Lakeshore residential development and growth of largemouth bass

(Micropterus salmoides): a cross-lakes comparison

by

Jereme William Gaeta

A thesis submitted in partial fulfillment of the requirements for a degree of

MASTER OF SCIENCE

(Limnology and Marine Science)

at the

UNIVERSITY OF WISCONSIN – MADISON

2009

i

Table of Contents

PAGE

Abstract ……………………………………………………………… ii

Acknowledgements ……………………………………………………………… iv

Introduction ……………………………………………………………… 1

Methods ……………………………………………………………… 3

Results ……………………………………………………………… 7

Discussion ……………………………………………………………… 9

Literature Cited ……………………………………………………………… 14

List of Tables ……………………………………………………………… 22

Tables ……………………………………………………………… 23

Figure Captions ……………………………………………………………… 26

Figures ……………………………………………………………… 31

List of Appendices ……………………………………………………………… 36

Appendices ……………………………………………………………… 37

ii

Lakeshore residential development and growth of largemouth bass (Micropterus salmoides):

a cross-lakes comparison

Jereme W. Gaeta†

Under the supervision of Professor Stephen R. Carpenter

at the University of Wisconsin – Madison

Abstract

Lakeshore residential development is associated with decreases in riparian zone vegetation

and littoral zone structure, and increased angling effort. Depending upon the species and

their associated body size, fishes may respond differently to these changes. Responses may

be particularly difficult to predict for species that undergo marked changes in habitat use and

diet over ontogeny, such as the popular sportfish largemouth bass (Micropterus salmoides).

To test for a relationship between lakeshore residential development and largemouth bass

growth across ontogeny, we compared largemouth bass size-specific growth rates across 16

lakes that span the regional range of lakeshore residential development (0 to 45.8 buildings

km -1

) in Wisconsin’s Northern Highland Lake District. Using a longitudinal multilevel

model, the relationship between lakeshore residential development and largemouth bass

growth was identified and related to body size. Largemouth bass growth rates were

positively correlated with lakeshore residential development for fish smaller than 164 mm

† Co-authors: Matthew J. Guarascio, Greg G. Sass, and Stephen R. Carpenter

iii

and were negatively correlated for larger sizes. While we cannot infer mechanism from our

study, we believe the relationship between lakeshore residential development and largemouth

bass growth is likely the result of angling-induced behavioral changes or angling-induced

selection pressure.

iv

Acknowledgements

Although this thesis has only my name on the cover, the work behind it was most

certainly a collaborative effort. Sampling for this project started in 2006 while I was off in

the Bering Sea and had as of then never even heard of limnology. Thus, I would like to give

a special thank you to my co-authors Matthew J. Guarascio and Greg G. Sass, as well as our

colleague Tyler Ahrenstorff for getting this project off the ground and collecting all of this

awesome data. Nicholas Preston, Shawn Devlin, and Gretchen Hansen - thanks for all of the

statistical advise. Drs. James Kitchell and Jake Vander Zanden thank you for your assistance

on an earlier version of this thesis. Jeff Maxted and Graham MacDonald - thanks for your

help with GIS. On a more personal note, I would like to express sincere gratitude to another

co-author and my advisor, Dr. Stephen Carpenter. Thank you Steve for everything you have

done to help me get where I am now, and thank you in advance for the next 2.5 (if you have

your way) to 3.5 years of guidance.

With all of life’s greatest journeys and stories, the end is but a small part of the

experience. To really reflect on what a personal achievement this thesis is, and to truly thank

everyone who has helped me along the way I must go back well before graduate school.

Thank you God for opportunity and for giving me the courage to push – to push my limits,

my fears, my dreams. I would like to thank my family: Mom, Dad and Theresa – thank you

for giving me my stubborn determination, for always telling me to do what my heart knew

was right, and for the endless support. Josh – thank you for teaching me to fight and for

showing me how to find inner strength when all seems lost. Mike – thank you for always

reminding me of who I am, even when I don’t want to be reminded. Leia, my love, thank

you for helping me “go, and come back radiant.” I know I don’t make it easy dealing with

my mad scientist’s personality and work schedule, but here’s to Sydney!!!!!

Dr. Gobalet, what can I say? You might be the most important person in my career.

There is no way I would have done what I have done without you giving me my foundation

in science. You are much more than merely an important person in my career though Doc.

Thank you for being a great friend in probably the toughest time of my life. Thank you for

never believing I was reaching my potential and for always making me give a little more.

v

Thank you for looking me in the eye and telling me to work harder because you knew I could

do better. Tough love can be the best kind.

I have to send a special shout out to all my boys in Madison that have kept me sane

along the way. Zach Lawson, thanks for teaching me how to fish and hunt. It would have

been a very tough first few years without a guide. A special thanks goes out to my office

mates Brian Weidel, Matt Diebel, and Steve Powers. Thanks for taking the time to answer

the most random fisheries or R question when you were knee deep in your own work. My

boys at the 415, Brian and Nico Preston, thank you for all of the enlightening kitchen

conversations, and Brian, “…let’s go together.” Noah Lottig, my TLS room mate, viva la

Bac-T!!!! Thanks for being a great friend, for always taking the time to talk to me when I

had no clue what the hell I was doing, and for getting me into cycling. Scott Higgins, thanks

for always taking the time to help me work through science and for always being down to

ride. Shawn Devlin - thanks for making my TLS and ESA experiences priceless. Matts 3

and 4 (Kornis and Fuller) wasn’t Sapelo something? Thanks for all the mud, beers, nearly

flipped golf carts, nearly captured mini deer, and for throwing me under the Kitchell bus…

that was way too funny. Kornis and Fuller, thanks for adopting me into your cohort. Hey

Gretchen, BIOCOM!!!!!

I would like to thank all of my undergrads, although they didn’t help collect these

data, they rock: Zach Raw-Dog Lawson, Matt Rounds, Danielle Haak, Erik the Viking

Kopperud, Sammer Healy, and Nicholas Rusticus Heredia. I would like to especially thank

all of the staff at the CFL and at TLS. Pam Fashingbauer you saved me more times than I

can count – Thank you so much! And Tim Kratz, thank you for understanding stupid

mistakes. Thanks to my Carpenter lab-mates and all of the grad students on this journey that I

didn’t mention. Steve, thanks again for being a great advisor and for all the great fly-fish and

deer hunting adventures. I am jacked about the next few years.

Sincerely,

Jereme William Gaeta

October 22, 2009

1

Introduction

Humans and the environment in which they live are tightly coupled; consequently,

human actions can considerably influence ecosystem dynamics. Lakes and humans, in

particular, are linked through the variety of ecosystem services ranging from the aesthetic to

the economically significant (Riera et al. 2001). Still, humans are extensively altering, and in

some cases, degrading these aquatic habitats. For instance, recreational fisheries have

caused shifts in aquatic food webs and altered lake ecosystems (Lewin et al. 2006).

Additionally, lakeshore residential development is associated with changes in lake riparian

zones (Christensen et al. 1996, Riera et al. 2001), littoral zones (Kratz et al. 2002, Jennings et

al. 2003, Marburg et al. 2006), and biota (Schindler et al. 2000, Sass et al. 2006, Brauns et al.

2007).

In lakes, anglers act as top predators exploiting many species and size classes of

fishes (Johnson and Staggs 1992, Kitchell and Carpenter 1993). Angling effects aquatic

communities in a variety of ways ranging from extreme cases of collapsed recreational

fisheries (Post et al. 2002) to, more commonly, altered fish population size structures and

densities (Olson and Cunningham 1989, Lewin et al. 2006). These angling-induced shifts in

size structure can change brood sizes, juvenile densities, juvenile growth rates, and length

and age at maturity (Conover and Munch 2002, Suski and Philipp 2004, Reznick and

Ghalambor 2005). Angling has also been shown to target more aggressive individuals and

results in angling-induced behavioral selection of less aggressive individuals potentially

altering fisheries (Jorgensen et al. 2007, Uusi-Heikkila et al. 2008, Philipp et al. 2009).

In many areas of the United States, such as Wisconsin’s Northern Highland Lake

District (NHLD), recreational fisheries are a pillar of the regional economy (Penaloza 1991,

2

Postel and Carpenter 1997, Peterson et al. 2003). Therefore, understanding and quantifying

whether or how humans alter these important fisheries is essential. Members of the sunfish

family (Centrarchidae), including bluegill (Lepomis machrochirus) and largemouth bass

(Micropterus salmoides), are two important sportfish in the NHLD. Schindler et al. (2000)

identified a significant inverse relationship between bluegill growth and lakeshore residential

development across all fish sizes classes tested (60 mm, 100 mm, 140 mm) in lakes from the

NHLD. A trend of lower growth with lakeshore residential development for the largest size

class (400 mm) of largemouth bass was also identified, although no statistically significant

effects of lakeshore residential development were found for largemouth bass in that study

(Schindler et al. 2000).

Our study used a cross-lakes comparison of 16 lakes spanning the full regional

gradient of lakeshore residential development (0 to 45.8 buildings km -1

) in the NHLD to test

for a relationship between lakeshore residential development and largemouth bass growth

across ontogeny. We used a longitudinal, multilevel approach to estimate growth responses

across a range of fish sizes. Our results suggest that the relationship between lakeshore

residential development and largemouth bass growth varies across ontogeny. Small fish (~

75 mm) had a strong positive growth response to lakeshore residential development, while

the strength of the response decreased with fish size and became increasingly negative for

fish longer than 164 mm.

3

Methods

Study Area

Largemouth bass (LMB) size-specific growth rates were surveyed across the regional density

gradient of lakeshore residential development in 16 lakes from the NHLD in Vilas County,

Wisconsin (Figure 1). The NHLD is a formerly glaciated, lake-rich region spanning about

5,330 km2 with around 7,600 lakes (Peterson et al. 2003, Carpenter et al. 2007) and is

vegetated by upland conifer-hardwood forests (Stearns 1951, Brown and Curtis 1952).

Human population densities in the region have increased nearly five fold in the last half-

century (Carpenter et al. 2007). Since the 1960’s, the majority of development in Vilas

County has occurred on lake shorelines (Schnaiberg et al. 2002) and, as of the early 2000’s,

the county had about 16,500 buildings within 100 m of lake shorelines (Riera et al. 2001).

We selected study lakes to span the regional gradient of lakeshore residential

development: zero to over 45.5 buildings km-1

(Table 1) within 100 m of lake shorelines.

Building densities were previously surveyed during the summers of 2001-2004 and data were

archived in the North Temperate Lakes Long-Term Ecological Research online database

(Carpenter and Kratz 2001). Esocidae species (e.g. muskellunge Esox masquinongy,

northern pike Esox lucius), predators of LMB, were common or abundant in all but three

lakes in our study: Camp Lake, Little Rock Lake, and Day Lake (Wisconsin Department of

Natural Resources 2005). Additionally, we chose lakes with a low density of coarse woody

habitat to reduce potential confounding effects of coarse woody habitat and building density.

Selected lakes ranged from 0 logs km-1

to 125 logs km-1

, spanning only 13% of the observed

regional gradient (Christensen et al. 1996).

Fish Sampling

4

We sampled largemouth bass between June and August of 2006 primarily via

electrofishing along the lake perimeter. We used angling to collect fish when lake

conductivity was not conducive to electrofishing. Twenty-seven to thirty fish were collected

from each lake to determine size specific growth rates. Fish length (mm) was recorded and

five to 10 scales were collected from each fish from the area posterior to a depressed pectoral

fin. Scales were sonicated and pressed between two slides. Growth rate (mm year-1

) was

determined using Fraser-Lee’s method of back calculation with Carlander’s recommended

constant of 20 mm for LMB (Carlander 1982). These calculations were made from non-

regenerated scales read into a digital imaging system.

Statistical Analysis

Our data were hierarchically structured with repeated measures of annulus-specific

growth observations (mm) nested within individual fish growth rates (mm year-1

). Individual

fish were also nested in a specific lake with a unique building density. We thus designed our

analysis around the hierarchical nature of the data and tested for a relationship between

lakeshore residential development and LMB growth rate using a longitudinal (repeated

measures) multilevel model (Goldstein 1995, Ai 2002, Paradis et al. 2006, Wagner et al.

2006).

Growth and length were loge transformed prior to analysis. Individuals with less than

two annuli (i.e. young-of-the-year and yearlings) were removed from the analysis because at

least two annuli are needed to estimate model parameters. As a result, all lakes had at least

26 individuals included in the analysis (Table 1). We performed all analyses in the R-Cran

statistical package (R Development Core Team 2008); package: ‘lme4’ version 0.999375-

31). Multilevel modeling methods followed procedures outlined in Gelman and Hill (2008).

5

The slopes and intercepts of individual fish growth rates (mm year-1

) were allowed to vary as

random effects. Model fixed effects included an intercept, fish length, lake building density,

and the interaction between fish length and building density.

