter. W00 = the asymptotic

average weight (grams) was estimated from average-lake class Loo, at and ,6 so
there were no minumum and maximum values. .................................................... 86

Table 14. Average mortality estimates by lake class from 34 Wisconsin lakes during
1954—2003. The table includes the average, minimum and maximum
instantaneous total mortality (Z) from catch curves of the age frequency, survival
(.5) estimated from e’Z, total annual mortality (A) estimated from 1 — S, natural
mortality (M) estimated from Pauly’s equation (Pauly 1980), instantaneous
fishing mortality (F) estimated from Z — M, and exploitation (u) estimated from
F *A/Z ..................................................................................................................... 87

Table 15. Estimated harvest under 25, 10, and 5 daily creel limits estimated from 8,447
creel survey interviews where panfish species were harvested from 106 northern
Wisconsin lakes during 1998—2002. Because angler interviews were divided by
the number in the angling party to arrive at number per angler, there were
instances where between 0—1 fish were harvested (i.e., Number of Fish Harvested
= 0). ........................................................................................................................ 88

Table 16. Yield per recruit (g) and number harvested per 100 bluegill under varying
minimum length and daily creel limits for Wisconsin lakes. Maximum yield per

ix

recruit and number harvested per 100 for the minimum length limits and daily
creel limits evaluated for each lake class are indicated in bold. The 127 mm
length limit represents the yield per recruit and number harvested in the absence
of a minimum length limit. Age is the estimated time (years) to reach each
minimum length limit ............................................................................................ 90

LIST OF FIGURES

Figure 1. Sub-sample and sample mean length (inches) at ages 2 (top), 3 (middle) and 4
(bottom) for bluegill (BLG, left panel), largemouth bass (LMB, middle panel) and
yellow perch (YEP, right panel) in Michigan lakes during 1993—2002. .............. 20

Figure 2. Sub—sample and sample von Bertalanffy parameter estimates a) (in/year) [left

side] and 1.00 (in) [right side] for bluegill (BLG top), largemouth bass (LMB
middle) and yellow perch (YEP bottom) in Michigan lakes during 1993-2002.. 22

Figure 3. Locations of 256 Wisconsin lakes used to examine variability in species
richness among lake classes (asterisks) and 358 Michigan and 410 Wisconsin
lakes used to examine variability in bluegill, largemouth bass, walleye and yellow
perch mean length at age 4 among Wisconsin lake classes (solid diamonds). ..... 57

Figure 4. The mean number of species (j 95% confidence interval) by Wisconsin lake
class in summer mini-fyke net and fall electrofishing surveys on 256 Wisconsin
lakes during 1998—2004. X axis labels include D = drainage, S = seepage. H =
high landscape position, L = low landscape position, SH = shallow, and DE =
Deep. ..................................................................................................................... 58

Figure 5. Regression tree mobile diagram example predicting the number of species by
Logc Surface Area, Loge Maximum Depth, and Latitude for 256 Wisconsin lakes
during 1998-2004. Results from regression tress analyses are presented in tabular
format rather than mobile diagrams. ..................................................................... 59

Figure 6. Bluegill (upper left panel), largemouth bass (upper right panel), walleye (lower
left panel), and yellow perch (lower right panel) mean length at age 4 (mm; :
95% confidence interval) for six Wisconsin lake classes. .................................... 60

Figure 7. The mean number of species (: 95% confidence interval) by new lake class in
summer mini-fyke net and fall electrofishing surveys on 256 Wisconsin lakes
during 1998—2004. ................................................................................................ 61

Figure 8. Bluegill (upper left panel), largemouth bass (upper right panel), walleye
(lower left panel), and yellow perch (lower right panel) mean length at age 4
(mm; : 95% confidence interval) for new lake classes. ....................................... 62

Figure 9. Locations of surveys used in this study. Different symbols represent each type
of estimate (mortality estimates (triangles), creel surveys (circles), weight-length
estimates (asterisks), and von Bertalanffy growth parameter estimates from mean
lengths at age (plus). ............................................................................................. 91

Figure 10. Figure 10. The frequency of anglers harvesting between 1 and 25 panfish in
Wisconsin lakes during 1998—2003 ...................................................................... 92

xi

Figure 11. Yield per recruit (g) by lake class under 5 minimum length limit scenarios
(127 mm (heavy —), 152 mm (-——), 178 mm (— — —), 203 mm (- - -), and 229 mm

(— - — ). Vertical solid line indicates average exploitation rate (It) for the lake
class. Lake class G upper left), lake class H (upper right), lake class 1 (lower left),
and lake class J (lower right) ................................................................................. 93

Figure 12. Number harvested per 100 by lake class under 4 minimum length limit
scenarios (127 mm (heavy —),152 mm (—), 178 mm (— — -—), 203 mm (- - -). and

229 mm (— - — ). Vertical solid line indicates average exploitation rate (u) for the
lake class. Lake class G (upper left), lake class H (upper right), lake class I
(lower left), and lake class I (lower right). ........................................................... 94

xii

INTRODUCTION

Individual lakes, within and across broad geographic areas, differ greatly in
inherent internal physical, chemical, and biological attributes, local and regional
landscape settings, and fish community assemblages. Within individual lakes, fish
population levels may fluctuate over time due to biotic and abiotic factors, such as food
availability, predation, competition, lake level, drought, and storm events (Van Den
Avyle 1993). The complex forces that continually shape fish populations operate at
different spatial and temporal scales, which can add to management uncertainty and risk

when setting fisheries management policy for an entire state.

Often fisheries management policies relating to angling regulations are
established for large geographic areas (e. g., states or regions within states), so factors that
affect fish populations across large geographic areas need to be understood. However,
most investigations of fish populations and community dynamics in inland lakes tend to
be small scale, either following populations in one lake through time (e. g. Fomey 1977;
Hansen et al. 1998), or multiple populations in several lakes in a small geographic area
(e. g. Tonn and Magnuson 1982, Carpenter and Kitchell 1993). These studies provide
insight into the temporal dynamics and current status and trends of fish populations in
close geographic proximity, but may not sufficiently explain spatial variability among
lakes, status and trends of a statewide resource, or the impact of a statewide fisheries
management strategy. Many ecosystem patterns and processes only emerge at larger
scales (Levin 1992). Further, variability of fish population growth, recruitment, or
mortality may be erroneously attributed to spatial variability among lakes if considered

separately and independently from temporal changes (Wiley et al. 1997). Therefore,

OOstudies that are large both in spatial and temporal scale are needed to accurately
partition fish population variability into its respective components. Through this
integration, underlying processes that affect survival, growth, and recruitment may be

more easily and accurately identified.

Large-scale research in time and space is not common because of the time and
expense involved in data collection. However, Michigan Department of Natural
Resources Fisheries Division and Wisconsin Department of Natural Resources, Bureau of
Fisheries Management and Habitat Protection personnel have conducted fisheries surveys
on lakes throughout the states of Michigan and Wisconsin. Historic survey data collected
as part of routine surveys across two states were not collected as part of a standardized or
organized sampling program, but due to the large spatial and temporal coverage, may still
provide insights into patterns and processes that operate at large spatial and temporal
scales. With some careful investigation of potential biases, historic fisheries survey data
may be used to classify lakes to account for lake-to-lake variability in fish populations,
explore long-term changes, characterize the scope of inter-annual variability, and
ultimately generate statewide management strategies. Ideally, the classification scheme
could then be used to predict the productive potential and sustainability of fish

populations in lakes for which recent survey information is lacking.

Many lake classification schemes have been developed to explain patterns in fish
community assemblages and species abundance by exploring relationships with measured
within-lake physical, chemical, or biological characteristics at the local scale (Schneider
1975; Johnson et al. 1977; Tom et al. 1983; Matuszek and Beggs 1988; Dolman 1990;

Schupp 1992). In fact, a large portion of variability in species assemblages and

abundance among lakes has been explained through classification based on measured
lake habitat characteristics. Part of the variability in species assemblages or abundance
left unexplained in these classification schemes may be attributed to temporal changes in
fish population dynamics or to processes operating at larger spatial scales (but not
included in the classification). These confounding factors may cloud our ability to see
clear relationships between fish population characteristics and habitat variables.
Therefore, by including temporal changes in population dynamics within the context of a
classification scheme that accounts for spatial variability among lakes at several spatial
scales, factors that affect fish population dynamics may be more accurately identified.
Further, classification schemes that account for variability in ecological processes like
growth or life history strategies, which indirectly may influence abundance, are less
prevalent but may provide additional framework for making management decisions or for
research into causal mechanisms (Shuter et al. 1998). Ecologically similar lakes are

expected to respond similarly to management actions (Emmons et al.1999).

The goal of this research was to explore the use of historic fisheries survey
information to explain variability in fish growth rates in relation to physical and
landscape characteristics of Michigan and Wisconsin inland lakes. To do so, I focused on
three main objectives 1) to determine if von Bertalanffy growth parameters and mean
lengths at age estimated from length-stratified sub-samples were biased in relation to
parameters estimated from full samples for game fishes surveyed routinely in Michigan
lakes, 2) to determine if an existing lake classification scheme explained variability in
fish species richness and mean length at age from historic survey data collected on

Michigan and Wisconsin lakes, and 3) to demonstrate how lake classification may be

used in management by estimating how bluegill yield per recruit and number harvested
varied as a function of minimum age entering the fishery and fishing mortality rate

among four classes of Wisconsin inland lakes.

CHAPTER 1
COMPARISON OF GROWTH PARAMETERS FROM LENGTH STRATIFIED

SUB—SAMPLES AND SAMPLES IN MICHIGAN LAKES

Abstract. —— Estimates of mean length at age, and of von Bertalanffy growth
parameters estimated from mean length at age, can be biased if based only on age
information collected from length-stratified sub-samples. To evaluate the potential for
using historic length-stratified sub-samples to compare the average growth of fishes
across Michigan lakes, I quantified the sub-sample bias in estimates of mean lengths at
ages 2—4 and von Bertalanffy growth parameters for bluegill, largemouth bass, and
yellow perch. In regression analysis of sub-sample based estimates versus sample based
estimates, only 4 of 15 possible tests indicated significant bias in sub-samples relative to
samples. Von Bertalanffy growth parameters indicated that yellow perch early growth

(0)) tended to be higher when estimated from sub-samples than from samples, but

indicated no bias for bluegill L00 and for largemouth bass to and Lao. While regression

tests showed that sub-sample estimates were generally linearly related to estimates from
samples, the magnitude of the variability in bias around von Bertalanffy estimates across
lakes was large for all species tested. My findings suggest that mean lengths at ages 2—4
for bluegill, largemouth bass, and yellow perch estimated from length-stratified sub-
samples were not generally biased, but that the large variation in von Bertlanffy
parameter estimates limits their utility for use in among-lake comparisons. Therefore,
historic growth data sets that lack complete sampling information (e.g., sample length
frequencies) can still provide meaningful comparisons of growth among lakes or over

time if comparisons are limited to mean lengths at ages 2—4.

Introduction

Fish growth affects many aspects of fish population dynamics. Body size
influences age at maturity (Trippel 1995), fecundity (Baccante and Reid 1988),
recruitment (Madenjian et al. 1996), and mortality (Madenjian and Carpenter 1991).
Growth rates during early life stages can directly affect size-dependent mortality by
determining the timing and duration of vulnerability to predators (Osenberg and
Mittelbach 1988; Olson 1996), and size at the end of the first growing season can
influence over-winter survival in some species (Madenjian and Carpenter 1991; Ludsin

and DeVries 1997).

Fish growth can be an important indicator of environmental conditions within
lakes (Diana 1995), so growth may be useful for classifying variability in fish populations
among lakes for the purpose of management. Lakes that are ecologically similar should
support populations with similar growth rates, and therefore respond similarly to
management actions such as angler regulations. The success of many angler regulations
that seek to increase the size structure or numbers of large fish in lakes depends largely
on the growth potential of the population. For example, for minimum size limits to be
effective, growth potential should be sufficiently high for fish to reach the minimum size
vulnerable to harvest (Van Den Avyle 1993). If growth potential is low and abundance
of small fish is high, preventing juvenile harvest with minimum size limits may

exacerbate poor size structure.

Fish growth is often estimated from mean length at age and fisheries biologists
often sub-sample a fixed number of fish per length category for estimating mean length at

age for a population (Murphy and Willis 1996). Sub-sampling a fixed number of fish per

length category and using the sub-sample to estimate the mean length and variance at age
may result in biased estimates of mean length, because the distribution of lengths around
any one age is often skewed (Goodyear 1995; Bettoli and Miranda 2001). Biased
estimates of mean length at age will inflate standard error estimates, thereby leading to an
increase in type II errors (failure to detect an effect; Bettoli and Miranda 2001). Bias is
more likely for ages that are not well represented in the sample, often the youngest or
oldest ages (Haddon 2001). Further, growth parameter estimates from the von
Bertalanffy growth equation are greatly influenced by the youngest and oldest ages in the
sample, which tend to be the least well represented (Ricker 1975; Haddon 2001). The
use of length-stratified sub-sample mean lengths at age to estimate von Bertalanffy

growth parameters for comparative analyses among lakes should be evaluated.

My objective was to determine if mean lengths at age and von Bertalanffy growth
parameters estimated from length-stratified sub-samples were biased in relation to
parameters estimated from full samples for game fishes surveyed routinely in Michigan
lakes. I quantified the bias between estimates of mean lengths at age and von Bertalanffy
growth parameters from length-stratified sub-samples of fish and sample mean lengths at
age reconstructed using an age—length key. My goal was to evaluate the potential for
using historic mean length at age summaries from length—stratified sub-samples to

characterize average growth of fishes and to compare growth parameters among lakes.
Methods
Lake Selection Criteria

Fisheries survey data were obtained from the Michigan Department of Natural

Resources, Fish Collection System (FCS) in June 2003. Fishery surveys were conducted

on Michigan lakes between 1993 and 2002. Fishery surveys were included if (1) the data
entry and results were approved through an established quality assurance protocol, (2) the
full sample length frequency included 2 100 fish, (3) at least five age classes were present
in the sub-sample, and (4) the sub-sampling was length-stratified (i.e. a specified number
of fish per length class were sub-sampled for age estimation). Surveys from 126 lakes
satisfied these criteria, including 111 surveys for bluegill Lepomis macrochirus, 22
surveys for largemouth bass Micropterus salmor’des, and 24 surveys for yellow perch

Percaflavescens.
Statistical Analysis

To quantify bias in estimates of mean lengths at age 2, 3, and 4, and von
Bertalanffy growth parameters from length-stratified sub-samples and samples, mean
length at age was estimated directly from sub-samples of fish, and from samples of fish
using age-length keys (Ricker 1975; Haddon 2001). Age length keys were constructed
from length-stratified sub-sample length and age information and were used to
reconstruct the age distribution from the length frequency of each sample (Ricker 1975).
Sample mean lengths at age were then calculated from the new age frequency
distribution. Mean length at age for the sample was taken as the sum of products
between mid-point lengths of each length interval and the number of that age in each
length interval, divided by the total number of fish in the age class. Mean length at age
from the sub-sample was taken as the arithmetic mean of the lengths at each age. Ages
2—4 were selected because these ages were likely to be estimated correctly for all species
and to make explicit comparisons to results from Betolli and Miranda (2001). Length-

stratified sub-samples were collected during the same survey, thereby avoiding potential

bias in using an age-length key to estimate age-frequency when sub-sample information

was not collected during the same period (Westrheim and Ricker 1978).
The Gallucci and Quinn (1979) modification of the von Bertalanffy growth
equation was used to estimate early growth rate (co) and the mean asymptotic length (Loo)

for each species in each lake from mean lengths at age for sub-samples and samples:

L =L 1_e—[w/Lw][t—t0]

I up

The modified von Bertalanffy model describes how length (L) at age (t) increases from

the hypothetical age at zero length (to) toward the average asymptotic length (L 00) at rate

a). The modified von Bertalanffy model reduces the correlation between the traditional
von Bertalanffy parameters K and Lao, provides a rate of growth in tangible units (e. g., a)

= inches per year), and provides a superior fit to grth data (Chen et al. 1992).
Parameters and their asymptotic standard errors were estimated using nonlinear
regression, with additive errors. Parameter estimates were excluded if the estimated

asymptotic length exceeded the state record by 10%.
I tested for differences in growth indices (mean length at ages 2, 3, and 4, and von

Bertalanffy growth parameters a) and Lao) between sub-samples and full samples using

geometric mean functional regression (GM regression; Sokal and Rohlf 1995).
Geometric mean functional regression is appropriate when the independent variable is

subject to natural variability (Ricker 1975). The t statistic was adjusted to test whether

the slope differed significantly from one by taking (1 — slope)/SE. Bias was inferred if
the slope of the regression was significantly greater than or less than one or if the

intercept was significantly less than or greater than zero (P 5 0.05).

Because I was interested in using sub-sample growth indices for among-lake
comparisons, I quantified the magnitude of the bias in growth indices from sub-samples
relative to the actual variation in unbiased growth indices from samples, across all
surveys as the “relative sub-sample bias”:

Z<Gsub - G full )2

surveys

2(0 full -5fuu)2

surveys

 

In this equation, Gsub is the growth index from the sub-sample (mean length at ages 2, 3,
and 4, and von Bertalanffy growth parameters a) and Lao), Gfu” is the growth index from

the sample, and G full is the mean of the growth index for samples. If the relative sub-

sample bias was greater than 10%, we considered the growth index to be unreliable for

among-lake comparisons.
Results

Regression tests indicated little bias in mean lengths at ages 2—4 estimated from
sub-samples. Mean lengths at age estimated from sub-samples and samples did not differ
for bluegill of ages 3—4, largemouth bass of ages 2—4, or yellow perch of ages 2—4 (slope
= l and intercept = 0; Table 1 Figure 1). In contrast, bluegill mean length at age 2 was

overestimated for small fish from sub-samples and underestimated for larger fish from

10

sub-samples (slope < l and intercept > 0; Table 1; Figure 1). Absolute differences
between mean lengths at age from sub-samples and samples ranged from 0.0 to 0.9 in for

bluegill, from 0.0 to 0.7 in for largemouth bass, and from 0.0 to 1.0 in for yellow perch.
Similarly, regression tests indicated little bias in von Bertalanffy growth

parameters estimated from sub-samples. For bluegill, L00 did not differ between sub-

samples and samples, and for largemouth bass, to and Lao did not differ significantly

between sub-samples and samples (Figure 2). In contrast, a) for bluegill and Lao for

yellow perch were lower when estimated from sub-samples than when estimated from
samples, and the absolute bias increased with the magnitude of the sample parameter
estimate from a y-intercept that was not different from zero (Table 2; Figure 2). In
addition, (I) for yellow perch was consistently over-estimated from sub-samples, because

the y-intercept was greater than zero and the slope was not significantly different fi'om
1.0 (i.e., absolute bias was constant and positive; Table 2). Absolute differences in Lao

from sub-samples and samples ranged from 0.003 to 3.755 in for bluegill, from 0.057 to

6.751 in for largemouth bass, and from 0.012 to 4.748 in for yellow perch.

