Temporal Instability of Lake Charr Phenotypes: Synchronicity of Growth Rates And

Home , Synchronicity (Grimm)

bioRxiv preprint doi: https://doi.org/10.1101/2020.08.13.249557; this version posted August 14, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-ND 4.0 International license.

1 Temporal instability of lake charr phenotypes: synchronicity of growth rates and

2 morphology linked to environmental variables?

4 Chavarie, L.1,2, Steve Voelker3, M.J. Hansen4, C.R., Bronte5, A.M., Muir6, M.S. Zimmerman7,

5 C.C. Krueger2

6 1University of British Columbia, Beaty Biodiversity Research Center, Vancouver, Canada

7 2Center for Systems Integration and Sustainability, Department of Fisheries and Wildlife,

8 Michigan State University, East Lansing, Michigan, USA

9 3 Utah State University, Department of Plants, Soils, and Climate, Logan, Utah, USA

10 4U.S. Geological Survey (retired), Hammond Bay Biological Station, Michigan, USA

11 5U.S. Fish and Wildlife Service, Green Bay Fish and Wildlife Conservation Office, New

12 Franken, Wisconcin, USA

13 6Great Lakes Fishery Commission, Ann Arbor, Michigan, USA

14 7Coast Salmon Partnership, Aberdeen, Washington, USA

16 *Corresponding author:

17 Email:[email protected],

18 Keywords: Allometry, developmental stability, morphological modulation, plasticity,

19 temporal changes, otolith, dendrochronology, Salvelinus namaycush

20 Abstract:

21 Pathways through which phenotypic variation arises among individuals arise can be

22 complex. One assumption often made in relation to intraspecific diversity is that the stability or

23 predictability of the environment will interact with expression of the underlying phenotypic

24 variation. To address biological complexity below the species level, we investigated variability

25 across years in morphology and annual growth increments between and within two sympatric lake

26 charr ecotypes in Rush Lake, USA. We found a rapid phenotypic shift in body and head shape

27 within a decade. The magnitude and direction of the observed phenotypic change was consistent

28 in both ecotypes, which suggests similar pathways caused the temporal variation over time. Over

29 the same time period, annual growth increments declined for both lake charr ecotypes and

30 corresponded with a consistent phenotypic shift of each ecotype. Despite ecotype-specific annual

31 growth changes in response to winter conditions, the observed annual growth shift for both

32 ecotypes was linked, to some degree, with variation in the environment. Particularly, a declining

33 trend in regional cloud cover was associated with an increase of early stage (age 1-3) annual growth

34 for lake charr of Rush Lake. Underlying mechanisms causing reduced growth rates and constrained

35 morphological modulation are not fully understood. An improved knowledge of the biology hidden

36 within the expression of phenotypic variation promises to clarify our understanding of temporal

37 morphological diversity and instability.

41 2

42 Introduction:

43 A rapidly changing climate can have wide-ranging effects on organisms across ecosystems,

44 which fosters a need to understand how ecosystems will respond to this variation in terms of

45 structure and function (Montoya José & Raffaelli 2010; Pacifici et al. 2015). Contemporary

46 climate change, which includes rapid increases in global temperatures, represents one of the most

47 serious and current challenges to ecosystems, not only by threatening ecosystems directly (Norberg

48 et al. 2012), but also by contributing to cumulative and additive effects with other perturbations

49 (e.g., industrial development, pollution, overhavest, non-native species; CAFF 2013; Poesch et al.

50 2016). Ecosystems are mosaics of different habitats, climate change combined with abiotic and

51 biotic variation across these habitats may lead to major eco-evolutionary responses (Grimm et al.

52 2013; Ware et al. 2019). Rapid biological responses to variation associated with climate change

53 have already been detected at all levels, from individuals to species, communities, and ecosystems

54 (Heino et al. 2009).

55 The importance of phenotypic variability has been emphasised in evolutionary and

56 ecological population dynamics (Kinnison & Hairston Jr 2007; Schoener 2011), because variation

57 fuels evolutionary change (Stearns 1989). Pathways through which phenotypic variation arises

58 among individuals can be complex. Intraspecific trait variation as expressed within phenotypes,

59 such as morphology or growth, can affect population dynamics through reproductive and mortality

60 pathways (Bolnick et al. 2011). Furthermore, the magnitude of plasticity in the variation of trait

61 expression differs among populations and ecotypes within a population (Skúlason et al. 2019). A

62 conceptual framework to predict evolutionary and ecological consequences of climate change is

63 currently limited by the scarcity of empirical data demonstrating phenotypic changes over time

64 among individuals within ecosystems. The causes, patterns, and consequences of ecological and 3

65 evolutionary responses to differences induced by environmental variability need to be quantified

66 across species, space, and time.

67 The effect of growth rate heterogeneity on morphology modulation (e.g., heterochonry,

68 allometry; Klingenberg 2014) is not a mechanism fully understood, despite being observed to

69 constrain or enhance morphological differences in several fish species (Olsson et al. 2006; Heino

70 2014; Jacobson et al. 2015). Variation in developmental rate associated with juvenile growth rates

71 has been demonstrated to have an effect on the origin of some ecotypes (Alexander & Adams

72 2004; Helland et al. 2009; McPhee et al. 2015). Yet how variation in early development and

73 juvenile growth rate influence later morphology remains ambiguous, with almost no attention

74 focused on among-individual variation within an ecotype. Models have often assumed simple

75 population (and ecotype) dynamics, with populations (and ecotypes) commonly showing a stable

76 ecological evolutionary equilibrium (Svanbäck et al. 2009; Skúlason et al. 2019).

77 Persistence of phenotypes within a system is generally associated with a relatively stable

78 ecological environment that favors canalized phenotypes (e.g., reduction of phenotypic variance

79 at either the within-genotype-among-individual or within-individual levels) over plastic

80 phenotypes (Schluter 2000; Nosil 2012; Westneat et al. 2015; Skúlason et al. 2019). The

81 propensity for growth has both environmental and heritable components with variability in

82 phenotypic expression (van der Have & de Jong 1996; Kingsolver et al. 2004), even for canalized

83 phenotypes. Given rapid environmental change occurring within aquatic ecosystems, phenotypic

84 changes among individuals must not be taken as negligible, because phenotypic variation is driven

85 by switches along developmental pathways (West-Eberhard 2003), in cases where developmental

86 system flexibility can adjust immediately to environmentally variable conditions (Japyassú &

87 Malange 2014). Thus, temporal variation in phenotypes within an ecotype may not be ecologically

88 trivial in a rapidly changing world.

89 Records of organismal growth patterns provide long-term data sets that reflect

90 environmental variation (Pereira 1995). The diversity of species and ecosystems for which growth

91 chronologies exist (e.g., trees, bivalves, corals, fishes; Black 2009) provide robust datasets for

92 assessment of temporal and spatial environmental variation and its ecological consequences across

93 ecotypes and species, trophic levels, and communities that link ecosystems and biomes. The lake

94 charr (Salvelinus namaycush) is a suitable fish species to link growth chronologies with

95 environmental variation because their longevity that can exceed 60 years (Smith et al. 2008;

96 Chavarie et al. 2016 ). Lake charr are also known to display intraspecific variation, mainly

97 diversifying along a depth gradient, with shallow- versus deep-water ecotypes exploiting different

98 prey resources within a lacustrine system (Chavarie et al. accepted). By selecting a case of co-

99 existing shallow- and deep-water ecotypes of lake charr in Rush Lake (Fig. A1; Chavarie et al.

100 2016), we examined the response of lake charr to environmental variation below the species level.