Model Structure

The multilevel model was composed of three levels: (1) a fish level in which

individual fish were represented by repeated measures of annulus-specific growth (mm year-

1) observations at a given fish length (n=3540); (2) a lake level in which lakes were

composed of individual fish (n=458) each with unique a growth rate; and (3) a regional level

in which lakes (n=16) spanned the regional gradient of building densities (buildings km-1

).

In our model, growth was defined as a function of fish length and building density. An

interaction term was included to allow the effect of building density to vary across fish

length.

Level 1: Fish level; within a fish within a lake

!

yi ~ " #0 j[ i] + #

1 j[ i]xi , $ y

2( ) , for i =1,...,n observations (Model 1.1)

Here,

!

yi is the growth for observation i in fish j at length

!

xi,

!

"0 j[ i] is the intercept (or growth

at length zero) of observation i in fish j,

!

"1 j[ i] is the growth parameter of observation i in fish

j and

!

" y

2

is the residual variance of

!

yi (growth) of observation i in fish j at length

!

xi. The

notation N(µ,!2) refers to a normal or multivariate normal distribution with mean vector µ

and covariance matrix !2. In the case of equation 1.1, the model is a linear regression of loge

transformed growth rate on loge transformed body size.

Level 2: Lake level; among fish within lake

6

!

"0 j

"1 j

#

$ %

&

' ( ~ )

*0k[ j ]

* 1k[ j ]

#

$ %

&

' ( ,

+0"2

,+0"2 +

1"2

,+

0"2 +

1"2

+1"2

#

$ %

&

' (

#

$ % %

&

' ( ( , for j =1,...,J fish (Model 1.2)

Here,

!

"0k[ j ] is the mean intercept (or growth at length zero) for fish j in lake k and has a

variance of

!

"0#

2 ,

!

" 1k[ j ] is the mean growth rate for fish j in lake k and has a variance of

!

"1#

2 ,

and

!

"#0$

2 #1$

2 is the covariance among

!

"0#

2 and

!

"1#

2 . Equation 1.2 relates growth parameters

!0j and !1j of an individual fish to the means for a lake and the covariance matrix among fish

within a lake.

Level 3: Regional level; among lakes

!

"0k

"1 k

#

$ %

&

' ( ~ )

*00k

+ *01kzk

* 10k

+ *11kxizk

#

$ %

&

' ( ,

+0"2

,+0"2 +

1"2

,+

0"2 +

1"2

+1"2

#

$ %

&

' (

#

$ % %

&

' ( ( , for k =1,...,K lakes (Model 1.3)

Here,

!

"00k

is the mean intercept (or growth at length zero) for lake k and has a variance of

!

"0#

2 ,

!

" 01k

is the effect of building density

!

zk on the intercept,

!

" 10k

is the mean growth rate for

the lake k and has a variance of

!

"1#

2 ,

!

"11k

is the effect of the interaction of fish length

!

xi and

building density

!

zk on the mean growth rate for lake k, and

!

"#0$

2 #1$

2 is the covariance

between

!

"0#

2 and

!

"1#

2 . Equation 1.3 is a bivariate regression of lake-specific growth

parameters "0k and "1k on the fixed effect of building density.

The effect of building density on growth rate for fish of a given length was calculated

from the derivative of the model with respect to building density with units of

!

ln(mm year-1)

building km-1:

01 11building effect i

k k i

k

xx

z

!" "

!

# $= = +% &' (

7

Standard deviation of the building effect was calculated from the covariance matrix of model

parameters using standard error propagation formulae (Meyer 1975). The building effect

was interpreted as the average change in growth rate of fish of a given length with the

addition of one building km-1

of shoreline.

Results

A total of 3,540 annulus observations (n = 458 fish) were made from 16 lakes in the NHLD.

A wide range of fish lengths were represented for each lake in our dataset (Figure 2). The

maximum growth rate observed was 133.8 mm year-1

at annulus (age) one from a fish in

Little St. Germain Lake (LSG). The lowest growth rate observed was 5.83 mm year-1

at

annulus 12 in a fish from Camp Lake (CP).

Although growth at length varied greatly within a lake, an apparent trend at small

sizes was observed in all lakes (Figure 2). This trend, however, is an artifact of how growth

was calculated. The growth rate at annulus one, or how much a fish grew in the first year, is

equal to the fish length at annulus one, thus producing a 1:1 relationship between growth rate

and length for observations at annulus one. Most lakes had annulus-specific observations

spanning the entire range of fish lengths (Table 2). After young-of-the-year and yearlings

were removed from the analysis, we captured fish from a range of over 300 mm and 15

years. The average maximum size at capture was 371 mm across all lakes. Two lakes had a

sparse sample of large individuals; the largest individuals from Little John and Little

Crooked Lakes were a 249 mm 8 year old and a 310 mm 11 year old bass, respectively.

Model Fit

8

Largemouth bass growth rate was successfully fit as a function of fish length,

building density, and the interaction between fish length and building density. Model

predicted growth rates were closely clustered around observed growth rates (Figure 3a).

Residuals were evenly distributed across predicted growth rate (Figure 3b) and across fish

length (Figure 3c). Model fit also appeared to be unbiased across the regional gradient of

building density (Figure 3d). As an unavoidable consequence of how growth rates are

calculated, growth rate and length at annulus one are equal creating a conspicuous trend for

observations at annulus one (clustered observations in the upper right of figure 3a, on the

right side of figure 3b, and on the left side of figure 3c). However, it was evident that the

artifact of growth calculations at annulus one did not bias our model results when the

residuals of observations at 65 – 90 mm (4.17 – 4.5 loge(mm); Figure 3e) and at 90 – 116 mm

(4.5 – 4.75 loge(mm); Figure 3f) were regressed across the observed building density

gradient. In the results that follow, the minimum fish size discussed is 75 mm due to the

small number of observations below 65 mm (4.17 loge (mm)). We were similarly cautious

for larger fish; although fish lengths of up to 420 mm were represented in the data, the largest

fish size discussed here is 375 mm (5.93 loge (mm)).

Model Results

Regionally, our analysis identified a significant relationship between building density

and LMB size specific growth rate that varied across fish length (Figures 4 and 5). Growth

rates of small LMB were positively correlated with building density, ranging from 75 mm

year-1

at zero buildings km-1

to 103 mm year-1

at 45.8 buildings km-1

(Figure 3a). Inversely,

growth rates of large LMB were negatively correlated with building density, declining from

12 to 8 mm year-1

across the building gradient (Figure 3c). No change in growth rate was

9

observed across the regional building density gradient at intermediate lengths (Figure 3b).

The change in growth rate across the regional building density gradient was much larger for

75 mm LMB than 375 mm LMB (Figure 3) because growth rate decreases with fish length

(Figure 2). However, the relative changes in growth rates were nearly the same. A 75 mm

LMB exhibited a 38% increase in growth rate across the building density gradient, while a

375 mm LMB exhibited a 30% decrease in growth rate across the gradient. These relative

differences were illustrated by taking the derivative of the model with respect to building

density, isolating the building density effect and identifying average change in growth rate at

a given length with the addition of one building km-1

of shoreline (Figure 4). The model

indicated that LMB growth rate was positively correlated with building density for LMB

sizes smaller than 163.8 mm and was negatively correlated for all larger sizes. However,

when error is taken into account, we can only say with confidence that LMB 112.2 mm or

smaller have a positive correlation with building density and LMB 228.1 mm or greater are

negatively correlated with building density. Growth rates showed the largest changes across

the gradient of building densities for the largest and smallest sizes classes.

Discussion

The relationship between LMB growth rate and lakeshore residential development

across fish length was determined using a longitudinal multilevel model. Our model

accounted for both repeated measures of annulus observations and the hierarchical structure

of the sample design. Small LMB growth rates were found to correlate with lakeshore

residential development, while growth rates of large LMB were inversely related to lakeshore

residential development. The mechanisms driving the observed relationship between LMB

10

growth rate and lakeshore residential development are likely complex and highly variable.

Comparative studies such as ours are powerful for testing patterns or associations across

lakes, but inferences about mechanisms must be cautious (Cole et al. 1991). Nonetheless, our

observed relationship between growth and lakeshore residential development may be

explained by several alternative hypotheses, which invite discussion.

Angling-induced selection for slower growing, less aggressive LMB associated with

lakeshore residential development may explain patterns of LMB growth in Wisconsin’s

NHLD. Individual fish aggression correlates with growth in several species of fish (Biro and

Stamps 2008). Largemouth bass aggression levels are known to vary within natural

populations and more aggressive, likely faster growing, individuals are more vulnerable to

angling once individuals reach a catchable size (Biro and Post 2008). Recreational angling

may target more vulnerable individuals, and therefore, has the potential to select for less

aggressive, slower growing fish (Biro and Post 2008, Uusi-Heikkila et al. 2008, Philipp et al.

2009). In a 24 year-long, four-generation study, Philipp et al. (2009) showed that removal of

more aggressive LMB shifted the population toward a higher frequency of less vulnerable

LMB. This selection pressure would be exacerbated if angling occurs during spawning, due

to increased vulnerability of male LMB while nest guarding (Suski and Philipp 2004, Cooke

et al. 2007).

Harvesting during spawning has the potential for rapid removal of aggressive, faster

growing individuals from a population. Even catch-and-release angling during spawning

would reduce the fitness of these more aggressive and vulnerable individuals due to egg

predation (Uusi-Heikkila et al. 2008) and nest abandonment while fish are handled by anglers

(Siepker et al. 2009). Whether anglers are harvesting LMB throughout the season or

11

practicing catch-and-release fishing during spawning, angling-induced selection is likely to

occur in the more developed lakes in our survey. Surveys suggest angler effort increases with

lakeshore residential development in the NHLD (Jorgensen and Stedman 2001, Stedman

2003, Jorgensen et al. 2006), and therefore, angler-induced selection may be more likely to

increase with lakeshore residential development. This process would increase survivorship of

slower-growing individuals, and thereby, lead to higher proportions of slower-growing

individuals in the larger size classes of LMB, as observed in our study. At the same time,

these responses of larger LMB would result in indirect impacts on smaller LMB consistent

with our results.

The interpretation above posits that angling-induced selection results in a reduction of

aggressive males associated with lakeshore residential development. Largemouth bass

aggression and associated vulnerability to angling is positively related to brood size (Suski

and Philipp 2004). Therefore, angling-induced selection can decrease the density of

aggressive males with larger broods and thereby decrease mean male fecundity in more

heavily-developed lakes. Studies in both the NHLD (Post et al. 1998) and Michigan (Olson

et al. 1995) have shown that YOY LMB growth rate is inversely related to YOY LMB

density. Thus a reduction in fecundity associated with lakeshore residential development

would decrease YOY LMB densities and thereby increase growth rates of small LMB.

Alternatively, our observed trends for growth rates of large LMB could be explained

by reduced fish growth due to physiological stressors and shifts in behavior associated with

handling during catch and release fishing. Influences of catch and release fishing on fish

growth vary among species and studies on the same species (Arlinghaus et al. 2007). To

date, only two studies have investigated the effects of catch and release fishing on LMB

12

growth and they yielded opposing outcomes (Arlinghaus et al. 2007, Siepker et al. 2007). A

40-day experiment performed in Texas showed that catch and release fishing had no effect on

LMB mass in the species optimum temperature range (Pope and Wilde 2004). Conversely, a

6-week experiment on LMB in the Midwest identified significant changes in foraging

behavior and greater mass loss associated with catch and release fishing (Siepker et al. 2006).

Using observations of altered foraging behavior associated with catch and release fishing,

Siepker et al. (2006) used bioenergetics models of longer-term (135 day) effects of catch and

release fishing on growth, corroborated their experimental results, and identified growth rates

of angled LMB to be 12-44% lower than non-captured fish. If the impacts of catch and

release fishing on LMB growth in the NHLD are similar to those observed in the Siepker et

al. (2006) study, we would expect growth rates of large, catchable individuals to be inversely

related to lakeshore residential development as indicated by our analysis.

Another scenario that could explain accelerated growth of small LMB in developed lakes

is an earlier ontogenetic shift to piscivory due to reduced growth rates of prey fishes.

Schindler et al. (2000) found growth rates from all size classes of bluegill, a common LMB

prey source, to be inversely related to lakeshore residential development in the NHLD. If

this trend holds for bluegill and other prey fishes in our study lakes, YOY LMB could make

an earlier ontogenetic shift to piscivory and thereby increase their growth rates. Such a

relationship between YOY growth rates of LMB and bluegill was demonstrated by Olson

(1996) in a study of four Michigan lakes.

Any of these possible mechanisms could explain our observed trends. Indeed, these

mechanisms are not exclusive and several known and unknown mechanisms could interact to

lead to our observed patterns. Experiments investigating the relationships between LMB

13

growth and aggression, and the impacts of catch and release fishing on LMB foraging

behavior and growth, along with surveys directly investigating the relationship between

residential lakeshore development and angling effort in the NHLD are needed to further

evaluate alternative hypotheses for the relationship between LMB growth and lakeshore

residential development.

