Connecting habitat to species abundance: the role of light and temperature on the abundance of walleye in lakes

Abstract

Walleye (Sander vitreus) are an ecologically important species managed for recreational, tribal, and commercial harvest. Walleye prefer cool water and low light conditions, and therefore changing water temperature and clarity potentially impacts walleye habitat and populations across the landscape. Using survey data collected from 1993 to 2018 from 312 lakes in Minnesota, we evaluated the relationship between thermal-optical habitat and the relative abundance of small (0–300 mm), medium (300–450 mm), and large (450 + mm) walleye. Thermal-optical habitat was positively correlated with the relative abundance of small and medium walleye but not large walleye. Walleye were more abundant in larger, naturally reproducing lakes opposed to smaller, stocked lakes. Thermal-optical habitat changed in 59% of lakes since 1980 (26% increasing and 33% decreasing) and appears to be driven primarily by changes in water clarity and thus optical habitat area. Our study provides important insights into local and regional drivers that influence walleye populations that can be used to assist fisheries managers in setting population goals and managing harvest.

Full text

OPEN ACCESS | Article
Connecting habitat to species abundance: the role of light
and temperature on the abundance of walleye in lakes
Shad Mahlum a, Kelsey Vitensea, Hayley Corson-Doschb, Lindsay Plattb, Jordan S. Readb, Patrick J. Schmalzc,
Melissa Tremlc, and Gretchen J. A. Hansen a
aDepartment of Fish, Wildlife, and Conservation Biology, University of Minnesota, St. Paul, MN 55188, USA; bU.S. Geological Survey,
Madison, WI 53705, USA; cMinnesota Department of Natural Resources, Section of Fisheries, St. Paul, MN, USA, 55155
Corresponding author: Shad Mahlum (email: shadmahlum@gmail.com)
Abstract
Walleye (Sander vitreus) are an ecologically important species managed for recreational, tribal, and commercial harvest. Wall-
eye prefer cool water and low light conditions, and therefore changing water temperature and clarity potentially impacts wall-
eye habitat and populations across the landscape. Using survey data collected from 1993 to 2018 from 312 lakes in Minnesota,
we evaluated the relationship between thermal-optical habitat and the relative abundance of small (0–300 mm), medium (300–
450 mm), and large (450 +mm) walleye. Thermal-optical habitat was positively correlated with the relative abundance of small
and medium walleye but not large walleye. Walleye were more abundant in larger, naturally reproducing lakes opposed to
smaller, stocked lakes. Thermal-optical habitat changed in 59% of lakes since 1980 (26% increasing and 33% decreasing) and
appears to be driven primarily by changes in water clarity and thus optical habitat area. Our study provides important insights
into local and regional drivers that influence walleye populations that can be used to assist fisheries managers in setting
population goals and managing harvest.
Key words: hierarchical models, walleye, thermal-optical habitat, safe operating space, water clarity, temperature
1. Introduction
In dynamic ecosystems, management regimes that ac-
count for fluctuations in habitat availability can increase
resilience to disturbance (Hansen et al. 2015). However,
managed aquatic resources often encompass broad spatial
extents and can comprise of hundreds if not thousands of in-
dividual units that vary in their ecological characteristics and
management objectives (e.g., the lake-rich landscape across
Minnesota). Aquatic ecosystems are often managed across
individual units. Managing thousands of individual lakes
can be dicult or impossible when resources are limited
(Lester et al. 2003,2021). Therefore, it is necessary to identify
key characteristics that drive populations at spatial scales
that allow agencies to eciently manage species in multiple
habitat units simultaneously, even in the absence of site-
specific population monitoring (Lester et al. 2003;Heino et
al. 2021). At the same time, landscape approaches to manage-
ment can also facilitate replication of management actions,
experimental approaches, and monitoring that can enable
faster learning than individual management approaches,
especially in an era of rapid change (Hansen et al. 2015).
Recreational fisheries provide economic and cultural ben-
efits to society (Cooke et al. 2015;Lynch et al. 2016). For
recreational fisheries, harvest management (i.e., individual
bag limits and length-based restrictions on harvest) is the pri-
mary lever through which managers can maintain ecosystem
services (Carpenter et al. 2017). Sustainable harvest of fish-
eries is influenced by the status of the fish population, angler
behavior (Hunt et al. 2011), and environmental conditions
(Gutowsky et al. 2019). Population abundance, as well as the
level of harvest allowed that maintains long-term sustainabil-
ity, may vary depending on habitat conditions (Ryder 1965;
Jacobson et al. 2016;Hansen et al. 2019). Here, habitat often
sets the upper limit of a population’s potential, while other
factors such as biotic interactions, harvest, or variability in
environmental conditions mean that potential abundance is
not attained at all sites (VanDerWal et al. 2009;Embke et
al. 2019). Yet, habitat characteristics associated with healthy
ecosystems often show nonlinear relationships with fauna
(Koch et al. 2009), and management strategies that ignore the
natural variability of the ecosystem (Paukert et al. 2016)may
result in overharvest of important fish species where cumu-
lative anthropogenic stressors act simultaneously (Embke et
al. 2019). Therefore, testing for a relationship between popu-
lations and habitat can set expectations regarding potential
abundance and carrying capacity (Fulton et al. 2016)andalso
identify locations where habitat is sucient for a healthy
population to persist or where bottlenecks could occur at
critical life history stages where habitat is limited (Sass et al.
2017;Cantin and Post 2018;Gostiaux et al. 2022).
Walleye (Sander vitreus) are an important subsistence fish-
ery for Native American and First Nations people and a
Can. J. Fish. Aquat. Sci. 80: 273–286 (2023) | dx.doi.org/10.1139/cjfas-2022-0109 273