While regression tests indicated little overall bias in growth indices from sub-
samples (i.e., significant linear relationships), relative sub-sample bias was large for some
growth indices. Relative sub-sample bias in mean lengths at age were less than 10% for
all species and ages except for yellow perch mean length at age 3 (Table 3). In general,
relative sub-sample bias in mean length at age was small and ranged from 3.1% to 14.1%

(Table 3). In contrast, relative sub-sample bias in von Bertalanffy growth parameters was

11

large and ranged from 16.9% to 51.1% (Table 3). Estimates of Lao were more variable
than (I) for all species (Table 3).
Discussion

I found that mean lengths at age and von Bertalanffy growth parameters estimated
from length-stratified sub-samples were linearly related to parameters estimated from full
samples for bluegill, largemouth bass, and yellow perch. Only 4 of 15 possible tests
indicated significant bias in sub-samples relative to samples. In general, when evident,
bias was larger for von Bertalanffy estimates than for mean length at age estimates. For
largemouth bass, mean length at age and von Bertalanffy growth parameters estimated
from sub-samples were not significantly biased, whereas for bluegill and yellow perch,
some mean length at age and von Bertalanffy growth parameters estimated from sub-
samples were biased. For example, bluegill mean length at age-2 tended to be

overestimated from sub-samples for small age-2 bluegill, and underestimated from sub-

samples for larger age-2 bluegill. Bluegill (D and yellow perch Loo were underestimated

from sub-samples (regression slope < 1), whereas yellow perch to was consistently

overestimated from sub-samples.

I found that mean lengths at age of bluegill, largemouth bass, and yellow perch
from sub-samples were generally no different than mean lengths at age from samples
(highly correlated 1:1 relationships), in contrast to studies by Betolli and Miranda (2001)
and Goodyear (1995) who found that mean lengths at age from length-stratified sub-
samples were biased in relation to random samples. Bettoli and Miranda (2001) showed

that mean size at age 2 for a single largemouth bass population was less when estimated

12

from a length-stratified sub-sample than when estimated from the full sample, whereas
the mean size at age 4 was greater when estimated from the sub-sample than when
estimated from the full sample. Similarly, Goodyear (1995) used computer simulations
to evaluate the effect of sub-sampling on estimates of mean length at age of red grouper
Epinephelus morio, and found that mean length of young grouper was underestimated
and mean length of older grouper was overestimated by sub-samples. Bettoli and
Miranda (2001) cautioned against using mean lengths at age from non-random stratified
sub-samples to estimate mean length at age for a population because of an increased risk
of type II error or failure to detect an effect. My approach differed in that I evaluated the
bias between sub-samples and samples across numerous populations using existing
surveys (not simulations). Although individual survey differences between sub-sample
and sample estimates were occasionally large, I found few consistent biases across

surveys, so comparisons among lakes may provide meaningful relative comparisons.

I found that von Bertalanffy growth parameters estimated from sub-samples for
bluegill, largemouth bass, and yellow perch were linearly related to estimates from
samples, though highly variable. The similarities between sub-sample and sample von
Bertalanffy grth parameters estimated from mean length at age data were unexpected
because length-stratified sub-sampling of ages may lead to biased estimates of mean
length at age (Goodyear 1995; Betolli and Miranda 2001). Since the age-length key was
first introduced by Fredriksson (1934), the usefulness of non-random and random sub-
sampling has been debated without resolution (Ketchen 1949; Kimura 1977; Goodyear
1995; Betolli and Miranda 2001). Random sub-sampling may provide similar estimates

of mean lengths at age to samples, whereas non-random sub-sampling may be

13

problematic in this regard (Kimura 1977; Goodyear 1995; Betolli and Miranda 2001).
However, the extent of the problem has not been thoroughly evaluated. Non-random or
length-stratified sub-sampling is a common practice and ensures that relatively rare size
classes are represented. In my analyses of mean length at age estimated from length-
stratified sub-samples, I excluded rare age classes that were most likely to be biased by
limiting the analyses to ages 2—4. Mean lengths at other ages from length-stratified sub-
samples may be more biased; therefore von Bertalanffy growth parameters estimated
from biased mean lengths at age may also be biased. The effect of including mean
lengths at age from all age classes in my estimation of von Bertalanffy grth parameters
from length-stratified sub-samples and samples may have contributed to high variability

in parameter estimates.

I did not consistently detect bias when comparing a large number of von
Bertalanffy growth parameter estimates from sub-samples and samples from multiple
lakes, though the variability in growth parameter estimates was large enough to warrant
caution. While my results differ from those of other studies, I do not suggest that bias
caused by non-random sub-sampling does not exist. Rather, the application of growth
information should dictate whether length-stratified sub-sample data should be used.
Here, I was interested in using length-stratified sub-sample data for statewide
comparisons of growth potential across lakes and assessing the degree to which relative
comparisons were possible. Because of the high variability in of von Bertalanffy growth
parameter estimates from sub-samples and samples, the use of these estimates for among-

lake comparisons is problematic.

l4

Estimates of mean length at age and associated growth parameters can also be
biased by non-random samples of the population, age estimation error, gear selectivity,
and size-selective mortality. Age estimation error results in biased estimates of growth
and mortality (Campana 2001). Young and especially old ages are usually the most
difficult to estimate accurately, and are also the most influential on estimates of growth,
mortality, or longevity (Campana 2001). If the age of older fish is underestimated,
mortality will be overestimated and the importance of strong year-classes can be masked
(Beamish and McFarlane 1983). For example, age estimation error resulted in
overestimation of abundance of younger ages, overestimation of mean weight at age, and
up to 19% overestimation of yield per recruit (Tyler et al. 1989). Gear selectivity also
biases estimates of growth and mortality (Ricker 1975). If large fishes are more
vulnerable to capture, older and larger fish will be over-represented, thereby resulting in
over-estimation of survival and growth (Ricker 1975). Size-selective mortality biases
estimates of growth, mortality, and yield, because larger individuals in each age class

often suffer higher mortality than smaller individuals in each age class (Ricker 1969).

Understanding growth potential is important for fishery management, but
collecting fish growth information is expensive and time consuming (Murphy and Willis
1996). In many states, growth data have not been readily accessible for large numbers of
lakes across large geographic scales, thereby limiting the potential for use in lake
classification. Recently, management agencies have begun to compile historic fisheries
survey information in large databases, thereby facilitating access to growth information
such as mean length at age (Hayes et al. 2003). Use of historic growth information for

purposes other than the intent of the original data collection has not been thoroughly

15

tested. Often, historic records are limited to summary information such as mean length at
age, rather than lengths and ages of all individual fish that are needed for constructing
age-length keys. Differences in sampling methods and other sources of potential bias
should be evaluated before historic data of this type are used. If shown to be unbiased,
historic grth data such as mean length at age summaries may be useful for developing
large—scale lake classification schemes that could ultimately be used as a basis for
developing fishery regulations. For Michigan lakes, I found that relative comparisons

among lakes using mean length at ages 2—4 estimated from sub-sample data are robust.

16

Table 1. Regression parameters for bluegill, largemouth bass, and yellow perch mean

lengths at ages 2—4 from sub-samples (dependent) and samples (independent) in

Michigan lakes during 1993—2002. N represents the number of surveys. Lengths were

measured in inches.

 

 

Species and age Coefficient SE t P
Bluegill Age 2 (N = 78; R = 0.971)

Intercept 0.245 0.088 2.786 0.007
Mean Length Sample 0.937 0.025 -2.493 0.015
Bluegill Age 3 (N = 104; R = 0.955)

Intercept 0.001 0.141 0.008 0.994
Mean Length Sample 0.992 0.029 -0.275 0.784
Bluegill Age 4 (N = 110; R = 0.958)

Intercept 0.157 0.158 0.991 0.324
Mean Length Sample 0.962 0.027 -1.436 0.154
Largemouth Bass Age 2 (N = 22; R = 0.986)

Intercept -0.303 0.280 -1.080 0.293
Mean Length Sample 1.037 0.039 0.942 0.357
Largemouth Bass Age 3 (N = 22; R = 0.986)

Intercept 0.231 0.343 0.673 0.509
Mean Length Sample 0.970 0.036 -0.828 0.418
Largemouth Bass Age 4 (N = 22; R = 0.974)

Intercept 0.359 0.549 0.654 0.520
Mean Length Sample 0.970 0.495 -0.060 0.953
Yellow Perch Age 2 (N = 20; R = 0.965)

Intercept 0.448 0.302 1.481 0.156
Mean Length Sample 0.923 0.057 -1 .341 0.197
Yellow Perch Age 3 (N = 22; R = 0.929)

Intercept 0.468 0.519 0.901 0.378
Mean Length Sample 0.938 0.078 -0.791 0.438
Yellow Perch Age 4 (N = 22; R = 0.971)

Intercept 0.735 0.389 1.889 0.074
Mean Length Sample 0.917 0.049 -1.696 0.105

 

l7

Table 2. Regression parameters for bluegill, largemouth bass, and yellow perch von

Bertalanffy growth parameters L00 (in) and a) (in/year) from sub-samples (dependent)
and samples (independent) in Michigan lakes during 1993—2002. N represents the
number of surveys.

 

 

Parameter Coefficient SE t P
Bluegill a) (N = 111 ; R = 0.806)

Intercept 0.248 0.141 1.757 0.082
Sample (I) 0.871 0.049 2.610 0.010
Bluegill Loo (N = 111 ; R = 0. 729)

Intercept 0.807 0.629 1.284 0.202
Sample Loo 0.930 0.061 1.153 0.251
Largemouth Bass u) (N = 22 ; R = 0.893 )

Intercept 0.379 0.385 0.984 0.337
Sample a) 0.903 0.091 1.065 0.300
Largemouth Bass Loo (N = 22 ; R = 0. 794)
Intercept 2.537 2.340 1.084 0.291
Sample Loo 0.868 0.118 1.121 0.275
Yellow Perch (n (N = 24; R = 0.933 )

Intercept 0.517 0.217 2.377 0.027
Sample a) 0.897 0.069 1.494 0.149
Yellow Perch Loo (N = 24 ; R = 0.915)

Intercept 1.843 0.974 1.893 0.072
Sample Loo 0.818 0.070 2.597 0.016

 

l8

Table 3. Relative sub-sample bias for bluegill, largemouth bass, and yellow perch mean

length at ages 2—4 (in) and von Bertalanffy growth parameters Lao (in) and a) (in/year)
from sub—samples and samples in Michigan lakes during 1993—2002.

 

Relative Sub-sample Bias

 

Species Growth Index Across Surveys (%)
Bluegill Mean Length Age 2 5.8
Mean Length Age 3 9.1
Mean Length Age 4 8.8
von Bertalanffy a) 36.1
von Bertalanffy Lao 51.1
Largemouth BassMean Length Age 2 3.3
Mean Length Age 3 3.1
Mean Length Age 4 5.3
von Bertalanffy a) 20.4
von Bertalanffy Lao 37.4
Yellow Perch Mean Length Age 2 7.4
Mean Length Age 3 14.1
Mean Length Age 4 6.5
von Bertalanffy a) 16.9
von Bertalanffy Loo 20.0

 

19

Sub-Sample

Sub-Sample

Sub—Sample

 

 

 

Full Sample

 

 

 

Full Sample

 

Full Sample

20

Sub—Sample

Sub-Sample

Sub-Sample

 

 

15-i

 

(n—

I r I '* I " I ' ’I

7 9 11 13 15

Full Sample

 

 

I'I'I'I‘
79111315

-

Full Sample

 

-

0'1

7 9 11 13 15

Full Sample

12-j

10-1

Sub-Sample

 

I ' I ' I fl I ' I ' I
2 4 6 8 10 12
Full Sample

Sub-Sample

 

 

 

I D I ' I I I ' I ' F
2 4 6 8 1O 12
Full Sample

Sub-Sample

 

 

 

é'i'éiérr'o'r'z

Full Sample
Figure l. Sub-sample and sample mean length (inches) at ages 2 (top), 3 (middle)
and 4 (bottom) for bluegill (BLG, left panel), largemouth bass (LMB, middle panel)
and yellow perch (YEP, right panel) in Michigan lakes during 1993-2002.

21

 

 

 

 

 

 

32

Q.

E

(U

m

I:

:1

U)
I ' I f I ' I I
0 2 4 6 10

Full Sample

9

Q

E

m

a)

b

3

a)
I I I I I
0 2 4 6 10

Full Sample

.9

O.

E

i

3

m

 

 

 

0 2 4 6
Full Sample

N
0|
1

N
O
llLll

l

J

Sub-Sample
6‘:

lllllllllll

 

 

10
5-
5 1O 15 20 25
Full Sample
25 —
20

Sub-Sample
at

 

lllllllJlelllAlllll

 

 

10
5
IUIUIIYVVVI'T'I'YYUVI
5 10 15 20 25
FullSample
25~
205

1111

Sub-Sample
a:
l

A
O
l

 

 

 

'YU'V'IYVVIUUYIIVI'U'

5 10 15 20 25
Full Sample

Figure 2. Sub-sample and sample von Bertalanffy parameter estimates a) (in/year) [left
side] and Lw(in) [right side] for bluegill (BLG top), largemouth bass (LMB middle) and
yellow perch (YEP bottom) in Michigan lakes during 1993—2002.

22

CHAPTER 2
A LAKE CLASSIFICATION SCHEME FOR FISH SPECIES RICHNESS AND FISH

GROWTH IN MICHIGAN AND WISCONSIN LAKES

Abstract. — Lake classification schemes can account for diversity in physical,
chemical, or biological attributes of lakes, and so may be useful in managing fisheries in
inland lakes. I evaluated a lake classification scheme consisting of 6 lake classes
differing in landscape position, connectivity, and maximum depth using analysis of
variance to determine if lake classes accounted for variability in fish species richness in
256 Wisconsin lakes and in mean lengths at age 4 of bluegill, largemouth bass, walleye,
and yellow perch in 358 Michigan and 410 Wisconsin lakes. Fish species richness and
mean lengths of age 4 bluegill, largemouth bass, and walleye varied among lake classes,
but not all lake class pairs differed. Map-based variables related to temperature and
habitat area were then evaluated using regression tree analyses to refine the classification
for each response variable. Lake surface area was an important predictor of fish species
richness, and mean lengths of age 4 bluegill and largemouth bass, whereas latitude was
an important predictor of mean lengths of age 4 yellow perch and walleye. Regression
tree results revealed species-specific separation points for identifying categories of lake
size, and for stratifying the region into northern and southern zones. The revised
classifications demonstrate the importance of incorporating regional, lake-type, and local
spatial scales for explaining variation among lakes. These classifications can serve as a
statewide framework for management, based on patterns of ecological similarity among
lakes. Angling regulations could be established for lake classes, rather than individual

lakes, thereby reducing the complexity of angling regulations.

23

Introduction

Lakes are inherently diverse in their physical, chemical, and biological
characteristics because internal and external forces continually shape lake ecosystems and
the processes that lead to predictable patterns. Lakes, particularly in the upper mid-
westem United States, are not only diverse in character and origin, but are also numerous
and extend across a broad geographic extent. For example, Wisconsin has nearly 15,000
lakes that range in surface area from 0.04 to 55,730 ha (Lillie and Mason 1983) and
Michigan has over 35,000 lakes that range in surface area from < 2.47 to 51,891 ha
(Humphrys and Colby 1962; Schneider 1975). Given the complexity and broad spatial
scale of lake resources, monitoring the status and trends of lake ecosystem health (e. g.
documenting temporal change in fish communities or water quality) or fisheries is
challenging. Monitoring ecosystem integrity and fisheries are not mutually exclusive
objectives, but are often tackled independently by different branches of management
agencies (e. g., water quality versus fisheries management). Monitoring all lakes in a
single year is impossible, even if many agencies are involved. Therefore, monitoring
programs should incorporate statistically rigorous sampling designs that stratify lakes
according to ecological similarity to obtain unbiased samples of the populations of lakes

(Hayes et al. 2003; Lester et al. 2003).

In this paper, I seek to evaluate and improve on an existing classification scheme,
developed and used in Wisconsin for use in monitoring, that is based on habitat
characteristics. I assume that ecologically similar lakes should respond similarly to
management actions or environment stressors (Marshall and Ryan 1987; Emmons et a1.

1999). By first accounting for variability in physical, chemical, or biological

24

characteristics of lakes, detection of change brought on by environmental stress or
management actions should be more easily identified because expected conditions are
known (Hawkins et al. 2000). Classifications should account for sufficient natural
variation in an indicator of interest (e.g., fish abundance, species richness, or fish growth
rates) that a comparison between observed and expected conditions will reveal subtle
changes, thereby minimizing type I and II errors (Hawkins et al. 2000). For example,
stratification based on a habitat classification model for Texas reservoirs decreased the
statewide variance in electro-fishing relative abundance estimates up to 43% for bluegill
Lepomis macrochirus and 23% for largemouth bass Micropterus salmoides (Dolman

1990).

Classification schemes that explicitly account for variability in ecological
processes such as growth or life history strategies, which may indirectly influence fish
abundance, are less prevalent, but can provide a framework for management decisions or
for research into causal mechanisms of variation in fish population demographics (e.g.,
Shuter et al. 1998). Understanding factors that affect fish growth, in particular, is
important in fisheries management because growth can respond to changes in population
size (e.g., Diana 1983; Healey 1980), so therefore may be an important indicator of
change brought on by environmental stress, environmental change, or angling pressure
(Trippel 1995). In addition, fish growth underlies, in large part, the response of fish
populations to management regulations, such as minimum size limits (Noble and Jones
1993). The role of regulations in fisheries management will increase in the future, as the

demand on fisheries increases as the number of licensed-anglers increases (Noble and

Jones 1993).

25

Fish growth is an important indicator of environmental and physiological
conditions in lakes (Diana 1995). The environment primarily regulates fish growth
through temperature and food availability (Magnuson et al. 1979; Diana 1995).
Therefore, fish grth should vary among lakes according to physical features that
dictate temperature, and the physical and environmental forces that influence nutrients,
productivity, and ultimately food availability. Factors that influence food availability,
such as predation and competition, tend to be highly variable, more difficult to measure,
less easily predicted, and therefore less desirable for use in classification. Factors that
influence temperature in lakes, such as latitude, surface area, and depth, are less variable.
By accounting for physical features that determine temperature in a classification scheme,
I seek to build a template for understanding the more dynamic biological interactions
such as predation and competition that influence food availability. To do so, I test the
ability of an existing lake classification scheme, developed for northern Wisconsin lakes,
to explain variation in fish species composition and growth among Wisconsin and
Michigan lakes, and evaluate the ability of factors that influence temperature in lakes

(latitude and surface area) to improve the explanatory power of the classification.

Ideally, variables used to develop classification schemes should be available for a
large number of lakes, capture the full range of variability in lake characteristics in the
population of lakes of interest, and enable the establishment of a classification scheme for
an entire state or region (Hawkins et al. 2000). Use of temporally stable or map-based
attributes in classification minimizes the need to visit the site to classify new lakes, and
enables the classification of a large number of lakes at one time. For example, the lake

classification scheme currently used in Wisconsin classifies lakes according to landscape

26

position, connectivity and depth. The classification distinguishes lakes that are connected
to other lakes in a drainage network from lakes that are land-locked. Connected lakes are
further separated by their general landscape position (high or low) in the drainage
network. Lakes that are deep and potentially stratified are separated from lakes that are
shallow. Each of these characteristics is known or easily estimated for all lakes in the

state from existing data sources.