Even though important mechanistic questions remain to be answered, we have shown that

lakeshore residential development is associated with altered growth rates in an economically

important sportfish, which has implications for fisheries. According to our model, LMB take

2 ! years longer to enter the fishery (356 mm) in highly developed lakes (45.8 buildings km-

1) relative to undeveloped lakes (given an initial length of 83 mm, the average size at age

one; Figure 6). Highly developed lakes hold populations of smaller, slower growing LMB.

Moreover, angling may also be selecting for less aggressive individuals, which may be

harder to catch (Uusi-Heikkila et al. 2008). If carried to an extreme, such changes to the

fishery could be detrimental to areas such as Wisconsin’s NHLD, which are economically

dependent upon recreational fisheries. Therefore, the scope of fisheries management must not

only include predator-prey interactions, but also selective effects of fisheries (Jorgensen et al.

2007) and considerations beyond the shoreline to include building practices and riparian land

use.

14

Literature Cited

Ai, X. 2002. Gender differences in Growth in Mathematics Achievement: Three-Level

Longitudinal and Multilevel Analyses of Individual, Home, and School Influences.

Mathematical Thinking and Learning 4(1): 1-22.

Arlinghaus, R., Cooke, S.J., Lyman, J., Policansky, D., Schwab, A., Suski, C., Sutton, S.G.,

and Thorstad, E.B. 2007. Understanding the complexity of catch-and-release in recreational

fishing: An integrative synthesis of global knowledge from historical, ethical, social, and

biological perspectives. Rev. Fish. Sci. 15(1-2): 75-167. doi: 10.1080/10641260601149432.

Biro, P.A., and Post, J.R. 2008. Rapid depletion of genotypes with fast growth and bold

personality traits from harvested fish populations. Proc. Natl. Acad. Sci. U. S. A. 105(8):

2919-2922. doi: 10.1073/pnas.0708159105.

Biro, P.A., and Stamps, J.A. 2008. Are animal personality traits linked to life-history

productivity? Trends Ecol. Evol. 23(7): 361-368. doi: 10.1016/j.tree.2008.04.003.

Brauns, M., Garcia, X.-F., Walz, N., and Pusch, M.T. 2007. Effects of human shoreline

development on littoral macroinvertebrates in lowland lakes. Journal of Applied Ecology

44(6): 1138-1144.

Brown, R.T., and Curtis, J.T. 1952. The Upland Conifer-Hardwood Forests of Northern

Wisconsin. Ecological Monographs 22(3): 217-234. doi: doi:10.2307/1943566.

15

Carlander, K.D. 1982. Standard intercepts for calculating lengths from scale measurements

for some centrarchid and percid fishes. Transactions of the American Fisheries Society 111:

332-336.

Carpenter, S.R., Benson, B.J., Biggs, R., Chipman, J.W., Foley, J.A., Golding, S.A.,

Hammer, R.B., Hanson, P.C., Johnson, P.T.J., Kamarainen, A.M., Kratz, T.K., Lathrop, R.C.,

McMahon, K.D., Provencher, B., Rusak, J.A., Solomon, C.T., Stanley, E.H., Turner, M.G.,

Vander Zanden, M.J., Wu, C.H., and Yuan, H.L. 2007. Understanding regional change: A

comparison of two lake districts. Bioscience 57(4): 323-335. doi: 10.1641/b570407.

Carpenter, S.R., and Kratz, T.K. 2001. Biocomplexity; Cross Lake Comparison. Available

from https://secure.limnology.wisc.edu/lterquery/index_new.jsp?project_id=BIOCOM1.

Christensen, D.L., Herwig, B.R., Schindler, D.E., and Carpenter, S.R. 1996. Impacts of

Lakeshore Residential Development on Coarse Woody Debris in North Temperate Lakes.

Ecological Application 6(4): 1143-1149.

Cole, J.J., Lovett, G., and Findlay, S. 1991. Comparative Analysis of Ecosystems: Paterns,

Mechanisms, and Theories. Springer-Verlag, New York, NY.

Conover, D.O., and Munch, S.B. 2002. Sustaining fisheries yields over evolutionary time

scales. Science 297(5578): 94-96.

Cooke, S.J., Suski, C.D., Ostrand, K.G., Wahl, D.H., and Philipp, D.P. 2007. Physiological

and behavioral consequences of long-term artificial selection for vulnerability to recreational

angling in a teleost fish. Physiol. Biochem. Zool. 80(5): 480-490. doi: 10.1086/520618.

16

Gelman, A., and Hill, J. 2008. Data Analysis Using Regression and Multilevel/Heirarchical

Models. Cambridge University Press, New York, NY.

Goldstein, H. 1995. Multilevel Statistical Models. Hodder Arnold, London.

Jennings, M.J., Emmons, E.E., Hatzenbeler, G.R., Edwards, C., and Bozek, M.A. 2003. Is

Littoral Habitat Affected by Residential Development and Land Use in Watersheds of

Wisconsin Lakes? Lakes and Reservoir Management 19(3): 272-279.

Johnson, B.M., and Staggs, M.D. 1992. The Fishery. In Food Web Management: A Case

Study of Lake Mendota. Edited by J.F. Kitchell. Springer-Verlag, New York, NY. pp. 353-

375.

Jorgensen, B.S., Nowacek, D., Stedman, R.C., and Brasier, K. 2006. People in a forested lake

district. In Long-Term Dymanics of Lakes in the Landscape. Edited by J.J. Magnuson, T.K.

Kratz and B.J. Benson. Oxford University Press, Inc., New York, NY. pp. 214-235.

Jorgensen, B.S., and Stedman, R.C. 2001. Sense of place as an attitude: lakeshore owners

attitudes toward their properties Journal of Environmental Psychology 21(3): 233-248.

Jorgensen, C., Enberg, K., Dunlop, E.S., Arlinghaus, R., Boukal, D.S., Brander, K., Ernande,

B., Gardmark, A., Johnston, F., Matsumura, S., Pardoe, H., Raab, K., Silva, A., Vainikka, A.,

Dieckmann, U., Heino, M., and Rijnsdorp, A.D. 2007. Ecology - Managing evolving fish

stocks. Science 318(5854): 1247-1248. doi: 10.1126/science.1148089.

17

Kitchell, J.F., and Carpenter, S.R. 1993. Variability in lake ecosystems: complex responses

by the apical predator. In Humans as Components of Ecosystems. Edited by M.J. McDonnell

and S.T.A. Pickett. Springer-Verlag, New York, NY. pp. 111-124.

Kratz, T.K., Bolgrien, D.W., Giblin, S., and O'Leary, J. 2002. Human influences on the

spacial distribution of coarse woody debrs in lakes in Northern Wisconsin, USA. Verh.

Internat. Verin. Limnol. 28: 1-4.

Lewin, W.C., Arlinghaus, R., and Mehner, T. 2006. Documented and potential biological

impacts of recreational fishing: Insights for management and conservation. Rev. Fish. Sci.

14(4): 305-367. doi: 10.1080/10641260600886455.

Marburg, A.E., Turner, M.G., and Kratz, T.K. 2006. Natural and anthropogenic variation in

coarse wood among and within lakes. Journal of Ecology 94: 558-568.

Meyer, S.L. 1975. Data Analysis for Scientists and Engineers. Wiley, New York, NY.

Olson, D.E., and Cunningham, P.K. 1989. Sport-Fisheries Trends Shown by an Annual

Minnesota Fishing Contest over a 58-Year Period. North Am. J. Fish Manage. 9(3): 287-297.

Olson, M.H. 1996. Ontogenetic niche shifts in largemouth bass: variability and consequences

for first-year growth. Ecology 77(1): 179-190.

Olson, M.H., Mittelbach, G.G., and Osenberg, C.W. 1995. Competition between predator

and prey: resource-based mechanisms and implications for stage-structured dynamics.

Ecology 76(6): 1758-1771.

18

Paradis, Y., Bertolo, A., Perot, A., Debas, S., and Magnan, P. 2006. Do benthivory and

piscivory result in similar growth in walleye? J. Fish Biol. 69(5): 1317-1329. doi:

10.1111/j.1095-8649.2006.01187.x.

Penaloza, L.J. 1991. Boating Pressure on Wisconsin's Lakes and Rivers. Department of

Natural Resources Technical Bulletin No. 174.

Peterson, G.D., Beard, T.D., Beisner, B.E., Bennet, E.M., Carpenter, S.R., Cumming, G.S.,

Dent, C.L., and Havlicek, T.D. 2003. Assessing future ecosystem services a case study of the

Northern Highlands Lake District, Wisconsin. Conserv. Ecol. 7(3): 24. doi: 1.

Philipp, D.P., Cooke, S.J., Claussen, J.E., Koppelman, J.B., Suski, C.D., and Burkett, D.P.

2009. Selection for Vulnerability to Angling in Largemouth Bass. Transactions of the

American Fisheries Society 138(1): 189-199. doi: 10.1577/t06-243.1.

Pope, K.L., and Wilde, G.R. 2004. Effect of catch-and-release angling on growth of

largemouth bass, Micropterus salmoides. Fisheries Management and Ecology 11(1): 39-44.

Post, D.M., Kitchell, J.F., and Hodgson, J.R. 1998. Interactions among adult demography,

spawning date, growth rate, predation, overwinter mortality, and the recruitment of

largemouth bass in a northern lake. Can. J. Fish. Aquat. Sci. 55(12): 2588-2600.

Post, J.R., Sullivan, M., Cox, S., Lester, N.P., Walters, C.J., Parkinson, E.A., Paul, A.J.,

Jackson, L., and Shuter, B.J. 2002. Canada's recreational fisheries: The invisible collapse?

Fisheries 27(1): 6-17.

19

Postel, S., and Carpenter, S.R. 1997. Freshwater ecosystem services. In Nature's Services.

Edited by G. Daily. Island Press, Washington, DC. pp. 195-214.

R Development Core Team. 2008. R: A Language and Environment for Statistical

Computing. R Foundation for Statistical Computing, Vienna, Austria.

Reznick, D.N., and Ghalambor, C.K. 2005. Can commercial fishing cause evolution?

Answers from guppies (Poecilia reticulata). Can. J. Fish. Aquat. Sci. 62: 791-801.

Riera, J., Voss, P.R., Carpenter, S.R., Kratz, T.K., Lillesand, T.M., Schnaiberg, J.A., Turner,

M.G., and Wegener, M.W. 2001. Nature, society and history in two contrasting landscapes in

Wisconsin, USA - Interactions between lakes and humans during the twentieth century, pp.

41-51.

Sass, G.G., Kitchell, J.F., Carpenter, S.R., Hrabik, T.R., Marburg, A.E., and Turner, M.G.

2006. Fish community and food web responses to a whole-lake removal of coarse woody

habitat. Fisheries 31(7): 321-330.

Schindler, D.E., Geib, S.I., and Williams, M.R. 2000. Patterns of Fish Growth along a

Residential Development Gradient in North Temperate Lakes. Ecosystems 3: 229-237.

Schnaiberg, J., Riera, J., Turner, M.G., and Voss, P.R. 2002. Explaining human settlement

patterns in a recreational lake district: Vilas County, Wisconsin, USA. Environmental

Management 30(1): 24-34.

Siepker, M.J., Cooke, S.J., Wahl, D.H., and Philipp, D.P. 2009. Individual Reproductive

Success of Largemouth Bass and Smallmouth Bass Subjected to Different Components of

20

Competitive Angling Events. Transactions of the American Fisheries Society 138(4): 818-

825. doi: 10.1577/t07-269.1.

Siepker, M.J., Ostrand, K.G., Cooke, S.J., Philipp, D.P., and Wahl, D.H. 2007. A review of

the effects of catch-and-release angling on black bass, Micropterus spp.: implications for

conservation and management of populations. Fisheries Management and Ecology 14: 91-

101.

Siepker, M.J., Ostrand, K.G., and Wahl, D.H. 2006. Effects of angling on feeding by

largemouth bass. J. Fish Biol. 69(3): 783-793.

Stearns, F. 1951. The Composition of the Sugar Maple-Hemlock-Yellow Birch Association

in Northern Wisconsin. Ecology 32(2): 245-265. doi: doi:10.2307/1930417.

Stedman, R.C. 2003. Is it really just a social construction? The contribution of the physical

environment to sense of place. Soc. Nat. Resour. 16(8): 671-685. doi:

10.1080/08941920390217627.

Suski, C.D., and Philipp, D.P. 2004. Factors affecting the vulnerability to angling of nesting

male largemouth and smallmouth bass. Transactions of the American Fisheries Society

133(5): 1100-1106.

Uusi-Heikkila, S., Wolter, C., Klefoth, T., and Arlinghaus, R. 2008. A behavioral perspective

on fishing-induced evolution. Trends Ecol. Evol. 23(8): 419-421. doi:

10.1016/j.tree.2008.04.006.