Canadian Science Publishing
274 Can. J. Fish. Aquat. Sci. 80: 273–286 (2023) | dx.doi.org/10.1139/cjfas-2022-0109
popular sport fish across North America that require cool
temperatures (Christie and Regier 1988) and low light condi-
tions (Ryder 1977). Known as the thermal-optical habitat area
(TOHA; Lester et al. 2004), walleye habitat can be defined as
the area of a lake where preferred thermal and optical condi-
tions overlap in space and time. Across Ontario lakes, TOHA is
predictive of walleye production and yield and can be used to
manage harvest in lakes with varying habitat quality (Lester
et al. 2004). Within a lake, walleye habitat is driven by envi-
ronmental factors that aect water clarity and temperature,
and the magnitude and direction of the eects vary depend-
ing on a lake’s baseline condition and morphometry (Geisler
et al. 2016;Tunney et al. 2018;Hansen et al. 2019). For ex-
ample, increased water clarity in a shallow lake in Minnesota
resulted in loss of TOHA that was linked to population decline
of walleye (Hansen et al. 2019).
Walleye are managed across thousands of lakes throughout
North America often with complex fishing regulations (Lester
et al. 2003). Although several studies have grouped lakes
based on their physical and chemical similarities to facili-
tate multi-lake management (Heiskary et al. 1987;Jacobson
et al. 2016), few have linked these lake characteristics to fo-
cal species (see Cross and McInerny 2005;Rypel et al. 2019;
Dassow et al. 2022) and it is unknown to what extent that
variation in TOHA has on walleye abundance across their dis-
tribution. This ultimately inhibits our ability to shift from
single lake management to broader multi-lake management
strategies (Lester et al. 2003;van Poorten and Camp 2019;
Lester et al. 2021).
Here, we quantify spatial and temporal dynamics of
walleye habitat across 312 lakes in Minnesota and test for a
relationship between walleye relative abundance and habitat
area. Our primary objective was to test for a relationship
between walleye habitat (thermal habitat area [THA], optical
habitat area [OHA], and TOHA) and the relative abundance
of walleye at various length classes across diverse lake types.
We used the change in TOHA to compare dierences in
modeled and observed walleye populations to help identify
(i) robust lakes where walleye abundances are higher than
predicted based on habitat area and in which increased wall-
eye harvest could possibly be sustained and (ii) lakes where
walleye populations are not meeting their potential based
on habitat area and in which additional management could
be considered to increase walleye abundances. Quantifying
the performance of each lake relative to expectations based
on available habitat, which is sensitive to changes in water
clarity and temperature, will allow managers to set realistic
goals for walleye harvest and identify lakes that are likely to
be resilient to future changes.
2. Methods
2.1. Site selection and biological sampling
In Minnesota, over 1000 lakes have regulated walleye pop-
ulations and most lakes are only surveyed once every 3–
10 years (J.F. Hansen, Minnesota Department of Natural Re-
sources [MNDNR], personal communication). We restricted
our analysis to lakes with at least 5 years of walleye survey
data from 1993 to 2018 and with corresponding habitat area
estimates (Table 1 ) for a total of 312 lakes (Fig. 1).
Walleye relative abundance was indexed as catch per unit
eort (CPUE, fish/net) from standard monitoring surveys us-
ing demersal gill nets from June to October. Sampling ef-
fort (number of gill nets) depends on the surface area of the
lake (Table 1 of the Minnesota 112 Department of Natural Re-
sources (MNDNR) 2017). The gill nets are 76 m in length with
a height of 1.8 m. Each net has five panels constructed with
multifilament knotted-nylon mesh with sizes ordered from
small to large (mesh size 1.9, 2.5, 3.2, 3.8, and 5.1 cm). We col-
lated data on individual walleye caught in gillnets, including
total length (mm). For each lake, we identified a lake-specific
3-week window that had the most samples taken across sam-
pling years for that lake. This allowed us to maximize the
number of years of data for each lake and control for seasonal
dierences in catchability within the lake. Thus, on average,
lakes were surveyed seven times for a total of 2227 lake-years
in our analysis (Table 1 ).
We calculated the CPUE of walleye in three length cate-
gories: small (0–300 mm), medium (300–450 mm), and large
(>450 mm). These categories reflect a balance of being fully
selected to the sampling gear and vulnerability to harvest. Se-
lectivity varies with fish length in a non-monotonic manner
(Radomski et al. 2020), and Walleye harvest regulations vary
among lakes. In general, these length categories represent (1)
small fish that are mostly unaected by harvest but not yet
fully recruited to the sampling gear, (2) medium fish that are
well sampled by gillnets although not fully recruited to the
gear (relative selectivity ∼0.8; Radomski et al. 2020)andsub-
ject to harvest in most lakes, and (3) large fish that are for the
most part fully selected to the gear but whose populations are
most impacted by harvest. In instances where subsampling
of walleye occurred during surveys (<3% of all lake-years), an
expansion factor was applied to 802 individuals (0.3% of all
walleye caught) with no length data based on the proportion
of the subsample of individuals in each length bin. Prelim-
inary analysis showed that removing these samples did not
aect the results and overall trends observed in this study.
Walleye stocking is a common management action in Min-
nesota lakes. To account for the potential influence of stock-
ing in general, lakes were classified by area managers into
four Natural Recruitment Categories (NRcat) based on the
stocking intensity relative to the natural production of the
lake: 1 =walleye populations are exclusively naturally re-
producing; 2 =stocking occurs to some degree, but in the
absence of stocking, the walleye population would persist;
3=natural reproduction is sporadic and the population can-
not persist without stocking; and 4 =the population is sus-
tained solely through stocking.
2.2. Thermal-, optical-, and thermal-optical
habitat area
Walleye habitat is defined by cool temperatures and low
light conditions. Thermal habitat area is the area of a lake
where temperatures are between 11 and 25 ◦C, correspond-
ing to optimal growth temperatures for walleye (Christie and
Regier 1988;Lester et al. 2004). Optical habitat area is the area

Canadian Science Publishing
Can. J. Fish. Aquat. Sci. 80: 273–286 (2023) | dx.doi.org/10.1139/cjfas-2022-0109 275
Table 1. Summary statistics for lake characteristics and catches of walleye.
Variable Median Minimum Maximum
Lake characteristics
Area (acres) 988 59 1 074 60
Depth (m) 12.5 1.8 68.7
Habitat (proportion of benthic area)
Thermal-optical habitat 0.023 0.001 0.093
Optical habitat 0.061 0.008 0.178
Thermal habitat 0.413 0.131 0.53
Sampling
Lake-years (n)6526
Walleye CPUE
Small 0.75 0 111
Medium 2.27 0 60
Large 1.67 0 31
Fig. 1. Study area of the lakes (black points) across Minnesota
(Datum =NAD83, Projection =UTM zone 15 N).
of the lake where light intensity is between 8 and 68 lx, as-
sociated with peak foraging activity (Ryder 1977;Lester et al.
2004). Thus, TOHA is the benthic area of a lake in which tem-
perature and light fall within their optimal ranges.
Daily measurements of water clarity and temperature pro-
files across depths are rare. Therefore, we used a combination
of observed data, hydrodynamic modeling, and statistical
modeling to reconstruct daily temperature and light profiles
for individual lakes from which we calculated daily OHA,
THA, and TOHA from 1980 to 2018 (Read et al. 2019). Meth-
ods for calculating TOHA were like those described for a
single lake in Hansen et al. (2019). To estimate THA, daily
lake temperature profiles (0.5 m depth increment) were sim-
ulated using a process-guided deep learning model of water
temperatures (Read et al. 2019), where deep learning models
were pre-trained using output from General Lake Model ver-
sion 3.1 (GLM; Hipsey et al. 2019) and incorporated a physical
constraint for energy conservation as a loss term (see Sup-
plementary Information S1.1 for deviations from Read et al.
[2019]). The predicted temperature profiles performed well
across seasons and depths with a root mean square error
(RMSE) of 1.72 ◦C(Read et al. 2021). We used modeled daily
estimates of water clarity to calculate optical habitat area
(Supplementary Information S1.2; Vitense and Hansen 2021).
Finally, we used hypsographic data to estimate the daily THA,
OHA, and TOHA that fell within the desired temperature and
clarity thresholds for walleye. To calculate annual habitat
areas, daily estimates of THA, OHA, and TOHA were averaged
from April 1 in the year of interest to March 31 of the follow-
ing year. These annual area estimates were then converted to
the proportion of the maximum potential habitat area (m2)
for a given year by dividing the total benthic area (m2;Hansen
et al. 2019). Finally, following Hansen et al. (2019) we used
a 3-year rolling average (e.g., two previous years plus year of
sampling) of THA, OHA, and TOHA to relate to the CPUE of
walleye.
2.3. Model selection, habitat comparison, and
predictive strength
To select the best model and to test its predictive quali-
ties, we split the data into two groups. The in-sample group
(n=1603) was used to fit the model, identify important co-
variate relationships, and to compare dierent habitat met-
rics. Here, we identified 10 a priori models (Table 2 ), for small,
medium, and large walleye, that included TOHA as the mea-
sure of available habitat in combination with the following

Canadian Science Publishing
276 Can. J. Fish. Aquat. Sci. 80: 273–286 (2023) | dx.doi.org/10.1139/cjfas-2022-0109
Table 2. Model selection for small (<300 mm), medium (300–450 mm), and large walleye
(450 +mm). A priori variables include thermal-optical habitat (TOHA), Lake area, natural recruit-
ment (NRcat), and date of sampling (ordinal date). Bold text indicates top models within DIC <5
and pD is the eective number of parameters.
Model pD DIC DIC
Small
∼TOHA +LakeArea +NRcat 88.26 89.08 0
∼TOHA +NRcat +TOHA∗NRcat +LakeArea 88.27 94.08 5
∼TOHA +NRcat 95.47 94.43 5
∼TOHA +NRcat +TOHA∗NRcat 96.32 98.45 9
∼TOHA +LakeArea +NRcat +ordinal date 79.47 3701.2 12
∼TOHA +LakeArea +ordinal date 83.25 04.87 16
∼TOHA +NRcat +ordinal date 82.07 3712.54 23
∼TOHA +NRcat +TOHA∗NRcat +ordinal date 84.99 3713.56 24
∼TOHA 98.12 3720.08 31
∼TOHA +ordinal date 83.23 3734.77 46
Medium
∼TOHA +NRcat +TOHA∗NRcat +LakeArea 77.92 6915.21 0
∼TOHA +LakeArea +NRcat 74.86 6915.96 1
∼TOHA +LakeArea +ordinal date 70.6 6923.99 9
∼TOHA +LakeArea +NRcat +ordinal date 64.22 6924.64 9
∼TOHA +NRcat +TOHA∗NRcat 78.96 6930.6 15
∼TOHA +NRCat 75.62 6931.81 17
∼TOHA +NRcat +ordinal date 57.99 6945.38 30
∼TOHA 78.92 6951.78 37
∼TOHA +ordinal date 65.65 6957.09 42
∼TOHA +NRcat +TOHA∗NRcat +ordinal date 51.39 6984.15 69
Large
∼TOHA +LakeArea +NRcat 68.19 5207.74 0
∼TOHA +LakeArea +ordinal date 66.71 5209.04 1
∼TOHA +NRcat +TOHA∗NRcat +LakeArea 68.58 5209.96 2
∼TOHA +NRCat 66.52 5210.75 3
∼TOHA +LakeArea +NRcat +ordinal date 62.84 5211.97 4
∼TOHA 69.52 5214.05 6
∼TOHA +NRcat +ordinal date 60.68 5214.11 6
∼TOHA +NRcat +TOHA∗NRcat 67.38 5214.76 7
∼TOHA +ordinal date 64.49 5214.94 7
∼TOHA +NRcat +TOHA∗NRcat +ordinal date 58.82 5222.5 15
covariates: lake surface area (log transformed), NRcat, and or-
dinal date of sampling (Table 2 ). We used the deviance infor-
mation criterion (DIC; Spiegelhalter et al. 2002) to compare
the dierent models (DIC; dierence in DIC compared to
the model with the smallest DIC), and to select an appropri-
ate model given the data (models with DIC <5 from the top
model were considered; Spiegelhalter et al. 2003). Then, to
identify which measure of available walleye habitat was best,
we compared three models for each walleye length category
(DIC <5) that contained either TOHA, OHA, or THA and an
additional null model without an available habitat metric.
The out-of-sample group consisted of two randomly selected
samples from each lake (n=624) and was used to assess the
utility of TOHA in predicting walleye abundance using the
RMSE of the final model.
We used a hierarchical Bayesian model (INLA; Rue et al.
2009) to test for a relationship between OHA, THA, TOHA
(habitatij,eq. 1b) and selected covariates to the CPUE of wall-
eye in each length category. We used a gamma distribution
to model the CPUE (eq. 1a), which required that we add a
small number to each catch value (0.01) to eliminate zeros.
Initial models included a random intercept for lake; however,
during model evaluation it became clear that sites were spa-
tially correlated because of unaccounted broadscale drivers
(i.e., violation of independence, Supplementary Information
S2: Fig. S1). Therefore, we included a random spatial term
(ui;eqs. 1b,1c) in the model to account for the distinct spatial
dependency that was present beyond the site level (see Sup-
plementary Information S2 on the gaussian markov random
field [GMRF] of eqs. 1c). We also included a random intercept