The literature contains support for using landscape position, connectivity and lake
depth to group similar lakes. The basis for the landscape position component of the
Wisconsin classification scheme was based on research that suggests that a lake’s
position in the landscape constrains its physical, chemical, and biological attributes
(Kratz et al. 1997; Riera et a1. 2000; Soranno et al. 1999; Webster et al. 1996, 2000). For
example, lake surface area, ionic concentrations, and fish species richness have been
shown to vary predictably with landscape position, with lakes lower in the drainage
network typically being larger and having higher surface area, ionic concentration, and
fish species richness than lakes higher in the drainage network in northern Wisconsin
(Kratz et al. 1997; Riera et al. 2000; Webster et al. 1996, 2000). Connectivity can also
explain variation among lakes. For example, Tom and Magnuson (1982) differentiated
two fish assemblage types primarily on the basis of connectedness and minimum oxygen
conditions in 18 northern Wisconsin lakes. Presence of an inlet or outlet may provide a
winter refuge from low oxygen concentrations or large piscivores, thereby supporting
greater species diversity (Tonn and Magnuson 1982). Finally, thermal stratification,
which is largely determined by lake area and depth, may influence fish species

distribution, the amount of species-specific suitable habitat and interactions in lakes, and

27

nutrient cycling (MacLean and Magnuson 1977; Magnuson et al. 1979; Crowder and
Magnuson 1983; Fee et al. 1994). Therefore, a classification scheme with landscape
position, depth, and connectivity is likely to explain variability in a biological response

like fish species richness.

Many recent landscape position studies were conducted in northern Wisconsin
lake districts within a homogeneous geomorphologic and land—use setting (e. g., Riera et
a1. 1997; Webster et al. 1996). However, across the Upper Great Lakes region, geology,
climate, and land-use factors are more variable. Larger-scale, regional filters like climate
may constrain species richness or other patterns at the local or lake-district scale (Barbour
and Brown 1974; Tonn 1990; Minns and Moore 1995). Therefore, when considering a
biological response like fish growth across the Upper Great Lakes region, the amount of
length attained in one growing season may vary with latitude because the length of the
growing season tends to vary along a latitudinal gradient (Colby et al. 1979; Colby and
Nepszy 1981; Beverton 1987; Quist et al. 2003). In addition, lake surface area, a factor
not explicitly accounted for in the current classification, has been linked to fish species
richness in many studies (Barbour and Brown 1974; Matuszek and Beggs 1988; Minns
1989; Magnuson et al. 1998), and affects solar radiation input and circulation patterns
within lakes (i.e., the thermal structure;Wetzel 1975; Lampert and Sommer 1997).
Therefore, before using a classification scheme based solely on landscape position,
connectivity, and maximum depth as the basis for a larger statewide or regional
management and monitoring framework, the extent to which biological attributes vary

with existing lake classes across a larger spatial area should be determined, and additional

28

map-based factors that have been shown to influence biological patterns should be

explored as potential grouping variables.

My objectives were to determine: (1) if the current Wisconsin lake classification
scheme explains variability in fish species richness in Wisconsin lakes, (2) if the current
Wisconsin lake classification scheme explains variability in fish growth in Michigan and
Wisconsin lakes, (3) if additional map-based variables influencing climate, habitat
volume, and habitat complexity can further refine the existing classification scheme, and
(4) if a refined classification scheme explains more variability in fish species richness and
fish growth. Given the large spatial scale of this study, I expected that mean length at age
would vary with latitude, and that surface area would be an important additional predictor
of fish species richness. The over-arching goal of this research is to develop the
foundation for building a sampling, monitoring, and management framework that can be

used at the statewide or regional level.

Methods
Study Area

The study area includes 358 inland lakes in Michigan and 410 inland lakes in
Wisconsin between 42 and 47 degrees north latitude (Figure 3). The states of Michigan
and Wisconsin share a similar geologic history, with landscapes shaped primarily by
glaciers 10,000—12,000 years ago (Scheider 1975; Becker 1983). Surface topology and
soils in the glaciated regions of both states are complex and range from areas of thick

glacial till to bedrock outcrops (Scheider 1975; Becker 1983).

29

Lake Morphometry and Landscape Features

Lakes were assigned to Wisconsin lake classes based on landscape position,
connectivity, and maximum depth (Table 4). Landscape position, connectivity, and
maximum depth were obtained from the Michigan Landscape Context GIS database
(Cheruvelil 2004) for Michigan lakes, and the Surface Water Inventory File (Wisconsin
Department of Natural Resources, 1967—1983) and Lakes Book (Wisconsin Department
of Natural Resources) for Wisconsin lakes. Water body type designations (e.g., seepage,
drainage, drained, and spring) were used to determine connectivity for Wisconsin lakes
(Wisconsin Department of Natural Resources, 1995). Drained, drainage, and spring lakes
had either an inlet or outlet, so were considered connected and grouped under the general
connectivity category of drainage lakes. Seepage lakes had no inlet or outlet, so were
grouped under the general connectivity category of seepage lakes. For Wisconsin lakes,
landscape position was quantified using watershed area. Lakes with watershed areas <
1,036 ha were considered high in the landscape, while lakes with watershed areas 2 1,036
ha were considered low in the landscape. Because watershed areas were not available for
all Michigan lakes, landscape position and connectivity were obtained from the Michigan
Landscape Context GIS database. Michigan and Wisconsin lakes were categorized as
deep (> 5.5 m maximum depth) or shallow (< 5.5 m maximum depth) to roughly separate

mixed from thermally-stratified lakes (Lillie and Mason 1983).
Fish Species Richness

Fish species richness estimates were not available for Michigan lakes, so I limited
this analysis to Wisconsin lakes. In Wisconsin lakes, species richness was estimated

from summer mini-fyke net and fall electro-fishing surveys in which all species were

30

targeted during 1998—2004. Because the effort was standardized among lakes, relative
comparisons of estimates of species richness are possible. Fish species richness was
defined as the number of unique species caught in a particular lake for a given year from

both sampling events combined.
Fish Growth

Fish mean length at age summaries from inland lake sampling programs were
compiled from historic and current databases in Michigan and Wisconsin. To standardize
for sampling season, I limited the analyses to surveys that occurred between January and
May, before the start of the growing season. For lakes with more than one survey, a
single survey was randomly selected from all surveys for that lake, so each lake was
represented once in the final dataset. The final dataset included surveys conducted during

1953—2003 for Michigan lakes and during 1947—2003 for Wisconsin lakes.

Various sampling gears and combinations were used in both Michigan and
Wisconsin lakes to sample fish during these periods. Approximately 62% of all surveys
were from single gear types, 28% were from multiple gears, and 10% were unknown gear
types. Of the surveys in which a single gear type was used, most were fyke-netting
surveys (50%), electro-fishing surveys (20%), or trap-netting surveys (15%). I used all
gear types and combinations in the analysis because I sought to obtain a large sample on

a broad spatial scale.

I chose mean length at age 4 for bluegill, largemouth bass, walleye, and yellow
perch as a growth index because age 4 was fully recruited to the sampling gears used in
both states and was the most commonly encountered age class in surveys for all species.

Age 4 is also likely to be an age that is accurately estimated, thereby minimizing age-

31

estimation error. Principal components analysis of surveys with mean lengths at ages 3-6
revealed that the first principal component explained 79-89 % of the variability in mean
lengths at age for all species, thereby making the first principle component a good
descriptor of overall growth among lakes. For all species, the first principal component
was highly correlated with mean lengths at age 4 (Pearson correlation coefficients
ranging from 0.92 — 0.96). Therefore, to ensure large sample sizes, mean length at age 4
was selected (instead of principal component factor scores) as a growth index for analysis
because there were many more surveys with mean length at age 4, then had a complete
matrix of ages 3-6 (as would be required for principle components analyses). The four
species were selected because they were commonly present in samples, represented warm

water and cool water species, and were important to angling fisheries in both states.
Statistical Analysis

To determine if the Wisconsin lake classification scheme explains variability in
fish species richness and fish growth, I tested for differences in fish species richness and
fish growth among lake classes using one-way analysis of variance (ANOVA) with lake
class as the factor. In lakes where the results of the ANOVA indicated significant
differences among lake classes, Bonferroni multiple comparison tests were used to
determine which lake classes differed from one another in the response variable. The
Bonferroni method is applicable for unequal sample sizes and when all pair-wise
comparisons are of interest (N eter et al. 1996). General patterns in significant pair-wise
comparisons were examined across all responses tested to infer the relative importance of
each categorical component of the classification scheme (i.e. connectivity, landscape

position, or depth categories).

32

Regression tree analysis was used to determine if additional map-based variables,
likely related to lake thermal regime, such as latitude and lake surface area could refine
the existing classification scheme. Latitude and lake surface area were obtained from the
Michigan Landscape Context GIS database (Cheruvelil 2004) for Michigan lakes, and the
Surface Water Inventory File (Table 5; Wisconsin Department of Natural Resources,
1967-1983) for Wisconsin lakes. Regression tree analysis is a flexible nonparametric
multivariate analysis method that allows categorical, ratio, and continuous or threshold
data types, accounts for interactions among explanatory variables, and does not assume
linearity among variables (Breiman et al. 1984; Magnuson et al. 1998). The least-squares
loss function with Automatic Interaction Detection (AID) was used because AID
incorporates interactions among predictor variables. The AID algorithm begins with a
single cluster of cases and searches the explanatory variables for a way to split the cluster
into two clusters with the smallest overall within-cluster sum of squares (Morgan and
Sonquist 1963; Wilkinson 2000). By repeatedly splitting the explanatory variables, in
this case latitude, surface area, and maximum depth, into pairs of subsets, the AID
algorithm produces a dichotomous key (Wilkinson 2000). Interactions are represented by
branches from the same node with different predictors splitting further down the tree

(Wilkinson 2000).

In the regression tree analyses, maximum depth was included as a continuous
variable to allow for interactions among latitude, lake surface area, and maximum depth,
and because poorer fits are likely to result when continuous predictor variables are
categorized prior to analysis (Wilkinson 2000). First, lake types were included,

regardless of connectivity or landscape position. Then, based on findings from the initial

33

ANOVAs, regression tree analyses were conducted separately for all seepage and then
for all drainage lakes. Proportional reduction in error (PRE), analogous to a squared-
multiple correlation coefficient, was used to judge the overall fit of the regression tree
models (Wilkinson 2000). Refined lake classes were then developed based on the
outcomes of the regression tree analyses. Differences in fish species richness and fish
growth among lake classes were tested using one-way analysis of variance (ANOVA)

with new lake class as the factor, followed by Bonferroni multiple comparison tests.
Results
Fish Species Richness

Fish species richness varied among lake classes in 256 Wisconsin lakes (Table 6),
but not all pair-wise comparisons of lakes classes were significantly different (Table 7).
The number of fish species ranged from 3 to 30 species across all lake classes. Fish
species richness in seepage lakes was lower than in deep drainage lakes, either high or
low in the landscape (Figure 4). The mean number of species in seepage lake classes was
10, while the mean number of species in drainage classes (classes 1—4) was 14. Fish
species richness was lower in high, shallow drainage lakes than in low, deep drainage
lakes (Figure 4). Overall, comparisons indicate that depth and connectivity may be more
important than landscape position in determining species richness. Of 15 possible
multiple comparison tests, only 1 test involving landscape position (contrasting a high
drainage class to a low drainage class) revealed a significant difference (Table 7). The
significant contrast involved both landscape position and depth categories. In contrast, 4
of 15 multiple comparison tests contrasting shallow or deep seepage and deep drainage

categories were significantly different (Table 7).

34

For all lakes regardless of connectivity or landscape position category, regression
tree analyses of latitude, maximum depth, and surface area revealed that fish species
richness was lower in lakes smaller than 226 ha, and small lakes had fewer species south
of 45.4°N latitude than small lakes north of 45.4°N latitude based on regression tree
analyses (Table 8; Figure 6). However, the low proportion reduction in error in the
overall regression tree model (PRE = 0.277) suggested that much of the variability in fish

species numbers was unexplained (Table 8).

Because landscape position did not emerge strongly in the ANOVA, I evaluated
connectivity categories (e. g. seepage versus drainage) in the next expanded regression
tree analyses for species richness. For seepage lakes, fish species richness was lower in
lakes smaller than 81 ha than in lakes larger than 81 ha (Table 9). For small seepage
lakes (< 81 ha), another split at 38 ha separated medium from small-sized lakes, with
mean species numbers being lowest in the smallest lake group (Table 9). Larger seepage
lakes (3 81 ha) had more species in deep lakes than in shallow lakes (Table 9), suggesting
that habitat volume may be important in determining species richness patterns. For
drainage lakes, fish species richness was lower in lakes shallower than 13 m than in lakes
deeper than 13 m (Table 9). The proportion reduction in error in the separate regression
tree models for seepage and drainage lakes ranged from 0113—0401, with more of the
variability in fish species numbers being explained in some connectivity lake types than

in the model of all lake types combined (Table 9).
Fish Growth
Mean length at age 4 varied among lake classes for bluegill, largemouth bass, and

walleye (Table 6), but not all pair-wise comparisons of lake classes differed (Table 7).

35

For bluegill, mean length at age 4 was greater in deep drainage lakes low in the landscape
(class 4) than in deep drainage lakes high in the landscape (class 2) or deep and shallow
seepage lakes (classes 5 and 6; Table 7; Figure 5). For largemouth bass, mean length at
age 4 was greater in deep drainage lakes low in the landscape (class 4) than in shallow
drainage lakes high in the landscape or deep seepage lakes (class 6; Table 7; Figure 5).
For walleye, mean length at age 4 was lower in deep and shallow drainage lakes low in
the landscape (classes 3 and 4) than in deep drainage lakes high in the landscape (class 2)
or deep and shallow seepage lakes (classes 5 and 5; Table 7; Figure 5). For yellow perch,
mean length at age 4 did not differ among lake classes (Table 6). Bluegill, yellow perch,
and largemouth bass were most frequently sampled in deep seepage lakes (class 6), where
sizes at age were also relatively low, whereas walleye were most frequently sampled in
deep lowland drainage lakes (class 4), where sizes at age were relatively low. In general,
shallow drainage lakes high in the landscape were least represented (fewest surveys) for
all tests. Only 11 of 60 possible multiple comparison test were significant (Table 7). Of
the 11 significant comparisons, 2 clearly separated high (class 2) versus low (classes 3
and 4) landscape categories, 2 contrasted deep and shallow as well as high and low
landscape classes (i.e., differences can be attributed to either depth or landscape position),

and 5 separated seepage from drainage classes (Table 7).

Based on regression tree analyses for all lakes regardless of connectivity or
landscape position category, mean lengths at age 4 for bluegill, largemouth bass, walleye,
and yellow perch were best predicted by latitude or lake surface area. For bluegill, mean
length at age 4 was higher in lakes larger than 446 ha than in lakes smaller than 446 ha

(Table 8). For largemouth bass, mean length at age 4 was higher in lakes larger than 528

36

ha than in lakes smaller than 528 ha (Table 8). For walleye, mean length at age 4 was
higher in lakes south of 45.2°N latitude than in lakes north of 45.2°N latitude (Table 8).
For yellow perch, mean length at age 4 was higher in lakes south of 42.4°N latitude than
in lakes north of 42.4°N latitude and in lakes north of 45.4°N latitude (Table 8). The low
proportion reduction in error in the regression tree models for bluegill, largemouth bass,
walleye and yellow perch (PRE = 0.10—0.32) suggested that much of the variability in

fish growth was unexplained (Table 8).

Because landscape position did not emerge as strongly as connectivity in the
ANOVA, 1 evaluated connectivity categories (e. g. seepage versus drainage) in the next
expanded regression tree analyses for mean length at age. For lakes of different
connectivity categories, mean length at age 4 exhibited similar patterns (as in the
previous analysis including all lake connectivity categories), with either lake surface area
or latitude emerging at the primary split for all responses tested (Table 9). Maximum
depth also emerged at secondary or tertiary splits in several models, including walleye in
seepage and high drainage lakes, and yellow perch in seepage and low drainage lakes
(Table 9). While lake surface area or latitude were consistently important in explaining
variation in mean length at age 4 for each fish species, the separation points varied
between each connectivity model and among species. For both bluegill and largemouth
bass, when lake surface area entered at the first split, the separation point was typically
lower for seepage lakes than for drainage lakes (Table 9). For walleye and yellow perch,
the separation point for latitude was typically farther south for yellow perch than for
walleye, except for drainage lakes (Table 9). The proportion reductions in error were

higher (PRE = 005—048) for 5 of 8 models that distinguished among connectivity lake

37

categories (Table 9) than for four models that did not distinguish among lake categories
(Table 8). Therefore, revised lake classes were created in the next step based on the

outcomes of the expanded regression tree analyses by connectivity category (Table 9).
New Lake Classes

New lake classes were created based on the separation points identified in the
regression tree analyses for species richness and bluegill, largemouth bass, walleye, and
yellow perch mean lengths at age (Table 9). Species richness and mean length at age
varied among new lake classes (Table 10; Figures 7 and 8), but not all lake class pairs
differed (Table 11). Of 58 possible multiple comparison tests, 32 were significantly
different. In all cases, the amount of variation explained by the new lake classes was
greater than the amount of variation explained by the Wisconsin lake classes (Table 6;
Table 10). The new lake classification scheme for species richness still resulted in 6
classes, but rather than landscape position, surface area and maximum depth were used to
separate seepage and drainage lakes respectively (Table 9; Figure 7). The new lake
classifications for bluegill, largemouth bass, walleye, and yellow perch mean length at

age resulted in between 4 and 7 lake groups depending on the species.
Discussion
Fish Species Richness

I found that the Wisconsin classification scheme based on landscape position,
connectivity, and maximum depth explained a moderate about of variability in fish
species richness in 256 Wisconsin lakes. The largest differences in fish species richness

were between seepage lakes and deep lowland drainage lakes, which correspond to the

38

extreme ends of the Riera et al. (2000) lake-order scheme. My results suggest that
connectivity may be a particularly important determinant of fish species richness. My
results may also suggest that lake area affects fish species richness, because seepage lakes
tended to have small surface areas, and lakes low in a drainage are larger in surface area

than seepage lakes or drainage lakes high in a drainage (Kratz et al. 1997; Riera et al.

2000).

The fact that landscape position did not consistently emerge as a strong predictor
of fish species richness may suggest the importance of explicitly considering lake surface
area when evaluating fish species richness patterns among lakes. Lake surface area has
been strongly associated with species richness in several studies (Barbour and Brown
1974; Rahel and Magnuson 1983; Matuszek and Beggs 1988; Minnsl989; Minns and
Moore 1995; Magnuson et al. 1998). Local extinction variables like lake area,
conductivity, pH, and depth were more important than isolation variables related to
connectivity in predicting species richness and community composition in Wisconsin and
Finnish lakes (Magnuson et al. 1998). Riera et al. (2000) detected significant differences
in lake surface area among lake order classes (i.e., landscape position determined lake

surface area), but surface areas for each lake order class were highly variable.