101 The effect of environmental variation, such as for temperature, on growth rates of aquatic

102 organisms varies, especially across depths, in part because deep-water habitats are usually more

103 environmentally stable (Thresher et al. 2007; Murdoch & Power 2013; Jeppesen et al. 2014). By

104 exploiting different prey items (Chavarie et al. 2016), invertebrates (deep-water ecotype) vs. fish

105 (shallow-water ecotype), lake charr ecoytpes in Rush Lake are likely to integrate diverse signals

106 of climate variation across multiple trophic levels (Black et al. 2013a; Black et al. 2013b). Finally,

107 Rush Lake is located near the southern edge of the lake charr’s species range in North America

108 where variation in growth rate would likely be more sensitive to climate variation than for more

109 populations closer to the center of the range and thereby may yield useful predictive responses to

110 climate change.

111 In this study, we tested whether phenotypic expression of two Rush Lake lake charr

112 ecotypes remained stable or changed over time. We predicted that if a rapid change in phenotypic

113 expression occurred and this change was associated with environmental variation, the shallow-

114 water ecotype would display a greater magnitude of variation than the deep-water ecotype due to

115 the shallow-water habitat being more responsive and sensitive to contemporary climate change.

116 We measured phenotypic expression in terms of morphology and growth chronology. The

117 objectives of our study were to: 1) quantify the morphological variation between and within lake

118 charr ecotypes over a ten-year period, 2) determine if patterns of growth chronologies expressed

119 by ecotypes were temporally synchronous with each other and associated with the phenotypic

120 variation displayed within each ecotype; and 3) examine if annual growth rates of lake charr

121 ecotypes were related to environmental variation by using tree-ring cross-dating techniques.

122 Material and Methods:

123 Rush Lake is a small lake (1.31 km2) that contains deep-water (>80m) and is less than 2

124 km from Lake Superior (Chavarie et al. 2016). Rush Lake provided the first documented example

125 of sympatric lake charr ecotypes in a small lake, diverging along a depth gradient. Two co-existing

126 ecotypes of lake charr were found; a large streamlined-bodied shallow-water lake charr (lean

127 ecotype) and a small plump-bodied deep-water ecotype (huronicus ecotype) analogous to the

128 humper lake charr ecotype from Lake Superior (Hubbs 1929; Chavarie et al. 2016). The lean

129 ecotype has a long head, long maxillae, and a posterior eye position, which are all characteristics

130 for piscivorous feeding (Proulx & Magnan 2004; Keeley et al. 2007; Janhunen et al. 2009). The

131 huronicus ecotype, with smaller gape and higher eye position than the lean ecotype, appears

132 adapted for low-light vison and as a vertical migrating predator feeding on the invertebrate

133 opossum shrimp (Mysis spp.) as its main prey (Hrabik et al. 2006; Muir et al. 2014).

134 Assignment of lake charr ecotypes

135 Lake charr ecotypes were caught with bottom-set gillnets set in June 2007 and September

136 2018. Gillnets were deployed at depths from 10 to 83 m. Sets were made using 183 m long by 1.8

137 m high multifilament nylon gillnets consisting of stretch- mesh sizes from 50.8 to 114.3 mm, in

138 12.7 mm increments. Date, time, GPS location, and minimum and maximum water depth were

139 recorded for each net set. All fish caught were photographed in lateral view (Muir et al. 2012) and

140 sagittal otoliths were removed for analysis of age and growth.

141 Lake charr from each year were morphologically assessed and assigned to ecotype

142 according to methods described by Muir et al. (2014). Lake charr caught in 2007 were previously

143 assigned to ecotype (Chavarie et al. 2016) and assignments for lake charr caught in 2018 used the

144 same methodology. Twenty sliding semi-landmarks and six homologous landmarks were digitized

145 from each image to characterize head shape and 16 homologous and four sliding semi-landmarks

146 were digitized from whole-body images to characterize body shape. Landmarks and semi-

147 landmarks were digitized as x and y coordinates using TPSDig2 software

148 (http://life.bio.sunysb.edu/ecotype). Digitized landmarks and semi-landmarks were processed in a

149 series of Integrated Morphometrics Programs (IMP) version 8

150 (http://www2.canisius.edu/;sheets/ecotypesoft), using partial warp scores, which are thin-plate

151 spline coefficients. Morphological methods and programs are described by Zelditch et al. (2012)

152 and specific procedures were described by Chavarie et al. (2013). All morphological

153 measurements were size-free by using centroid sizes (Zelditch et al. 2012). Ecotypes were assigned

154 to each individual using a combination of Bayesian cluster analyses using head and body shape

155 information (MCLUST; Fraley & Raftery 2009) and a visual identification by experienced lake

156 charr biologists (M.J. Hansen & C. C. Krueger). Disagreements between model and visual

157 assignments were settled using decision rules described by Muir et al. (2014).

158 Temporal morphological variation between ecotypes and years

159 All data from 2007 and 2018 were combined to align all samples in the same shape space

160 and partial warps for temporal morphological analyses between ecotypes and years. Principal

161 component analysis (PCA) of body- and head-shape data were used to visualize morphological

162 variation between and within lake charr ecotypes and years using PCAGen8 (IMP

163 software). Canonical variate analyses (CVA) and validation procedures on body and head shape

164 data were used to assess temporal differences within and between ecotypes (ecotype*year) using

165 CVAGen (IMP software). Jacknife validation procedures included a test of assignment, with 1000

166 jackknife sets using 20% of our data as unknowns (Zelditch et al. 2012). Single-factor permutation

167 multivariate analysis of variance (MANOVA) with 10000 permutations of CVAGen was used to

168 test whether body and head shape differed between and within (i.e, years) ecotypes. If MANOVA

169 indicated differences among groups (α < 0.05), procrustes distance means were calculated for

170 pairwise comparisons using TWOGROUP from the IMP software as post-hoc tests (García-

171 Rodríguez et al. 2011). Procrustes distance for each pairwise comparison described the magnitude

172 of difference between ecotypes and years. A bootstrapped Goodall’s F statistic (N = 4900

173 bootstraps; full Procrustes based) was used to determine if pairwise comparison differed.

174 Allometric trajectories in body and head shape were compared between ecotypes and years by

175 regressing PC1 scores (size-free data) against centroid size (i.e., measure of size) (e.g., variation 8

176 in developmental pathways can result in allometric trajectory patterns that can be parallel,

177 divergent, convergent, or common; Simonsen et al. 2017); an allometric relationship occurred if

178 the slope differed from 0.

179 Relative body condition was compared between ecotypes and years in a 2-way analysis of

180 variance (ANOVA), with main effects for ecotype and year and the interaction between ecotypes

181 and years (Zar 1999). To correct for size-related trends in body condition, relative body condition

182 was defined as residuals from the power relationship between log10 (W) and log10 (TL). If the

183 ecotype*year interaction was significant, years were compared within ecotypes and ecotypes were

184 compared within years in 1-way ANOVAs. To visualize the results, least-squares means (+ SE)

185 from the ANOVA were back-transformed from logarithms into original units of measure.