Canadian Science Publishing
Can. J. Fish. Aquat. Sci. 80: 273–286 (2023) | dx.doi.org/10.1139/cjfas-2022-0109 277
for the sampling year (at;eqs. 1b,1d). Default diuse normal
priors were used for the fixed eects. For σyear (Eq. 1d), we
used a penalized complexity prior (σ=0.5, σ=0.05; Simpson
et al. 2017).
walleyei,t∼Gamma (μi,t,r)
(1a)
log (μi,t)=intercept +β1×Habitati,t+ ··· + βk
×Covariatek,i,t+ui+at
(1b)
ui∼GMRF (0,
)
(1c)
at∼N0,σ2
year
(1d)
2.4. Application to walleye management
We compared model-predicted walleye CPUE to observed
CPUE to test the performance of individual lake-years based
on the associated covariates. Using the out-of-sample data set,
we estimated a prediction interval for each lake-year by ran-
domly generating 1000 predicted CPUE values. We then com-
pared observed lake-year specific CPUE to these prediction in-
tervals. Observed values within the 95% prediction intervals
were considered to meet expectations of walleye abundance
relative to habitat area. Observed walleye CPUE less than the
95% prediction intervals were classified as underperforming,
and samples that observed walleye catch rates more than
the 95% prediction intervals were classified as overperform-
ing relative to available habitat. To summarize our results,
we calculated the percentage of lake-years that met expecta-
tions, underperformed, or overperformed. At the lake level,
we summarized the percentage of lakes that had no under-
performing lake-years (0%), 50% underperforming, and 100%
underperforming lake-year samples for the respective lake.
Finally, we tested for temporal changes in available OHA,
THA, and TOHA from 1980 to 2018 to provide context into
whether populations may be shifting and to help direct wall-
eye management. For this, we evaluated lake-specific habitat
trends using a non-parametric estimation of slope (Sen 1968).
All statistical analysis was done in the R statistical environ-
ment (version 4.1.1; R Core Team 2021).
3. Results
3.1. Model selection of covariates
The relative abundance of small and large walleye was
best explained by including the covariates of lake surface
area and natural recruitment status (Table 2 ). However, the
CPUE of medium-length walleye was best explained by lake
surface area and an interaction between TOHA and natural
recruitment (Table 2 ). For the final analysis, we included
natural recruitment and lake surface area as covariates
when comparing dierent habitat measures to walleye CPUE
because these covariates were included in the model (small
and large), or the model was within DIC 1 of the top
model (medium; Table 2 ).
Table 3. Models comparing the use of thermal-, optical-,
and thermal-optical habitat (THA, OHA, and TOHA re-
spectively) for small (<300 mm), medium (300–450 mm),
and large walleye (450 +mm). Bold text indicates the
top model (DIC <5) and pD is the eective number of
parameters.
Model pD DIC
Small
TOHA +LakeArea +NRcat 200.26 3707.38
THA +LakeArea +NRcat 209.83 3716.32
OHA +LakeArea +NRcat 200.69 3717.07
LakeArea +NRcat 211.03 3717.90
Medium
TOHA +LakeArea +NRcat 185.13 6924.01
THA +LakeArea +NRcat 187.34 6929.53
LakeArea +NRcat 185.60 6933.72
OHA +LakeArea +NRcat 185.02 6936.27
Large
LakeArea +NRcat 171.12 5212.01
OHA +LakeArea +NRcat 172.80 5213.90
THA +LakeArea +NRcat 171.87 5214.34
TOHA +LakeArea +NRcat 172.77 5215.60
3.2. Drivers of walleye relative abundance
Models including TOHA outperformed those including the
component parts and the NULL model for small and medium
length walleye (i.e., NULL, OHA and THA; DIC >5; Tab le 3 )
where TOHA had a positive correlation with walleye CPUE
(Figs. 2 and 3). However, for large walleye, available habitat
was not correlated with CPUE and therefore no support given
for a top model with OHA, THA, and TOHA relative to the
NULL model, which included only lake area and recruitment
class (Table 3 ). Based on the out-of-sample data, the RMSE for
small and medium length walleye were 2.34 and 3.85 fish/net,
respectively. Because large walleye did not show a relation-
ship with any of the habitat measures, we did not evaluate
model performance.
Lake surface area was positively correlated with the den-
sity of walleye across all length categories regardless of which
habitat type was included in the model (Figs. 2 and 4). Dier-
ences among NRcat levels were clearer for small and medium
length walleye, with higher catch rates in lakes with natural
recruitment (NRcat =1 and 2) compared to lakes with lim-
ited or no natural recruitment and supported primarily by
stocking (NRcat =3and4;Figs. 2 and 5). Conversely, the rela-
tionships among NRcat levels were less clear for large wall-
eye, although catch rates were highest in lakes supported
only by natural recruitment (NRcat =1), with no significant
dierences between all other recruitment categories (Figs. 2
and 5).
Catch rates of walleye of all length classes were spatially
correlated (Supplementary Information S2: Fig. S2). Specifi-
cally, sites located in the central part of Minnesota extend-
ing to the eastern border were more likely to display lower
catches of walleye compared to sites in the southwestern

Canadian Science Publishing
278 Can. J. Fish. Aquat. Sci. 80: 273–286 (2023) | dx.doi.org/10.1139/cjfas-2022-0109
Fig. 2. Panel A: Posterior distribution (β, points) and 95% credibility estimates (whiskers) for small, medium, and large walleye.
Credible intervals that overlap zero (vertical horizontal line) are not considered important predictors of walleye CPUE. Panel
B: Dierences among natural recruitment categories (NRcat). Points with 95% credibility estimates (whiskers) indicate the
median dierence between NRcat and comparisons that overlap zero (horizontal line) indicate non-significant dierences
between NRcat categories.
Fig. 3. Relationship between thermal optical habitat and walleye CPUE (fish/gill net) of dierent size classes across lakes and
years.
corner of the state (Supplementary Information S2 Figs. 2A,
C, E). Across the study area, spatial correlation (=0.1) was
found among sites ranging from 43 km (small walleye; Sup-
plementary Information S2: Fig. S2.B) to 69 km apart (large
walleye; Supplementary Information S2: Fig. S2.F).
3.3. Performance across lake-years and
lakes
Overall, the model predicted walleye CPUE well. Most lake-
years of the out-of-sample data set met expectations for small
and medium length walleye (90% and 87%, respectively) with

Canadian Science Publishing
Can. J. Fish. Aquat. Sci. 80: 273–286 (2023) | dx.doi.org/10.1139/cjfas-2022-0109 279
Fig. 4. Relationship between lake surface area and walleye CPUE (fish/gill net) with 95% credibility intervals (gray area).
Fig. 5. Distribution of CPUE (fish/gill net) for small, medium, and large walleye across the four natural recruitment categories:
1=naturally reproducing/no stocking, 2 =naturally reproducing/stocking occurs, 3 =irregular natural reproducing/stocking,
and 4 =no natural reproduction/stocking. Points are the predicted means with the 95% credibility intervals represented by
the whiskers. Matching letters indicate non-significant dierences between natural recruitment categories.
similar proportions of lake-years split between the overper-
forming and underperforming categories (Table 4 ). Lake-years
classified as underperforming displayed similar or higher
TOHA values compared to lake-years that performed well or
overperformed (Fig. 6). For most individual lakes, both ran-
dom samples from the out-of-sample data set were classified
as meeting expectations or overperforming. In 7% and 14% of
lakes, one of the two random samples was classified as un-
derperforming for small and medium walleye, respectively
(Table 4 ). For 1.3% and 1.9% of the lakes, both random sam-
ples of the out-of-sample data set were classified as underper-
forming.