21

Wagner, T., Hayes, D.B., and Bremigan, M.T. 2006. Accounting for multilevel data

structures in fisheries data using mixed models. Fisheries 31(4): 180-187.

Wisconsin Department of Natural Resources. 2005. Wisconsin Lakes PUB-FH-800. Bureau

of Fisheries and Habitat Management. pp. 142-156.

22

List of Tables

PAGE

Table 1. Lake Characteristics ………………………………………………… 23

Table 2. Largemouth Bass Lengths and Ages ……………………………… 24

Table 3. Model Summary ………………………………………………… 25

23

Table 1. Summary of physical lake characteristics, building density (no. km-1

), and

largemouth bass (Micropterus salmoides) (n; young-of-the-year and yearlings removed) from

each lake.

Lake Lake

Code

Perimeter

(km) Area (ha)

Maximum

Depth (m)

Building

Density

(no. km-1

)

Largemouth

Bass (n)

Allequash AL 10.2 165.3 24 0.0 30

Arrowhead AR 3.5 40.1 43 45.8 29

Black Oak BO 12.0 230.1 85 18.0 27

Brandy BR 3.5 45.1 44 30.1 29

Camp CP 2.9 17.6 31 0.0 30

Day DY 5.5 47.3 48 0.2 30

Found FD 6.4 139.3 21 16.6 30

Johnson JN 3.6 34.7 42 26.2 27

Little Crooked LC 4.8 63.8 20 5.5 30

Little John LJ 5.3 63.4 19 2.1 30

Little Rock LR 1.4 8.1 10 0.0 30

Little St. Germain LSG 23.3 402.2 53 19.8 27

Moon MN 3.4 54.4 38 15.0 27

Round RD 3.7 71.5 25 0.3 26

Upper Buckatabon UB 13.2 211.4 47 12.6 30

White Sand WS 9.3 304.6 71 5.8 26

24

Table 2. Summary of largemouth bass (Micropterus salmoides) lengths and ages across all

lakes.

Description Min Max Mean Standard

Deviation

Length at capture across all

lakes (mm) 118 420 264.537 72.283

Max length per lake (mm) 249 420 371.375 44.739

Min length at age one per lake

(mm) 52 77 64.336 7.068

Age at capture across all lakes

(years) 2 16 7.729 3.843

Max age per lake (years) 8 16 13.625 2.247

25

Table 3. Longitudinal multilevel model random and fixed effects.

Model Details Random Effects Fixed Effects

Model Level Sampled

Unit

Sample

Size Parameter

Standard

Deviation Correlation

Coefficient

Parameter

Coefficient

Estimate

Coefficient

Standard

Deviation

Level 1: Fish Annuli 3540 Residual 0.270

Intercept 0.979 Level 2: Lake Fish 458

Slope 0.192 -0.997

Intercept 9.2358 0.4471 Intercept 1.254

Building Density 0.0466 0.0250

ln(length) -1.1416 0.0862 Level 3: Region Lake 16

Slope 0.241

-0.994

Interaction -0.0092 0.0048

26

Figure Captions

Figure 1. Map of study lakes in the Northern Highlands Lake District located in Vilas

County, Wisconsin, USA.

Figure 2. Growth rate (mm year-1

) at length (mm) for each observed annulus (n= 3540)

obtained from back-calculated largemouth bass (Micropterus salmoides) scales (n=458).

Refer to Table 1 for lake codes.

Figure 3. Longitudinal multilevel model fit: fish level random effects. Model predicted

ln(growth rate) (ln(mm year-1

)) plotted against (a) observed ln(growth rate) and (b) model

residuals. Model residuals across sampled (c) fish length (ln(mm)) and (d) building density

(no. km-1

). Model residuals across sampled building density (no. km-1

) of fish sizes (e) 4.17

– 4.5 ln(mm) and (f) 4.5 – 4.75 ln(mm). Points between vertical dotted lines in (c) are shown

in (e) and (f). Gray line indicates 1:1 line (a) or zero residual line (b-f). Residuals are

jittered when plotted against building density (d-f).

Figure 4. Longitudinal multilevel lake level random effects model fit. Size-specific

individual random effects fish growth rate (mm year -1

; loge transformed) means and standard

deviations for each lake plotted across the regional building density gradient (no. km-1

) for

three sizes of largemouth bass (Micropterus salmoides) (a) 75 mm, (b) 164 mm, and (c) 375

mm; plotted with longitudinal multilevel model regression lines.

27

Figure 5. The effect of building density across the observed gradient of largemouth bass

(Micropterus salmoides) length (mm). A value above the zero line indicates a size with a

positive relationship between building density and growth rate; a value below the zero line

indicates a negative relationship. Dotted lines indicate ± one standard deviation calculated

from the covariance matrix of model parameters using standard error propagation formulae.

The building effect is interpreted as the average change in growth rate of fish of a given

length if one building km-1

is added to the shoreline; units are

!

ln(mm year-1)

building km-1.

Figure 6. Model predicted growth trajectories given an initial length of 83 mm, the mean

sampled length at age one, for a largemouth bass (Micropterus salmoides) in a lake with no

development and a largemouth bass in a lake with high development (45.8 buildings km-1

).

Gray dotted line represents legal length limit in the NHLD, 356 mm (14 inches).

28

Figure 1.

Map of study lakes in the Northern Highlands Lake District located in Vilas County,

Wisconsin, USA.

29

Figure 2.

Growth rate (mm year-1

) at length (mm) for each observed annulus (n= 3540) obtained from

back-calculated largemouth bass (Micropterus salmoides) scales (n=458). Refer to Table 1

for lake codes.

30

Figure 3.

Longitudinal multilevel model fit: fish level random effects. Model predicted ln(growth rate)

(ln(mm year-1

)) plotted against (a) observed ln(growth rate) and (b) model residuals. Model

residuals across sampled (c) fish length (ln(mm)) and (d) building density (no. km-1

). Model

residuals across sampled building density (no. km-1

) of fish sizes (e) 4.17 – 4.5 ln(mm) and

(f) 4.5 – 4.75 ln(mm). Points between vertical dotted lines in (c) are shown in (e) and (f).

Gray line indicates 1:1 line (a) or zero residual line (b-f). Residuals are jittered when plotted

against building density (d-f).

31

Figure 4.

Longitudinal multilevel lake level random effects model fit. Size-specific individual random

effects fish growth rate (mm year -1

; loge transformed) means and standard deviations for

each lake plotted across the regional building density gradient (no. km-1

) for three sizes of

largemouth bass (Micropterus salmoides) (a) 75 mm, (b) 164 mm, and (c) 375 mm; plotted

with longitudinal multilevel model regression lines.

32

Figure 5.

The effect of building density across the observed gradient of largemouth bass (Micropterus

salmoides) length (mm). A value above the zero line indicates a size with a positive

relationship between building density and growth rate; a value below the zero line indicates a

negative relationship. Dotted lines indicate ± one standard deviation calculated from the

covariance matrix of model parameters using standard error propagation formulae. The

building effect is interpreted as the average change in growth rate of fish of a given length if

one building km-1

is added to the shoreline; units are

!

ln(mm year-1)

building km-1.

33

Figure 6.

Model predicted growth trajectories given an initial length of 83 mm, the mean sampled

length at age one, for a largemouth bass (Micropterus salmoides) in a lake with no

development and a largemouth bass in a lake with high development (45.8 buildings km-1

).

Gray dotted line represents legal length limit in the NHLD, 356 mm (14 inches).

34

List of Appendices

PAGE

Appendix I. Mean Length at Age …………………………...……………..….. 35

Appendix II. Standard Deviations of Mean Length at Age …………………….. 36

Appendix III. Number of Annuli per Mean Length at Age …………………….. 37

Appendix IV. Model R – Code …………………………………………...…… 38

Appendix V. Model Summary in Matrix Notation ……….……………….…… 39

Appendix VI. Building Density Effect Error Propagation ….…………...…….. 40

Appendix VII. Model Selection of Potential Covariates ….…..………….…….. 41

Appendix VIII. Building Density and Coarse Woody Habitat Density ………… 44

Appendix IX. Coarse Woody Habitat Density Effect …………………...……... 45

35

Appendix I. Mean length at age. Refer to Table 2 for lake names.

Lake

ID Annuli

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16

AL 87.15 131.32 164.38 189.84 214.78 237.49 259.14 278.80 298.10 316.78 333.82 347.85 363.79 382.41 393.65 405.99

AR 87.99 130.96 158.00 178.65 196.97 212.55 223.15 235.44 254.37 273.65 296.98 309.39 321.14 317.53

BO 79.18 122.57 165.09 200.39 233.17 259.32 289.83 312.08 334.94 339.65 357.19 367.29

BR 85.51 123.81 151.68 176.11 199.40 220.24 238.18 255.37 273.36 288.29 303.75 337.05 361.89 392.38

CR 87.77 126.16 158.61 182.33 203.20 222.50 239.49 255.66 270.52 285.01 299.80 316.99 327.55 341.36

DY 67.57 107.94 139.10 165.15 189.99 211.83 233.87 251.64 267.81 285.13 299.31 316.29 330.62 345.74 343.19 359.33

FD 74.90 115.10 150.69 178.59 201.92 222.16 240.68 257.23 272.88 287.41 301.83 319.98 332.70 349.51 396.49 407.64

JN 86.90 130.16 164.32 191.51 216.79 236.76 254.47 270.88 286.99 310.90 330.95 340.10

LC 75.74 110.73 146.90 174.84 203.21 239.11 255.09 266.99 279.11 284.21 295.90

LJ 76.78 117.58 147.60 171.33 189.99 206.07 219.58 232.99

LR 92.07 128.17 157.53 183.83 204.89 224.84 244.46 261.72 279.06 305.76 322.18 339.91 351.78 348.87 360.85 373.08

LSG 98.54 138.17 170.65 198.60 222.66 245.36 264.32 276.56 291.19 303.66 335.65 371.80

MN 95.61 147.18 184.12 213.65 238.29 257.47 271.40 286.62 299.99 316.14 331.12 343.36 364.70 387.79

RD 81.82 126.99 159.65 181.68 206.11 229.30 248.66 267.11 281.32 293.25 304.01 316.53 330.31 340.95

UB 83.47 115.15 143.07 169.15 193.07 212.19 232.30 248.71 270.03 287.47 309.33 326.76 345.92

WS 64.68 95.56 125.28 154.95 184.01 208.78 231.54 250.81 267.43 282.76 295.26 312.61 328.72 347.84 378.07 387.88

36

Appendix II. Standard deviations of mean length at age shown in Appendix I. Refer to Table 2 for lake names

Lake

ID Annuli

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16

AL 9.46 11.29 12.08 12.85 12.66 13.25 14.21 13.54 13.63 14.33 15.05 15.24 15.93 7.14 9.01 NA

AR 14.12 15.03 13.96 13.08 12.23 12.18 13.00 13.62 16.27 13.45 9.84 17.11 16.05 NA

BO 11.77 17.69 15.66 18.24 19.82 20.76 21.54 23.38 27.04 5.01 6.18 6.71

BR 9.27 11.63 10.52 10.98 10.53 11.07 11.05 12.98 15.30 18.11 22.14 22.11 20.54 NA

CR 11.70 13.77 13.97 14.08 14.47 14.72 15.41 16.13 17.33 17.00 18.76 18.38 7.54 NA

DY 7.41 8.88 11.14 13.21 16.94 16.89 17.17 16.93 17.53 19.34 19.16 18.67 21.56 22.76 9.55 NA

FD 9.02 11.80 9.63 11.31 10.93 12.44 11.57 10.76 13.76 16.88 21.62 21.77 26.52 44.45 NA NA

JN 11.29 15.48 14.39 9.51 9.81 8.53 9.50 11.00 12.81 7.06 8.98 NA

LC 9.91 13.67 13.74 18.21 22.93 10.87 10.51 9.80 10.85 2.43 2.25

LJ 8.36 10.97 10.04 10.35 7.73 8.64 1.40 0.61

LR 9.59 11.63 15.76 17.31 18.36 17.90 16.12 18.09 20.08 17.30 15.80 11.76 14.99 NA NA NA

LSG 12.69 12.30 14.49 14.57 11.45 14.04 16.39 18.22 18.08 18.60 29.66 NA

MN 12.10 14.68 13.89 14.17 13.59 15.04 12.85 12.77 14.05 16.84 17.66 18.61 17.70 6.81

RD 8.71 9.19 10.50 10.09 2.57 3.96 6.52 5.29 6.55 8.64 NA NA NA NA

UB 10.22 11.85 14.60 14.70 15.54 10.27 10.48 10.33 9.04 7.21 NA NA NA

WS 6.67 10.07 12.73 15.07 19.05 21.11 24.23 23.67 21.10 17.69 14.24 13.50 12.90 14.97 NA NA

37

Appendix III. Number of annuli per mean length at age in Appendix I. Refer to Table 2 for lake names.