186 Age assignment and growth increments from otoliths

187 Sagittal otoliths were used to estimate lake charr age and growth increments for fish

188 sampled in 2007 and 2018. Otolith thin sections have been validated for age estimation of lake

189 charr to an age of at least 50 years (Campana et al. 2008). Lake charr otoliths were prepared as

190 described by Campana et al. (2008). One otolith from each fish was embedded in epoxy. A Buehler

191 Isomet 1000 Precision Saw was used to remove a thin transverse section (400 µm) containing the

192 nucleus perpendicular to the sulcus. Sections were mounted on glass slides and polished. Digital

193 images of otolith sections were captured for age and growth assessment. Criteria for annulus

194 demarcation followed those of Casselman and Gunn (1992). Age estimates were used to inform

195 demarcation of growth increments, measured from the nucleus to the maximum ventral radius of

196 the otolith, and radial measurements at each annulus were used to back-calculate length-at-age

197 using the biological intercept back-calculation model (Campana 1990). The biological intercept

198 (sagittal otolith radius = 0.137 mm; age-0 lake trout length = 21.7 mm; Hansen et al. 2012) was

199 based on equations describing relationships between length, age in days, and sagittal otolith width

200 of age-0 lake charr (Bronte et al. 1995).

201 Back-calculated length at age from otoliths: growth patterns displayed by ecotypes through time

202 Otolith growth measurements can be used for several different purposes to gain ecological

203 insight, but often need different analytical techniques to answer different questions. Towards this

204 end, we used three analytical techniques and have provided a summary of the advantages and

205 disadvantages of each (Table A1). To determine if patterns of growth chronologies expressed by

206 ecotypes were temporally synchronous with each other, growth in length with age was modeled

207 using a parameterization of the Von Bertalanffy length-age model (Gallucci & Quinn 1979; Quinn

208 & Deriso 1999):

−(휔⁄ 퐿∞)(푡−푡0) 209 퐿푡 = 퐿∞(1 − 푒 ) + 휀

210 The length-age model describes back-calculated length Lt (mm) at age t (years) as a function of

211 age at length = 0 (t0 = years; incubation time of embryos from fertilization to hatching), early

212 annual growth rate (ω = L∞×K = mm/year; Gallucci & Quinn 1979), theoretical maximum length

213 (L∞ = mm), and residual error (ε). Parameters ω and L∞ were estimated using nonlinear mixed-

214 effect models, package ‘nlme’ (R Core Team 2016), with a fixed population effect (the average

215 growth curve for the population from which individual fish were sampled), individual as random

216 effect (growth curves for individual fish sampled from the population), and sex (male or female),

217 ecotype (lean or huronicus), or sampled year (2007 or 2018) as fixed factors (to compare average

218 growth curves between sexes, ecotypes, and years; Vigliola & Meekan 2009). Mixed-effects

219 models are appropriate for modelling the within-group correlation of longitudinal, auto-correlated,

220 and unbalanced data, such as back-calculated growth histories (Pinheiro & Bates 2000). Five

221 models of varying complexity were compared using AIC statistics (Burnham and Anderson 2002):

222 (1) ecotypes and sample years both included, (2) ecotypes only included, (3) years only included,

223 (4) sexes only included, and (5) neither ecotypes nor years included.

224 Annual growth increments were modeled using a linear mixed-effects model (Weisberg et

225 al. 2010), wherein annual growth increments were modeled as a function of a fixed age effect (age

226 of the fish when the increment formed), a random year effect (year in which the increment formed),

227 a random fish effect (unique identifier for individual fish), and residual variation. The fixed age

228 effect accounts for the fact that growth increments decline with age approximating a negative

229 exponential curve. The random year effect reflects average increment width associated with each

230 year of growth, after accounting for age effects (i.e., growth increment declines with age). The

231 random year effect accounts for year-to-year environmental effects as random draws from a normal

232 distribution, with a different draw for each year. The random fish effect accounts for fish-to-fish

233 variation in growth as random draws from a normally distributed population with a different draw

234 for each fish. The last source of variation is unmodeled residual variation. Differences between

235 sexes (male or female), ecotypes (lean or huronicus) and sample years (2007 or 2018) were tested

236 by including sex, ecotype, and sample year as fixed effects.

237 Otolith increment cross-dating: growth of lake charr in relation to environmental variation

238 In recent decades, advancements in dendrochronological techniques have been

239 increasingly applied to sagittal otoliths, leading to novel insights on how broad-scale climate

240 variation can impact both freshwater and marine systems (Black et al. 2005; Matta et al. 2010;

241 Black et al. 2013b). The use of dendrochronological methods (i.e., cross-dating techniques)

242 ensures that specific growth annuli are assigned to an exact year (Black et al. 2005; Black et al. 11

243 2012; Black et al. 2013a). In turn, this process enhances connection to common environmental

244 signals across fish, which have syncronously limited growth in certain years. To assign otolith

245 growth increments to exact calendar years, transverse sections of sagittal otoliths were aligned by

246 calendar year and cross-dated visually using the list method (Yamaguchi 1991) to confirm years

247 when particularly large or small otolith increments would be expected. Thereafter, visual cross-

248 dating was statistically confirmed using COFECHA software (Holmes 1983). In using

249 COFECHA, otolith time-series with series intercorrelation values (i.e., Rbar) lower than 0.20 with

250 the initial master chronology were removed and placed in a separate group that included more than

251 one third of all fish sampled. In a previous study of lake charr otolith variation, Rbar values ranged

252 from 0.42 to 0.97 (Black et al. 2013a), supporting the assumption that otolith width series with

253 Rbar < 0.20 included anomalous inter-annual growth variation that did not match the initial master

254 chronology. Otolith increment time series with Rbar < 0.20 underwent a second round of cross-

255 dating with COFECHA, separately from those that matched better with the initial master

256 chronology. Because otolith increment data grouped together by ecotype without a priori

257 knowledge of ecotype, all subsequent chronologies and analyses were conducted separately based

258 on ecotype assignment from the morphological analyses described above.

259 Dendrochronological detrending methods generally attempt to remove growth variation

260 and emphasize inter-annual variation in growth controlled by climate (Fritts 1976). The first

261 approach to detrending used the ARSTAN program (Cook and Krusic 2014) to fit cubic splines of

262 various rigidity based on fish age. Thereafter, autoregressive modeling was used to enhance inter-

263 annual growth variability and the resulting indexed time-series averaged within calendar years

264 using a bi-weight robust mean. Thereafter, we undertook a second regional chronology

265 standardization (RCS) approach known to better enhance low frequency signals compared to the

266 cubic spline method (Table A1; Briffa & Melvin 2011). This detrending method divided each raw

267 growth increment value by that expected from the mean growth increment for each ecotype and

268 age combined (Fig. A2). These ratios were then multiplied by 100 and averaged within calendar

269 years to yield a percentage change in growth for each year. In subsequent analyses of

270 environmental influence on growth chronologies, data from each lake charr were truncated to

271 feature only growth during young ages (age 1-3); which were excluded in earlier age-effect

272 analyses but are a critical stage to phenotypic variation linked to environmental differences

273 (Angilletta et al. 2004; Georga & Koumoundouros 2010; Ramler et al. 2014). The approach

274 employed for these comparisons corrected for age directly rather than using ARSTAN detrending

275 (see Methods above and Table A1). We also limited the analysis to calendar years where each

276 combination of ecotype and collection period included otolith data from at least seven fish. This

277 constrained the calendar years investigated to 1986 to 2012 for huronicus and 1988 to 2010 for

278 leans.

279 Otolith increments, detrended with the ARSTAN program, were calculated as means

280 within a year for both ecotypes and were initially compared against monthly resolution climate

281 data for the corresponding year. Based on a priori knowledge of fish biology and lake-effect

282 climate phenomena, temperature, precipitation, and cloud cover were selected as environmental

283 variables (Chavarie et al. 2018; Voelker et al. 2019). Interpolated climate data (e.g., air

284 temperature and precipitation) were obtained from ClimateNA version 5.6 software (Wang et al.