Canadian Science Publishing
280 Can. J. Fish. Aquat. Sci. 80: 273–286 (2023) | dx.doi.org/10.1139/cjfas-2022-0109
Table 4. Sample performance is the percentage of lake-
years that are considered to be meeting expectations
(within 95% prediction intervals), overperforming (>95%
prediction intervals), and underperforming (<95% pre-
diction intervals) for small (<300 mm) and medium
length walleye (300–450 mm). Lake performance is
the number of lakes that have 0%, 50%, and 100% of
lake-years within the out-of-sample data set that are
underperforming.
Small Medium
Sample performance (%)
Meeting Expectations 90.9 86.9
Overperforming 4.3 4.1
Underperforming 4.8 9.0
Lakes underperforming (%)
0% 91.7 84.0
50% 7.1 14.1
100% 1.2 1.9
Fig. 6. Violin plot comparing the thermal optical habitat to
the relative performance of the out-of-sample data set. The
performance categories are defined as (1) observed values
that were within the 95% prediction intervals were consid-
ered “meeting expectations,” (2) observed walleye catch rates
less than the 95% prediction intervals were classified as “un-
derperforming”, and (3) observed walleye catch rates (fish/gill
net) that exceeded the 95% prediction intervals were classi-
fied as “overperforming”. The solid vertical line within each
violin plot is the 50% quantile.
3.4. Temporal trends in available habitat
Optical habitat changed over time in 69% of lakes (14%
increasing [decreasing clarity] and 55% decreasing [increas-
ing clarity]) from 1980 to 2018 (Fig. 7). Thermal habitat also
changed in the majority (53%) of lakes, with 50% increasing
and 3% decreasing. However, the magnitude of change was
generally higher for OHA; overall changes in OHA ranged
from a 43.5% decrease to a 488% increase, whereas changes
in THA ranged from −12% to 45%. Thermal-optical habitat
also changed in most of the lakes; TOHA increased in 26% of
lakes and decreased in 33% of lakes since 1980 (Fig. 7). For
TOHA, the median habitat increase was 58.9% (range =6.1%–
1009%) whereas median habitat loss was 46.6% (range =7.5%–
102%; Fig. 7). Optical habitat area was highly correlated to
changes in TOHA but no such relationship was observed be-
tween changes in THA and TOHA (Fig. 8).
4. Discussion
Evaluating the relationship between available habitat and
walleye population abundance can help set realistic fishery
management expectations and to anticipate responses to en-
vironmental change. Here, we found that the proportion of
a lake in which the preferred temperature and light condi-
tions coexist for walleye was predictive of gillnet catches of
small and medium length walleye and was a better predictor
than either thermal- or optical-habitat alone. Our results sug-
gest that habitat area could be used to predict the abundance
of small and medium length walleye when survey data are
unavailable and to set expectations for walleye abundance
given habitat conditions. In contrast, none of the measures
of habitat area we tested related to catches of larger walleye.
In addition, relative abundances of walleye were correlated to
the level of natural recruitment and the surface area of the
lake. We documented increasing and decreasing trends for
TOHA in 59% of walleye lakes with changes in OHA primarily
driving TOHA trends. We also found a significant spatial de-
pendency in walleye relative abundance that covariates alone
could not explain. Our results can assist managers in identi-
fying lakes that have limited potential as sustainable walleye
fisheries and would benefit by shifting fishery management
to alternate species (e.g., resist, accept, or direct; Thompson
et al. 2021).
4.1. Predictors of walleye relative abundance
Available habitat was an important driver of small and
medium length walleye. This result is in-line with previous
studies showing that TOHA was positively correlated with
CPUE of smaller walleye (Hansen et al. 2019). However, we
documented no relationship between habitat and larger wall-
eye. Large walleye often experience a range of factors that
contribute to mortality over their life cycle. For example,
fishing pressure on these age classes may mask the relation-
ship between catches and available habitat. Fishing mortal-
ity can vary, depending on gear type (Payer et al. 1989), but
has been shown to be as high as 27% (Herbst et al. 2016). Al-
ternatively, larger, more mature walleye may use habitat dif-
ferently than smaller, immature walleye and thus may not
overlap spatially with the demersal habitat being sampled
during surveys. While the literature is limited, this does not
appear to be the case as gill nets that fished near the bottom
of the lake had a higher likelihood of capturing larger adult
walleye compared to smaller walleye (Gorman et al. 2019). In

Canadian Science Publishing
Can. J. Fish. Aquat. Sci. 80: 273–286 (2023) | dx.doi.org/10.1139/cjfas-2022-0109 281
Fig. 7. Temporal trends (left panels) of optical (OHA), thermal (THA), and thermal-optical habitat (TOHA) from 1980 to 2018.
Values greater than 0 (dashed line) indicate increasing habitat with significant increases indicated in green. Values less than 0
indicate decreasing habitat with significant decreases in yellow. Slope values were calculated using a Sen’s slope. The % change
(right panels) in OHA, THA, and TOHA from 1980 to 2018. To improve the figure readability, we set the x-axis limit to 200.
addition, we used a 3-year rolling average of TOHA to
compare relative abundances and this correlated well with
catches of small and medium walleye. Still, the early growth
conditions associated with large walleye are several years re-
moved from this 3-year average. Thus, dierent temporal res-
olutions of TOHA may relate better to large walleye and po-
tentially to the dierent size classes in general (e.g., 2- vs. 4-
year rolling average).