The relative importance of isolation and extinction variables in Wisconsin and
Finnish lakes differed based on the local geomorphic setting (Magnuson et al. 1998).
Similarly, Webster et al. (2000) studied landscape controls on lake chemical responses to
drought in four lake districts in Wisconsin and Ontario ranging between 45 and 51° N
latitude. The degree to which chemical properties of individual or small groups of lakes

responded to drought varied with local hydrologic setting (groundwater versus surface-

39

water dominated) and landscape position. Therefore, local and landscape position
variables alone may not explain all variability in physical, chemical, or biological
characteristics of lakes. Consideration of the hydrologic and geomorphologic setting is
also important, which could account for my inability to differentiate species richness
among all lake class pairs in the ANOVA of the Wisconsin lake classification scheme,
the relatively low predictive power of the Wisconsin classification scheme, and the

emergence of latitude as an important predictor in the later regression tree analyses.

The Wisconsin lake classes used in this analysis were derived using general
approximations for landscape position (high and low) and connectivity (seepage and
drainage), based on watershed area for Wisconsin lakes and visual inspection of maps for
Michigan lakes. The lake order classes used by Riera et al. (2000) were based on the type
and strength of connections between a lake and the surface drainage network, and were
determined from stream order designations from stream network information. However,
Riera et al. (2000) presented convincing evidence that lakes higher in the landscape
consistently had smaller catchments. Therefore, the watershed area approximation of less
than 1036 ha used in this analysis for Wisconsin lakes is likely to be useful in separating
lakes high in the landscape from lakes low in the landscape. Still, more accurate
determinations of landscape position could help to further refine the Wisconsin
classification scheme. Whether approximated or measured directly, landscape position is
thought to constrain not just physical attributes of lakes, such as lake surface area, but
also chemical attributes of lakes such as concentrations of ions (Kratz et al. 1997; Riera et
al. 2000; Soranno et al. 1999; Webster et al. 1996, 2000). Because water chemistry data

are important, but less available for large numbers of lakes, the use of map-based

40

variables such as landscape position to characterize lake productivity is appealing from a

classification perspective.
Fish Growth

1 found that the Wisconsin classification scheme based on landscape position,
connectivity, and maximum depth explained a moderate amount of variability (R2 0.02—
0.12) mean length at age 4 of bluegill, largemouth bass, and walleye. As with species
richness, deep seepage lakes were consistently different than lowland drainage lakes for
bluegill, largemouth bass, and walleye growth. Largemouth bass and bluegill growth was
lower in deep seepage lakes than in low drainage lakes. In contrast, walleye growth was
higher in deep or shallow seepage lakes than in low drainage lakes. Walleye prefer cool
water, whereas largemouth bass and bluegill prefer warm water. Seepage lakes such as
kettle lakes in glaciated areas like Michigan and Wisconsin, are typically bowl shaped,
and may lack expansive warm littoral areas typically used by centrarchids. Seepage lakes
in Wisconsin tend to have greater water clarity, lower nutrients, lower alkalinity, and are

less eutrophic than drainage lakes (Lillie and Mason 1983).

The distribution of lakes surveyed was not even among lake classes in my
analysis. Of all lakes included in my analysis, only 16 percent of shallow lakes (classes
1, 3, and 5) had survey information, which is likely due to the tendency of resource
management agencies to focus sampling efforts on large, deep lakes, or to the fact that
the distribution of lake types within the region is uneven. Except for species richness and
walleye growth in shallow lowland drainage lakes, other shallow Wisconsin lake classes

had sample sizes less than 30 for each variable tested. Because sampling was typically

41

not randomly distributed and the targeted species were unknown, inferences about

statewide patterns in species presence or absence by lake class are not possible.
New Lake Classes

Lake surface area and latitude explained significant amounts of variation in
species richness and fish growth. The consistent emergence of latitude and lake area in
regression tree analyses suggests that both local lake-type and regional factors influenced
fish species richness and growth in my study lakes. Barbour and Brown (1974) found
that latitude explained 64% of the variation in species numbers in 14 lakes across North
America, a much larger spatial scale than my study. In my study, for all lake types,
latitude was an important predictor of walleye and yellow perch growth, but did not
emerge as a predictor of bluegill or largemouth bass growth. Many studies have shown
that fish growth varies with latitude (Colby et al. 1979; Colby and Nepszy 1981;
Beverton 1987; Quist et al. 2003). Because fish species differ in their thermal
requirements, growth potential, production potential, and presence in a lake may be
directly related to thermal structure of that lake, which is largely determined by lake
morphometry and climate (Wetzel 1975; Johnson et al. 1977; Tonn et al. 1983; Marshall
and Ryan 1987; Christie and Regier 1988). In this region, temperature may be a more

important driver for percid growth, than for centrarchid growth.

In addition to setting up the thermal properties of lakes, lake morphometry also
influences nutrient cycling in lakes and primary productivity. Annual rates of
phytoplankton photosynthesis per unit of lake surface area have been shown to increase
as a function of lake size (Fee et al. 1992). Nutrient recycling may be more efficient in

large lakes because the intensity of turbulence within the mixed layer and the depth of the

42

mixed layer increase with increasing lake size (Fee at al. 1994). Through these processes,
the mixed layer retains particles longer, thus increasing the likelihood of phytoplankton
uptake and minimizing nutrient loss to the sediments. Because of increased nutrient
recycling efficiency, large lakes may be more productive than smaller stratified lakes,
which would also influence growth potential. Likewise, lakes that do not stratify and
continually mix may have different nutrient cycling than lakes that stratify (Lillie and
Mason 1983). Stratification patterns are a function of lake surface area, fetch, and
maximum depth (Wetzel 1977). Fish yield, a response variable not included in this
analysis, has been linked to measures of primary productivity (Ogelsby 1977; Ogelsby et
al. 1987), but has also been linked to thermal habitat volume (Christie and Regier 1988).
Fish abundance and harvest have been predicted from lake surface area in several studies
(Hansen 1989; Baccante and Colby 1996; Nate et al. 2000; Nate et al. 2001). Therefore,
including latitude and lake surface area in the classification scheme may expand the
utility of the classification scheme beyond monitoring of species richness to management

of fish populations.

Ricklefs (1987) suggested that ecology should focus on macro-scale explanations
for local phenomena. Here, I used regional, lake-type (in the sense of Tom 1990), and
local features to explain two local phenomena, species richness and fish growth, of
interest to inland lake resource management agencies. The importance of lake-type
variables was demonstrated because the influence of latitude, lake surface area, and
maximum depth varied with lake connectivity. Forces that influence species richness and
fish population dynamics in inland lakes occur at multiple spatial and temporal scales

(Minns 1986; Minns and Moore 1995; Hershey et al. 1999; Beard et al. 2003; Lester et al.

43

2003), therefore multiple scales should be considered when developing classification
schemes for monitoring or management programs. Multiple spatial scale classifications
and monitoring programs not only help us to understand the present condition, but also
aid in monitoring the status and trends of aquatic ecosystems in the face of future
environmental stressors. For example, changes in species distributions as a result of
global climate change will be most evident at the regional scale because species ranges
shift in response to warmer climates, changes in precipitation patterns, changes in soil
moisture content, and frequency of fire (Minns and Moore 1995). In a synthesis of
classification papers presented to the North American Benthological Society, Hawkins et
al. (2000) suggested that a tiered approach to classification, where local habitat features
form the foundation of the classification and large-scale features are used to refine the
classification, should lead to greater precision and accuracy in predicting biological
conditions. The approach is supported by the fact that local extinction factors like lake
surface area and depth were found to be more important than isolation factors like
connectivity in determining species richness in northern Wisconsin and Finnish lakes

(Tonn and Magnuson 1982; Magnuson et al. 1998).

When interpreting biological patterns among lakes such as fish species richness or
growth among species, each measure includes among-site spatial, within-site spatial,
temporal, and measurement error components of total variability. In classification, we
seek to explain among-site spatial variability, which can only be successful if among-site
spatial variability is the largest component of the total variability (Hawkins et al. 2000).
My study was unique in that it encompassed a large geographic scale, and incorporated

data from an approximate 60-year time period. Patterns and processes affecting aquatic

44

ecosystems operate at various spatial and temporal scales (O’Neill 1989; Roughgarden
1989; Tonn 1990). By expanding the temporal scale, more variability in time is
introduced, which typically should reduce our ability to classify among—site variability
(Hawkins et al. 2000). However, because the number of lake surveys was large, I had
the statistical power to see the underlying spatial pattern. By confining my analyses to
available, temporally stable, map-based variables, I evaluated a classification that could
be implemented for most lakes in the region with minimal effort. Because the predictor
variables were stable in time, they could be used to predict biological attributes across
large time periods, thereby increasing my sample of lakes and expanding the spatial

scale, which ultimately revealed an underlying spatial pattern.

My research did not explicitly account for chemical or biological differences
among lake classes, which may also explain the relatively low predictive ability of some
of my models. Other variables not tested here are likely to be important predictors of
species richness or mean length at age, and could be used to further enhance the
classification schemes if they become available for a large number of lakes or can be
predicted from map-based variables like landscape position. For example, species
richness has been shown to vary with pH (Matuszek and Beggs 1988; Minns 1989) and
alkalinity (Rahel 1984), and bluegill mean lengths at age have been correlated with
Secchi depth and alkalinity (Tomcko and Pierce 2001). More research is needed to
quantify the effects of other factors on species richness, fish growth, and other biological
responses, and to expand our ability to extrapolate aspects of nutrient status to un-
sampled lakes using map-based or remote sensing technologies. Because my

classification does not currently contain a water chemistry variable to indicate lake

45

productivity or nutrient status, landscape position may still provide an important indicator
of ionic concentrations (Webster et al. 1996; Kratz et al. 1997; Soranno et al. 1999; Riera

et al. 2000; Webster et al. 2000).
Management Implications

My overall goal was to develop a management tool for inland lakes that may
ultimately be used to generate statewide or regional management frameworks for
sampling and managing fish populations and for monitoring aquatic ecosystem health.
The new classification scheme developed herein is more complex (increasing from from
6 to 26 lake classes), but shows a marked improvement over the existing Wisconsin
classification scheme (i.e., more variability is explained by the new lake classes). By
accounting for inherent differences among lakes through a classification scheme that
incorporates physical and landscape features at multiple spatial scales, we provide a
template for understanding lake-to-lake variability in fish species richness and growth of
fish populations. Through an improved understanding of the “expected state” for a given
lake type, underlying processes that affect survival and recruitment of fish populations
may also be more easily and accurately identified. Ideally, the classification scheme
could then be used to 1) predict the productive potential and sustainability of fish
populations in lakes for which recent survey information is lacking, 2) serve as a basis for
stratifying organized, statewide random sampling efforts, 3) serve as a basis for

experimental regulations, and 4) refine stocking practices.

A common focus of state fisheries resource management agencies has been to
conduct intensive surveys of game fish populations on individual lakes, while ignoring

non-game-fishes (Hayes et al. 2003). Lake surveys are often conducted in response to

46

public complaints of poor fishing, rather than as part of an organized, statistically-based
sampling design that seeks to obtain information on all aspects of the fish community in a
comparable manner. Local fisheries managers often focus on responding to the public,
while placing less emphasis on larger-scale statewide issues such as reporting on the
status and trends of the fishery (e. g., monitoring), stocking and regulation evaluations, or
the status and trends of aquatic ecosystem health. Nonetheless, larger-scale, statewide
issues are part of the mission of fisheries management agencies (Hayes et al. 2003).
Standardized statewide monitoring strategies, should enable both local and statewide
questions to be addressed with a single sampling effort, but are often met with resistance,
primarily because these programs supposedly “compete” with local management needs
(Bonar and Hubert 2002; Hayes et al. 2003). However, standardized random sampling
can provide a means to extrapolate to lakes that have not been sampled, thereby
providing tools to address local management issues (e.g., Hansen et al. 1989) and a sound
basis for comparing conditions among similar lake types (Bonar and Hubert 2002). The
new lake classification scheme presented here defines lake types, thereby providing a

sound basis for comparisons for both monitoring and management activities.

47

Table 4. Wisconsin lake classification scheme based on maximum depth (deep: 2 5.5 m,
shallow <55 m), landscape position (high: watershed area < 1036 ha, low: watershed
area 2 1036 ha), and connectivity (drainage: inlet or outlet present, seepage: no inlet or
outlet present).

 

 

Wisconsin

Classification Description
1 Shallow, High, Drainage
2 Deep, High, Drainage
3 Shallow, Low, Drainage
4 Deep, Low, Drainage
5 Shallow, Seepage
6 Deg), Segpage

 

48

Table 5. Summary of lake morphology and latitude used in regression tree analyses to
predict fish species richness in 256 Wisconsin lakes during 1998-2004 and mean length
at age 4 for bluegill, largemouth bass, walleye, and yellow perch in 358 Michigan and

410 Wisconsin lakes during 1947—2003.

Species Richness
N Minimum Maximum
42.5082

Latitude 256
Maximum Depth (m) 256
Surface Area (ha) 256

46.5801

Mean LenLth at Age 4
N Minimum Maximum

768 41.7772 47.4636
768 l 37
768 6 8124

 

49

Table 6. Analysis of variance of species richness and mean length at age 4 for bluegill,
largemouth bass, walleye, and yellow perch among six classes of Michigan and
Wisconsin lakes during 1947—2003.

 

 

Response N F P R2
Species Richness 256 12.783 5 0.001 0.204
Bluegill 412 4.309 0.001 0.050
Largemouth Bass 370 5.712 _<_ 0.001 0.073
Walleye 376 9.995 _<_ 0.001 0.119
Yellow Perch 368 2.189 0.0549 0.029

 

50

Table 7. Bonferonni matrix of pairwise comparison probabilities for species richness
and mean length at age 4 for bluegill, largemouth bass, walleye, and yellow perch among
the six classes of Michigan and Wisconsin lakes during 1947—2003. Bold indicates
significant difference.

 

 

 

Lake 1 2 3 4 5 6
Class High High Low Low Seepage Seepage
Shallow Deep Shallow Deep Shallow Deep
Fish Species Richness
1 l
2 0.840 1
3 1 0.904 1
4 0.041 1 0.006 1
5 1 0.006 0.556 5. 0.001 1
6 l 5 0.001 0.384 5 0.001 1 l
Bluegill
1 1
2 1 1
3 l l l
4 1 0.004 1 l
5 l 1 0.414 0.002 1
6 1 l 1 0.006 1 1
Largemouth Bass
1 2 3 4 5 6
l 1
2 0.403 1
3 0.203 1 l
4 0.006 0.064 1 1
5 1 1 1 0.161 1
6 1 0.656 0.674 5 0.001 1 1
Walleye
1 1
2 1 1
3 1 0.038 1
4 1 _<_ 0.001 1 1
5 1 1 0.008 0.002 1
6 l 1 0.010 5 0.001 1 1
Yellow Perch *
* ANOVA not significant

51

Table 8. Regression tree results using latitude (decimal degrees), surface area (ha) and
maximum depth (m) to predict fish species richness and mean length at age 4 for bluegill,
largemouth bass, walleye, and yellow perch for 358 Michigan and 410 Wisconsin lakes
during 1947—2003. PRE = proportional reduction in error, Values = back-transformed
hectares or meters, and latitude in decimal degrees, Means = the mean value (number of
species or length at age 4) for each split (group), N = the number of lakes in each group,
and New Class = the new response-specific class assignment based on regression tree
results.

 

 

Number
PRE Split Variable Value Mean of Lakes
Species Richness
0.277 1 Surface Area < 228 12 196
2 Latitude < 45.4046 10 72
2 Latitude > 45.4046 13 124
3 Surface Area < 57 10 40
3 Surface Area > 57 14 84
1 Surface Area > 228 16 60
Bluegill
0.101 1 Surface Area < 446 146 373
1 Surface Area > 446 174 39
Largemouth Bass
0.077 1 Surface Area < 528 287 341
1 Surface Area > 528 327 29
Walleye
0.316 1 Latitude < 45.1999 442 74
l Latitude > 45.1999 359 48
Yellow Perch
0.150 1 Latitude < 42.3858 208 53
1 Latitude > 42.3858 181 315
2 Latitude < 45.8479 186 204
2 Latitude > 45.8479 171 1 1 l

 

52

Table 9. Regression tree results using latitude (decimal degrees), Log.z surface area (ha)
and Loge maximum depth (m) to predict fish species richness in 256 Wisconsin lakes and
fish mean length at age 4 for Michigan and Wisconsin lakes during 1947—2003. PRE =
proportional reduction in error, Values = back-transformed hectares or meters, and
latitude in decimal degrees, Means = the mean value (number of species or length at age
4) for each split (group), N = the number of lakes in each group, and New Class = the
new response-specific class assignment based on regression tree results.

 

 

Connectivity New
Categry PRE Split Variable Value Mean N Class
Fish Species Richness
Seepage 0.401 1 Surface Area < 81 8 46
2 Surface Area < 38 6.9 22 A
2 Surface Area 3 38 9.3 24 B
1 Surface Area 3 81 12 24
2 Maximum Depth < 11 10 14 C
2 Maximum Depth 3 ll 15 10 D
Drainage 0.1 13 1 Maximum Depth < 18 13 161 E
1 Maximum Depth 3 18 18 25 F
Bluegill
Seepage 0.081 1 Surface Area < 12 112 8 G
1 Surface Area 3 12 147 184 H
Drainage 0.177 1 Surface Area < 446 147 189 I
1 Surface Area 3 446 177 31 J
Largemouth Bass
Seepage 0.052 1 Surface Area < 19 246 9 K
1 Surface Area 3 19 284 157 L
Drainage 0.115 1 Surface Area < 686 292 186 M
1 Surface Area 3 687 338 18 N
Walleye
Seepage 0.475 1 Latitude < 45.3646 377 74
2 Surface Area < 53 480 11 O
2 Surface Area 3 53 430 20 P
1 Latitude 3 45.3646 448 31
2 Surface Area < 30 420 9 Q
2 Surface Area 3 30 371 65
3 Maximum Depth < 5 428 5 R
3 Maximum Depth 3 5 366 60 S
Drainage 0.324 1 Latitude < 45.1055 439 45 T
1 Latitude 3 45.1055 352 226 U
Yellow Perch

53

Seepage

Drainage

0.180

0.137

v—ip—INNt—ib—I

Latitude
Latitude
Maximum Depth
Maximum Depth
Latitude
Latitude

54

< 43.3319
3 43.3319
< 5
3 5
< 42.2289
3 42.2289

201
177
155
180
216
181

50
124
15
109
23
171

<

N~<><€

Table 10. Analysis of variance of species richness and fish mean length at age 4 among

new lake classes developed from regression tree analyses for Michigan and Wisconsin
lakes during 1947—2003.

 

 

Response df F P R2
Species Richness 5;250 20.553 _<_ 0.001 0.291
Bluegill 3;408 19.931 5 0.001 0.128
Largemouth Bass 3;366 15.810 S 0.001 0.115
Walleye 6;269 39.230 S 0.001 0.389
Yellow Perch 4;363 15.351 5 0.001 0.145

 

55

Table 11. Bonferonni matrix of pairwise comparison probabilities for species richness
and fish mean length at age 4 among new classes developed from regression tree analyses
for Michigan and Wisconsin lakes during 1947—2003. Bold indicates significant
difference.