Lake

ID Annuli

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16

AL 30 30 29 27 27 27 27 27 27 26 24 23 19 6 3 1

AR 29 29 29 28 28 27 21 20 7 5 3 2 2 1

BO 27 27 23 21 18 16 10 8 5 4 3 2

BR 29 29 28 23 23 22 22 21 20 17 13 7 5 1

CR 30 30 30 30 30 30 29 28 22 21 16 11 5 1

DY 30 30 29 29 29 29 27 26 26 21 18 13 9 6 2 1

FD 30 30 29 29 29 29 28 22 13 11 7 6 5 2 1 1

JN 26 26 25 24 24 23 15 15 13 8 4 1

LC 30 30 30 15 11 7 7 5 4 2 2

LJ 30 30 28 23 17 15 6 2

LR 30 30 29 29 29 27 25 21 15 8 7 6 3 1 1 1

LSG 27 27 13 7 7 7 6 5 5 5 2 1

MN 27 27 26 25 25 25 23 21 20 14 11 11 6 3

RD 26 26 16 8 4 4 4 3 3 3 1 1 1 1

UB 30 30 20 19 17 12 11 9 7 3 1 1 1

WS 27 27 22 21 19 17 15 15 14 13 10 9 8 5 1 1

38

Appendix IV. R-code for longitudinal multilevel model of largemouth bass size-specific

growth across the Wisconsin’s Northern Highland Lake District regional gradient of human

development. Model parameters include growth (mm year-1

), length (mm), and building

density (no. km-1

). Model includes two group levels: individual and lake.

lmer( ln(growth) ~ ln(length)*(building density) + (1 + ln(length) | lake / individual),

control=list(gradient = FALSE, niterEM = 0), method="ML")

39

Appendix V. Longitudinal multilevel analysis results in matrix notation.

!

yi is ln(growth)

(mm year-1

) at observation i,

!

xi is the ln(length) at observation i, and

!

zk is the building

density (no. km-1

) in lake k.

Level 1: Fish level; within a fish within a lake

!

yi ~ " #0 j[ i] + #

1 j[ i]xi , 0.2702( ) , for i =1,...,3540 observations

Level 2: Lake level; between fish within lake

!

"0 j

"1 j

#

$ %

&

' ( ~ )

*0k[ j ]

* 1k[ j ]

#

$ %

&

' ( ,

0.9792

+0.997 +0.997

0.1922

#

$ %

&

' (

#

$ % %

&

' ( ( , for j =1,...,458 fish

Level 3: Regional level; between lakes

!

"0k

"1 k

#

$ %

&

' ( ~ )

9.236 + 0.047zk

*1.142 + *0.009xizk

#

$ %

&

' ( ,

1.2542

*0.994 *0.994

0.2412

#

$ %

&

' (

#

$ % %

&

' ( ( , for k =1,...,16 lakes

40

Appendix VI. Building density effect error propagation with R – Cran script

The following Taylor series expansion was used to propagate error in the building density

effect analysis.

!

Building Density Effect = "2 + "3L ± #

# = var "2( ) + L2 * var "3( ) + 2L * cov "2,"3( )

cov "2,"3( ) = cor "2,"3( ) var "2( ) * var "3( )

Where :

"2 = building density coefficient

"3 = interaction coefficient

L = fish length

R- Cran script:

se_mod=sqrt(diag(vcov(M1))) #coefficient standard errors; where M1 is the main model

se_build=se_mod[3]

se_int=se_mod[4]

cov_B_int=-0.994*sqrt((se_build^2)*(se_int^2)) #cov=(correlation coefficient of beta2 and

beta3 AKA r)*sqrt(var(beta2)*var(beta3))

var_E=((se_build^2)+(ln_l_range^2)*(se_int^2)+(2*(ln_l_range)*(cov_B_int)))

sd_E=sqrt(var_E)

upper_E=E_build+sd_E

lower_E=E_build-sd_E

41

Appendix VII. Model selection of potential covariates: reduced, stepwise, and global models

To account for the potential influence of the covariates on LMB growth, we initially

used three models in the analysis: a reduced model based on the sample design in which

growth was defined as a function of length, building density, and the interaction between

length and building density (Model 1, reduced model); a stepwise model produced from

stepwise selection in which covariates were systematically added or removed to evaluate

covariate effect, covariates that improved model fit were added to the reduced model (Model

2, stepwise model); and a global model in which all the covariates were added to the reduced

model (Model 3, global model).

Model fixed effects included an intercept, fish length, lake building density, the

interaction term, and, in the case of models two and three, the covariates (Table A1).

Table A1. Covariates included in global model and used in forward stepwise selection in

addition to building density (no. km-1

) and length (mm). All covariate data was obtained

from the North Temperate Lakes Long-Term Ecological Research online database (Carpenter

and Kratz 2001).

Covariate Units

Coarse woody habitat density

no. km-1

Conductance !S/cm

Maximum lake depth m

Secchi depth m

Chlorophyll a concentration !g/L

Dissolved organic carbon mg/L

Shoreline morphometry index**

!

perimeter

2 " area( )

Area ha

** index from Scheuerell and Schindler (2004) and Wetzel (2001)

The stepwise model, in which covariates that could plausibly improve model fit were

added to the reduced model, followed the same model structure as the reduced model (Model

42

1). Model levels one and two of the stepwise model are identical to model one (Model 1.1

and 1.2). The covariates (Table A1) were added as fixed effects at the third level:

!

"0k

"1 k

#

$ %

&

' ( ~ )

*00k

+ *01kzk

+ X1k

+ ...+ Xnk

* 10k

+ *11kxizk

#

$ %

&

' ( ,

+0"2

,+0"2 +

1"2

,+

0"2 +

1"2

+1"2

#

$ %

&

' (

#

$ % %

&

' ( ( , for k =1,...,K lakes(Model 2.3)

Where

!

X1k

+ ...+ Xnk

are the lake k specific effects of stepwise selected covariates 1 through

n on the intercept. Equation 2.3 is a bivariate multiple regression of lake-specific growth

parameters !0k and !1k on the fixed effects of building density and the covariates.

The global model, in which all covariates were added to the reduced model, followed

the same model structure as the above models. Model levels one and two of the global model

are identical to model one (Model 1.1 and 1.2); all of the covariates were added as fixed

effects at the third level:

!

"0k

"1 k

#

$ %

&

' ( ~ )

*00k

+ *01kzk

+ X1k

+ ...+ X8k

* 10k

+ *11kxizk

#

$ %

&

' ( ,

+0"2

,+0"2 +

1"2

,+

0"2 +

1"2

+1"2

#

$ %

&

' (

#

$ % %

&

' ( ( , for k =1,...,K lakes

(Model 3.3)

Where

!

X1k

+ ...+ X8k

are the lake k specific effects of the eight covariates (Table 1) on the

intercept. Equation 3.3 has the same form as equation 2.3, but in equation 3.3 all covariates

were retained in the model.

Stepwise selection of models was based on a conservative ! DIC (deviance

information criterion, (Gelman and Hill 2008)) cutoff of 4; i.e. if the addition of a covariate

did not decrease the DIC by at least 4, the covariate was not included at that step. We

selected this cutoff based on the discussion of model selection using information statistics in

Burnham and Anderson (1998). Because we recognize that any criterion for model selection

is arbitrary, we compared the stepwise outcome (model 2) to the model with no covariates

(model 1) and the model with all covariates (model 3). Because all covariates were fixed-

effects, models were fit by maximizing the log-likelihood not the restricted log-likelihood

(method = “ML” in lmer function).

Stepwise selection indicated the fit of the reduced model could not be improved with

the addition of any one covariate. The stepwise selection process ended at the first step with

a maximum DIC reduction of only 2.5 when dissolved organic carbon was added to the

model, well below the cutoff value of 4 per additional covariate. Therefore, model 2 was

identical to model 1. Likewise, adding all eight covariates into a global model (Model 3)

43

only reduced the DIC value by 6.6, well below the cutoff value. Consequently we adopted

model 1, the reduced model, as the best model.

44

Appendix VIII. The relationship between building density and coarse woody habitat density

for 61 lakes from the NHLD (data from Carpenter and Kratz 2001). Black circles represent

our study lakes.

45

Appendix IX. The effect of coarse woody habitat (CWH) density across the observed

gradient of largemouth bass length (mm). A value above the zero line indicates a size class

with a positive relationship between CWH density and growth rate; a value below the zero

line indicates a negative relationship. Dotted lines indicate ± one standard deviation

calculated from the covariance matrix of model parameters using standard error propagation

formulae. Calculations are based on the analog of Model 1 with CWH replacing building

density. The CWH density effect is interpreted as the average change in growth rate of fish of

a given length if one piece of CWH km-1

is added to the shoreline; units are

!

ln(mm year-1)

CWH km-1.

Lakeshore residential development and growthof largemouth bass (Micropterus salmoides):a cross-lakes comparison

Introduction

Critical lake littoral habitat, riparian habitat, andecosystem function are altered as a result of lakeshoreresidential development (LRD) (Engel & Pederson1998; Francis & Schindler 2009). Fallen trees in lakelittoral zones, for instance, serve as an importantrefuge for fish (Sass et al. 2006; Roth et al. 2007), asubstrate for invertebrate prey production (VanderZanden & Vadeboncoeur 2002; Roth et al. 2007), andas fish nesting habitat (Hunt & Annett 2002), but thiscoarse woody habitat is negatively correlated withLRD (Christensen et al. 1996; Jennings et al. 2003;Francis & Schindler 2006). Future inputs of coarsewoody habitat into the littoral zone are greatly reducedby reductions in riparian vegetation associated withLRD (Francis & Schindler 2006; Marburg et al. 2006).LRD is inversely correlated with littoral macrophyterichness (Bryan & Scarnecchia 1992) and cover of

floating leaf and emergent vegetation (Jennings et al.2003; Radomski 2006). Shifts in macroinvertebratecommunities (Brauns et al. 2007; Rosenberger et al.2008) and reduced organic sediments in the littoralzone (Francis et al. 2007) are also associated withLRD. Likewise, exploitation rates of game fishes areexpected to increase with LRD (NRC 1992). Thesechanges associated with LRD have the potential toramify through both aquatic and terrestrial food webs(Engel & Pederson 1998).

Altered habitat structure and ecosystem functionassociated with LRD may drive changes in fishecology. A fish diet survey by Francis & Schindler(2009) found a negative correlation between LRD andenergetically favourable food sources, likely due toaltered riparian habitat. Within developed lakes, blackcrappie (Pomoxis nigromaculatus) nest adjacent toundeveloped sections of shoreline and associate withmacrophytes, which were less abundant adjacent to

Ecology of Freshwater Fish 2011: 20: 92–101Printed in Malaysia Æ All rights reserved

! 2010 John Wiley & Sons A/S

ECOLOGY OFFRESHWATER FISH

Gaeta JW, Guarascio MJ, Sass GG, Carpenter SR. Lakeshore residentialdevelopment and growth of largemouth bass (Micropterus salmoides):a cross-lakes comparison.Ecology of Freshwater Fish 2011: 20: 92–101. ! 2010 John Wiley &Sons A ⁄S

Abstract – Lakeshore residential development is associated with changesin littoral habitat, riparian habitat, and ecosystem function with potentialimpacts ramifying through aquatic food webs. Effects of these changes oneconomically important game fishes may vary with fish size. Weinvestigated largemouth bass (Micropterus salmoides) size-specific growthrates across 16 lakes spanning the range of lakeshore residentialdevelopment in Wisconsin’s Northern Highland Lake District using alongitudinal multilevel model. Growth rates of small fish had a strongpositive relationship with lakeshore residential development. The strengthof the relationship decreased with length and became increasingly negativefor fish longer than 210 mm. This pattern may be driven by a release fromdensity-dependent growth, shifts in available prey sources, reducedmacrophyte cover, or angling-induced selection pressures. Regardless ofthe mechanism, our results indicate, relative to undeveloped lakes,largemouth bass in highly developed lakes take 1.5 growing seasons longerto enter the fishery (356 mm).

J. W. Gaeta1, M. J. Guarascio2,G. G. Sass3, S. R. Carpenter11Center for Limnology, University of Wisconsin –Madison, Madison, WI, USA, 2Department ofZoology, University of Wisconsin – Madison,Madison, WI, USA, 3Illinois River BiologicalStation, Illinois Natural History Survey, Instituteof Natural Resource Sustainability, University ofIllinois at Urbana-Champaign, Havana, IL, USA

Key words: largemouth bass; Micropterussalmoides; fish growth; lakeshore residentialdevelopment; longitudinal multilevel model;longitudinal hierarchical model

JeremeW. Gaeta, Center for Limnology, Universityof Wisconsin – Madison, 680 North Park St,Madison,WI 53706, USA; e-mail: jgaeta@wisc.edu

Accepted for publication October 8, 2010

92 doi: 10.1111/j.1600-0633.2010.00464.x

developed shorelines (Reed & Pereira 2009). Thesame trend was identified for largemouth bass(Micropterus salmoides) although a mechanism wasnot identified. Reductions in littoral habitat are thesuspected driver of the negative correlation betweenLRD and small fish (£100 mm) aggregations (Scheue-rell & Schindler 2004). Bluegill (Lepomis macrochi-rus) growth rates negatively correlate with LRD(Schindler et al. 2000). Schindler et al. (2000) alsoidentified a marginally significant negative relation-ship between LRD and growth rates of the largestsize class (400 mm) of largemouth bass but did notidentify a conclusive relationship for smaller sizeclasses. These shifts in behaviour and growth asso-ciated with LRD may have implications for recrea-tional fisheries.