Canadian Science Publishing
282 Can. J. Fish. Aquat. Sci. 80: 273–286 (2023) | dx.doi.org/10.1139/cjfas-2022-0109
Fig. 8. Linear relationships between thermal-optical habitat
(TOHA) changes in thermal habitat (top figure, THA) and opti-
cal habitat (bottom, OHA). The dierent colors among points
indicate significant increases in TOHA (green), lakes with sta-
ble changes in TOHA (purple), and lakes where TOHA is sig-
nificantly decreasing (yellow).
Walleye catch rates for small- and medium-sized walleye
were higher in lakes with larger surface area. The positive re-
lationship between walleye relative abundance and lake sur-
face area is like previous studies that assessed the eects of
lake size (Nate et al. 2000;Wehrly et al. 2012;Hansen et al.
2017). For instance, Nate et al. (2000,2001) found a significant
relationship with lake size and the recruitment of age-0 and
adult walleye. Larger lakes may provide more diverse habitats
(Wehrly et al. 2012) or prey species. Although we also found a
significant relationship with large walleye, the eect size ap-
peared to be minimal compared to small- and medium-sized
walleye classes, again suggesting that other factors are limit-
ing the carrying capacity of the lake for older age classes.
Catch rates of small- and medium-sized walleye are related
to the level of natural recruitment and stocking. Following
trends observed in Nate et al. (2000), we found increased
small- and medium-sized walleye abundances between lakes
that are believed to sustain naturally reproducing walleye
populations (regardless of the presence of stocking) com-
pared to lakes with limited or no natural recruitment. Also
similar to Nate et al. (2000), we found that dierences be-
tween walleye abundance attributable to natural recruitment
were not as clear for large walleye. Here, lakes with sus-
tainable natural reproduction and no stocking (NRcat =1)
showed increased catch rates for large walleye compared to
lakes where stocking occurred (NRcat =2–4). These results
support the idea that lakes with natural reproduction can sus-
tain higher walleye densities than lakes supported by stock-
ing, and that such results hold up even when accounting for
dierences in habitat area.
4.2. Spatial and temporal dynamics of habitat
Walleye habitat was spatially correlated, with sites within
70 km showing high degrees of correlation. This correlation
is on a similar scale to that found by Myers et al. (1997),where
spatial synchrony for recruitment among freshwater species
was found to be <50 km. While studies evaluating synchrony
among walleye populations are limited, many recent studies
evaluating other lake species have found synchrony to ex-
tend up to 400 km among dierent populations (Coregonus
hoyi,Bunnell et al. 2010;Coregonus artedi,Myers et al. 2015;
Perca flavescens,Dembkowski et al. 2016). The scale of the ex-
planatory variables may explain the dierence between our
study and previous studies. For instance, we focused on lake-
specific explanatory variables that would not be expected to
extend beyond the lake boundaries rather than climatic vari-
ables that work on broader spatiotemporal extents (Bunnell
et al. 2010;Myers et al. 2015;Dembkowski et al. 2016). Our
results highlight the importance of incorporating the spa-
tially explicit nature of the landscape during modeling to
account for unknown processes that can drive ecosystems
(Hoeting 2009). Paleoecological scales, such as geomorphol-
ogy and substrate composition, may be driving some of the
spatial patterns observed in our study (see Cross and McIn-
erny 2005 as an example for bluegill [Lepomis macrochirus]).
However, based on visual evaluation of residuals relative to
ecoregion patterns, the spatial patterns of walleye catches
do not appear to correlate well with the natural large-scale
spatial patterns across the study area (Heiskary et al. 1987).
The spatial patterns also appear to radiate from the urban
Minneapolis–St.Paul metro area with lower relative abun-
dances of walleye close to the metro area and thus could be
the result of increased fishing pressure on lakes closer to ur-
ban populations compared to lakes that require more eort
to reach (Hunt et al. 2011).
Temporal trends of TOHA exhibited a range of responses
among the lakes included in our study. For instance, changes
in TOHA can result from changes in water clarity, temper-
ature, or both. Water clarity trends in lakes in the upper
Midwestern United States are both increasing and decreas-
ing depending on the lake (Oliver et al. 2017;Olmanson et
al. 2020). Furthermore, the response of OHA to changing
water clarity varies among lakes due to dierences in lake

Canadian Science Publishing
Can. J. Fish. Aquat. Sci. 80: 273–286 (2023) | dx.doi.org/10.1139/cjfas-2022-0109 283
morphometry (Lester et al. 2004;Geisler et al. 2016). Optical
habitat area increased over time in lakes across our study
area, and these changes in OHA were correlated with changes
in TOHA. Consequently, increased TOHA could be due to de-
clining water clarity, driven by eutrophication (Brezonik et
al. 2019), precipitation, and drought dynamics in certain
kinds of lakes (Lisi and Hein 2019). Conversely, lakes where
TOHA has decreased could be related to increased water
clarity due to zebra mussel (Dreissena polymorpha) invasions
(Higgins and Zanden 2010;Geisler et al. 2016)orbetterwater
quality management.
Suitable THA increased in most lakes in our study. Many
lakes are warming (O’Reilly et al. 2015) and can have di-
rect consequences on the thermal structure of the lake
(McCormick 1990;Snucins and John 2000;Keller 2007). Still,
the influence of climate change on THA depends on lake
morphometry and climate conditions (Kraemer et al. 2015).
Earlier springs and later falls increase the length of the
growing season, meaning more days in which water temper-
atures are not colder than the lower limit of walleye optimal
growth temperatures (11 ◦C). Because water temperatures in
Minnesota lakes rarely exceed the upper thermal tolerance
of walleye, we expect a net increase in thermal habitat for
many lakes (Hansen et al. 2019,Honsey et al. 2020). As water
temperatures continue to increase, available thermal habitat
will likely play a more important role in determining walleye
abundance as temperatures begin to exceed the upper limit
of walleye tolerance more frequently. This is especially true
in smaller, shallower lakes that lack the thermal refugia
often seen in larger lakes or in lakes at the southern extent
of walleye distributions.
4.3. Application to walleye management
To manage natural resources, it is important to understand
the limitations of the ecosystem to support the fauna that
depend on it. Static measures of habitat are often used to
assess the carrying capacity of an ecosystem, which may re-
sult in unrealistic expectations in a rapidly changing envi-
ronment. Thermal-optical habitat oers a dynamic measure-
ment of walleye habitat that is susceptible to changes in the
environment and reflects how walleye respond to dierent
environmental drivers such as temperature and water clarity
(Tunney et al. 2018;Hansen et al. 2019). TOHA is a relatively
simple metric with which to set realistic boundaries to as-
sess the potential productivity of a lake to support walleye.
In the absence of long-term monitoring data, using TOHA as
a proxy for walleye populations could help set the lower and
upper bounds of what would be expected and with which to
evaluate appropriate management strategies.
Integrating dynamic information on habitat availability
(Milanesi et al. 2020) into management frameworks (e.g., safe
operating space; Carpenter et al. 2017) will allow managers
to account for potential changes in fisheries productivity
in the face of environmental change. For instance, harvest
restrictions are the primary mechanism available to man-
agers to account for changing habitats. Thus, setting harvest
limits within a safe operating space of the ecosystem (e.g.,
what the habitat can support; Carpenter et al. 2017;Hansen
et al. 2019) will ensure populations are not overfished and
therefore more robust to ecosystem change. Furthermore,
understanding how lakes perform relative to their physical,
biological, and management attributes can feed into manage-
ment decisions and where to allocate recourses (e.g., resist,
accept, direct framework; Thompson et al. 2021;Feiner et al.
2022;Rahel 2022;Shultz et al. 2022). For instance, most wall-
eye populations met expectations given the available habitat,
lake size, and level of natural recruitment. Yet, underper-
forming walleye lake-years had similar or consistently higher
TOHA values compared to lake-years that were categorized
as meeting expectations or overperforming. Such disparities
would indicate that factors, outside available habitat, are
driving some walleye populations (e.g., overexploitation;
Embke et al. 2019). Depending on the underlying cause of
the decline, these lakes may provide an opportunity to resist
change by promoting management actions that allow the
populations to grow and meet expectations. Conversely,
many lakes within the study area had limited available
habitat and were supported solely through stocking eorts.
These lakes might benefit by either accepting that the wall-
eye population has limited potential to be self-sustaining
and directing conservation eorts to alternate species that
are more resilient to the current lake conditions (e.g.,
largemouth bass [Micropterus salmoides]; Hansen et al. 2018).
The dierences in catchability across sampling gears and
methods may limit our understanding of how walleye relate
to habitat metrics outside of Minnesota. Mesh size and fish
length both aect the selectivity of gillnets for indexing the
abundance of walleye (Radomski et al. 2020). Mortality of fish
due to harvest and other causes unrelated to habitat is also
related to fish size, with larger individuals subject to multiple
sources of mortality throughout their lives that would decou-
ple their abundance from habitat availability. We examined
multiple size groups of fish to attempt to strike a balance
between selectivity and harvest pressure, although no size
group is both fully selected to the gear and tightly coupled
to habitat. In addition, Giacomini et al. (2020) found that the
North American (NA) standard net had a 20% higher CPUE
compared to the Fall Walleye Index Netting (FWIN) used by
the Ontario Ministry of Natural Resources. The gillnet used
in this study uses a narrower mesh size range compared to
the NA and FWIN nets (Giacomini et al. 2020), and we expect
that the nets used in this study would likely underestimate
CPUE relative to both these nets. Thus, caution is warranted
when directly compare the absolute CPUE across sampling
protocols. However, because we observed similar trends to
Lester et al. 2004, we would expect similar trends between
TOHA and gear-specific CPUE. Therefore, to be useful in man-
agement decisions we recommend that agencies evaluate the
relationship between TOHA and CPUE based on their specific
sampling protocols.
The mechanisms that drive walleye population are often
complex and are the result of multiple factors acting in con-
cert with each other and independently across timescales
(Jackson et al. 2021). A key attribute that we were not able
to address was to try and identify thresholds of TOHA that
can be used to identify lakes that may benefit in manage-
ment objectives beyond walleye. Long-term monitoring data