 

 

Lake
Class 1 2 3 4 S 6 7
Fish Species Richness
1 1.000
2 1.000 1.000
3 0.591 1.000 1.000
4 5 0.001 0.004 0.066 1.000
5 5 0.001 0.001 0.150 1.000 1.000
6 5 0.001 5 0.001 5 0.001 1.000 5 0.001 1.000
Bluegill

1 1.000
2 0.007 1.000
3 0.008 1.000 1 .000
4 5 0.001 5 0.001 5 0.001 1.000

Largemouth Bass
1 1.000
2 0.015 1.000
3 0.002 0.333 1 .000
4 5 0.001 _<_ 0.001 5 0.001 1.000

Walleye
1 1.000
2 0.466 1.000
3 0.368 1.000 1.000
4 1.000 1.000 1.000 1.000
5 5 0.001 5 0.001 0.029 0.094 1.000
6 0.998 1.000 1.000 1.000 5 0.001 1.000
7 5 0.001 5 0.001 5 0.001 0.007 0.689 5 0.001 1.000
Yellow Perch

1 1.000
2 5 0.001 1.000
3 5 0.001 0.194 1.000
4 0.321 5 0.001 5 0.001 1.000
5 5 0.001 0.093 1.000 5 0.001 1.000

 

56

 

Figure 3. Locations of 256 Wisconsin lakes used to examine variability in species
richness (asterisks) and 358 Michigan and 410 Wisconsin lakes used to examine
variability in bluegill, largemouth bass, walleye and yellow perch mean length at age 4
(solid diamonds).

57

18

 

m 16
g 14 + I
(I)
Q 12 +
‘0 10 l .l.
q—
E 8
8 6
g 4
z 2
O I I T I I I I

 

D D D D S S
H H L L
SH DE SH DE SH DE

Lake Class
Figure 4. The mean number of species (j: 95% confidence interval) by Wisconsin lake
class in summer mini-fyke net and fall electrofishing surveys on 256 Wisconsin lakes

during 1998—2004. X axis labels include D = drainage, S = seepage, H = high landscape
position, L = low landscape position, SH = shallow, and DE = Deep.

58

 

Mean=12.88
SD=5.19
N=256

 

 

 

 

 

 

l
LNAREA<15.42

 

 

Mean=11.68 Mean=16.38
SD=4.77 SD=4.89
N=196 N=60

1
l

 

 

 

 

 

 

 

I
LAT<45I404598

 

 

Mean=9.76 Mean=12.80
SD=4.16 SD=4.76
N=72 N=124

 

 

 

 

 

 

 

 

 

F
LNAREAT4201 323

 

 

Mean=9.97 Mean=14.15
SD=4.04 SD=4.50
N=40 N=84

 

 

 

 

 

 

Figure 5. Regression tree mobile diagram example predicting the number of species by

Log, Surface Area, Logc Maximum Depth, and Latitude for 256 Wisconsin lakes during
1998-2004. Results from regression tress analyses are presented in tabular format rather
than mobile diagrams.

59

200
180
160
140
120

500

450

Mean Length (mm)

400

350

300

100 r

 

 

 

 

400
350
.l. + I + + 300 :
250
I I I I I I 200
1 2 3 4 5 6
1 250
3 200
j " + + 150
i i +
1
fl I I I I I 100
1 2 3 4 5 6
Lake Class

q
—
cl

111

111141111

 

 

1111411]

IJJLII

 

 

1

23456

Figure 6. Bluegill (upper left panel), largemouth bass (upper right panel), walleye (lower
left panel), and yellow perch (lower right panel) mean length at age 4 (mm; 3: 95%
confidence interval) for six Wisconsin lake classes.

60

25

 

 

3 E
'6 20: +
(D 1
o. -
CD15] "
“5 g +
(1510: +
e :l
3 59
Z :
O- I I I I I I

 

A B C D E F
Lake Class

Figure 7. The mean number of species (: 95% confidence interval) by new lake class in
summer mini-fyke net and fall electrofishing surveys on 256 Wisconsin lakes during
1998—2004.

61

 

 

 

 

 

 

 

 

 

200: 400:
180§ ‘1 3503 +
160-3 E
: 300-
1405 + + f + I
120-? 2503 ‘Ir
E 100: . . . 1 200‘ . r . .
:5 G H I J K L M N
U)
5
_I 500: 250:
incl 3 +
2 g + .. ‘l' 2005 -|-
4005 1 + + +
E -|- 1503
350: ‘I' :
3 Z
300 T I I I r I I 100 I I I I I
o P Q R s T u v w x Y 2

Lake Class

Figure 8. Bluegill (upper left panel), largemouth bass (upper right panel), walleye (lower
left panel), and yellow perch (lower right panel) mean length at age 4 (mm; 3: 95%
confidence interval) for new lake classes.

62

CHAPTER 3

EVALUATING ANGLING REGULATIONS FOR BLUEGILL IN WISCONSIN

Abstract. — 1 determined expected ranges of bluegill growth and mortality for
different lake classes in Wisconsin from existing survey data to determine the best
combination of minimum length limits (152, 178, 203, and 229 mm) and daily creel
limits (25, 10, 5) to maximize yield per recruit and number of bluegill harvested per 100
individuals for each lake class. For each combination of minimum length limit and daily
creel limit, yield per recruit increased with lake size, from the smallest lake class to the
largest lake class, and ranged 4—30 grams per recruit among regulation scenarios. Under
a 25-fish daily creel limit, a 152-mm minimum length limit produced the highest yield
per recruit for small and large seepage lakes and small drainage lakes, whereas a l78-mm
minimum length limit produced the highest yield per recruit for large drainage lakes.
With the exception of large drainage lakes, higher minimum length limits (203 or 229
mm) at the 25 daily creel limit yielded fewer grams per recruit than reduced daily creel
limits at smaller minimum length limits (152 or 178 mm). The number of fish harvested
at or above quality length (152 mm) ranged 7-31 per 100 among lake classes. Higher
minimum length limits under the 25 daily creel limit often yielded similar numbers of
quality-sized fish as reduced daily creel limits under 152 or 178 mm minimum length
limits. My results provide ranges of bluegill growth, natural mortality, and fishing
mortality for lake classes that may be used to evaluate future regulation proposals and to
identify candidate lakes for experimental regulations. Regulations like minimum length
limits may be more effective in increasing the number of quality-sized fish harvested in

lake classes (e. g. New Lake Classes 1 or J) that have the highest growth and exploitation.

63

Introduction

Bluegill (Lepomis macrochirus) are among the most highly sought fish species by
anglers in Wisconsin (McClanahan 2003) and are typically managed with liberal seasons,
no minimum length limit, and high daily creel limits (Becker 1983; Coble 1988).
Bluegill are prolific and rarely limited by recruitment, so angling is generally not
perceived to have long-term impact on bluegill populations (Coble 1988). However,
angling can directly affect bluegill population size structure through selective removal of
large fish (Goedde and Coble 1981;Cob1e 1988; Beard and Kampa 1999), which
indirectly shifts life history strategies to favor early maturation, decreased growth, and
reduced maximum size (Drake et al.1997; Jennings et a1 1997). Typically, high daily
creel limits such as 25 per day are intended to distribute harvest among anglers, but are
not intended to directly address poor population size structure caused from angling or life

history changes (Noble and Jones 1993).

In contrast, minimum length limits have recently been evaluated as a means to
improve size structure of bluegill populations (Beard et al. 1997a; Paukert et al. 2002).
Minimum length limits prevent harvest of fish shorter than a specified length, and thereby
may reduce mortality on vulnerable populations at low abundance, or reduce exploitation
of immature fish before they spawn (Noble and Jones 1993). Minimum length limits
may be ineffective where harvest has little effect on total mortality (Noble and Jones
1993). Minimum length limits have been implemented as a means to increase size
structure for other centrarchids such as largemouth bass Micropterus salmoides (Wilde
1997; Newman and Hoff 2000) and black crappie Pomoxis nigromaculatus or white

crappie Pomoxis annularis (Maceina et al. 1998; Hale et al. 1999) with mixed results.

64

For example, minimum length limits were most likely to improve crappie size structure
and yield in North American lakes if growth was fast and natural mortality was low
(Allen and Miranda 1995). Paukert et al. (2002) suggested that natural mortality was
more important than growth for determining success of a 200-mm minimum length limit
for bluegill in 18 Nebraska lakes. Delaying harvest is not likely to increase size structure
for populations with high natural mortality (Coble 1988). Overall, the relative
importance of growth, natural mortality, and fishing mortality in determining the success
of minimum length limits should depend on the species, and on the growth and mortality

experienced by the population.

The growth and mortality experienced by a fish population likely vary among
lakes, because lakes vary greatly in morphometry, fish assemblages, and productivity.
Consequently, regulation changes are often proposed for individual lakes, or small
numbers of lakes, after analysis of time-series data or simulations of population dynamics
in response to a proposed set of regulations (e.g., Maceina et al. 1998). The weakness of
this approach is that the number of experimental units is usually small (e. g., one to a few
lakes), and the methods used for data collection and analyses are typically not
standardized among lakes. Therefore, the information obtained from studying individual
lakes before and after regulation changes cannot be extrapolated to other potential
candidate lakes for the same regulations, or for studying the effects of state- or region-
wide regulation of fish populations. Lake classification provides a means to account for
natural variation among lakes, and thereby provides a framework for implementing large-
scale management actions, such as minimum length limits or daily creel limits, on many

lakes at the same time.

65

When evaluating length limits, lake classes based on growth potential may
provide a framework for understanding other dynamic rates such as natural and fishing
mortality, age-at-maturity, fecundity, and recruitment, because growth is linked to all of
these rates and processes. For example, growth parameters and environmental
temperature are strongly related to the rate of natural mortality among many fish stocks
(Pauly 1980; Beamesferder and North 1995). High fishing mortality may lead to
decreased density, increased growth, decreased age and size at maturity, and decreased
fecundity (Trippel 1995). Large parents may spawn earlier and produce offspring that
experience a longer growing season and therefore may have higher over-winter survival
(Miranda and Muncy 1987; Ridgway et a1. 1991). Therefore, growth can influence
recruitment. Decreased age at maturity and increased growth rates as a result of high
exploitation, have been shown for northern pike Esox Iucius (Diana 1983), brown trout
Salmo trutta (Hegge et al. 1991), Atlantic cod Gadus morhua (Beacham 1983), and arctic
char Salvelinus alpinus (Hegge et al. 1991). Size at maturity is often positively related to
asymptotic size (Chamov 1993), estimated from the von Bertalanffy growth equation
(von Bertalanffy 1938), so early growth rates may affect population size structure.
Consequently, growth, which is relatively easy to estimate from data collected during
routine fisheries surveys, may be more useful for categorizing lakes for fishery regulation
than other fish population attributes, such as recruitment, mortality, or density, which are

more difficult to estimate.

My objective was to estimate how bluegill yield per recruit and number harvested
varied as a function of minimum age entering the fishery and fishing mortality rate

among four classes of Wisconsin inland lakes. 1 used these estimates to determine the

66

minimum length and daily bag limits that maximize yield per recruit and number of
quality-sized or larger bluegill harvested per 100 recruits for each lake class. Lake
classes for bluegill were developed based on growth and habitat characteristics of lakes in
Wisconsin and Michigan as part of a previous study (Chapter 2), but had not been
evaluated as a framework for regulating bluegill fisheries. First, I determined expected
ranges of bluegill growth and mortality for each lake class. Next, I estimated yield per
recruit and number of bluegill harvested across four minimum length limits and three
daily creel limits, based on the expected ranges of growth and mortality for each lake
class. Given the strong linkage between grth and other population characteristics,
such as recruitment and mortality, 1 expected that lake classes would provide a useful
framework for setting angling regulations, especially minimum length limits, for bluegill
fisheries in Wisconsin. I also expected that regulations that maximized yield per recruit
or number of fish harvested might vary among lake classes due to differences in growth

and mortality rates among lake classes.
Methods
Study Area and Data Sources

My study area was the population of all inland lakes in Wisconsin, because my
interest was in using lake classification as a basis for evaluating potential angling
regulations for bluegill on the state’s inland lakes. Lake classes based on bluegill growth
potential were established for Wisconsin lakes as part of a previous study, based on
categories of connectedness (seepage and drainage) and lake surface area (Table 12;
Chapter 2). The lake classification scheme was developed based on analysis of patterns

in bluegill mean length at age 4 for inland lakes in Wisconsin and Michigan (Chapter 2),

67

but has not been evaluated as a basis for setting regulations, or for its ability to explain
variation in other population characteristics, such as mortality, which should be

considered when setting regulations.

My interest was in using the lake classification scheme to evaluate bluegill
regulation strategies that could be set on a statewide basis for a class of lakes, so data
requirements for fisheries population information were large (e. g. statewide) and different
from the previous study (Chapter 2) in that I focused on Wisconsin lakes only. I used
existing fisheries survey data from Wisconsin Department of Natural Resources, Bureau
of Fisheries Management and Habitat Protection databases to develop new datasets
specific to this analysis. Lakes were typically sampled as part of regional or local
sampling priorities, so final datasets were not comprised of randomly selected lake
surveys. However, sampling was extensive in both space and time, so I assumed that
lakes sampled provided a representative sample of bluegill population characteristics in

inland lakes in Wisconsin (Figure 9).

To identify expected ranges of bluegill growth and mortality among classes of
Wisconsin lakes, 1 identified subsets of lakes in the database with fyke-netting and
electrofishing surveys that included mean length at age and weight-length data. From
mean length at age data, I estimated growth parameters, age frequency, total mortality,

and natural mortality. From weight-length data, I estimated weight-length relationships
and asymptotic average weight (W...) for yield/recruit models. Most surveys did not

contain all information needed (e.g., length, weight, and age), so I developed separate
datasets for length-age, weight-length, and age structure (Figure 9) and assumed that each

dataset was a representative sample of each population attribute for each lake class. I

68

tested for differences in growth and mortality parameters among lake classes. However,
because my intent was to demonstrate how lake classification could be used in
management and not all lake classes had large sample sizes for each parameter, I
proceeded with minimum length limit and daily creel limit analyses by lake class,
regardless of the outcome of the tests for differences in growth or mortality estimates

among lake classes.
Growth

Growth parameters were estimated for the von Bertalanffy growth model using

non-linear regression (von Bertalanffy 1938):
Lt = Loo(1—e(-.K(t_t0)));

where L, = the mean length at age t, LO0 = the asymptotic average length, K = the

instantaneous rate at which L, approaches Loo, and t0 = the hypothetical age at which

length is zero (Haddon 2001). Von Bertalanffy growth parameters were estimated from
mean lengths at ages 1—7 for each lake survey, and then parameter estimates were
averaged for each lake class. Growth parameters and their asymptotic standard errors
were estimated using nonlinear regression, with additive errors. Parameter estimates
were excluded if the estimated asymptotic length exceeded the state record by 10%.

Weight-length relationships were estimated for the allometric weight-length model to

obtain an estimate of W00 for the yield per recruit model:

WI 2 aLiIB;

69

where W,- = the mean weight at length L,, a is a scaling constant, and fl is the allometric

shape parameter (Haddon 2001). Parameters of weight-length relationships were

estimated for each lake survey from the log-transformed equation, and then averaged

across all lake surveys within each lake class. Asymptotic average weight W00 was

estimated by substituting Loo into the averaged weight-length model. The assumption that

growth parameters differed by lake class was tested with analysis of variance with lake

class as the factor.
Mortality

Age frequencies were estimated using age-length keys and catch curves were then
used to estimate instantaneous total mortality (Z) from the age frequency (Ricker 1975).
Survival (S) was estimated from e72, and total annual mortality (A) equaled 1 — S.

Natural mortality (M) was estimated using the Pauly equation (1980):

LogeM = —0.0152 - 0.279LogeL00 + 0.6543LogeK + 0.4634LogeT;

where LOO and K are von Bertalanffy grth parameters, and T = the mean annual

temperature (C) experienced by the stock. Because mean annual water temperature was
not available, mean annual air temperature was obtained from the National Oceanic and
Atmospheric Administration website (http://climvis.ncdc.noaa.gov/). Mean annual air
temperature roughly corresponds to mean annual surface water temperature for
freshwater systems (Pauly 1980; Shuter et al. 1983). Statewide mean annual air
temperatures were averaged for a 60-year period corresponding to the period of fishery

survey information. Instantaneous fishing mortality (F) was estimated by subtracting

70

instantaneous natural mortality (M) from instantaneous total mortality (Z = F + M).
Exploitation was estimated from F, A, and Z using Baranov's equation (u = F *A/Z). In
Wisconsin, the aggregate panfish bag limit was reduced from 50 to 25 in 1998.
Differences in mortality rates between periods (pre—1988 and post—1988) were tested
with analysis of covariance to determine if mortality rates could be combined, and
averaged across periods with loge number at age as the dependent variable, age as the
(covariate) independent variable, and period (pre or post) as the treatment variable. The
assumption that total instantaneous mortality estimates differed by lake class was tested

with analysis of variance with lake class as the factor.
Exploitation

Harvest limit reductions were incorporated into the yield per recruit model by
estimating the percent reduction in exploitation for each of two reduced creel-limit
scenarios (from 25 to 10 bluegill per day and from 25 to 5 bluegill per day) using angler
interviews from recent creel surveys (1998—2002). Creel surveys for northern Wisconsin
lakes were conducted during the angling season (first Saturday in May through March 1
of the following year) as part of a larger random set of northern Wisconsin walleye lakes
using a random stratified roving access design (Pollack et al. 1994; Beard et al. 1997b).
The current creel limit in Wisconsin is an aggregate panfish limit that includes bluegill,
pumpkinseed Lepomis gibbosus, black crappie, white crappie, yellow perch Perca
flavescens, green sunfish Lepomis cyanellus, warmouth Lepomis gulosus, and
orangespotted sunfish Lepomis humilis, so I used panfish rather than bluegill to estimate
reduced exploitation levels. First, the number of panfish harvested per interview was

divided by the number of anglers in the party to derive the number of panfish harvested

71

by each individual (e.g., if two anglers in a party harvested 10 panfish, they were treated
as if each angler harvested 5 panfish). Second, the total number of anglers who harvested
panfish at each level of harvest per angler was tabulated across all creel surveys (i.e., for
all creel surveys, the number of anglers who harvested 1, 2, 3, n panfish). Third, the
frequency of individual anglers who harvested each number of panfish was multiplied by
the number harvested to derive the total harvest for each harvest limit across all surveys.
For reduced creel limits of 5 and 10 fish per day, harvest numbers were reduced to 5 for
any number harvested that exceeded 5, or to 10 for any number harvested that exceeded
10, respectively, and then the total harvest was recalculated. The percent reduction in
total harvest was estimated as the reduced total harvest divided by the initial total harvest
(under the current regulation). I assumed the reduced panfish harvest would be

representative of reduced bluegill harvest.
Yield per Recruit and Number Harvested

To estimate how bluegill yield per recruit varied among ages vulnerable to harvest

and fishing mortality rates, I used a Beverton Holt yield per recruit model (Ricker 1975):

-Kt —t —2Kt -t —3Kt -t
Eng—Marrow, L36 (r 0),.3e (" o>_. (" 0) .
N0 Z Z+K Z+2K Z+3K

 

 

where yield per recruit, Y i/NO, is a function of growth and mortality estimates for each

lake class (defined above), across a range of hypothetical fishing mortality rates, F, and

ages of vulnerability to harvest, t,. The Beverton Holt yield per recruit model examines

yield as a function of growth, age at first capture, and fishing mortality to evaluate trade-

offs in biomass harvested per individual recruited, between capturing a large number of

72

young fish and capturing a smaller number of fish later in their life-span (King 1995).
For each lake class, yield per recruit was estimated for minimum length limits of 127 mm
(5 in), 152 mm (6 in), 178 mm (7 in), 203 mm (8 in), and 229 mm (9 in) and daily creel
limits of 25, 10, and 5 representing a range of exploitation rates (u = FA/Z). Quality
length for bluegill was defined as 152 mm (6 inches; Willis et al.1993), so I used 152 mm
as the lower end of the minimum length limits evaluated, but provided a smaller
minimum length limit (127 mm) as a means to assess yield per recruit and number

harvested per 100 in the absence of a minimum length limit.