285 2016). Cloud cover climate data from airports within 7 km of Lake Superior were obtained at daily

286 resolution from the NOAA Great Lakes Environmental Research Laboratory

287 (https://www.glerl.noaa.gov/) and summarized by month.

288 Pearson correlation values were calculated between each ecotype-specific growth

289 chronology and monthly climate variables for the corresponding and previous two year (i.e., to

290 detect lag effect). After initial inspection of correlations between annual growth increments and

291 monthly climate variables, the number of potential explanatory variables were consolidated into

292 seasonal means, whereas winter to spring was defined as December through the next April for a

293 corresponding-year, summer was defined as June to August, and fall was defined as September to

294 November. Autocorrelation was expected to be present in otolith increment time series and

295 resulting chronologies due to year-to-year lags in growth owing to fat storage and subsequent

296 metabolic withdrawals, skip spawning effects, and climate and climatic-effects on water

297 temperature. Robust assessments and modeling of autocorrelation on short time series is

298 statistically impossible. Thus, we quantified what proportional weighting of climate data among

299 the corresponding and two previous years produced the largest gains in Pearson correlations

300 between otolith growth increment and seasonal climate data from an individual year to provide a

301 window into how climate signals are incorporated into fish growth.

302 The influence of weighted seasonal variables on otolith growth increment was assessed

303 using forward selection multiple regression models, package “lm” (R Core Team 2016). Final

304 regression models were constructed for each ecotype separately to determine which individual

305 seasonal climate variables as well as combinations of two or more seasonal variables were

306 significant (α < 0.05) and resulted in higher Akaike Information Criterion values (Burnham &

307 Anderson 2004).

308 Results:

309 Temporal morphological variation between ecotypes and years

310 In total, 107 lake charr were sampled, including 39 huronicus and 20 leans in 2007 and 27

311 huronicus and 21 leans in 2018. For both ecotypes, lake charr caught in 2007 had deeper bodies

312 than lake charr caught in 2018 (Fig. 1). The first two principal components explained 48.9% of the

313 variation in lake charr body shape from Rush Lake (Fig. 1a). Body shape differed between years

314 within each lake charr ecotype (CVA, Axis 1 ƛ = 0.015, p < 0.01 and Axis 2 ƛ = 0.29, p < 0.01;

315 Fig. 1b). Jackknife classification of body shape had a 54.3% rate of correct year and ecotype

316 assignment (i.e., ecotypes and years as different factors). Body shape means differed between

317 ecotypes and years (Permutation MANOVA, F = 11.7, df = 3, p ≤0 .01), and the magnitude of

318 these differences was slightly larger between ecotypes than between years. Pairwise body shape

319 comparisons differed between ecotypes for both 2007 and 2018 (F-tests; p ≤ 0.05). For lean and

320 huronicus lake charr sampled in 2007, the Goodall’s F was 21.1 and distance between means was

321 0.030 ± 0.0028 (SE), whereas for the lean and huronicus sampled in 2018, the Goodall’s F was

322 18.2 and distance between means was 0.032 ± 0.0031 (SE). Pairwise body shape comparisons also

323 differed between years for both ecotypes (F-tests; p ≤ 0.05). For huronicus 2007 vs 2018, the

324 Goodall’s F was 10.8 and distance between means was 0.020 ± 0.0012 (SE), whereas for the lean

325 2007 vs 2018, the Goodall’s F was 10.4 and distance between means was 0.025 ± 0.0024 (SE).

326 Allometric trajectories in body shape did not differ between 2007 and 2018, except for leans

327 sampled in 2018 (R2 = 0.56, p ≤ 0.01).

328 For both ecotypes, lake charr from 2007 had longer and deeper heads than lake charr from

329 2018 (Fig. 2). The first two principal components explained 61.2% of the variation in lake charr

330 head shape (Fig. 2a). Head shape differed between years within each lake charr ecotype (CVA,

331 Axis 1 ƛ = 0.012, p < 0.01 and Axis 2 ƛ = 0.13, p < 0.01; Fig. 2b). Jackknife classification on head

332 shape had a 49.5% rate of correct year and ecotype assignment (i.e., ecotypes and years as factors).

333 Head shape differed between ecotypes and years (Permutation MANOVA, F = 21.75 df = 3, p ≤

334 0.01; Fig 3), and the magnitude of these differences was slightly larger between years than between

335 ecotypes. Pairwise comparisons of head shape differed between ecotypes for both 2007 and 2018

336 (F-tests; p ≤ 0.05). For lean and huronicus sampled in 2007, Goodall’s F was 21.7 and distance

337 between means was 0.040 ± 0.0035 (SE), whereas for the lean and huronicus sampled 2018,

338 Goodall’s F was 21.0 and distance between means was 0.048 ± 0.0044 (SE). Pairwise head shape

339 comparison also differed between years for both ecotypes (F-tests; p ≤ 0.05). For the huronicus,

340 Goodall’s F was 54.9 and distance between means was 0.063 ± 0.0036 (SE), whereas for the lean,

341 Goodall’s F was 38.6 and distance between means was 0.063 ± 0.0041 (SE). Allometric trajectories

342 in head shape did not differ except for lean lake charr in 2018 (R2 = 0.36, p < 0.01; Fig. 3).

343 Relative body condition differed between ecotypes and years (ecotype*year: F1, 134 = 15.0;

344 P < 0.01; Fig. 4.). Within years, relative body condition of the huronicus ecotype was higher than

345 the lean ecotype in 2007 (F1, 66 = 25.4; P < 0.01) but similar to the lean ecotype in 2018 (F1, 68 =

346 0.2; P = 0.7). Within ecotypes, relative body condition of the huronicus was higher in 2007 than

347 in 2018 (F1, 80 = 55.1; P < 0.01) but the lean ecotype did not differ between years (F1, 54 = 0.4; P =

348 0.5).

349 Otoliths back-calculated growth: growth patterns displayed by ecotypes through time

350 Length-at-age of lake charr was best described by separate models for ecotypes and sample

351 years (Table 1). Lean lake charr grew faster to a longer asymptotic length than huronicus lake charr

352 in 2007 and 2018 (Fig. 5). Lean and huronicus lake charr sampled in 2018 grew faster at early age

353 than those sampled in 2007, whereas both ecotypes sampled in 2007 grew to longer asymptotic

354 length than those sampled in 2018. The early growth rate of lean lake charr sampled in 2007 was

355 similar rate to huronicus lake charr sampled in 2018. In contrast, the asymptotic length of lean lake

356 charr sampled in 2018 was similar to the asymptotic length of huronicus lake charr sampled in

357 2007.

358 Average otolith growth increments (corrected for age) differed between lean and huronicus

359 ecotypes (F1, 1927 = 16.9; P < 0.01), but not between males and females (F1, 1927 = 0.4; P = 0.52;

360 Table 2) or between sampling years (F1, 1927 = 0.08; P = 0.77; Table 2). Average annual growth

361 increments of huronicus and lean lake charr fluctuated without a specific trend prior to calendar

362 year 2009 and then declined steadily between 2009 and 2018 (Fig. 6). For huronicus lake charr,

363 average annual growth increments varied without temporal trend from 1977 until 1988, increased

364 slowly and erratically between 1989 until 2009, and then declined steadily between 2009 and 2018.