Canadian Science Publishing
284 Can. J. Fish. Aquat. Sci. 80: 273–286 (2023) | dx.doi.org/10.1139/cjfas-2022-0109
are often limited to larger more robust lakes and thus we
lacked some of the representative data across the range of
lake types to identify thresholds between TOHA and walleye
catches. Although our study provides useful information to
help guide walleye management, future research could also
focus on incorporating additional drivers (e.g., both anthro-
pogenic and natural) of walleye populations beyond changes
in TOHA. This will allow us to better understand how avail-
able habitat relates to walleye relative to these stressors (e.g.,
invasive species).
Inland recreational fisheries are influenced by complex
and dynamic interactions among a range of ecosystem and
population drivers that directly and indirectly influence
aquatic ecosystems. To eectively manage aquatic ecosys-
tems, it is necessary to identify relationships between the fo-
cal species and available habitat. Here, we demonstrate that
walleye populations were correlated to available TOHA and
recommend that understanding habitat trends through time
can assist managers to set realistic expectations and guide re-
sources. Furthermore, evaluating walleye populations at the
broad spatial extents observed in our study can assist in shift-
ing from a single lake management scheme to the manage-
ment of multiple habitat units simultaneously. Information
is constantly increasing on the similarities of walleye pop-
ulations across the landscape and how dierent biological
(e.g., OHA; Vitense and Hansen 2021) and anthropogenic pres-
sures (e.g., boater movements: Kao et al. 2021; watermilfoil:
Thomas et al. 2021) interact across space. This allows man-
agers to identify population thresholds that allow for the
implementation of management schemes that integrate for-
ward thinking management approaches that account for con-
tinually changing ecosystem dynamics.
Acknowledgements
We acknowledge with gratitude the hundreds of MNDNR em-
ployees who collected data that form the basis of this re-
search. Special thanks to C. Geving and J. Hansen for assist-
ing with data collation and interpretation. We also acknowl-
edge the input of the MNDNR Fisheries Research Unit for
assisting in project development and model interpretation.
Also, thanks to Fisheries Systems Ecology Lab members and
A. Appling for help with modeling and providing feedback for
previous versions of the manuscript. The MNDNR provided
funding for KV and SM. This work was supported in part by
grant G20AC00096 from the USGS Midwest Climate Adapta-
tion Climate Science Center. GH acknowledges the support
of the USDA National Institute of Food and Agriculture Hatch
project MIN-41–101. Any opinions, findings, conclusions, or
recommendations expressed in this publication are those of
the author (s) and do not necessarily reflect the view of the Na-
tional Institute of Food and Agriculture or the United States
Department of Agriculture but do represent the views of the
U.S. Geological Survey. Any use of trade, firm, or product
names is for descriptive purposes only and does not imply en-
dorsement by the U.S. Government. Finally, we would like to
thank the anonymous five reviewers that provided construc-
tive feedback on the manuscript, which greatly improved the
overall content of our study.
Article information
History dates
Received:30May2022
Accepted: 9 September 2022
Version of record online: 10 January 2023
Copyright
© 2023 Copyright remains with the author(s) or their insti-
tution(s). This work is licensed under a Creative Commons
Attribution 4.0 International License (CC BY 4.0), which per-
mits unrestricted use, distribution, and reproduction in any
medium, provided the original author(s) and source are cred-
ited.
Data availability
Data are not yet provided but will be made available through
the University of Minnesota data repository (https://conserva
ncy.umn.edu/drum).
Author information
Author ORCIDs
Shad Mahlum https://orcid.org/0000-0002-2663-2677
GretchenJ.A.Hansenhttps://orcid.org/0000-0003-0241-7048
Author contributions
Conceptualization: KV, GJAH
Data curation: SM, KV, HC-D, LP, JSR, PJS, MT, GJAH
Formal analysis: SM
Funding acquisition: GJAH
Investigation: GJAH
Methodology: SM, KV, HC-D, LP, JSR, PJS, MT, GJAH
Project administration: GJAH
Resources: GJAH
Supervision: JSR, GJAH
Visualization: SM
Writing – original draft: SM, KV, HC-D, LP, JSR, PJS, GJAH
Writing – review & editing: SM, KV, HC-D, LP, JSR, PJS, MT,
GJAH
Competing interests
The authors declare there are no competing interests.
Supplementary material
Supplementary data are available with the article at https:
//doi.org/10.1139/cjfas-2022-0109.
References
Brezonik, P.L., Bouchard, R.W., Jr, Finlay, J.C., Grin, C.G., Olmanson,
L.G. Anderson, J.P., et al. 2019. Color, chlorophyll a, and suspended
solids eects on Secchi depth in lakes: implications for trophic
state assessment. Ecol. Appl. 29: e01871. doi:10.1002/eap.1871. PMID:
30739365.
Bunnell, D.B., Adams, J.V., Gorman, O.T., Madenjian, C.P., Riley, S.C.,
Roseman, E.F., and Schaeer, J.S. 2010. Population synchrony of a
native fish across three Laurentian Great Lakes: evaluating the ef-
fects of dispersal and climate. Oecologia 162: 641–651. doi:10.1007/
s00442-009- 1487-6. PMID: 19888603.