The number of bluegill reaching a minimum length limit that were harvested by
anglers was estimated as the difference between the number present with and without

fishing:

Ct = N06_M(t—t0) _NOe—Z(t—t0);

where C, = the number estimated to be harvested at age t, N0 = the number of recruits

entering the population at age 0, and M and Z are as defined above. For simulating the

number of bluegill that would be harvested under each combination of minimum length

limit and daily creel limit, I set N0 = 100. The number harvested was analogous to the

biomass harvested from the yield per recruit analysis, except that the number harvested

was numbers per 100 recruits rather than grams per individual recruit.

73

Results
Growth

Bluegill grew at different rates both within and among lake classes, but generally,

average asymptotic average length (Loo) and instantaneous growth rate (K) increased as

lake size increased from small seepage lakes (Class G) to large drainage lakes (Class J).

Based on 170 lake surveys during 1948—2003, the asymptotic average length (L00) ranged

255—281 mm and did not differ among lake classes (F = 1.29; df= 3, 166; P = 0.28). The
average instantaneous growth rate (K) ranged 0.247—0.255/year and did not differ among
lake classes (F =0.01; df= 3, 166; P = 1.00; Table 13). Based on 43 lake surveys during

1944—2002, the allometric shape parameter (0) ranged 299—3. 14 and did not differ
among lake classes (F = 0.22; df= 3, 39; P = 0.88). The asymptotic average weight (Woo)

ranged 312—549 grams across lake classes, and both parameters tended to increase with

increasing lake size (Table 13).
Mortality

Based on 34 surveys during 1954—2003, total mortality of bluegill varied greatly
among lake classes, whereas natural mortality of bluegill varied little among lake classes.
Instantaneous total mortality did not differ significantly between periods, before or after
the 1998 creel-limit reduction (F = 3.68; df= 1, 163; P = 0.06), so I combined estimates
between the two periods for each lake class. Estimates of instantaneous total mortality
(Z) ranged 0.26—1.61 and did not differ among lake classes (F = 0.42; df= 3, 30,; P =
0.74). Instantaneous fishing mortality (F) estimates were lowest in small seepage lakes

and highest in small drainage lakes (Table 14). Instantaneous natural mortality (M) did

74

not vary as much as total or fishing mortality, but were highest in small and large seepage
lakes and lowest in large drainage lakes (Table 14). In all lake classes, instantaneous
fishing mortality (F) was higher than instantaneous natural mortality. Average

exploitation (u) ranged 027—047 (Table 14).
Exploitation

Panfish species were harvested from 105 lakes where creel surveys were
conducted during 1998—2003. Of 8,447 anglers interviewed, less than 3% of anglers
harvested their daily creel limit of 25 fish or more (Table 15; Figure 10). Under the
current 25 fish daily creel limit, 50,131 panfish were harvested, whereas a reduced daily
creel limit of 10 fish (a 60% reduction in creel) would reduce harvest by 21% (39,728
panfish), and a reduced daily creel limit of 5 fish (an 80% reduction in creel) would

reduce harvest by 45% (27,788 panfish).
Yield per Recruit and Number Harvested

For each combination of minimum length limit and daily creel limit, estimated
yield per recruit increased with lake size, from the smallest lake class (G) to the largest
lake class (J), and ranged 4—30 grams per recruit across regulation scenarios (Table 16;
Figure 11). Under a 25-fish daily creel limit, a 152-mm minimum length limit produced
the highest yield per recruit for small and large seepage lakes and small drainage lakes,
whereas a 178-mm minimum length limit produced the highest yield per recruit for large
drainage lakes. For all lake classes, yield per recruit declined as the daily creel limit was
reduced from 25 fish to 5 fish, which was expected because creel limits were directly
related to the fishing mortality rate. With the exception of large drainage lakes (J), higher

minimum length limits (203 or 229 mm) at the 25 daily creel limit yielded fewer grams

75

per recruit than reduced daily creel limits at smaller minimum length limits (152 or 178
mm). The number of fish harvested at or above the minimum length limit ranged 7—31
per 100 for all lake classes (Table 16; Figure 11). Higher minimum length limits under
the 25 daily creel limit often yielded similar numbers of quality-sized fish (>152
mm)than reduced daily creel limits under 152 or 178 mm minimum length limits, though

the fish would be larger under the higher minimum length limits.

For small seepage lakes (G), yield per recruit was maximized by a 25-fish daily
creel limit and a 152-mm minimum length limit (Table 16; Figure 11). However, at low
exploitation rates, such as the current lake class average of 16.7%, the difference in yield
per recruit between the 152-mm minimum length limit and the 178-mm or 203-mm
minimum length limits for the 25 fish daily creel limit was small (1-4 g). Under lower
exploitation rates (10 or 5 fish daily creel limits), the 152-mm and 178-mm minimum
length limits produced similar yields, and the 203-mm and 209-mm minimum length
limits produced the lowest yield per recruit. Numbers harvested ranged 7—1 7 fish, and
was maximized by a 25-fish daily creel limit and 152-mm minimum length limit (17 fish

per 100 recruits; Table 16; Figure 12).

For large seepage lakes (H) under a 25-fish daily creel limit, yield per recruit was
slightly higher under a 152-mm minimum length limit than under a 178-mm minimum
length limit, and a 10 fish daily creel limit and 152-mm minimum length limit produced
the second highest yield per recruit (Table 16; Figure 11). Under reduced daily creel
limits of 10 fish and 5 fish, yield per recruit was highest under a 152-mm minimum
length limit, slightly lower under a 178-mm minimum length limit, and lowest under 203-

mm and 229-mm minimum length limits. Numbers harvested ranged 9—25 fish, and was

76

maximized by a 25-fish daily creel limit and 152-mm minimum length limit (25 fish per
100 recruits; Table 16; Figure 12). A 203-mm minimum length limit and 25-fish daily
creel limit produced similar numbers of fish as a S-fish daily creel limit and 152-mm or

178-mm minimum length limits.

For small drainage lakes (1) under a 25-fish daily creel limit, yield per recruit was
slightly higher with a 152-mm minimum length limit than a 178-mm minimum length
limit, and much higher than 203-mm or 229-mm minimum length limits (Table 16;
Figure 11). Under reduced daily creel limits of 10 fish and 5 fish, yield per recruit was
highest for 152-mm and 178-mm minimum length limits. The decline in yield per recruit
was larger under higher minimum length limits (203—229 mm) than under reduced daily
creel limits (10 or 5 fish). Number harvested ranged 15—31, and was maximized by a 25-
fish daily creel limit and 178—mm minimum length limit (31 fish per 100 recruits; Table
16; Figure 12). Similarly, the decline in numbers harvested under higher minimum

length limits (203—229 mm) was greater than under lower daily creel limits (10 or 5 fish).

For large drainage lakes (J) under a 25-fish daily creel limit, yield per recruit was
maximized by 152-mm and 178-mm minimum length limits, slightly lower (3 g) for a
203-mm minimum length limit, and lowest (9 g) for a 229-mm minimum length limit
(Table 16; Figure 11). Under a 10 fish daily creel limit, a 152-mm minimum length limit
produced the highest yield per recruit, but yield per recruit under 152-mm or 178-mm
minimum length limits was slightly greater than under a 25-fish daily creel limit with
203-mm or 229-mm minimum length limits. At a S-fish daily creel limit, a 229-mm
minimum length limit produced the lowest yield per recruit, whereas 152-mm and 178-

mm minimum length limits produced the highest yield per recruit. Numbers harvested

77

were highest for a 25-fish daily creel limit and 152-mm or 178-mm minimum length
limits, and lowest for a 5-fish daily creel limit and 229-mm minimum length limit (Table
16; Figure 12). A reduced daily creel limit of 10 fish yielded similar numbers of fish as

higher minimum length limits (203 or 229 mm) at the current 25-fish daily creel limit.

Discussion
Growth
I found that von Bertalanffy asymptotic average length and instantaneous growth
rate (K) tended to increase as lake size increased from small seepage lakes to large

drainage lakes. Lake class average asymptotic length (Loo = 255—281 mm) and

instantaneous growth rate (K = 0.247—0.255/year) were similar to those reported by Beard

et al. (1997a), who found that L00 ranged between 213 and 268 mm and K ranged

between 0.24 and 0.29/year for slow and fast growing bluegill populations in northern
Wisconsin. Beard et al. (1997a) did not provide the number of lake surveys for their
estimates, or describe physical characteristics of lake types used, which may limit
comparisons to my findings obtained from 170 lake surveys in southern and northern

Wisconsin. Ranges in growth parameters for bluegill in 18 Nebraska lakes were 213—311

mm for Loo and 0.21—0.46/year for K during 1998-1999, similar to my LOO estimates, but

much larger than my range in growth (Paukert et al. 2002). However, my range in
growth parameters was the range among 4 lake class averages, whereas Paukert et al.

2002 reported ranges across lakes.

78

Mortality

My natural mortality estimates were similar among lake classes, despite
differences in mean asymptotic weights and instantaneous grth rates for bluegill
among lake classes. The Pauly equation used for estimating instantaneous natural
mortality (M) was developed through a meta-analysis of 175 different freshwater and
marine fish stocks, including 84 species, though bluegill was not among the species
included in the analysis (Pauly 1980). However, Beamesderfer and North (1995) found
that natural mortality was correlated with latitude, air temperature, and degree-days for
another centrarchid, largemouth bass, across 40 populations in North America. My use
of statewide mean annual air temperature averaged across a 60-year period likely
diminished the sensitivity of the estimate, and therefore could be improved by estimating
regional mean annual air temperature or by measuring surface water temperature on
lakes. Another approach would be to estimate yield across a range of natural mortality
rates (e.g., Maceina et al. 1998). My estimates of M were similar among lake classes, so
natural mortality was nearly constant and therefore did not strongly influence predicted

differences in yield per recruit or numbers harvested among lake classes.

My estimates of average annual mortality obtained from catch curves were lower
than others reported in the literature. For example, Ricker (1945) studied three Indiana
bluegill populations and estimated annual mortality between 60% and 87%, whereas my
annual mortality estimates were between 43% and 61%. Mean annual air temperature in
Wisconsin is likely lower than in Indiana, so my estimates of natural mortality, and
presumably annual mortality, should be lower than for more southerly states like Indiana.

Snow (1978) reported bluegill annual mortality in Murphy Flowage Wisconsin to be

79

59%, and Paukert et al. (2002) reported annual mortality rates in 18 Nebraska lakes
between 15 and 53%, similar to my estimates. Mortality estimates derived from catch
curves may be inaccurate when survival rate is not constant with age, the population is
not sampled randomly, or recruitment fluctuates greatly (Ricker 1975; Allen 1997).
However, Maceina et al. (1998) compared total annual mortality estimated from catch
curves to four other methods for estimating total annual mortality and found that all five

methods provided consistent estimates of total annual mortality for crappie in Weiss

Lake, Alabama.
Exploitation

I found that exploitation rates ranged 17—39% and increased with lake size for
small and large seepage lakes and small drainage lakes, which has also been shown for
other fisheries. For example, angler harvest was positively related to lake size for lake
trout Salvelinus namaycush namaycush (Goddard et al. 1987), walleye Stizostedion
zander (Baccante and Colby 1996), and panfish (Wisconsin Department of Natural
Resources unpublished data), so lake classes that are descriptive of lake size may be
useful for evaluating harvest regulation strategies that depend on exploitation rates.
Exploitation rates in my study were similar to those reported for Indiana lakes (u = 20-
45%; Ricker 1945) and Murphy Flowage Wisconsin (u = 3—35%; Snow 1978). In a
literature review of 46 different studies, Coble (1988) estimated an average bluegill

exploitation rate of 27%.
Yield per Recruit and Number Harvested

High minimum length limits for bluegill generally reduced yield per recruit and

number harvested, similar to results of other studies of panfish. Paukert et al. (2002)

80

evaluated a ZOO-mm length limit to increase bluegill size structure for Nebraska lakes and
found that anglers would have to sacrifice yield and number harvested to gain a small
increase in numbers of fish longer than 8 inches. Minimum length limits were more
successful for increasing crappie size structure in lakes where growth was high, natural
mortality was low, and angler exploitation was high (Allen and Miranda 1995). In
Wisconsin lakes, natural mortality was not high and did not differ greatly among lake
classes. Growth rates did not differ greatly among lake classes and may not have been
high enough overall for a minimum length limit to drastically increase numbers of large
fish in most lakes. Large seepage and small drainage lakes had the highest exploitation
rates, so may be the best candidate lakes for minimum length limits, but exploitation rates
were not as high as those reported by Allen and Miranda (1995) for southern crappie
populations. In fact, current exploitation rates may not be sufficiently high enough to
maximize yield per recruit or numbers harvested for any of the lake classes I examined

for Wisconsin.

Reduced daily creel limits at the lowest minimum length limit often produced
similar or greater yield per recruit than high minimum length limits (e.g. 229 mm) at a
higher daily creel limit, indicating that at least in some cases, reduced daily creel limits
may be more effective for increasing yield per recruit of quality-sized or larger bluegill
than high minimum length limits. Similarly, numbers harvested per 100 recruits under a
5 fish daily creel limit and 152-mm minimum length limit were greater than numbers
harvested under a 25 fish daily creel limit and 229-mm minimum length limit for three of

the four lake classes. However, reducing the daily creel limit means reducing

81

exploitation from current levels, which generally did not maximize yield per recruit or

numbers harvested.

Reduced daily creel limits in combination with minimum length limits may be
effective for decreasing fishing mortality on spawning adult bluegill, which may be more
necessary in small seepage lakes than in large drainage lakes. Daily creel limit
reductions are often proposed for protection of bluegill in small lakes, which are assumed
to be more vulnerable to over-exploitation in Wisconsin. If exploitation is excessive in
small lakes, then growth is likely to be high and age at maturity is likely to be low,
especially in seepage lakes where immigration is not possible. I did not see evidence of
high exploitation in two small seepage lakes for which total mortality could be estimated,
and age to reach 152-mm was higher (3.6 years) than all other lake classes. However, my
sample size was small, so more detailed studies of small lakes should identify

representative ranges of growth and mortality to justify a reduced daily creel limit.

Changes in bluegill angling regulations require support from the angling public.
For example, Reed and Parsons (1999) found that most anglers (61%) on four Minnesota
lakes supported a proposed daily creel limit reduction from 30 to 20 fish, but would not
support further reduction. My results suggest that reduced creel limits (from the current
25 to 10 or 5 fish) would not increase yield per recruit or numbers harvested per 100
recruits, and that reduced daily creel limits may not reduce harvest as much as anglers
perceive. Reed and Parsons (1999) also found that 56% of anglers opposed the idea of
bluegill length limits, but that the most acceptable minimum length limit was 6 inches.
Among the four minimum length limits 1 evaluated, a 152-mm (6-inch ) minimum length

limit generally increased yield per recruit and numbers harvested per 100 recruits at

82

current levels of estimated exploitation in Wisconsin. Paukert et al. (2002) reported that
67% of anglers interviewed on a Nebraska lake during 1998—2001 were there to harvest
bluegill to eat and were not concerned with catching large bluegill. Therefore, increasing
size structure may not be a concern to most bluegill anglers. However, in Minnesota,
anglers interested in catching larger bluegill may make more trips if size structure
increased, suggesting that some anglers are interested in catching quality-sized bluegill
(Reed and Parsons 1988). In general, I found that benefits of minimum length limits

(e. g., higher yield per recruit or numbers available for harvest) were more pronounced at
higher levels of exploitation, as was found by Allen and Miranda (1995) for crappie in

North America.
Management Implications

My results provide ranges of bluegill growth, natural mortality, and fishing
mortality for lake classes that may be used to evaluate future regulation proposals. My
estimates of yield per recruit and numbers harvested may be used to predict the effects of
proposed regulation changes on angling harvest of bluegill. My estimates of yield per
recruit and numbers harvested should be refined as more data become available, to
improve our understanding of the trade-offs between growth and mortality in managing

bluegill fisheries.

The impact of statewide angling regulations such as creel and size limits on fish
populations requires analyses on a statewide scale. Ideally, information used in modeling
angling regulations would be collected using standard methods and a statistically rigorous
study design. More frequently, the only long-term information on a large spatial scale is

collected using non-standard methods and non-random sampling, which often minimizes

83

the utility of the information (Bonar and Hubert 2002; Hayes et al. 2003). However,
using historic fisheries survey and creel survey data, Beard and Kampa (1999) were able
to convincingly demonstrate a decline in the mean size and proportion of large bluegill in
fyke net surveys in Wisconsin during 1967—1991. While these historic datasets are often
fraught with challenges, large statewide datasets may offer the only information
available, and through their use, may reveal obvious shortcomings in statewide sampling
or data capture efforts, as was the case with this analysis. For example, most historic
age-length information was limited to summary information, such as mean length at age
summaries, rather than individual fish lengths and ages. Consequently, I was only able to
construct age frequencies and estimate mortality on 34 lakes, even though I queried all
existing databases over a 60-year period. Individual age-length data probably exist for
more lakes in field office files, but unless such data are archived in a centralized
database, it remains largely inaccessible and therefore cannot be used for retrospective
analyses. Estimating the correct range in mortality rates is critical when evaluating
regulations. In addition, I was able to estimate von Bertalanffy growth parameters for
170 surveys from summary, length-stratified, sub-sampled mean lengths at age, but these
estimates may be biased because length-stratified sub—sampled mean lengths at age do not
constitute a random sample of lengths at age (Betolli and Miranda 2001, Chapter 1).
Complete raw data, rather than summary data, should be entered into centralized

statewide databases.

84

Table 12. Lake classification criteria for predicting bluegill mean length at age 4
(Chapter 2).

 

 

Connectivity Surface Area Lake
Category (ha) Class
Seepage < 12 G
3 12 H
Drainage < 446 I
3 446 J

 

85

Table 13. Minimum, maximum, and average Von Bertalanffy grth parameters
estimated for bluegill in four classes of Wisconsin inland lakes surveyed during 1948—

2003, where N = number of lake surveys, L00 = the asymptotic average length (m), K =

the instantaneous rate at which L, approaches L.Jo (l/year), t0 = the hypothetical age at
which length is zero (years), and average weight—length coefficients estimated for
bluegill in four classes of Wisconsin inland lakes surveyed during 1944—2002, where N =

number of lake surveys, Loge (a) = a scaling constant, and ,6 = the allometric shape
parameter. W<30 = the asymptotic average weight (grams) was estimated from average-lake

class Loo, a, and ,6 so there were no minumum and maximum values.