365 Average annual growth increments of huronicus lake charr were smallest in 2015-2018. For lean

366 lake charr, average annual growth increments varied erratically from 1984 through 1990, declined

367 from 1991 through 1995, increased from 1996 through 2009, and then declined between 2009 and

368 2018. Average annual growth increments of lean lake charr were nearly as small in 2015-2018 as

369 in 1991-1995. Over the entire period, mean annual growth increments were 44% more variable for

370 lean than for huronicus ecotypes (i.e., growth varied more among years for leans than huronicus

371 overall; Table A2). From 2009 to 2018, mean increment width declined 20% faster for leans than

372 huronicus (i.e., growth of leans declined faster after 2009 than growth of huronicus). Prior to 2009,

373 mean increment width was 76% higher for leans than huronicus (i.e., leans grew faster before 2009

374 than huronicus).

375 Otolith increment cross-dating: growth of lake charr in relation to environmental variation

376 Inter-annual climate variation, particularly when including lagged effects, was correlated

377 with fish growth for both ecotypes, as demonstrated by Pearson correlations regularly exceeding 17

378 0.2 (Fig. 7). For both lake charr ecotypes and year-corresponding and lagged effects, annual otolith

379 growth increments were positively correlated with summer air temperatures (except for the lean

380 corresponding year, slightly negative) and fall precipitation (i.e., more growth with warmer

381 temperatures and more precipitation) and negatively correlated with summer and fall cloud cover

382 (i.e., more growth with less cloud cover). In comparison, the direction of the relationship with

383 winter to spring temperatures and precipitation as snow differed between the two ecotypes (Fig. 7

384 and A3). For each set of seasonal variables, inclusion of weighted climate data from previous

385 years, to accommodate lagging effects, tended to strengthen correlations. Overall, based on the

386 forward selection multiple regression models, the total amount of variation in otolith annual

387 growth increments explained by climatic variables was greater for the lean ecotype than huronicus

388 ecotype (R2 = 0.56 vs 0.35; Table A3). These regression models confirmed that growth increments

389 of huronicus ecotypes were most strongly associated with winter to spring temperatures and

390 precipitation as snow and secondarily with summer temperature, whereas the lean ecotype was

391 most strongly associated with winter to spring temperatures and previous fall precipitation (Table

392 A3). Among-ecotype differences in the relationship of annual growth increment with winter and

393 spring temperatures and with precipitation as snow were then examined more closely, where

394 regression analyses confirmed these differences (Fig. 7 and A3). Specifically, differences in annual

395 growth increments between ecotypes for any given year indicated that cold and snowy winters

396 tended to favor growth for the huronicus ecotype whereas warmer winters with less snow favored

397 growth for the lean ecotype (Fig. 7 and A3).

398 For the early life-stages of both ecotypes, defined here as ages 1-3, annual growth

399 increments were correlated with cloud cover only. A negative relationship between early life stage

400 growth and summer cloud cover (i.e., more growth with less cloud cover) occurred for both the

401 huronicus (R2 = 0.19, p = 0.01) and the lean (R2 = 0.43, p < 0.01; Fig. 8) ecotypes. Annual growth

402 of early life-stage lake charr appeared to be correlated with the temporal trend of summer cloud

403 cover. For example, cloud cover in July has decreased by up to 33% over the past three decades,

404 resulting in higher annual growth in early life-stage lake charr from 2007 to 2018 (Fig. 8).

405 Discussion

406 In this study, we demonstrated a rapid phenotypic shift that occurred for two lake charr

407 ecotypes in a small lake. Within the last decade, a major decline in annual growth increments has

408 occurred with a corresponding morphological shift in body and head shape for both lean and

409 huronicus ecotypes. Even though the lean (shallow-water) ecotype displayed a greater annual

410 growth variation over the years than the huronicus (deep-water) ecotype, both ecotypes displayed

411 similar magnitudes of morphological change in the most recent decade resulting in analogous

412 “sensitivity” in phenotypic change, meaning “responsiveness” independent of habitats. We

413 interpret these results to mean that the response threshold that determines individual sensitivity to

414 a particular cue (i.e., environmental variables) caused similar phenotypic changes by individuals

415 of both ecotypes (Moczek 2010; Baerwald et al. 2016), in spite of differences in sensitivity

416 between growth responses of the two ecotypes. A rapid phenotypic shift was somewhat surprising

417 because mobile predators, such as lake charr, they exploit a large range of prey, and thus, can

418 compensate rapidly to prey fluctuations and dampen oscillations or fluctuations of lower trophic

419 levels (McCann et al. 2005; McCann 2012) being caused by environmental variables.

420 One assumption often made in relation to intraspecific diversity, mostly tested

421 experimentally or modelled, is that a stable or predictable environment interacts with underlying

422 variation in expression of phenotypes (Wagner & Schwenk 2000; Svanbäck et al. 2009; Westneat

423 et al. 2015; Skúlason et al. 2019). In our study, the magnitude and direction of the observed

424 phenotypic shift in both annual growth and morphology over a single decade was consistent for

425 each ecotype and suggested similar pathways to which phenotypic variation was expressed. The

426 degree of phenotypic variation that occurs within an ecotype theoretically depends on the relative

427 strength and time of mechanisms that drive phenotypic change (Wood et al. 2020), the

428 environment is not necessary constant across cohorts (Johnson et al. 2014). In our case, the

429 observed phenotypic shift was relatively small, but nonetheless, detectable (i.e., cryptic eco-

430 evolutionary dynamics; Kinnison et al. 2015). Several questions arise from our results, but one of

431 interest is the organism’s capacity for phenotype acclimation to a changing environment (via

432 phenotypic plasticity and adaptation; Huey & Berrigan 1996; Gorsuch et al. 2010). Was this

433 phenotypic shift an isolated event or does this type of change occur frequently in this system?

434 Answers to the question of phenotypic acclimation and the frequency of its occurrence within and

435 among systems would require long data sets collected over multiples decades and would help to

436 fill an important knowledge gap about non-equilibrium population dynamics affecting

437 evolutionary dynamics.

438 Biological complexity is a key component of ecological variability and evolution (McShea

439 & Brandon 2010; McShea 2017). Complexity is a recurring idea in evolutionary biology that can

440 refer to a number of concepts at different levels (organism to ecosystem), but biological complexity

441 is defined herein as mechanisms producing (or limiting) variation in phenotypic expression (e.g.,

442 morphology; Heim et al. 2017; Duclos et al. 2019). In this study, a decline in annual growth

443 increment for lean and huronicus ecotypes occurred at the same time as a consistent phenotypic

444 shift in morphology and these changes occurred within the time frame of a decade. Growth rate in

445 fishes has been shown to drive intraspecific morphological differentiation (Tonn et al. 1994;

446 Olsson et al. 2007; Chivers et al. 2008). Growth rate seems to play a key role in regulating

447 morphological expression, but underlying mechanisms remain uncertain (Olsson et al. 2006;

448 Svanbäck et al. 2017; Franklin et al. 2018). A possible explanation is that at higher growth rates,

449 energy is allocated to somatic growth and morphology modulation in addition to metabolic

450 maintenance, but that at lower growth rates, energy is used almost exclusively for metabolic

451 maintenance and/or reproduction (Olsson et al. 2006; Svanbäck et al. 2017). Our results concur

452 with this latter mechanism because lake charr sampled in 2007 had higher annual growth rates and

453 less morphological overlap between ecotypes than lake charr sampled in 2018. An organism’s

454 response to change (abiotic and biotic) can include variation in the mean phenotype itself, but it

455 can also include differences in the phenotypic variance (O'Dea et al. 2019). To our knowledge,

456 few other field studies have demonstrated a temporal change in morphology and growth rate,

457 within an ecotype much less, temporal changes that were consistent between ecotypes (but see

458 Svanbäck et al. 2009 for an example of consistent phenotypic variation, in magnitude and direction

459 between ecotypes (PC1: 2003 to 2004)).