Canadian Science Publishing
Can. J. Fish. Aquat. Sci. 80: 273–286 (2023) | dx.doi.org/10.1139/cjfas-2022-0109 285
Cantin, A., and Post, J.R. 2018. Habitat availability and ontogenetic shifts
alter bottlenecks in size-structured fish populations. Ecology, 99:
1644–1659. doi:10.1002/ecy.2371. PMID: 29705987.
Carpenter, S.R., Brock, W.A., Hansen, G.J.A., Hansen, J.F., Hennessy, J.M.
Isermann, D.A., et al. 2017. Defining a safe operating space for in-
land recreational fisheries. Fish Fish. 18: 1150–1160. doi:10.1111/faf.
12230.
Christie, G.C., and Regier, H.A. 1988. Measures of optimal thermal habitat
and their relationship to yields for four commercial fish species. Can.
J. Fish Aquat. Sci. 45: 301–314. doi:10.1139/f88-036.
Cooke, S.J., Arlinghaus, R., Johnson, B.M., and Cowx, I.G. 2015. Recre-
ational fisheries in inland waters. Freshwater fisheries ecology. pp.
449–465. doi:10.1002/9781118394380.ch36.
Cross, T.K., and McInerny, M.C. 2005. Spatial habitat dynamics aecting
bluegill abundance in Minnesota bass–panfish lakes. North Am. J.
Fish. Manag. 25: 1051–1066. doi:10.1577/M04-072.1.
Dassow, C.J., Latzka, A.W., Lynch, A.J., Sass, G.G., Tingley Iii, R.W., and
Paukert, C.P. 2022. A resist-accept-direct decision-support tool for
walleye Sander vitreus (Mitchill) management in Wisconsin. Fish.
Manage. Ecol. 29: 378– 391, doi: 10.1111/fme.12548. PMID: 35942481.
Dembkowski, D.J., Willis, D.W., and Wuellner, M.R. 2016. Synchrony
in larval yellow perch abundance: the influence of the Moran ef-
fect during early life history. Can. J. Fish. Aquat. Sci. 73: 1567–1574.
doi:10.1139/cjfas-2015- 0310.
Embke, H.S., Rypel, A.L., Carpenter, S.R., Sass, G.G., Ogle, D. Cichosz, T.,
et al. 2019. Production dynamics reveal hidden overharvest of inland
recreational fisheries. Proc. Natl. Acad. Sci. U. S. A. 116: 24676–24681.
doi:10.1073/pnas.1913196116.
Feiner, Z.S., Shultz, A.D., Sass, G.G., Trudeau, A., Mitro, M.G. Dassow,
C.J., et al. 2022. Resist-accept-direct (RAD) considerations for cli-
mate change adaptation in fisheries: the Wisconsin experience. Fish.
Manag. Ecol. 29: 346–-353. doi: 10.1111/fme.12549. PMID: 35942481.
Fulton, C.J., Noble, M.N., Radford, B., Gallen, C., and Harasti, D. 2016.
Microhabitat selectivity underpins regional indicators of fish abun-
dance and replenishment. Ecol. Indic. 70: 222–231. doi:10.1016/j.
ecolind.2016.06.032.
Geisler, M.E., Rennie, M.D., Gillis, D.M., and Higgins, S.N. 2016. A pre-
dictive model for water clarity following dreissenid invasion. Biol.
Invasions, 18: 1989–2006. doi:10.1007/s10530-016-1146-x.
Giacomini, H.C., Lester, N., Addison, P., Sandstrom, S., Nadeau, D., Chu,
C., and de Kerckhove, D. 2020. Gillnet catchability of walleye (Sander
vitreus) : comparison of North American and provincial standards.
Fish. Res. 224: 105433. doi:10.1016/j.fishres.2019.105433.
Gorman, A.M., Kraus, R.T., Gutowsky, L.F.G., Vandergoot, C.S., Zhao, Y.
Knight, C.T., et al. 2019. Vertical habitat use by adult walleyes conflicts
with expectations from fishery-independent surveys. Trans. Am. Fish.
Soc. 148: 592–604. doi:10.1002/tafs.10150.
Gostiaux, C., Boehm, J.H.I.A., Jaksha, N.J., Dembkowski, D.J., Hennessy,
J.M., and Isermann, D.A. 2022. Recruitment bottlenecks for age-0
walleye in northern Wisconsin lakes. North Am. J. Fish. Manag 42:
507–522.
Gutowsky, L.F.G., Giacomini, H.C., de Kerckhove, D.T., Mackereth, R.,
McCormick, D., and Chu, C. 2019. Quantifying multiple pressure
interactions aecting populations of a recreationally and commer-
cially important freshwater fish. Global Change Biol. 25: 1049–1062.
doi:10.1111/gcb.14556. PMID: 30580472.
Hansen, G.J.A., Hennessy, J.M., Cichosz, T.A., and Hewett, S.W. 2015. Im-
proved models for predicting walleye abundance and setting safe har-
vest quotas in northern Wisconsin lakes. North Am. J. Fish. Manag.
35: 1263–1277. doi:10.1080/02755947.2015.1099580.
Hansen, G.J.A., Midway, S.R., and Wagner, T. 2018. Walleye recruitment
success is less resilient to warming water temperatures in lakes with
abundant largemouth bass populations. Can. J. Fish. Aquat. Sci. 75:
106–115. doi:10.1139/cjfas-2016- 0249.
Hansen, G.J.A., Read, J.S., Hansen, J.F., and Winslow, L.A. 2017. Projected
shifts in fish species dominance in Wisconsin lakes under climate
change. Global Change Biol. 23: 1463–1476. doi:10.1111/gcb.13462.
PMID: 27608297.
Hansen, G.J.A., Winslow, L.A., Read, J.S., Treml, M., Schmalz, P.J., and Car-
penter, S.R. 2019. Water clarity and temperature eects on walleye
safe harvest: an empirical test of the safe operating space concept.
Ecosphere, 10: e02737. doi:10.1002/ecs2.2737.
Heino, J., Alahuhta, J., Bini, L.M., Cai, Y., Heiskanen, A.S. Hellsten, S.,
et al. 2021. Lakes in the era of global change: moving beyond single-
lake thinking in maintaining biodiversity and ecosystem services.
Biol.Rev.Camb.Philos.Soc.96: 89–106. doi:10.1111/brv.12647. PMID:
32869448.
Heiskary, S.A., Wilson, C.B., and Larsen, D.P. 1987. Analysis of regional
patterns in lake water quality: using ecoregions for lake manage-
ment in Minnesota. Lake Res. Manage. 3: 337–344. doi:10.1080/
07438148709354789.
Herbst, S.J., Stevens, B.S., Hayes, D.B., and Hanchin, P.A. 2016. Estimat-
ing walleye (Sander vitreus) movement and fishing mortality using
state-space models: implications for management of spatially struc-
turedpopulations.Can.J.Fish.Aquat.Sci.73: 330–348. doi:10.1139/
cjfas-2015- 0021.
Higgins, S.N., and Zanden, M.J.V. 2010. What a dierence a species makes:
a meta–analysis of dreissenid mussel impacts on freshwater ecosys-
tems. Ecol. Monogr. 80: 179–196. doi:10.1890/09-1249.1.
Hipsey, M.R., Bruce, L.C., Boon, C., Busch, B., Carey, C.C. Hamilton,
D.P., et al. 2019. A general lake model (GLM 3.0) for linking with
high-frequency sensor data from the global lake ecological observa-
tory network (GLEON). Geosci. Model. Dev. 12: 473–523. doi:10.5194/
gmd-12- 473-2019.
Hoeting, J.A. 2009. The importance of accounting for spatial and tempo-
ral correlation in analyses of ecological data. Ecol. Appl. 19: 574–577.
doi:10.1890/08-0836.1. PMID: 19425418.
Honsey, A.E., Feiner, Z.S., and Hansen, G.J.A. 2020. Drivers of walleye
recruitment in Minnesota’s large lakes. Can. J. Fish. Aquat. Sci. 77:
1921–1933. doi:10.1139/cjfas-2019- 0453.
Hunt, L.M., Arlinghaus, R., Lester, N., and Kushneriuk, R. 2011. The ef-
fects of regional angling eort, angler behavior, and harvesting ef-
ficiency on landscape patterns of overfishing. Ecol. Appl. 21: 2555–
2575. doi:10.1890/10-1237.1. PMID: 22073644.
Jackson, M.C., Pawar, S., and Woodward, G. 2021. The temporal dynamics
of multiple stressor eects: from individuals to ecosystems. Trends
Ecol. Evol. 36: 402–410.
Jacobson, P.C., Cross, T.K., Dustin, D.L., and Duval, M. 2016. A fish habitat
conservation framework for Minnesota lakes. Fisheries, 41: 302–317.
doi:10.1080/03632415.2016.1172482.
Kao, S.-Y.Z., Enns, E.A., Tomamichel, M., Doll, A., Escobar, L.E. Qiao, H.,
et al. 2021. Network connectivity of Minnesota waterbodies and im-
plications for aquatic invasive species prevention. Biol. Invasions. 23:
3231–3242. doi: 10.1007/s10530-021- 02563-y.
Keller, W. 2007. Implications of climate warming for boreal shield
lakes: a review and synthesis. Environ. Rev. 15: 99–112. doi:10.1139/
A07-002.
Koch, E.W., Barbier, E.B., Silliman, B.R., Reed, D.J., Perillo, G.M.E. Hacker,
S.D., et al. 2009. Non-linearity in ecosystem services: temporal and
spatial variability in coastal protection. Front. Ecol. Environ. 7: 29–
37. doi:10.1890/080126.
Kraemer, B.M., Anneville, O., Chandra, S., Dix, M., Kuusisto, E. Living-
stone, D.M., et al. 2015. Morphometry and average temperature aect
lake stratification responses to climate change. Geophys. Res. Lett.
42: 4981–4988. doi:10.1002/2015GL064097.
Lester, N.P., Dextrase, A.J., Kushneriuk, R.S., Rawson, M.R., and Ryan,
P.A. 2004. Light and temperature: key factors aecting walleye abun-
dance and production. Trans. Am. Fish. Soc. 133: 588–605. doi:10.
1577/T02-111.1.
Lester, N.P., Marshall, T.R., Armstrong, K., Dunlop, W.I., and Ritchie,
B. 2003. A broad-scale approach to management of Ontario’s recre-
ational fisheries. North Am. J. Fish. Manag. 23: 1312–1328. doi:10.
1577/M01-230AM.
Lester, N.P., Sandstrom, S., de Kerckhove, D.T., Armstrong, K., Ball, H.
Amos, J., et al. 2021. Standardized broad-scale management and mon-
itoring of inland lake recreational fisheries: an overview of the On-
tario experience. Fisheries, 46: 107–118. doi:10.1002/fsh.10534.
Lisi, P.J., and Hein, C.L. 2019. Eutrophication drives divergent water clar-
ity responses to decadal variation in lake level. Limnol. Oceanogr. 64:
S49–S59. doi:10.1002/lno.11095.
Lynch, A.J., Cooke, S.J., Deines, A.M., Bower, S.D., Bunnell, D.B. Cowx,
I.G., et al. 2016. The social, economic, and environmental importance
of inland fish and fisheries. Environ. Rev. 24: 115–121. doi:10.1139/
er-2015- 0064.