In many areas of the United States, such asWisconsin’s Northern Highland Lake District(NHLD), recreational fisheries are a pillar of theregional economy (Penaloza 1991; Postel & Carpenter1997; Peterson et al. 2003). Therefore, understandingand quantifying whether humans alter these importantfisheries is essential. We investigated the relationshipbetween largemouth bass growth rate across both fishsize and LRD. Our study built upon the findings ofSchindler et al. (2000) by nearly doubling the samplesize of lakes and fish, spanning a larger gradient ofLRD, avoiding potential confounding effects of coarsewoody habitat, and employing a more sensitivehierarchical analysis method.

We performed a cross-lakes comparison of 16lakes spanning the full regional gradient of LRD(0–45.8 buildingsÆkm)1) in the NHLD to test for arelationship between LRD and largemouth bassgrowth across fish size. We used a longitudinal,

multilevel approach to estimate growth responsesacross a range of fish sizes and found that growthrates of both small and large largemouth bass weresignificantly related to LRD.

Methods

Study area

We surveyed largemouth bass size-specific growthrates among 16 lakes spanning the known gradient oflakeshore residential development in Wisconsin’sNHLD (Fig. 1). The NHLD is a formerly glaciated,lake-rich region spanning about 5330 km2 withapproximately 7600 lakes (Peterson et al. 2003;Carpenter et al. 2007) and is vegetated by uplandconifer-hardwood forests (Stearns 1951; Brown &Curtis 1952). Human population densities in theregion have increased nearly fivefold in the last halfcentury (Carpenter et al. 2007), and since the 1960s,the majority of that development has occurred on lakeshorelines (Schnaiberg et al. 2002). In the early 2000s,Vilas County, the county of our study lakes, had nearly16,500 buildings within 100 m of lake shorelines(Riera et al. 2001).

Lakeshore residential development of our studylakes ranged from 0 to 45.8 buildingsÆkm)1 within100 m of lake shorelines (Table 1). Predators oflargemouth bass from the Esocidae family (e.g.,muskellunge Esox masquinongy or northern pike Esoxlucius) were common or abundant in all but three lakesin our study: Camp Lake, Little Rock Lake, and DayLake (Wisconsin Department of Natural Resources2005). Study lakes were selected for low coarsewoody habitat densities (0–125 logsÆkm)1) and

Fig. 1. Map of study lakes in the NorthernHighlands Lake District located in VilasCounty, Wisconsin, USA.

Lakeshore residential development and growth of largemouth bass

93

spanned only 13% of the observed regional coarsewoody habitat gradient (Christensen et al. 1996) toreduce potential confounding effects of coarse woodyhabitat and lakeshore residential development.

Fish sampling

We sampled largemouth bass between June andAugust of 2006 primarily via electrofishing alongthe lake perimeter. Fish were collected via anglingwhen lake conductivity was not suitable for electro-fishing. Thirty fish were collected from each lake todetermine size-specific growth rates. Fish length (totallength; mm) was recorded, and 5 to 10 scales werecollected from each fish from the area posterior to adepressed pectoral fin. We removed young-of-the-yearfish from the analysis owing to the lack of annuli, andas a result, sample size varied between lakes(Table 1). Scales from yearling fish and older weresonicated and pressed between two slides. Nonregen-erated scales were read into a digital imaging system.Annual growth rates (mmÆyear)1) were determinedusing Fraser-Lee’s method of back calculation withCarlander’s recommended constant of 20 mm forlargemouth bass (Carlander 1982) as used inSchindler et al. (2000). It is possible that LRD couldhave changed during the lifetimes of the longer-livedbass in our study, especially because LRD boomedduring the 1990s but slowed substantially during the2000s (Carpenter et al. 2007). To eliminate anypotential effects of changing LRD levels, only theannual growth estimates from 2001 to 2005 wereincluded as repeated measures of annual growth foreach fish.

Estimating ages from hard structures such as scalesand otoliths is challenging (Buckmeier & Howells2003) especially for older largemouth bass (Maraldo& Maccrimmon 1979); however, largemouth basshave been successfully aged up to 16 years (Buckme-ier & Howells 2003). We acknowledge that back-calculated growth rates from any structure must berecognised as estimates with inherent errors (Maceinaet al. 2007), and for that reason, we used relativelylarge sample sizes (N = 27–30 per lake) to evaluategrowth responses in our study.

Statistical analysis

Our data were hierarchically structured with repeatedmeasures of annulus-specific growth observations(mm) nested within individual fish growth rates(mmÆyear)1), individual fish nested within lakes, andeach lake with a unique set of lake characteristics, suchas LRD. We designed our analysis around thehierarchical nature of the data and tested for arelationship between LRD and largemouth bassgrowth rate using a longitudinal (repeated measures)multilevel model (Goldstein 1995; Ai 2002; Wagneret al. 2006). Unlike least squares regression methodsmore commonly used to determine size-specificgrowth rates, the longitudinal multilevel modelapproach allows us to account for repeated measuresof annuli and to quantify, rather than lose, variation ingrowth at multiple levels (among fish and lakes).

We performed all analyses in R-Cran statisticalpackage (R Development Core Team 2010; package:‘lme4’ version 0.999375-33). Multilevel modellingmethods followed procedures outlined in Gelman &Hill (2008). Growth rates were loge-transformed priorto analysis. Likewise, fish length was loge-trans-formed and grand-mean-centred prior to analysis. Weallowed both slopes and intercepts to vary as randomeffects.

Based on sample design, we expected our model toinclude fish length and LRD; however, we performedforward stepwise selection with a suite of additionalcovariates to account for unexplained variance. Thecovariates included in the stepwise selection processwere LRD, coarse woody habitat, conductance, max-imum lake depth, Secchi depth, chlorophyll a concen-tration, dissolved organic carbon, area, and an index ofshoreline morphometry (Wetzel 2001). The initialmodel defined growth only as a function of length. Atevery step, model fit was assessed as each covariate aswell as the interaction of the covariate with fish lengthwas systematically added to the initial model as a fixedeffect. Models were fit by maximizing the loglikelihood. At each step, the covariate or interactionwith the greatest change in deviation informationcriterion (DIC) of four or more was included, as

Table 1. Summary of physical lake characteristics and largemouth bass(Micropterus salmoides) sample size (N; young-of-the-year fish removed).

LakeLakeCode

Perimeter(km)

Area(ha)

MaximumDepth (m)

BuildingDensity(no. km)1)

LargemouthBass (N )

Allequash AL 10.2 165.3 7.3 0.0 30Arrowhead AR 3.5 40.1 13.1 45.8 30Black Oak BO 12.0 230.1 25.9 18.0 30Brandy BR 3.5 45.1 13.4 30.1 30Camp CP 2.9 17.6 9.4 0.0 30Day DY 5.5 47.3 14.6 0.2 30Found FD 6.4 139.3 6.4 16.6 30Johnson JN 3.6 34.7 12.8 26.2 30Little Crooked LC 4.8 63.8 6.1 5.5 30Little John LJ 5.3 63.4 5.8 2.1 30Little Rock LR 1.4 8.1 6.5 0.0 30Little St.Germain

LSG 23.3 402.2 16.2 19.8 28

Moon MN 3.4 54.4 11.6 15.0 28Round RD 3.7 71.5 7.6 0.3 30UpperBuckatabon

UB 13.2 211.4 14.3 12.6 30

White Sand WS 9.3 304.6 21.6 5.8 27

Gaeta et al.

94

suggested by Burnham & Anderson (1998) for AICand by Spiegelhalter et al. (2002) for DIC. Allcovariate data, including LRD, were previouslysurveyed during the summers of 2001–2004 andarchived in the North Temperate Lakes Long-TermEcological Research online database (Carpenter &Kratz 2001).

Model structure

The multilevel model was composed of three levels:(1) the lowest measurement level in which annulus-specific observations of fish length were used to modelvariation in growth (2032 annuli observations); (2) thefish level in which the intercepts and slopes ofindividual fish growth trajectories were allowed tovary (473 individual fish); and (3) the lake level inwhich among-lake variation in growth was modelledusing lake characteristics (16 lakes).

Level 1: Annulus level: within fish, within lake

yi ! N"b0j#i$ % b1j#i$xi;r2y&;

for i ' 1; . . . ; n observations"Model 1:1&

Here, yi is the growth for observation i in fish j atlength xi, b0j#i$ is the intercept (or growth at the centredor average length) of fish j, b1j#i$ is the slope parameter(or growth–length relationship) of fish j, and r2

y is theresidual variance of yi (growth) of observation i in fishj at length xi The notation N (l,r2) refers to a normal(or, below, a multivariate normal) distribution withmean vector l and covariance matrix r2. In the case ofEq. (1.1), the model is a linear regression of loge-transformed growth rate on loge-transformed centredbody size.Level 2: Fish level: among fish, within lake

b0jb1j

!! N

c0k#j$c1k#j$

!;

r20b qr2

0br21b

qr20br

21b r2

1b

! !;

for j ' 1; . . . ; J fish

"Model 1:2&

Here, c0k#j$ is the mean intercept (or growth at averageor centred length) for lake k, c1k#j$ is the mean growthrate (or growth–length relationship) for lake k, r2

0b andr21b are the variation among fish-specific slopes

and intercepts, respectively, and qr20br

21b is the

covariance among r20b and r2

1b where b0j and b1j j

have correlation q. Equation (1.2) relates growthparameters b0j and b1j of an individual fish to thelake means and the covariance matrix among fishwithin a lake.Level 3: Lake level: among fish, among lakes

c0kc1k

! "! N

d00k % d01kz01k % ( ( ( % d0nkz0nkd10k % d11kxiz11k % ( ( ( % d1nkxiz1nk

! ";

!

r20c qr2

0cr21c

qr20cr

21c r2

1c

!!

; for k ' 1; . . . ;K lakes

"Model 1:3&

Here, d00k is the overall intercept (or grand meangrowth at average or centred length over all lakes) withr20y as the variance among lake intercepts, and d0nk is

the effect of covariate n with a value of z0nk on theintercept. d10k is the overall growth rate (or grand meangrowth–length relationship) with r2

1y as the variation ingrowth rates (or slopes) among lakes; d1nk is the effectof the interaction of fish length xi and covariate nwith avalue of z1nk on the overall growth rate. qr2

0yr21y is the

covariance between r20y and r2

1y where c0,k and c1,khave correlation q. Equation (1.3) is a bivariateregression of lake-specific growth parameters c0k andc1k on the fixed effects of the added covariates.

The relationship between LRD and the growth rate ofa fish of a given length, henceforth referred to as thebuilding density effect, was calculated as the derivativeof model-predicted growth rate with respect to LRDwith units of ln "mm(year)1&=building(km)1: Thestandard deviation of the building density effect wascalculated from the covariance matrix of model param-eters using standard error propagation formulae (Meyer1975). The building density effect can be interpreted asthe average change in growth rate of fish of a givenlength with the addition of 1 buildingÆkm)1 of shore-line.

Results

A total of 2032 annulus observations from 473 fishwere made from 16 lakes in the NHLD. The totalnumber of bass represented in each lake ranged from27 to 30 (Table 1), and fish lengths from 57 to408 mm were represented in our data set (Fig. 2). Themean loge length, or length on which the data werecentred, was 5.3 logeÆmm. The maximum growth rateobserved was 133.8 mmÆyear)1 at annulus (age) onefrom a fish in Little St. Germain Lake (LSG). Thelowest growth rate observed was 5.8 mmÆyear)1 atannulus 12 in a fish from Camp Lake (CP).

Although growth at length varied greatly within alake, an apparent trend at small sizes (approximately90 mm or )0.85 centred logeÆmm) was observed in alllakes (Fig. 3). This trend, however, is an artefact ofhow growth was calculated. The growth rate atannulus one, or how much a fish grew in the firstyear, is equal to the fish length at annulus one, thus

Lakeshore residential development and growth of largemouth bass

95

producing a 1:1 relationship between growth rate andlength for observations at annulus one.

Most lakes had annulus-specific observations span-ning the entire range of fish lengths (Fig. 3). Afteryoung-of-the-year fish were removed from the analy-sis, we captured fish from a range of over 300 mm and15 years. The average maximum size at capture was371 mm across all lakes. Two lakes had a sparsesample of large individuals; the largest individualsfrom Little John and Little Crooked Lakes were a 249-mm 8-year-old and a 310-mm 11-year-old largemouthbass, respectively.