460 Many examples of taxa exist where population abundance fluctuates (e.g., density) over

461 time as a result of variation in resource levels, often influenced by environmental changes (Grant

462 & Grant 1992; Persson & De Roos 2003; Svanbäck et al. 2009). In this study, some aspects of

463 body and head shape shifts may have been mediated by changes in growth rates and body

464 condition, which likely, were induced by variation in food availability and abiotic conditions. Body

465 condition can affect body shape in fish, and often reflects bulkiness of individuals, and

466 consequently body depth (Olsson et al. 2006; Borcherding & Magnhagen 2008; Jacobson et al.

467 2015; Svanbäck et al. 2017). Although both ecotypes were subjected to annual growth rate

468 declines, the huronicus ecotype was more affected by body condition changes than the lean

469 ecotype, which might suggest physiological differences between ecotypes in the dynamics of

470 energy processing (e.g., metabolism and reproduction) and storage (Goetz et al. 2013). For

471 example, in 2018, 30.8% of the huronicus females were in a resting reproductive stage compared

472 to only 8.3% of lean females. In contrast to females, all males in 2018 of both ecotypes were

473 reproductively mature (Unpublished data). Differences in energetic requirements associated to

474 reproductive output appear related to the apparent skipped spawning patterns observed by ecotype

475 and sex.

476 Spatial and temporal fluctuations of trophic resources in Rush Lake could have influenced

477 how individuals used these resources and the resulting annual growth rate patterns (influenced by

478 density-dependent fluctuations and intraspecific competition; Svanbäck et al. 2009; Jacobson et

479 al. 2015). Such spatial and temporal periodicity can occur at different scales (e.g., for temporal

480 periodicity: seasonal, inter-annual, decadal) and be driven by fluctuations in abiotic (e.g.,

481 temperature, precipitation, light, nutrients) and biotic processes (e.g., growth, reproduction, trophic

482 interactions), thereby shaping species behavior and “rewiring” food webs (McMeans et al. 2015;

483 Bartley et al. 2019). Developmental stability suppresses variation in phenotypic expression of a

484 single genotype within a single environment, thus it can modulate the magnitude and structure of

485 phenotypic variation among individuals (Willmore et al. 2006; Lazić et al. 2015). Organisms can

486 be affected by a single perturbation of the timing or rate in development, which has been perceived

487 to be a means to produce trait novelty (e.g., heterochrony; Parsons et al. 2011; Lazić et al. 2015;

488 Westneat et al. 2015). Allometry, the shape variation associated with size variation (Zelditch et al.

489 2012), is also seen as a canalized process and an interacting agent that can limit morphological

490 variation (Klingenberg 2010). At the very least, some of our results represent a case of plastic

491 allometry for the lean ecotype in 2018. Considering that the lean body condition did not markedly

492 differ between sampling years and were within the species range values (Hansen et al, In press),

493 we are confident that morphological differences were not due to starvation of individuals.

494 Altogether, the expression of phenotypic variation observed in this study was probably influenced

495 by more than one varying niche dimension (e.g., temperature, resources level, competition) with

496 concomitant effects (e.g., developmental stability and allometry).

497 Lake charr seem to show a fluctuating nature of phenotypic expression (adaptive or not),

498 closer to Eurasian perch (Perca fluviatilis) than other salmonids such as Arctic charr (Salvelinus

499 alpinus). Both lake charr and perch appear to show less discreteness between ecotypes than Arctic

500 charr but yet, a higher level of phenotypic variance (Chavarie et al., accepted; Jonsson & Jonsson

501 2001; Svanbäck et al. 2009). When compared to lake charr ecotypes from other lakes, the genetic

502 diversity and divergence of lake charr in Rush Lake is low for both ecotypes (Chavarie et al. 2016).

503 This low genetic diversity and divergence favor the hypothesis that phenotypic variation is the

504 result of phenotypic plasticity rather than genetic adaptations (although rapid genetic change

505 cannot be excluded). Epigenetically mediated biological complexity is known to be an important

506 process to tailor phenotypic reaction norms (e.g., linear and nonlinear) to selective environmental

507 pressures (Crozier & Hutchings 2014; Ramler et al. 2014; Duclos et al. 2019). A rapid change in

508 the environment can also induce changes in the phenotypic variance within an ecotype by exposing

509 previously hidden cryptic genetic variation or by inducing new epigenetic changes (O’Dea et al.

510 2016). It has been hypothesized that heritable epigenetic mechanisms can lead to phenotypic

511 variation generated by bet hedging strategies, whereas phenotypic variability buffers varying

512 environments (O’Dea et al. 2016). Environmental factors can induce a component of variation that

513 introduces a fined-grained variation around any coarse-scale temporal trends, resulting in year-to-

514 year variation in phenotypes but not in genotypes – because genetic changes are not expected to

515 be so fine-grained (Merilä & Hendry 2014). Thus, phenotypic variability, both in means and

516 variance, can have evolutionary scope for a population in the face of changing selection regimes

517 by affecting population dynamics and probabilities of extinction (Chevin et al. 2010; Reed et al.

518 2010; Johnson et al. 2014).

519 Synchronous change in growth rates of fishes have been observed at global scales, with

520 major declines in growth linked to climate change (Thresher et al. 2007; Sheridan & Bickford

521 2011; Baudron et al. 2014). However, little is known of how growth trajectories and their

522 associated phenotypic reaction norms that integrate environmental factors may differ within (e.g.,

523 cohorts) and among ecotypes in freshwater ecosystems (Heino et al. 2002; Johnson et al. 2014).

524 In Rush Lake, similarities and differences of the annual growth rates of lake charr ecotypes were

525 correlated to environmental variation. In our study, cloud cover was the main environmental

526 variable that had steadily decreased over the same time period that lake charr growth declined

527 (except for early stage) and morphology shifted in Rush Lake (Fig. 8). Environmental

528 heterogeneity is thought to have stronger effects on morphology at early life stages (Johnson et al.

529 2014; Morris 2014; Ramler et al. 2014), suggesting that the effects of the cloud cover on lake charr

530 might have been more significant at age 1-3 years than during later stages of life. The effect of

531 cloud cover on growth at age 1-3 was stronger in the lean than the huronicus ecotype, which might

532 explain why allometry was detected only for the lean ecotype in 2018. The relationship between

533 lake charr annual otolith growth and cloud cover could be related to how solar irradiance can co-

534 vary with other climate variables that may affect fish growth (Reist et al. 2006; Poesch et al. 2016).

535 For example, in Lake Superior, cloud cover and light penetration were shown to affect the degree

536 of heterotrophy (primary production < community respiration; Brothers & Sibley 2018). Rush

537 Lake is located only two km away from Lake Superior at its nearest point and this proximity

538 produces strong lake-effect climate phenomena during summer (Voelker et al. 2019). Primary

539 production can have cascading effects to top-predators, thereby influencing the level and quality

540 of available food ration (Hunter & Price 1992). Summer temperature and fall precipitation were

541 also positively associated with growth rates. When higher temperatures are accompanied by

542 suitable net addition of food ration (e.g., from direct and indirect effects of temperature and

543 precipitation factors), increases in growth could be expected up to the point of the optimum

544 temperature of the species (Elliott & Hurley 2003; Elliott & Elliott 2010).