Canadian Science Publishing
286 Can. J. Fish. Aquat. Sci. 80: 273–286 (2023) | dx.doi.org/10.1139/cjfas-2022-0109
McCormick, M.J. 1990. Potential changes in thermal structure and cy-
cle of Lake Michigan due to global warming. Trans. Am. Fish. Soc.
119: 183–194. doi:10.1577/1548-8659(1990)119%3c0183:PCITSA%3e2.
3.CO;2.
Milanesi, P., Rocca, F.D, and Robinson, R.A. 2020. Integrating dynamic
environmental predictors and species occurrences: toward true dy-
namic species distribution models. Nat. Ecol. Evol. 10: 1087–1092.
doi:10.1002/ece3.5938. PMID: 32015866.
Myers, J.T., Yule, D.L., Jones, M.L., Ahrenstor, T.D., Hrabik, T.R. Clara-
munt, R.M., et al. 2015. Spatial synchrony in cisco recruitment. Fish.
Res. 165: 11–21. doi:10.1016/j.fishres.2014.12.014.
Myers, R.A., Mertz, G., and Bridson, J. 1997. Spatial scales of interannual
recruitment variations of marine, anadromous, and freshwater fish.
Can. J. Fish. Aquat. Sci. 54: 1400–1407. doi:10.1139/f97-045.
Nate, N.A., Bozek, M.A., Hansen, M.J., and Hewett, S.W. 2000. Variation in
walleye abundance with lake size and recruitment source. North Am.
J. Fish. Manag. 20: 119–126. doi:10.1577/1548-8675(2000)020%3c0119:
VIWAWL%3e2.0.CO;2.
Nate, N.A., Bozek, M.A., Hansen, M.J., and Hewett, S.W. 2001. Variation
of adult walleye abundance in relation to recruitment and limnolog-
ical variables in northern Wisconsin lakes. North Am. J. Fish. Manag.
21: 441–447. doi:10.1577/1548-8675(2001)021%3c0441:VOAWAI%3e2.
0.CO;2.
O’Reilly, C.M., Sharma, S., Gray, D.K., Hampton, S.E., Read, J.S. Rowley,
R.J., et al. 2015. Rapid and highly variable warming of lake surface
waters around the globe. Geophys. Res. Lett. 42: 10–773. doi:10.1002/
2015GL066235.
Oliver, S.K., Collins, S.M., Soranno, P.A., Wagner, T., Stanley, E.H. Jones,
J.R., et al. 2017. Unexpected stasis in a changing world: lake nutri-
ent and chlorophyll trends since 1990. Global Change Biol. 23: 5455–
5467. doi:10.1111/gcb.13810. PMID: 28834575.
Olmanson, L.G., Page, B.P., Finlay, J.C., Brezonik, P.L., Bauer, M.E., Grif-
fin, C.G., and Hozalski, R.M. 2020. Regional measurements and spa-
tial/temporal analysis of CDOM in 10,000+optically variable Min-
nesota lakes using Landsat 8 imagery. Sci. Total Environ. 724: 138141.
doi:10.1016/j.scitotenv.2020.138141. PMID: 32247976.
Paukert, C.P., Glazer, B.A., Hansen, G.J.A., Irwin, B.J., Jacobson, P.C. Ker-
shner, J.L., et al. 2016. Adapting inland fisheries management to a
changing climate. Fisheries, 41: 374–384. doi:10.1080/03632415.2016.
1185009.
Payer, R.D., Pierce, R.B., and Pereira, D.L. 1989. Hooking mortality of
walleyes caught on live and artificial baits. North Am. J. Fish. Manag.
9: 188–192. doi:10.1577/1548-8675(1989)009%3c0188:HMOWCO%3e2.
3.CO;2.
Radomski, P., Anderson, C.S., Bruesewitz, R.E., Carlson, A.J., and
Borkholder, B.D. 2020. An assessment model for a standard gill net
incorporating direct and indirect selectivity applied to walleye. North
Am. J. Fish. Manag. 40: 105–124. doi:10.1002/nafm.10384.
Rahel, F.J. 2022. Managing freshwater fish in a changing climate: resist,
accept, or direct. Fisheries, 47:245–255. doi: 10.1002/fsh.10726. PMID:
35465292.
Read, J.S., Appling, A.P., Vitense, K., Oliver, S.K., and Hansen, G.J.A. 2021.
Walleye thermal optical habitat area (TOHA) of selected Minnesota
lakes. U.S. Geological Survey data release.
Read, J.S., Jia, X., Willard, J., Appling, A.P., Zwart, J.A. Oliver, S.K., et al.
2019. Process-guided deep learning predictions of lake water temper-
ature. Water Resour. Res. 55: 9173–9190. doi:10.1029/2019WR024922.
Rue, H., Martino, S., and Chopin, N. 2009. Approximate Bayesian infer-
ence for latent Gaussian models by using integrated nested Laplace
approximations. J. R. Stat. Soc. Ser. B Stat. Methodol. 71: 319–392.
doi:10.1111/j.1467-9868.2008.00700.x.
Ryder, R.A. 1965. A method for estimating the potential fish production
of north-temperate lakes. Trans. Am. Fish. Soc. 94: 214–218. doi:10.
1577/1548-8659(1965)94%5b214:AMFETP%5d2.0.CO;2.
Ryder, R.A. 1977. Eects of ambient light variations on behavior of year-
ling, subadult, and adult walleyes (Stizostedion vitreum vitreum). J.
Fish. Board Can. 34: 1481–1491. doi:10.1139/f77-213.
Rypel, A.L., Simonson, T.D., Oele, D.L., Grin, J.D.T., Parks, T.P. Seibel,
D., et al. 2019. Flexible classification of Wisconsin lakes for im-
proved fisheries conservation and management. Fisheries, 44: 225–
238. doi:10.1002/fsh.10228.
Sass, G.G., Rypel, A.L., and Staord, J.D. 2017. Inland fisheries habi-
tat management: lessons learned from wildlife ecology and a pro-
posal for change. Fisheries, 42: 197–209. doi:10.1080/03632415.2017.
1276344.
Sen, P.K. 1968. Estimates of the regression coecient based on Kendall’s
tau. J. Am. Stat. Assoc. 63: 1379–1389. doi:10.1080/01621459.1968.
10480934.
Shultz, A., Luehring, M., Ray, A., Rose, J.D., Croll, R. Gilbert, J., et al. 2022.
Case study: applying the resist–accept–direct framework to an Ojibwe
tribe’s relationship with the natural world. Fish. Manage. Ecol. 29:
392–-408. 10.1111/fme.12568. PMID: 35942481.
Simpson, D., Rue, H., Riebler, A., Martins, T.G., and Sørbye, S.H. 2017.
Penalising model component complexity: A principled, practical ap-
proach to constructing priors. Stat. Sci. 32(1): 1–28. doi:10.1214/
16-STS576.
Snucins, E., and John, G. 2000. Interannual variation in the thermal struc-
ture of clear and colored lakes. Limnol. Oceanogr. 45: 1639–1646.
doi:10.4319/lo.2000.45.7.1639.
Spiegelhalter, D.J., Best, N.G., Carlin, B.P., and Van Der Linde, A. 2002.
Bayesian measures of model complexity and fit. J. R. Stat. Soc. Ser. B
Stat. Methodol. 64: 583–639. doi:10.1111/1467-9868.00353.
Spiegelhalter, D.J., Thomas, A., Best, N., and Lunn, D. 2003. WinBUGS
version 1.4 user manual. MRC Biostatistics Unit, Cambridge. Available
from http://www. mrc-bsu.cam.ac.uk/bugs.
Thomas, S.M., Verhoeven, M.R., Walsh, J.R., Larkin, D.J., and Hansen,
G.J.A. 2021. Species distribution models for invasive Eurasian water-
milfoil highlight the importance of data quality and limitations of
discrimination accuracy metrics. Nat. Ecol. Evol. 11: 12567–12582.
doi:10.1002/ece3.8002. PMID: 34594521.
Thompson, L.M., Lynch, A.J., Beever, E.A., Engman, A.C., Falke, J.A. Jack-
son, S.T., et al. 2021. Responding to ecosystem transformation: resist,
accept, or direct? Fisheries, 46: 8–21. doi:10.1002/fsh.10506.
Tunney, T.D., McCann, K.S., Jarvis, L., Lester, N.P., and Shuter, B.J. 2018.
Blinded by the light? Nearshore energy pathway coupling and relative
predator biomass increase with reduced water transparency across
lakes. Oecologia, 186: 1031–1041. doi:10.1007/s00442-017- 4049-3.
PMID: 29388026.
van Poorten, B.T., and Camp, E.V. 2019. Addressing challenges common
to modern recreational fisheries with a buet-style landscape man-
agement approach. Rev. Fish. Sci. 27: 393–416.
VanDerWal, J., Shoo, L.P., Johnson, C.N., and Williams, S.E. 2009. Abun-
dance and the environmental niche: environmental suitability esti-
mated from niche models predicts the upper limit of local abun-
dance. Am. Nat. 174: 282–291. doi:10.1086/600087. PMID: 19519279.
Vitense, K., and Hansen, G.J.A. 2021. Data and R code supporting “Non-
linear water clarity trends and impacts on littoral area in Minnesota
lakes.”
Wehrly, K.E., Breck, J.E., Wang, L., and Szabo-Kraft, L. 2012. A landscape-
based classification of fish assemblages in sampled and unsampled
lakes. Trans. Am. Fish. Soc. 141: 414–425. doi:10.1080/00028487.2012.
667046.