Model selection

The forward stepwise selection process consideredboth the interaction of fish length and LRD and theinteraction of fish length and maximum depth ascandidate predictors. The initial model fitted growth asa function of length and had a DIC of 752.4. Addingthe interaction of LRD and length at step oneimproved the DIC by 5.3, and adding the interactionof length and maximum depth at step two improvedthe DIC by 5.8. No covariate improved model fit bythe minimum cut-off of 4 at step three, based oncriteria proposed for the AIC (Burnham & Anderson1998; Spiegelhalter et al. 2002). When compared tothe initial model, unconditional at levels 2 and 3, allthe standard deviations of the random effects of thefinal stepwise-selected model were within 0.001 or 1%of the initial model except for the slope at level 3. The

addition of the interactions of LRD and maximumdepth with length improved model fit and reduced thestandard deviation of the slope at level 3 by 0.116 or37%.

Model fit

Largemouth bass growth rate was successfully mod-elled as a function of fish length, the interaction of fishlength and LRD, and the interaction of fish length andmaximum depth. Predicted growth rates were closelyclustered around observed growth rates (Fig. 2a).Residuals were evenly distributed across predictedgrowth rate (Fig. 2b) and across fish length (Fig. 2c).Model fit was also unbiased across the regionalgradient of building density (Fig. 2d). As an unavoid-able consequence of how growth rates are calculated,growth rate and length at annulus one are equalcreating a conspicuous trend for observations atannulus one (clustered observations in the upper rightof Fig. 2a, on the right side of Fig. 2b, and on the leftside of Fig. 2c). Nonetheless, growth calculations atannulus one did not affect model results for laterannuli.

Model results

Our analysis identified a significant relationshipbetween LRD and largemouth bass size-specificgrowth rate that varied across fish length (Table 2).In general, we observed a more negative slope

(a) (b)

(c) (d)

Fig. 2. Longitudinal multilevel model fit.(a) Observed loge growth rate (mmÆyear)1)plotted against predicted loge growth rate(mmÆyear)1). Model residuals plotted against(b) predicted loge growth rate (mmÆyear)1),(c) centred fish length (logeÆmm), and(d) building density (no. km)1). Residualswere jittered when plotted against buildingdensity.

Gaeta et al.

96

between growth and length as LRD increased (Fig. 3),although this relationship was dampened with greatermaximum depth (Table 2; Fig. 3). The relationshipbetween LRD and growth across fish length was

illustrated by taking the derivative of the model withrespect to LRD (Fig. 4). This identified the averagechange in growth rate at a given length with theaddition of 1 buildingÆkm)1, or the building density

Fig. 3. Lake-specific longitudinal multilevel model fit. Largemouth bass (Micropterus salmoides) annulus-specific loge growth rate(mmÆyear)1) across centred fish length (logeÆmm) per lake (points) is shown with lake-specific model-predicted growth trajectories (black line)and random effects (grey lines). Lake codes are shown in upper right (refer to Table 1 for lake name and characteristics) with LRD(buildingsÆkm)1) in bottom left.

Table 2. Longitudinal multilevel model results. Parameter standard deviations (SD), correlations between the intercepts and slopes (corr), coefficients estimates(coef est), and coefficient standard errors (coef SE) of fixed and random effects at each level of the model. Parameter symbols in parenthesis correlate withparameter symbols in Models 1.1–1.3.

Model Details Random Effects Fixed Effects

Model Level Sample Unit Sample Size Parameter SD Corr Coef Parameter Coef Est Coef SE

Level 1 Annuli 2032 Residual 0.259 (ry)

Level 2 Fish 473 Intercept 0.126 (r0b) 0.496 (q)Slope 0.169 (r1b)

Level 3 Lake 16 Intercept 0.145 (r0c) 0.060 (q) Intercept 3.105 (d00k) 0.038Length )1.402 (d10k) 0.129

Slope 0.197 (r1c) LRD: length )0.018 (d11k) 0.005Maximum depth: length 0.029 (d12k) 0.010

Lakeshore residential development and growth of largemouth bass

97

effect. The model indicated that largemouth bassgrowth rate was positively correlated with buildingdensity for largemouth bass sizes smaller than210 mm and was negatively correlated for all largersizes. Growth rates showed the largest changes acrossthe gradient of building densities for the largest andsmallest size classes.

Discussion

The relationship between largemouth bass growthrate and LRD across fish length was determined usinga longitudinal multilevel model. This innovativeapproach allowed us to account for repeated measuresof annuli observations, include the hierarchical natureof the sample design, and incorporate variancebetween hierarchical levels. We found a stronglysignificant negative relationship between LRD and thegrowth rate of large sizes (>210 mm) of largemouthbass. Intriguingly, the opposite outcome was observedat smaller body sizes: a significant positive relation-ship was observed between LRD and the growth rateof small sizes (<210 mm) of largemouth bass. Theresults for small sizes of largemouth bass appear to benew. The findings for large sizes of largemouth basscorroborate trends reported by Schindler et al. (2000).Furthermore, our sample design and statistical meth-ods showed that the negative relationship of LRD andgrowth rates of large sizes of largemouth bass is

statistically robust. Thus, our findings confirm thetrends reported by Schindler et al. (2000).

Comparative studies such as ours are powerful fortesting patterns or associations across lakes, butinferences about mechanisms must be cautious (Coleet al. 1991). Nonetheless, we have identified anecologically important pattern that suggests LRDalters largemouth bass growth. The mechanismsdriving this relationship are likely complex and highlyvariable. For instance, release from density-dependentgrowth, shifts in available prey sources, reducedmacrophyte cover, and angling-induced selectionpressures are all potential mechanisms that could beacting independently or concurrently to drive theobserved trends in largemouth bass growth.

The observed trend of increased growth rates ofsmaller largemouth bass with LRD (Fig. 4) could bedriven by several mechanisms. For instance, reduc-tions in lake-wide aggregations of small fish(£100 mm) associated with LRD (Scheuerell &Schindler 2004) may release young-of-the-year andyearling largemouth bass from negative density-dependent growth responses (Olson et al. 1995; Postet al. 1998), resulting in increased growth rates ofthese small sizes with LRD. Reductions in vegetation,similar to those associated with LRD (Jennings et al.2003; Radomski 2006), have also been shown to resultin increased growth rates of small largemouth bass(Bettoli et al. 1992; Olson et al. 1998). Likewise,shifts in macroinvertebrate abundance associated with

Fig. 4. The effect of building density on growth rate across theobserved gradient of largemouth bass (Micropterus salmoides)length (black line). A value above zero (grey line) indicates apositive relationship between building density and growth rate; avalue below zero indicates a negative relationship. Dotted linesare ± one standard deviation calculated from the covariance matrixof model parameters using standard error propagation formulae.The building effect is interpreted as the average change in thegrowth rate of fish of a given length if 1 buildingÆkm)1 is added tothe shoreline; units are ln "mm year)1&=building km)1.

Fig. 5. Model-predicted growth trajectories for the average large-mouth bass in a lake without development and a lake with highdevelopment (45.8 buildingsÆkm)1). Simulations were initiated atthe mean observed length at age 1 in the three least and mostdeveloped lakes, 80.0 and 99.7 mm, respectively. Simulations wereperformed using the average maximum lake depth (12 m). Greydotted line shows the legal length limit in the NHLD, 356 mm(14 inches).

Gaeta et al.

98

LRD (Rosenberger et al. 2008) could result in anincreased availability of energetically beneficial foodsources with LRD for smaller largemouth bass inhighly developed lakes. An increase in energetic foodsources could potentially result in greater growth ratesfor small fish and induce earlier ontogenetic shift topiscivory with LRD (Olson 1996). Earlier ontogeneticshifts to piscivory would promote a rapid growthdivergence along a LRD gradient for small sizes oflargemouth bass that would likely diminish with sizeas individuals in undeveloped lakes undergo ontoge-netic shift later in the season, similar to the growthtrends we observed (Fig. 4).

Angling has also been shown to act as a strongselection pressure not only resulting in an increase ingrowth rate of smaller size classes, as seen in ourmodel (Fig. 4), but also causing a shift in maturitytowards smaller sizes and younger ages (Reznick et al.2001; Reznick & Ghalambor 2005; Lewin et al.2006). Reductions in size and age at maturity resultin a reallocation of energy from somatic to reproduc-tive growth of mature individuals causing reducedgrowth rates of larger sizes, similar to our results(Fig. 4). Furthermore, angling can impact populationsby sparing the less-vulnerable, slower-growing indi-viduals. Vulnerability to angling has been correlatedwith growth rate (Biro & Post 2008), and a quarter-century-long experiment showed that largemouth bassvulnerability is heritable and recreational fisheries canselect towards a less-vulnerable population (Philippet al. 2009). This suggests that exploitation of large-mouth bass associated with LRD (NRC 1992) couldresult in reduced growth rates. However, the selectionfor slower-growing individuals is not limited toharvested populations.

Faster-growing individuals are likely more vulner-able to angling, as mentioned above (Biro & Post2008; Philipp et al. 2009), and largemouth bass areparticularly vulnerable while nesting (Suski & Philipp2004). Therefore, catch-and-release practices duringspawning have the potential to promote egg loss andthereby reduce fecundity of these more vulnerable andfaster-growing individuals owing to rapid nest preda-tion (Uusi-Heikkila et al. 2008) and nest abandonmentafter being handled by anglers (Siepker et al. 2009),resulting in the removal of faster growth rates from thepopulation (Philipp et al. 2009). An increase inangling associated with LRD, therefore, has thepotential to alter growth rates through harvesting orcatch-and-release practices.

Regardless of the mechanism, the observed patternof largemouth bass size-specific growth rate versusLRD has implications for fish ecology and manage-ment. To determine how the observed growth patternmight impact largemouth bass fisheries acrossLRD, we used our model to predict the growth

trajectories of an average individual in both a lakewithout development and a lake with high develop-ment (45.8 buildingsÆkm)1) given the mean maximumdepth (Fig. 5). We initiated the model at the averagelength at age 1 in the three lakes with the lowest LRDand the three lakes with the highest LRD or 80.0 and99.7 mm, respectively. We found that the length at ageis greater for individuals in highly developed lakesuntil around age 9 or a length of 300 mm. Above thissize, the length at age of individuals in lakes withoutLRD is greater than that of individuals in highlydeveloped lakes. As a result, individuals in highlydeveloped lakes take about 1.5 growing seasons longerto reach the legal length limit of 356 mm. If individ-uals in lakes without LRD follow this growth trajec-tory, they will reach trophy lengths several yearsbefore individuals in high-LRD lakes will reach.

Ages at length for adult largemouth bass in ourstudy are greater than those observed in some otherregions (Bennett 1937; Jackson et al. 2008). However,long-term research on tagged largemouth bass popu-lations in the Northern Highland Lake District isconsistent with our findings. For example, taggedlargemouth bass have been observed growing at ratessimilar to our model in both Little Rock Lake, VilasCo. WI (e.g., 304 mm in 2001, 337 mm in 2005;201 mm in 2001, 330 mm in 2009; Gaeta, J.W.unpublished data), and Paul Lake, Gogebic Co. MI(e.g., 150 mm in 1988, 312 mm in 1997; 188 mm in1986, 342 mm in 1997; J.F. Kitchell, B. Weidel,J. Hodgson, T. Cline and S. Carpenter, unpublisheddata). Largemouth bass adults appear to persist despiterather slow growth rates in some lakes of the NorthernHighland Lake District. Regional differences in large-mouth bass growth rates in relation to habitat and otherfactors are an important topic for further research.

We determined that growth rates of small large-mouth bass (<210 mm) are positively correlated withLRD, while growth rates of large individuals(>210 mm) are inversely related to LRD as they grow(Fig. 4). Future work investigating potential mecha-nisms of the observed pattern should study the timingof ontogenetic shifts as well as the diets of young-of-the-year and yearling largemouth bass across LRD.Likewise, quantifying both species-specific retentionand catch-and-release rates across the NHLD andbetween the NHLD and other regions could provideinvaluable insight into this potential driver of growth.Researchers should also attempt to establish therelationship between growth rate and vulnerability oflargemouth bass to angling. Humans may choose todevelop lakes with inherent characteristics, such asfish community structure, that drive the observedpattern of largemouth growth versus LRD. Nonethe-less, we have identified an ecologically importantpattern of largemouth bass growth versus LRD.

Lakeshore residential development and growth of largemouth bass

99

Fisheries’ managers should note that largemouth bassin highly developed lakes take longer to enter thefishery and may reach trophy lengths more rapidly inundeveloped systems.

Acknowledgements

Our research was supported by the North Temperate Lakes LTERprogram, a GERS Fellowship to J. Gaeta, and a NSF ResearchExperience for Undergraduates award toM.Guarascio.We thankTyler Ahrenstorff for help with field data collection, GrahamMacDonald and Jeff Maxted for help with GIS, and ShawnDevlin, Gretchen Hansen, and Nicholas Preston for statisticaladvice. We thank James Kitchell, Jake Vander Zanden, and threeanonymous reviewers for helpful comments on earlier versions ofthis manuscript and three anonymous reviewers for helpfulcomments on this manuscript.