545 Response to winter climate appeared to vary between ecotypes, with environmental

546 variables (e.g., temperature and precipitation as snow), correlated in different directions with

547 growth rates (directly or indirectly). Why the deep-water ecotype (huronicus) would show higher

548 annual growth increments in years with warm winter temperatures and low snow cover (via direct

549 or lagging effects) and the shallow-water ecotype (lean) would have higher annual growth

550 increments in years with cold winter temperature and high snow level is unknown. These results

551 could be explained in part by reduced habitat partitioning between ecotypes during winter, along

552 with an increase of intraspecific competition, affecting energy storage (Amundsen et al. 2008).

553 Another explanation, which is not exclusive of the previous one, could be that each ecotype’s prey

554 types, density and quality (e.g., time response to environmental variable) is modified differently

555 by lagging effects from winter environmental conditions (e.g., ice cover duration, ice and snow

556 thickness). The lean ecotype is known to feed on forage fish whereas the huronicus ecotype mainly

557 feeds on the invertebrate Mysis (Chavarie et al. 2016b). Invertebrates can differ in response time

558 and magnitude to environmental changes compared to forage fishes (Heino et al. 2009, Wrona et

559 al. 2016, Wrona et al. 2006). This hypothesis of a differential response of prey items to

560 environmental conditions seems to be supported by correlations strengthened with lake charr

561 growth when weighted climate data from previous years were included. This result might be

562 expected for an organism in which growth rates may draw on a mixture of recently acquired and

563 stored resources or where climate variables in one year may affect the abundance and composition

564 of prey in subsequent years. For the most part, winter environmental conditions can play an

565 essential role in ecological and evolutionary processes that define life-history characteristics (e.g.,

566 somatic growth, size and age-at-maturity, reproduction investment, and longevity) of lacustrine

567 species (Shuter et al. 2012).

568 Conclusion

569 We demonstrated such rapid morphological change in Rush Lake lake charr, a change that

570 was expressed similarly by both lean and huronicus ecotypes over a decade of time. These shifts

571 in morphology shared by both ecotypes were synchronous with declining in cloud cover over the

572 same time frame. The mechanisms that connect annual growth increments with morphological

573 modulation are not fully understood (Olsson et al. 2006; Olsson et al. 2007; Svanbäck et al. 2009);

574 however, the biology underlying phenotypic variation can have major implications for populations

575 responding to climate change. Multidimensional phenotypic variability and its influence on

576 population dynamics patterns is a relatively poorly studied phenomena (Westneat et al. 2015), but

577 individual and population resistance and resilience to climatic changes may depend on this

578 variability (Johnson et al. 2014). Mean fitness can improve more rapidly through plasticity than

579 through genetic adaptation (at least on the short-term perspective; Hendry et al. 2011).

580 The similarity in phenotypic response expressed by both ecotypes raises the question

581 whether organisms in small lakes are more vulnerable to climate change than those in large lakes.

582 Small lakes generally sustain a higher degree of habitat coupling (e.g., littoral-pelagic; Schindler

583 & Scheuerell 2002; Dolson et al. 2009), which is critical to food-web dynamics. Thus, the degree

584 of habitat coupling found in each freshwater ecosystem might translate to its degree of

585 vulnerability to climate change. Field studies, such as ours, that focus on temporal phenotypic

586 instability within an aquatic ecosystem promise to clarify our understanding of how the interplay

587 among phenotypes, trophic dynamics, and environmental context influences both ecosystem and

588 evolutionary processes (Ware et al. 2019).

589

590 Acknowledgements

591 We thank the Huron Mountain Club for access to their lands, housing, and lakes and for sharing

592 their knowledge of the lake charr of Rush Lake. Special thanks to Kerry Woods, Director of

593 Research, Huron Mountain Club Wildlife Foundation for coordinating and supporting the project.

594 Financial support was provided by the Great Lakes Fishery Commission. Any use of trade, firm

595 or product names is for descriptive purposes only and does not imply endorsement by the U.S.

596 Government. The findings and conclusions in this article are those of the authors and do not

597 necessarily represent the views of the US Fish and Wildlife Service.

598 Literature:

599 Alexander, G.D. & Adams, C.E. (2004) Exposure to a common environment erodes inherited between- 600 population trophic morphology differences in Arctic charr. Journal of Fish Biology, 64, 253-257. 601 Amundsen, P.-A., Knudsen, R. & Klemetsen, A. (2008) Seasonal and ontogenetic variations in resource 602 use by two sympatric Arctic charr morphs. Environmental Biology of Fishes, 83, 45-55. 603 Angilletta, M.J., Jr., Steury, T.D. & Sears, M.W. (2004) Temperature, Growth Rate, and Body Size in 604 Ectotherms: Fitting Pieces of a Life-History Puzzle1. Integrative and Comparative Biology, 44, 605 498-509. 606 Baerwald, M.R., Meek, M.H., Stephens, M.R., Nagarajan, R.P., Goodbla, A.M., Tomalty, K.M.H., 607 Thorgaard, G.H., May, B. & Nichols, K.M. (2016) Migration-related phenotypic divergence is 608 associated with epigenetic modifications in rainbow trout. Molecular Ecology, 25, 1785-1800. 609 Bartley, T.J., McCann, K.S., Bieg, C., Cazelles, K., Granados, M., Guzzo, M.M., MacDougall, A.S., 610 Tunney, T.D. & McMeans, B.C. (2019) Food web rewiring in a changing world. Nature Ecology 611 & Evolution, 3, 345-354. 27