 

 

Lake

Class Parameter N Min Max Average

G LOC 10 163 365 269
K 10 0.074 0.764 0.247
to 10 -2.673 1.573 0.085
Loge (a) 2 -10.985 -10.982 -10.984
6 2 2.979 3.001 2.990
W00 2 312

H LOO 66 158 361 255
K 66 0.069 0.678 0.250
to 66 -2.942 1.333 -0.227
Loge (a) 15 -13.346 -8.618 -11.291
b 15 2.520 3.455 3.09
W00 15 342

I LoO 80 162 382 269
K 80 0.084 0.715 0.251
In 80 -3.049 1.950 -0.067
Loge (a) 19 —14.393 -7 .739 -11.434
6 19 2.417 3.728 3.137
W00 19 451

J L00 14 210 354 281
K 14 0.134 0.583 0.255
t0 14 -1.823 1.067 -0.532
Loge (a) 7 -13.027 -9.349 -11.304
b 7 2.733 3.457 3.123
W... 7 549

 

86

Table 14. Average mortality estimates by lake class from 34 Wisconsin lakes during
1954—2003. The table includes the average, minimum and maximum instantaneous total

mortality (Z) from catch curves of the age frequency, survival (S) estimated from e'Z, total

annual mortality (A) estimated from 1 — S, natural mortality (M) estimated from Pauly’s
equation (Pauly 1980), instantaneous fishing mortality (F) estimated from Z - M, and
exploitation (u) estimated from F *A/Z.

 

 

Lake Max

Class Z Min Z Z S A M F u
G 0.566 0.256 0.876 0.568 0.432 0.347 0.219 0.167
H 0.792 0.423 1.253 0.453 0.547 0.347 0.445 0.308
I 0.932 0.289 1.606 0.394 0.606 0.340 0.592 0.385
J 0.670 0.564 0.777 0.512 0.488 0.338 0.332 0.242

 

87

Table 15. Estimated harvest under 25, 10, and 5 daily creel limits estimated from 8,447
creel survey interviews where panfish species were harvested from 106 northern
Wisconsin lakes during 1998—2002. Because angler interviews were divided by the
number in the angling party to arrive at number per angler, there were instances where
between 0—1 fish were harvested (i.e., Number of Fish Harvested = 0).

 

 

Number 25 I0 5
of Fish Frequency Harvest Harvest Harvest
Harvested of anglers Percent Limit Limit Limit

0 900 0.1065 900 900 900
1 1234 0.1461 1234 1234 1234
2 1315 0.1557 2630 2630 2630
3 606 0.0717 1818 1818 1818
4 754 0.0893 3016 3016 3016
5 452 0.0535 2260 2260 2260
6 542 0.0642 3252 3252 2710
7 272 0.0322 1904 1904 1360
8 398 0.0471 3184 3184 1990
9 210 0.0249 1890 1890 1050
10 356 0.0421 3560 3560 1780
11 99 0.0117 1089 990 495
12 221 0.0262 2652 2210 1105
13 101 0.0120 1313 1010 505
14 108 0.0128 1512 1080 540
15 131 0.0155 1965 1310 655
16 102 0.0121 1632 1020 510
17 61 0.0072 1037 610 305
18 69 0.0082 1242 690 345
19 45 0.0053 855 450 225
20 112 0.0133 2240 1120 560
21 38 0.0045 798 380 190
22 38 0.0045 836 380 190
23 34 0.0040 782 340 170
24 38 0.0045 912 380 190
25 161 0.0191 4025 1610 805
26 16 0.0019 416 160 80
27 3 0.0004 81 30 15
28 3 0.0004 84 30 15
29 1 0.0001 29 10 5
30 8 0.0009 240 80 40
31 1 0.0001 31 10 5

88

 

32 3 0.0004 96 30 15
33 3 0.0004 99 30 15
37 1 0.0001 37 10 5
40 4 0.0005 160 40 20
41 2 0.0002 82 20 10
43 1 0.0001 43 10 5
45 1 0.0001 45 10 5
50 3 0.0004 150 30 15
Total 8447 1 50131 39728 27788

 

89

Table 16. Yield per recruit (g) and number harvested per 100 bluegill under varying
minimum length and daily creel limits for Wisconsin lakes. Maximum yield per recruit
and number harvested per 100 for the minimum length limits and daily creel limits
evaluated for each lake class are indicated in hold. The 127 mm length limit represents
the yield per recruit and number harvested in the absence of a minimum length limit.
Age is the estimated time (years) to reach each minimum length limit.

 

 

 

Yield Jer Recruit Number Harvested per 100
Lake 25 10 5 25 10
Class Length Age Creel Creel Creel Creel Creel 5 Creel
G 127 2.7 13.8 12.5 10.3 18 15 11
152 3.5 13.5 12.0 9.8 16 14 10
178 4.5 12.2 10.7 8.6 14 12 9
203 5.8 9.6 8.4 6.7 10 9 7
229 7.8 6.0 5.2 4.1 6 5 4
H 127 2.5 18.7 17.9 16.1 27 24 19
152 3.4 18.9 17.7 15.5 23 20 17
178 4.6 17.0 15.7 13.4 17 16 13
203 6.1 12.9 11.8 10.0 10 10 9
229 8.8 6.4 5.8 4.8 4 4 4
I 127 2.5 25.9 25.4 23.9 33 29 24
152 3.3 27.4 26.4 24.1 28 25 21
178 4.3 26.6 25.1 22.4 21 20 17
203 5.6 22.5 21.0 18.4 14 14 12
229 7.5 15.0 13.9 12.0 8 7 7
J 127 1.8 29.8 28.2 24.7 24 21 16
152 2.5 30.6 28.4 24.4 23 20 15
178 3.4 29.4 26.9 22.7 19 17 14
203 4.5 25.7 23.3 19.3 15 13 11
229 6.1 19.1 17.1 14.0 10 9 8

 

90

 

Figure 9. Locations of surveys used in this study. Different symbols represent each type
of estimate (mortality estimates (triangles), creel surveys (circles), weight-length
estimates (asterisks), and von Bertalanffy growth parameter estimates from mean lengths
at age (plus).

91

1400 ‘
1200
1000

800

600

0. , I I I'll IIIIJ [Ill
5 1O 15 20

O 25

Number Harvested

Frequency

 

 

Figure 10. The frequency of anglers harvesting between 1 and 25 panfish in Wisconsin
lakes during 1998—2003.

92

Yield/Recruit (grams)

Yield/Recruit (grams)

4o-
35-
30-
25-

 

 

 

 

0.0 0.1 0.2 0.3 0.4 0.5 0.6
Exploitation Rate (u)

40,
35‘
30-
25-
2o,
15-
1o-

5.

 

 

 

0

0.0 0.1 0.2 0.3 0.4 0.5 0.6

I

Exploitation Rate (u)

40-
35-

Yield/Recruit (grams)
N
O

 

 

__,/

__,-

 

0.0 0.1 0.2 0.3 0.4 0.5 0.6
Exploitation Rate (u)

40-

35-

Yield/Recruit (grams)

 

 

 

I

I

I

I

I

I

0.0 0.1 0.2 0.3 0.4 0.5 0.6

Exploitation Rate (u)

Figure 1 1. Yield per recruit (g) by lake class under 5 minimum length limit scenarios
(127 mm (heavy ——), 152 mm (———), 178 mm (— — —), 203 mm (- - -), and 229 mm (- - — ).
Vertical solid line indicates average exploitation rate (u) for the lake class. Lake class G
upper left), lake class H (upper right). lake class 1 (lower left), and lake class J (lower

right).

93

401 40-
351
304
25-

20-

Number Harvested
Number Harvested

 

 

 

 

0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.0 0.1 0.2 0.3 0.4 0.5 0.6
Exploitation Rate (u) Exploitation Rate (u)

Number Harvested
Number Harvested

   

 

 

0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.0 0.1 0.2 0.3 0.4 0.5 0.6
Exploitation Rate (u) Exploitation Rate (u)

Figure 12. Number harvested per 100 by lake class under 4 minimum length limit
scenarios (127 mm (heavy —),152 mm (—), 178 mm (— — —), 203 mm (- - -), and 229
mm (— - — ). Vertical solid line indicates average exploitation rate (11) for the lake class.
Lake class G (upper left), lake class H (upper right), lake class I (lower left), and lake
class J (lower right).

94

SUMMARY AND CONCLUSIONS

The goal of this research was to explore the use of historic fisheries survey
information to explain variability in fish grth rates in relation to physical and
landscape characteristics of Michigan and Wisconsin inland lakes. I focused on three
main objectives: (1) to determine if von Bertalanffy growth parameters and mean lengths
at age estimated from length-stratified sub-samples were biased in relation to parameters
estimated from full samples for game fishes surveyed routinely in Michigan lakes, (2) to
determine if an existing lake classification scheme explained variability in fish species
richness and mean length at age from historic survey data collected on Michigan and
Wisconsin lakes, and (3) to demonstrate how lake classification may be used in
management by estimating how bluegill yield per recruit and number harvested varied as
a function of minimum age entering the fishery and fishing mortality rate among four

classes of Wisconsin inland lakes.

In Chapter 1, 1 found that historic length-stratified sub-sample summary data
could be used to compare the average growth of fishes across Michigan lakes. Bluegill,
largemouth bass, and yellow perch mean lengths at age 4 estimated from length-stratified
sub-samples were generally not biased compared to samples. Regression tests showed
that length-stratified sub-sample von Bertalanffy growth parameter estimates were
generally linearly related to estimates from samples, but the magnitude of the variability
between sub-sample and sample across lakes was large for all species tested. Large
variation in von Bertalanffy parameter estimates from length-stratified sub-sample
summary data may limit their utility for use in among-lake comparisons. Historic growth

data sets that lack complete sampling information (e.g., sample length frequencies) can

95

still provide useful information for among lake comparisons, but mean length at age may
be a better growth index than von Bertalanffy growth parameters. Based on these

conclusions, I used mean lengths at age 4 as my index of growth for Chapter 2.

In Chapter 2, I found that an existing lake classification scheme consisting of 6
lake classes differing in landscape position, connectivity, and maximum depth explained
variation in fish species richness and mean lengths of age 4 bluegill, largemouth bass, and
walleye, but not all lake class pairs differed and overall, little variation was explained by
the classification. Seepage lake classes, regardless of depth category differed from
lowland deep drainage lakes for most responses tested, so drainage type was retained as I
refined the classification. Additional map-based variables related to temperature and
habitat area changed the classification in different ways for each response variable tested.
Based on regression tree results, lake surface area was an important predictor of fish
species richness, and mean lengths of age 4 bluegill and largemouth bass, whereas
latitude was an important predictor of mean lengths of age 4 yellow perch and walleye.
Regression tree results revealed species-specific separation points for identifying
categories of lake size (large and small lakes), and for stratifying the region into northern
and southern zones. The revised classifications may have the potential to serve as a
statewide framework for management recreational fisheries, based on patterns of
ecological similarity among lakes. I tested this claim in Chapter 3, where I used lake
classification criteria specific to bluegill and developed in Chapter 2, as the basis for

angling regulation simulations.

Finally, in Chapter 3, to evaluate the lake classification scheme developed in

Chapter 2 for use in management, I determined expected ranges of bluegill growth and

96

mortality for different lake classes in Wisconsin from existing survey data to determine
the best combination of minimum length limits (152, 178, 203, and 229 mm) and daily
creel limits (25, 10, 5) to maximize yield per recruit and number of bluegill harvested per
100 individuals for each lake class. For each combination of minimum length limit and
daily creel limit, yield per recruit increased with lake size, from the smallest lake class to
the largest lake class, and ranged 4—30 grams per recruit among regulation scenarios.
Under a 25-fish daily creel limit, a 152-mm minimum length limit produced the highest
yield per recruit for small and large seepage lakes and small drainage lakes, whereas a
178-mm minimum length limit produced the highest yield per recruit for large drainage
lakes. My results provide ranges of bluegill growth, natural mortality, and fishing
mortality for lake classes that may be used to evaluate future regulation proposals and to
identify candidate lakes for experimental regulations. With additional data, my estimates
of yield per recruit and numbers harvested could be refined to improve our understanding

of the trade-offs between growth and mortality in managing bluegill fisheries.

In conclusion, the integration of historic and current fisheries data provided a
means to expand the spatial and temporal scale of my research. By developing large
datasets from two states for an approximate sixty year period, I was able to increase the
statistical power to detect patterns amidst the many sources of error and potential bias.
Examining potential bias in historic datasets continues to be important, but through the
use of historic data, we can illucidate weaknesses in sampling strategies and develop

research questions that may be more formally tested.

97

LIST OF REFERENCES

98

LIST OF REFERENCES

Allen, M. S. 1997. Effects of variable recruitment on catch-curve analysis for crappie
populations. North American Journal of Fisheries Management 17:202-205.

Allen, M. S., and L. E. Miranda. 1995. An evaluation of harvest restrictions in managing
crappie fisheries. North American Journal of Fisheries Management 15:766—772.

Baccante, D. A. and P. J. Colby. 1996. Harvest, density and reproductive characteristics
of North American walleye populations. Annales Zoologici F ennici 33:601-615.

Baccante, D. A., and D. M. Reid 1988. Fecundity changes in two exploited walleye
populations. North American Journal of Fisheries Management 8:199—209.

Barbour, C. D., and J. H. Brown. 1974. Fish species diversity in lakes. American
Naturalist 108:473—489.

Beacharn, T. D. 1983. Variability in median size and age at sexual maturity of Atlantic
cod (Gadus morhua) on the Scotian Shelf in the Northwest Atlantic Ocean.
Fishery Bulletin 81 :303—321.

Beamesderfer, R. C. P., and J. A. North. 1995. Growth, natural mortality, and predicted
response to fishing for largemouth bass and smallmouth bass populations in North
America. North American Journal of Fisheries Management 15:688—704.

Beamish, R. J ., and W. N. McFarlane. 1983. The forgotten requirement for age
validation in fisheries biology. Transactions of the American Fisheries Society
112: 735—743.

Beard, T. D. Jr., S. P. Cox, and S. R. Carpenter. 2003. Impacts of daily bag limits on
angler effort in Wisconsin walleye lakes. North American Journal of Fisheries
Management 23: 1283—1293.

Beard, T. D. Jr., M. T. Drake, J. E. Breck, and N. A. Nate. 1997a. Effects of simulated
angling regulations on stunting in bluegill populations. North American Journal
of Fisheries Management 17:525—532.

Beard, T. D. Jr., S. W. Hewett, Q. Yang, R. M. King, and S. J. Gilbert. 1997b.
Prediction of angler catch rates based on walleye population density. North
American Journal of Fisheries Management 17: 621—627.

Beard, T. D. Jr., and J. M. Kampa. 1999. Changes in bluegill, black crappie, and yellow
perch populations in Wisconsin during 1967—1991. North American Journal of
Fisheries Management 19: 103 7—1 043.

Becker, G. C. 1983. Fishes of Wisconsin. The University of Wisconsin Press, Madison.

99

Betolli, P. W., and L. E. Miranda. 2001. Cautionary note about estimating mean length
at age with sub-sampled data. North American Journal of Fisheries Management
21: 425—428.

Beverton, R. J. H. 1987. Longevity in fish: some ecological and evolutionary
considerations. Pages 161—185 in A. D. Woodhead and K. H. Thompson, editors.
Evolution of longevity in animals: a comparative approach. Plenum, New York.

Bonar, S. A., and W. A. Hubert. 2002. Standard sampling of inland fish: benefits,
challenges, and a call for action. Fisheries 27:10—16.

Breiman, L., J. H. Friedman, R. A. Olshen, and C. J. Stone. 1993. Classification and
Regression Tress. Chapman and Hall/CRC, New York.

Campana, S. E. 2001. Accuracy, precision and quality control in age determination,
including a review of the use and abuse of age validation methods. Journal of
Fish Biology 59:197—242

Carpenter, S. R. and J. F. Kitchell, editors. 1993. The trophic cascade in lakes.
Cambridge University Press, Cambridge UK.

Chamov, E. L. 1993. Life history invariants: some explorations of symmetry in
evolutionary ecology. Oxford University Press, New York.

Chen, Y., D. A. Jackson, and H. H. Harvey. 1992. A comparison of von Bertalanffy and
polynomial functions for modeling fish growth data. Canadian Journal of
Fisheries and Aquatic Sciences 49:1228—1235.

Cheruvelil, K. S. 2004. Michigan Landscape Context GIS Database, Michigan State
University.

Christie, G. C., and H. A. Regier. 1988. Measures of optimal thermal habitat and their
relationship to yields of four commercial fish species. Canadian Journal of
Fisheries and Aquatic Sciences 45:301—3 14.

Coble, D. W. 1988. Effects of angling on bluegill populations: management
implications. North American Journal of Fisheries Management 8:277—283.

Colby, P. J. , R. E. McNicol, and R. A. Ryder. 1979. Synopsis of biological data on
walleye, Stizostedion v. vitreum (Mitchell 1818). F AO Fisheries Synopsis 119.

Colby, P. J. and S. J. Nepszy. 1981. Variation among stocks of walleye (Stizostedion
vitreum vitreum): management implications. Canadian Journal of Fisheries and
Aquatic Sciences 38:1818—1831.

Cook, M. F., T. J. Goeman, P. J. Radomski, J. A. Younk, and P. C. Jacobson. 2001.
Creel limits in Minnesota: a proposal for change. Fisheries 26:19—26.

100

Crowder, L. B. and J. J. Magnuson. 1983. Cost benefit analysis of temperature and food
resource use: a synthesis with examples from fishes. Pages 189—221 in W. P.
Aspey and S. I. Lustick, editors. Behavioral energetics: the cost of survival in
vertebrates. Ohio State University Press, Columbus.

Diana, J. S. 1995. Biology and ecology of fishes. Cooper Publishing Group, Carmel,
Indiana.

Diana, J. S. 1983. Growth, maturation, and production of northern pike in three
Michigan lakes. Transactions of the American Fisheries Society 112238—46.

Dolman, W. B. 1990. Classification of Texas reservoirs in relation to limnology and fish

community associations. Transactions of the American Fisheries Society
119:511—520.

Drake, M. T., J. E. Claussen, D. P. Philipp, and D. L. Pereira. 1997. A comparison of
bluegill reproductive strategies and growth among lakes with different fishing
intensities. North American Journal of Fisheries Management 17:496—507.

Emmons, E. E., M. J. Jennings, and C. Edwards. 1999. An alternative classification
method for northern Wisconsin lakes. Canadian Journal of Fisheries and Aquatic
Sciences 56:661—669.

Fee, E. J ., R. E. Hecky, S. E. M. Kasian, and D. R. Cruikshank. 1996. Effects of lake
size, water clarity, and climatic variability on mixing depths in Canadian Shield
lakes. Limnology and Oceanography 41(5):912—920.