References

Ai, X. 2002. Gender differences in Growth in MathematicsAchievement: Three-Level Longitudinal and MultilevelAnalyses of Individual, Home, and School Influences.Mathematical Thinking and Learning 4: 1–22.

Bennett, G.W. 1937. The Growth of the Large Mouthed BlackBass, Huro salmoides (Lacepede), in the Waters of Wiscon-sin. Copeia 1937: 104–118.

Bettoli, P.W., Maceina, M.J., Noble, R.L. & Betsill, R.K. 1992.Piscivory in Largemouth Bass as a Function of AquaticVegetation Abundance. North American Journal of FisheriesManagement 12: 509–516.

Biro, P.A. & Post, J.R. 2008. Rapid depletion of genotypeswith fast growth and bold personality traits from harvestedfish populations. Proceedings of the National Academy ofSciences of the United States of America 105: 2919–2922.

Brauns, M., Garcia, X.-F., Walz, N. & Pusch, M.T. 2007.Effects of human shoreline development on littoral macro-invertebrates in lowland lakes. The Journal of AppliedEcology 44: 1138–1144.

Brown, R.T. & Curtis, J.T. 1952. The Upland Conifer-Hardwood Forests of Northern Wisconsin. Ecological Mono-graphs 22: 217–234.

Bryan, M.D. & Scarnecchia, D.L. 1992. Species Rich-ness, Composition, and Abundance of Fish Larvae andJuveniles Inhabiting Natural and Developed Shorelines of aGlacial Iowa Lake. Environmental Biology of Fishes 35:329–341.

Buckmeier, D.L. & Howells, R.G. 2003. Validation of otolithsfor estimating ages of largemouth bass to 16 years. NorthAmerican Journal of Fisheries Management 23: 590–593.

Burnham, K.P. & Anderson, D.R. 1998. Model selection andinference a practical information-theoretic approach. NewYork: NY Springer Verlag New York Inc. 353 pp.

Carlander, K.D. 1982. Standard intercepts for calculatinglengths from scale measurements for some centrarchid andpercid fishes. Transactions of the American Fisheries Society111: 332–336.

Carpenter, S.R. & Kratz, T.K. 2001. Biocomplexity; coordi-nated field studies. https://secure.limnology.wisc.edu/lterque-ry/metadata_new.jsp?dataset_id=BIOLAKE1.

Carpenter, S.R., Benson, B.J., Biggs, R., Chipman, J.W., Foley,J.A., Golding, S.A., Hammer, R.B., Hanson, P.C., Johnson,P.T.J., Kamarainen, A.M., Kratz, T.K., Lathrop, R.C.,McMahon, K.D., Provencher, B., Rusak, J.A., Solomon,C.T., Stanley, E.H., Turner, M.G., Vander Zanden, M.J., Wu,C.H. & Yuan, H.L. 2007. Understanding regional change: acomparison of two lake districts. BioScience 57: 323–335.

Christensen, D.L., Herwig, B.R., Schindler, D.E. & Carpenter,S.R. 1996. Impacts of Lakeshore Residential Developmenton Coarse Woody Debris in North Temperate Lakes.Ecological Application 6: 1143–1149.

Cole, J.J., Lovett, G. & Findlay, S. 1991. Comparative analysisof ecosystems: patterns, mechanisms, and theories. NewYork, NY: Springer-Verlag.

Engel, S. & Pederson Jr, J.L. 1998. The construction, aesthetics,and effects of lakeshore development: a literature review. No.Research Report 177. Madison, WI: Wisconsin Departmentof Natural Resources.

Francis, T.B. & Schindler, D.E. 2006. Degradation of littoralhabitats by residential development: Woody debris in lakes ofthe Pacific Northwest and Midwest, United States. Ambio 35:274–280.

Francis, T.B. & Schindler, D.E. 2009. Shoreline urbanizationreduces terrestrial insect subsidies to fishes in North Amer-ican lakes. Oikos 118: 1872–1882.

Francis, T.B., Schindler, D.E., Fox, J.M. & Seminet-Reneau, E.2007. Effects of urbanization on the dynamics of organicsediments in temperate lakes. Ecosystems 10: 1057–1068.

Gelman, A. & Hill, J. 2008. Data analysis using regression andMultilevel ⁄Heirarchical Models. New York, NY: CambridgeUniversity Press. 625 pp.

Goldstein, H. 1995. Multilevel statistical models. London:Hodder Arnold.

Hunt, J. & Annett, C.A. 2002. Effects of habitat manipulationon reproductive success of individual largemouth bass in anOzark reservoir. North American Journal of Fisheries Man-agement 22: 1201–1208.

Jackson, Z.J., Quist, M.C. & Larscheid, J.G. 2008. Growthstandards for nine North American fish species. FisheriesManagement and Ecology 15: 107–118.

Jennings,M.J., Emmons, E.E., Hatzenbeler, G.R., Edwards, C.&Bozek, M.A. 2003. Is Littoral Habitat Affected by ResidentialDevelopment and Land Use in Watersheds of WisconsinLakes? Lakes and Reservoir Management 19: 272–279.

Lewin, W.C., Arlinghaus, R. & Mehner, T. 2006. Documentedand potential biological impacts of recreational fishing:insights for management and conservation. Reviews inFisheries Science 14: 305–367.

Maceina, M.J., Boxrucker, J., Buckmeier, D.L., Gangl, R.S.,Lucchesi, D.O., Isermann, D.A., Jackson, J.R. & Martinez,P.J. 2007. Current status and review of freshwater fish agingprocedures used by state and provisional fisheries agencieswith recommendations for future directions (vol 32, pg 329,2007), Fisheries, 32: 508.

Maraldo, D.C. &Maccrimmon, H.R. 1979. Comparison of agingmethods and growth-rates for largemouth bass, micropterus-salmoides lacepede, from northern latitudes. EnvironmentalBiology of Fishes 4: 263–271.

Marburg, A.E., Turner, M.G. & Kratz, T.K. 2006. Natural andanthropogenic variation in coarse wood among and withinlakes. Journal of Ecology 94: 558–568.

Gaeta et al.

100

Meyer, S.L. 1975. Data analysis for scientists and engineers.New York, NY: Wiley.

NRC 1992. Restoration of aquatic ecosystems: science, tech-nology, and public policy. Washington, DC: National Acad-emy Press.

Olson, M.H. 1996. Ontogenetic niche shifts in largemouth bass:variability and consequences for first-year growth. Ecology77: 179–190.

Olson, M.H., Mittelbach, G.G. & Osenberg, C.W. 1995.Competition between predator and prey: resource-basedmechanisms and implications for stage-structured dynamics.Ecology 76: 1758–1771.

Olson, M.H., Carpenter, S.R., Cunningham, P., Gafny, S.,Herwig, B.R., Nibbelink, N.P., Pellett, T., Storlie, C., Trebitz,A.S. &Wilson, K.A. 1998.Managing macrophytes to improvefish growth: a multi-lake experiment. Fisheries 23: 6–12.

Penaloza, L.J. 1991. Boating pressure on Wisconsin’s Lakesand rivers. Department of Natural Resources. TechnicalBulletin No. 174.

Peterson, G.D., Beard, T.D., Beisner, B.E., Bennet, E.M.,Carpenter, S.R., Cumming, G.S., Dent, C.L. & Havlicek,T.D. 2003. Assessing future ecosystem services a case studyof the Northern Highlands Lake District, Wisconsin. Con-servation Ecology 7: 24.

Philipp, D.P., Cooke, S.J., Claussen, J.E., Koppelman, J.B.,Suski, C.D. & Burkett, D.P. 2009. Selection for Vulnerabilityto Angling in Largemouth Bass. Transactions of the Amer-ican Fisheries Society 138: 189–199.

Post, D.M., Kitchell, J.F. & Hodgson, J.R. 1998. Interactionsamong adult demography, spawning date, growth rate,predation, overwinter mortality, and the recruitment oflargemouth bass in a northern lake. Canadian Journal ofFisheries and Aquatic Sciences 55: 2588–2600.

Postel, S. & Carpenter, S.R. 1997. Freshwater ecosystemservices. In: Daily, G., ed. Nature’s services. Washington,DC: Island Press, pp. 195–214.

R Development Core Team 2010. R: a language and environ-ment for statistical computing. Vienna, Austria: R Foundationfor Statistical Computing.

Radomski, P. 2006. Historical changes in abundance offloating-leaf and emergent vegetation in Minnesota lakes.North American Journal of Fisheries Management 26: 932–940.

Reed, J.R. & Pereira, D.L. 2009. Relationships BetweenShoreline Development and Nest Site Selection by BlackCrappie and Largemouth Bass. North American Journal ofFisheries Management 29: 943–948.

Reznick, D.N. & Ghalambor, C.K. 2005. Can commercialfishing cause evolution? Answers from guppies (Poeciliareticulata). Canadian Journal of Fisheries and AquaticSciences 62: 791–801.

Reznick, D., Buckwalter, G., Groff, J. & Elder, D. 2001. Theevolution of senescence in natural populations of guppies(Poecilia reticulata): a comparative approach. ExperimentalGerontology 36: 791–812.

Riera, J., Voss, P.R., Carpenter, S.R., Kratz, T.K., Lillesand,T.M., Schnaiberg, J.A., Turner, M.G. & Wegener, M.W.

2001. Nature, society and history in two contrastinglandscapes in Wisconsin, USA – interactions between lakesand humans during the twentieth century. Land Use Policy18: 41–51.

Rosenberger, E.E., Hampton, S.E., Fradkin, S.C. & Kennedy,B.P. 2008. Effects of shoreline development on the nearshoreenvironment in large deep oligotrophic lakes. FreshwaterBiology 53: 1673–1691.

Roth, B.M., Kaplan, I.C., Sass, G.G., Johnson, P.T., Marburg,A.E., Yannarell, A.C., Havlicek, T.D., Willis, T.V., Turner,M.G. & Carpenter, S.R. 2007. Linking terrestrial and aquaticecosystems: the role of woody habitat in lake food webs.Ecological Modelling 203: 439–452.

Sass, G.G., Gille, C.M., Hinke, J.T. & Kitchell, J.F. 2006.Whole-lake influences of littoral structural complexity andprey body morphology on fish predator-prey interactions.Ecology of Freshwater Fish 15: 301–308.

Scheuerell, M.D. & Schindler, D.E. 2004. Changes in thespatial distribution of fishes in lakes along a residentialdevelopment gradient. Ecosystems 7: 98–106.

Schindler, D.E., Geib, S.I. & Williams, M.R. 2000. Patterns ofFish Growth along a Residential Development Gradient inNorth Temperate Lakes. Ecosystems 3: 229–237.

Schnaiberg, J., Riera, J., Turner, M.G. & Voss, P.R. 2002.Explaining human settlement patterns in a recreational lakedistrict: Vilas County, Wisconsin, USA. EnvironmentalManagement 30: 24–34.

Siepker, M.J., Cooke, S.J., Wahl, D.H. & Philipp, D.P. 2009.Individual Reproductive Success of Largemouth Bass andSmallmouth Bass Subjected to Different Components ofCompetitive Angling Events. Transactions of the AmericanFisheries Society 138: 818–825.

Spiegelhalter, D.J., Best, N.G., Carlin, B.R. & van der Linde,A. 2002. Bayesian measures of model complexity and fit.Journal of the Royal Statistical Society Series B-StatisticalMethodology 64: 583–616.

Stearns, F. 1951. The Composition of the Sugar Maple-Hemlock-Yellow Birch Association in Northern Wisconsin.Ecology 32: 245–265.

Suski, C.D. & Philipp, D.P. 2004. Factors affecting thevulnerability to angling of nesting male largemouth andsmallmouth bass. Transactions of the American FisheriesSociety 133: 1100–1106.

Uusi-Heikkila, S., Wolter, C., Klefoth, T. & Arlinghaus, R.2008. A behavioral perspective on fishing-induced evolution.Trends in Ecology & Evolution 23: 419–421.

Vander Zanden, M.J. & Vadeboncoeur, Y. 2002. Fishes asintegrators of benthic and pelagic food webs in lakes.Ecology 83: 2152–2161.

Wagner, T., Hayes, D.B. & Bremigan, M.T. 2006. Accountingfor multilevel data structures in fisheries data using mixedmodels. Fisheries 31: 180–187.

Wetzel, R.G. 2001. Limnology: lake and river ecosystems. SanDiego, CA: Academic Press. 1006 pp.

Wisconsin Department of Natural Resources 2005. Wisconsinlakes PUB-FH-800. Bureau of Fisheries and Habitat Man-agement, Madison, pp. 142–156.

Lakeshore residential development and growth of largemouth bass

101