Fee, E. J ., R. E. Hecky, G. W. Regehr, L. L. Hendzel, and P. Wilkinson. 1994. Effects
of lake size on nutrient availability in the mixed layer during summer
stratification. Canadian Journal of Fisheries and Aquatic Sciences 51:2756—2768.

Fee, E. J ., J. A. Shearer, E. R. DeBruyn, and E. U. Schindler. 1992. Effects of lake size
on phytoplankton photosynthesis. Canadian Journal of Fisheries and Aquatic
Sciences 49:2445—2459.

Fomey, J. L. 1977. Evidence of inter- and intraspecific competition as factors regulating
walleye (Stizostedion vitreum vitreum) biomass in Oneida Lake, New York.
Journal of the Fisheries Research Board of Canada 34:1812-1820.

F ridriksson, A. 1934. On the calculation of age distribution within a stock of cod by
means of relatively few age determinations as a key to measurements on a large
scale. Journal du Conseil international pour l’Exploration de la Mer 86: 1—14.

Gallucci, V. F., and T. J. Quinn 11. 1979. Reparameterizing, fitting, and testing a simple
growth model. Transactions of the American Fisheries Society 108:14—25.

Goddard, C. I., D. H. Loftus, J. A. MacLean, C. H. Oliver, and B. J. Shuter. Evaluation
of the effects of fish community structure on observed yields of lake trout

101

(Salvelinus namaycush). Canadian Journal of Fisheries and Aquatic Sciences
44:239—248.

Goedde, L. E., and D. W. Coble. 1981. Effects of angling on a previously fished and
unfished warmwater fish community in two Wisconsin lakes. Transactions of the
American Fisheries Society 110:594—603.

Goodyear, C. P. 1995. Mean size at age: an evaluation of sampling strategies with
simulated red grouper data. Transactions of the American Fisheries Society
124:746—755.

Haddon, M. 2001. Modelling and quantitative methods in fisheries. Chapman and
Hall/CRC, Boca Raton, Florida.

Hale, R. S., M. E. Lundquist, R. L. Miller, and R. W. Petering. 1999. Evaluation of a
254-mm minimum length limit on crappies in Delaware Reservoir, Ohio. North
American Journal of Fisheries Management 19:804—814.

Hansen, M. J. 1989. A walleye population model for setting harvest quotas. Wisconsin
Department of Natural Resources, Fish Management Report 143, Madison.

Hansen, M. J., M. A. Bozek, J. R. Newby, S. P. Newman, M. D. Staggs. 1998. Factors
affecting recruitment of walleyes in Escanaba Lake, Wisconsin, 1958—1995.
North American Journal of Fisheries Management 18:764-774.

Hawkins, C. P., R. H. Norris, J. Gerritsen, R. M. Hughes, S. K. Jackson, R. K. Johnson,
and R. J. Stevenson. 2000. Evaluation of the use of landscape classifications for

the prediction of freshwater biota: synthesis and recommendations. Journal of
the North American Benthological Society 19:541-556.

Hayes, D., E. Baker, R. Bednarz, D. Boreson, Jr., J. Braunscheidel, J. Breck, M.
Bremigan, A. Hanington, R. Hay, R. Lockwood, A. Nuhfer, J. Schneider, P.
Seelbach, J. Wybrant, and T. Zorn. 2003. Developing a standardized sampling
program: the Michigan experience. Fisheries 28: 1 8—25.

Healey, MC. 1980. Growth and recruitment in experimentally exploited lake Whitefish
(Coregonus clupeaformis) populations. Journal of the Fisheries Research Board
of Canada 37:255—267.

Hegge, O., B. K. Dervo, and J. Skurdal. 1991. Age and size at sexual maturity of
heavily exploited Arctic char and brown trout in Lake Atnsjo, southeastern
Norway. Transactions of the American Fisheries Society 120: 141—149.

Hershey, A. E., G. M. Gettel, M. E. McDonald, M. C. Miller, H. Mooers, W. J. O’Brien,
J. Pastor, C. Richards, and J. A. Schuldt. 1999. A geomorphic-trophic model for
landscape controls of arctic lake food webs. BioScience 49:887—897.

102

Humphrys, C. R. and J. Colby. 1962. Summary of acreage analysis charts. Michigan
State University, Department of Resource Development Water Bulletin No. 15,
East Lansing.

Jennings, M. J ., J. E. Claussen, and D. P. Philipp. 1997. Effects of population size
structure on reproductive investment of male bluegill. North American Journal of
Fisheries Management 17:516—524.

Johnson, M. G., J. H. Leach, C. J. Minns, and C. H. Oliver. 1977. Limnological
characteristics of Ontario lakes in relation to associations of walleye (Stizostedion
vitreum), northern pike (Esox Iucius), lake trout (Salvelinus namaycush), and

smallmouth bass (Micropterus dolomieui). Journal of the Fisheries Research
Board of Canada 34:1592—1601.

Ketchen, K. 1949. Stratified subsampling for determining age distributions.
Transactions of the American Fisheries Society 79:205—212.

Kimura, D. K. 1977. Statistical assessment of the age-length key. Journal of the
Fisheries Research Board of Canada 34:317—324.

King, M. 1995. Fisheries Biology, Assessment, and Management. Blackwell Science,
Cambridge.

Kratz, T. K., K. E. Webster, C. J. Bowser, J. J. Magnuson, and B. J. Benson. 1997. The
influence of landscape position on lakes in northern Wisconsin. Freshwater
Biology 37:209—217.

Lampert, W., and U. Sommer. 1997. Limnoecology: the ecology of lakes and streams.
Oxford University Press, New York.

Lester, N. P., T. R. Marshall, K. Armstrong, W. I. Dunlop, B. Ritchie. 2003. A broad-
scale approach to management of Ontario’s recreational fisheries. North

American Journal of Fisheries Management 23:1312—1328.

Levin, S. A. 1992. The problem of pattern and scale in ecology: The Robert H.
MacArthur Award Lecture. Ecology 73:1943-1967.

Lillie, R. A. and J. W. Mason. 1983. Limnological characteristics of Wisconsin lakes.
Wisconsin Department of Natural Resources Technical Bulletin 138, Madison.

Ludsin, S. A., and D. R. DeVries. 1997. First-year recruitment of largemouth bass: the
inter-dependency of early life stages. Ecological Applications 7:1024—1038.

MacArthur, R. H. and E. O. Wilson. 1967. The theory of island biogeography.
Princeton University Press, Princeton, New Jersey.

103

Maceina, M. J ., O. Ozen, M. S. Allen, and S. M. Smith. 1998. Use of equilibrium yield
models to evaluate length limits for crappies in Weiss Lake, Alabama. North
American Journal of Fisheries Management 18:854—863.

MacLean, J ., and J. J. Magnuson. 1977. Species interactions in percid communities.
Journal of the Fisheries Resources Board of Canada 34:1941—1951.

Madenjian, C. P., and S. R. Carpenter. 1991. Individual-based model for grth of
young-of-the-year walleye: a piece of the recruitment puzzle. Ecological
Applications 1:268—279.

Madenjian, C. P., J. T. Tyson, R. L. Knight, M. W. Kershner, and M. J. Hansen. 1996.
First-year growth, recruitment, and maturity of walleyes in Western Lake Erie.
Transactions of the American Fisheries Society 125: 821—830.

Magnuson, J. J ., L. B. Crowder, and P. A. Medvick. 1979. Temperature as an ecological
resource. American Zoologist 19:331—343.

Magnuson, J. J ., and T. K. Kratz. 2000. Lakes in the landscape: approaches to regional
limnology. Verh. Intemat. Verein. Limnol. 27:74—87.

Magnuson, J. J., W. M. Tonn, A. Banerjee, J. Toivonen, O. Sanchez, M. Rask. 1998.
Isolation vs. extinction in the assembly of fishes in small northern lakes. Ecology
79:2941—2956.

Marshall, T. R., and P. A. Ryan. 1987. Abundance patterns and community attributes of
fishes relative to environmental gradients. Canadian Journal of Fisheries and
Aquatic Sciences 44(Supplement 2):l98—215.

Matuszek, J. E. and G. L. Beggs. 1988. Fish species richness in relation to lake area, pH,
and other abiotic factors in Ontario lakes. 45: 1931—1941.

McClanahan, D. R. 2003. Use of a mail survey to estimate statewide Wisconsin angler
harvest and effort. Master’s thesis, University of Wisconsin - Stevens Point.

Minns, C. K. 1989. Factors affecting fish species richness in Ontario lakes.
Transactions of the American Fisheries Society 118:533—545.

Minns, C. K. and J. E. Moore. 1990. Factors limiting the distributions of Ontario’s
freshwater fishes: the role of climate and other variables, and the potential
impacts of climate change. Pages 137—160 in R. J. Beamish, editor. Climate
change and northern fish populations. Canadian Special Publication in Fisheries
and Aquatic Sciences 121.

Miranda, L. E., and R. J. Muncy. 1987. Recruitment of young-of-the-year largemouth
bass in relation to size structure of the parent stock. North American Journal of
Fisheries Management 7: 131—137.

104

Morgan, J. N., and J. A. Sonquist. 1963. Problems in the analysis of survey data, and a
proposal. Journal of the American Statistical Association 58:415—434.

Murphy, B. R., and D. W. Willis. 1996. Fisheries techniques. American Fisheries
Society, Bethesda.

Nate, N. A., M. A. Bozek, M. J. Hansen, S. W. Hewett. 2000. Variation in walleye
abundance with lake size and recruitment source. North American Journal of
Fisheries Management 20: 107-1 14.

Nate, N. A., M. A. Bozek, M. J. Hansen, S. W. Hewett. 2001. Variation in walleye
abundance with lake size and limnological variables. North American Journal of
Fisheries Management 21 :441-447.

National Oceanic and Atmospheric Administration, 2004. (http://climvis.ncdc.nogagov/I

Neter, J., M. H. Kutner, C. J. Nachtsheim, and W. Wasserrnan. 1996. Applied linear
statistical models, fourth edition. Irwin, Chicago.

Newman, S. P., and M. H. Hoff. 2000. Evaluation of a 16-inch minimum length limit
for smallmouth bass in Pallette Lake, Wisconsin. North American Journal of
Fisheries Management 20:90—99.

Noble, R. L., and T. W. Jones. 1993. Managing fisheries with regulations. Pages 383—
402 in C. C. Kohler and W. A. Hubert editors. Inland Fisheries Management in
North America. American Fisheries Society, Bethesda.

O’Neill, R. V. 1989. Perspectives in hierarchy and scale. Pages 140—156 in J.
Roughgarden, R. M. May, and S. A. Levin, editors. Perspectives in ecological
theory. Princeton University Press, Princeton, New Jersey.

Oglesby, R. T. 1977. Relationships of fish yield to lake phytoplankton standing crop,
production, and morphoedaphic factors. Journal of the Fisheries Resources Board
of Canada 34:2271—2279.

Oglesby, R. T., J. H. Leach, and J. Fomey. 1987. Potential Stizostedion yield as a
function of chlorophyll concentration with special reference to Lake Erie.
Canadian Journal of Fisheries and Aquatic Sciences 44:166—170.

Olson, M. H. 1996. Predator-prey interactions in size-structured fish communities:
implications of prey growth. Oecologia 108:757—763.

Osenberg, C. W., E. E. Werner, G. G. Mittelbach, and D. J. Hall. 1988. Growth patterns
in bluegill (Lepomis macrochirus) and pumpkinseed (L. gibbosus) sunfish:
environmental variation and the importance of ontogenetic niche shifts. Canadian
Journal of Fisheries and Aquatic Sciences 45:17—26.

105

Paukert, C. P., D. W. Willis, and D. W. Gabelhouse. 2002. Effect and acceptance of
bluegill length limits in Nebraska natural lakes. North American Journal of
Fisheries Management 22:1306—1313.

Pauly, D. 1980. On the interrelationships between natural mortality, growth parameters,
and mean environmental temperature in 175 fish stocks. Journal du Conseil
international pour l’Exploration de la Mer 39: 1 75-1 92.

Pollack, K. H., C. M. Jones, and T. L. Brown. 1994. Angler survey methods and their
applications in fisheries management. American Fisheries Society, Special
Publication 25, Bethesda, Maryland.

Quist, M. C., C. S. Guy, R. D. Schultz, and J. L. Stephen. 2003. Latitudinal comparisons
of walleye growth in North America and factors influencing growth of walleyes in

Kansas reservoirs. North American Journal of Fisheries Management 23:677—
692.

Rahel, F. J. 1984. Factors structuring fish assemblages along a bog lake successional
gradient. Ecology 65:1276—1289.

Rahel, F. J., and J. J. Magnuson. 1983. Low pH and the absence of fish species in
naturally acidic Wisconsin lakes: references for cultural acidification. Canadian
Journal of Fisheries and Aquatic Sciences 40:3—9.

Reed, J. R., and B. G. Parsons. 1999. Angler opinions of bluegill management and
related hypothetical effects on bluegill fisheries in four Minnesota lakes. North
American Journal of Fisheries Management 19:515-519.

Ricker, W. E. 1945. Natural mortality among Indiana bluegill sunfish. Ecology 26:111—
121.

Ricker, W. E. 1969. Effects of size-selective mortality and sampling bias on estimates of
growth, mortality, production, and yield. Journal of the Fisheries Research Board
of Canada 26:479—541.

Ricker, W. E. 1975. Computation and interpretation of biological statistics. Bulletin 191
of the Fisheries Research Board of Canada.

Ricklefs, R. E., 1987. Community diversity: relative roles of local and regional
processes. Science 235: 1 67—1 71.

Ridgway, M. S., B. J. Shuter, and E. E. Post. 1991. The relative influence of body size
and territorial behavior on nesting asynchrony in male smallmouth bass

Micropterus dolomieui (Pisces: Centrarchidae). Journal of Animal Ecology
60:665—681.

106

Riera, J. L., J. J. Magnuson, T. K. Kratz, and K. E. Webster. 2000. A geomorphic
template for the analysis of lake districts applied to the Northern Highland Lake
District, Wisconsin, USA. Freshwater Biology 43:301—318.

Roughgarden, J. 1989. The structure and assembly of communities. Pages 203—226 in J.
Roughgarden, R. M. May, and S. A. Levin, editors. Perspectives in ecological
theory. Princeton University Press, Princeton, New Jersey.

Ryder, R. A. 1965. A method for estimating the potential fish production of north-
temperate lakes. Transactions of the American Fisheries Society 94:214—218.

Schneider, J. C. 1975. Typology and fisheries potential of Michigan lakes. Michigan
Academy 8:59—84.

Schupp, D. H. 1992. An ecological classification of Minnesota lakes with associated fish
communities. Minnesota Department of Natural Resources, Section of Fisheries,
Investigation Report 417, St. Paul.

Shuter, B. J ., D. A. Schlesinger, and A. P. Zimmerman. 1983. Empirical predictors of
annual surface water temperature cycles in North American lakes. Canadian
Journal of Fisheries and Aquatic Sciences 40:1838—1845.

Shuter, B. J., M. L. Jones, R. M. Korver, and N. P. Lester. 1998. A general, life history
based model for regional management of fish stocks: the inland lake trout

(Salvelinus namaycush) fisheries of Ontario. Canadian Journal of Fisheries and
Aquatic Sciences 55:2161—2177.

Snow, H. E. 1978. A 15-year study of the harvest, exploitation, and mortality of fishes in
Murphy F lowage, Wisconsin. Wisconsin Department of Natural Resources,
Technical Bulletin 103.

Sokal, R. R., and F. J. Rohlf. 1995. Biometry. W. H. Freeman and Company, New
York.

Soranno, P. A., K. E. Webster, J. L. Riera, T. R. Kratz, J. S. Baron, P. A. Bukaveckas, G.
W. Kling, D. S. White, N. Caine, R. C. Lathrop, and P. R. Leavitt. 1999. Spatial
variation among lakes within landscapes: ecological organization along lake
chains. Ecosytems 2:395—410.

Tomcko, C. M., and R. B. Pierce. 2001. The relationship of bluegill growth, lake
morphometry, and water quality in Minnesota. Transactions of the American
Fisheries Society 130:317—321 .

Tonn, W. M. 1990. Climate change and fish communities: a conceptual framework.
Transactions of the American Fisheries Society 119:337—352.

107

Tonn, W. M., and J. J. Magnuson. 1982. Patterns in the species composition and
richness of fish assemblages in northern Wisconsin lakes. Ecology 63:1149—
1166.

Tonn, W. M., J. J. Magnuson, and A. M. Forbes. 1983. Community analysis in fishery
management: an application with northern Wisconsin lakes. Transactions of the
American Fisheries Society 112:368—377.

Trippel, E. A. 1995. Age at maturity as a stress indicator in fisheries. BioScience
45:759—771.

Tyler, A. V., R. J. Beamish, and G. A. McFarlane. Implications of age determination
errors to yield estimates. Pages 27—35 in R. J. Beamish, and G. A. McFarlane,
editors. Effects of ocean variability on recruitment and an evaluation of
parameters used in stock assessment models. Canadian Special Publication of
Fisheries and Aquatic Sciences 108.

Van Den Avyle, M. J. 1993. Dynamics of exploited fish populations. Pages 105—135 in
C. C. Kohler and W. A. Hubert, editors. Inland fisheries management in North
America. American Fisheries Society, Bethesda, Maryland.

Von Bertalanffy, L. 1938. A quantitative theory of organic growth. Human Biology
10:181—213.

Webster, K. E., T. K. Kratz, C. J. Browser, J. J. Magnuson, and W. J. Rose. 1996. The
influence of landscape position on lake chemical responses to drought in Northern
Wisconsin. Limnology and Oceanography 41 :977—984.

Webster, K. E., P. A. Soranno, S. B. Baines, T. K. Kratz, C. J. Bowser, P. J. Dillon, P.
Campbell, E. J. Fee, and R. E. Hecky. 2000. Structuring features of lake
districts: landscape controls on lake chemical response to drought. Freshwater
Biology. 43: 499—515.

Westrheim, S. J., and W. E. Ricker. 1978. Bias in using an age-length key to estimate
age-frequency distributions. Journal of the Fisheries Research Board of Canada
35: 184—189.

Wetzel, R. G. 1975. Limnology. W.B. Saunders, Philadelphia, Pennsylvania.

Wilde, G. R. 1997. Largemouth bass fishery responses to length limits. Fisheries 22:14—
21.

Wiley, M. J ., S. L. Kohler, and P.W. Seelbach. 1997. Reconciling landscape and local
views of aquatic communities: lessons from Michigan trout streams. Freshwater
Biology 37:133—148.

Wilkinson, L. 2000. Classification and regression trees. Pages 35—51 in Systat I, SPSS,
Chicago.

108

Willis, D. W., B. R. Murphy, and C. S. Guy. 1993. Stock density indiceszdevelopment,
use, and limitations. Reviews in Fisheries Science 12203—222.

Wisconsin Department of Natural Resources. 1967—1983. Surface Water Resources of
selected Wisconsin Counties. Wisconsin Department of Natural Resources,
Madison.

Wisconsin Department of Natural Resources. 1995. Wisconsin Lakes Book. Wisconsin
Department of Natural Resources, Madison

109

      
     

IIIIIII

lilillllli]

I! RRRRRRRRRRRRRRR
3 129

 

lllilillllllllill

02 36 1165