612 Baudron, A.R., Needle, C.L., Rijnsdorp, A.D. & Tara Marshall, C. (2014) Warming temperatures and 613 smaller body sizes: synchronous changes in growth of North Sea fishes. Global Change Biology, 614 20, 1023-1031. 615 Black, B.A. (2009) Climate-driven synchrony across tree, bivalve, and rockfish growth-increment 616 chronologies of the northeast Pacific. Marine Ecology Progress Series, 378, 37-46. 617 Black, B.A., Biela, V.R., Zimmerman, C.E. & Brown, R.J. (2012) Lake trout otolith chronologies as 618 multidecadal indicators of high-latitude freshwater ecosystems. Polar Biology, 36, 147-153. 619 Black, B.A., Boehlert, G.W. & Yoklavich, M.M. (2005) Using tree-ring crossdating techniques to 620 validate annual growth increments in long-lived fishes. Canadian Journal of Fisheries and 621 Aquatic Sciences, 62, 2277-2284. 622 Black, B.A., Matta, M.E., Helser, T.E. & Wilderbuer, T.K. (2013a) Otolith biochronologies as 623 multidecadal indicators of body size anomalies in yellowfin sole (Limanda aspera). Fisheries 624 Oceanography, 22, 523-532. 625 Black, B.A., von Biela, V.R., Zimmerman, C.E. & Brown, R.J. (2013b) Lake trout otolith chronologies as 626 multidecadal indicators of high-latitude freshwater ecosystems. Polar Biology, 36, 147-153. 627 Bolnick, D.I., Amarasekare, P., Araújo, M.S., Bürger, R., Levine, J.M., Novak, M., Rudolf, V.H.W., 628 Schreiber, S.J., Urban, M.C. & Vasseur, D.A. (2011) Why intraspecific trait variation matters in 629 community ecology. Trends in Ecology & Evolution, 26, 183-192. 630 Borcherding, J. & Magnhagen, C. (2008) Food abundance affects both morphology and behaviour of 631 juvenile perch. Ecology of Freshwater Fish, 17, 207-218. 632 Briffa, K.R. & Melvin, T.M. (2011) A Closer Look at Regional Curve Standardization of Tree-Ring 633 Records: Justification of the Need, a Warning of Some Pitfalls, and Suggested Improvements in 634 Its Application. Dendroclimatology: Progress and Prospects (eds M.K. Hughes, T.W. Swetnam 635 & H.F. Diaz), pp. 113-145. Springer Netherlands, Dordrecht. 636 Bronte, C.R., Selgeby, J.H., Saylor, J.H., Miller, G.S. & Foster, N.R. (1995) Hatching, dispersal, and 637 bathymetric distribution of age-0 wild lake trout at the Gull Island Shoal complex, Lake Superior. 638 Journal of Great Lakes Research, 21, 233-245. 639 Brothers, S. & Sibley, P. (2018) Light may have triggered a period of net heterotrophy in Lake Superior. 640 Limnology and Oceanography, 63, 1785-1798. 641 Burnham, K.P. & Anderson, D.R. (2004) Multimodel Inference: Understanding AIC and BIC in Model 642 Selection. Sociological Methods & Research, 33, 261-304. 643 CAFF (2013) Status and trends in Arctic biodiversity. Arctic Biodiversity Assessment. Conservation of 644 Arctic Flora and Fauna, Akureyri. 645 Campana, S.E. (1990) How reliable are growth back-calculations based on otoliths? Canadian Journal of 646 Fisheries and Aquatic Sciences, 117, 59-71. 647 Campana, S.E., Casselman, J.M. & Jones, C.M. (2008) Bomb radiocarbon chronologies in the Arctic, 648 with implications for the age validation of lake trout (Salvelinus namaycush) and other Arctic 649 species. Canadian Journal of Fisheries and Aquatic Science, 65, 733-743. 650 Casselman, J.M. & Gunn, J.M. (1992) Dynamics in year-class strength, growth, and calcified-structure 651 size of native lake trout (Salvelinus namaycush) exposed to moderate acidification and whole- 652 lake neutralization. Canadian Journal of Fisheries and Aquatic Sciences, 49, 102-113. 653 Chavarie, L., Howland, K., Venturelli, P., Kissinger, B.C., Tallman, R. & Tonn, W. (2016 ) Life-history 654 variation among four shallow-water morphotypes of lake trout from Great Bear Lake, Canada. 655 Journal of Great Lakes Research, DOI:10.1016/j.jglr.2015.1007.1006. 656 Chavarie, L., Howland, K.L. & Tonn, W.M. (2013) Sympatric polymorphism in lake trout: the 657 coexistence of multiple shallow-water morphotypes in Great Bear Lake. Transactions of the 658 American Fisheries Society, 142, 814-823. 659 Chavarie, L., Muir, A.M., Zimmerman, M.S., Baillie, S.M., Hansen, M.J., Nate, N.A., Yule, D.L., 660 Middel, T., Bentzen, P. & Krueger, C.C. (2016) Challenge to the model of lake charr evolution:

661 shallow- and deep-water morphs exist within a small postglacial lake. Biological Journal of the 662 Linnean Society, n/a-n/a. 663 Chavarie, L., Reist, J.D., Guzzo, M.M., Harwood, L. & Power, M. (2018) Influences of environmental 664 variation on anadromous Arctic charr from the Hornaday River, NWT. Hydrobiologia. 665 Chevin, L.-M., Lande, R. & Mace, G.M. (2010) Adaptation, plasticity, and extinction in a changing 666 environment: towards a predictive theory. PLoS Biology, 8, e1000357-e1000357. 667 Chivers, D.P., Zhao, X., Brown, G.E., Marchant, T.A. & Ferrari, M.C.O. (2008) Predator-induced 668 changes in morphology of a prey fish: the effects of food level and temporal frequency of 669 predation risk. Evolutionary Ecology, 22, 561-574. 670 Crozier, L.G. & Hutchings, J.A. (2014) Plastic and evolutionary responses to climate change in fish. 671 Evolutionary Applications, 7, 68-87. 672 Dolson, R., McCann, K., Rooney, N. & Ridgway, M. (2009) Lake morphometry predicts the degree of 673 habitat coupling by a mobile predator. Oikos, 118, 1230-1238. 674 Duclos, K.K., Hendrikse, J.L. & Jamniczky, H.A. (2019) Investigating the evolution and development of 675 biological complexity under the framework of epigenetics. Evolution & Development, 21, 276- 676 293. 677 Elliott, J.M. & Elliott, J.A. (2010) Temperature requirements of Atlantic salmon Salmo salar, brown trout 678 Salmo trutta and Arctic charr Salvelinus alpinus: predicting the effects of climate change. Journal 679 of Fish Biology, 77, 1793-1817. 680 Elliott, J.M. & Hurley, M.A. (2003) Variation in the temperature preference and growth rate of individual 681 fish reconciles differences between two growth models. Freshwater Biology, 48, 1793-1798. 682 Fraley, C. & Raftery, A.E. (2009) MCLUST Version 3 for R: normal mixuture modeling and model- 683 based clustering. pp. 56. Department of Statistics, University of Washington, Seattle, 684 Washington. 685 Franklin, O.D., Skúlason, S., Morrissey, M.B. & Ferguson, M.M. (2018) Natural selection for body shape 686 in resource polymorphic Icelandic Arctic charr. Journal of Evolutionary Biology, 31, 1498-1512. 687 Gallucci, V.F. & Quinn, T.J. (1979) Reparameterizing, fitting, and testing a simple growth model. 688 Transactions of the American Fisheries Society, 108, 14-25. 689 García-Rodríguez, F.J., García-Gasca, S.A., Cruz-Agüero, J.D.L. & Cota-Gómez, V.M. (2011) A study of 690 the population structure of the Pacific sardine Sardinops sagax (Jenyns, 1842) in Mexico based on 691 morphometric and genetic analyses. Fisheries Research, 107, 169-176. 692 Georga, I. & Koumoundouros, G. (2010) Thermally induced plasticity of body shape in adult zebrafish 693 Danio rerio (Hamilton, 1822). Journal of Morphology, 271, 1319-1327. 694 Goetz, F., Jasonowicz, A., Johnson, R., Biga, P., Fischer, G. & Sitar, S. (2013) Physiological differences 695 between lean and siscowet lake trout morphotypes: Are these metabolotypes? Canadian Journal 696 of Fisheries and Aquatic Sciences, 71, 427-435. 697 Gorsuch, P.A., Pandey, S. & Atkin, O.K. (2010) Temporal heterogeneity of cold acclimation phenotypes 698 in Arabidopsis leaves. Plant, Cell & Environment, 33, 244-258. 699 Grant, P.R. & Grant, B.R. (1992) Demography and the Genetically Effective sizes of Two Populations of 700 Darwin's Finches. Ecology, 73, 766-784. 701 Grimm, N.B., Chapin Iii, F.S., Bierwagen, B., Gonzalez, P., Groffman, P.M., Luo, Y., Melton, F., 702 Nadelhoffer, K., Pairis, A., Raymond, P.A., Schimel, J. & Williamson, C.E. (2013) The impacts 703 of climate change on ecosystem structure and function. Frontiers in Ecology and the 704 Environment, 11, 474-482. 705 Hansen, M.J., Nate, N.A., Krueger, C.C., Zimmerman, M.S., Kruckman, H.G. & Taylor, W.W. (2012) 706 Age, growth, survival, and maturity of lake trout morphotypes in Lake Mistassini, Quebec. 707 Transactions of the American Entomological Society (Philadelphia), 141, 1492-1503. 708 Heim, N.A., Payne, J.L., Finnegan, S., Knope, M.L., Kowalewski, M., Lyons, S.K., McShea, D.W., 709 Novack-Gottshall, P.M., Smith, F.A. & Wang, S.C. (2017) Hierarchical complexity and the size 710 limits of life. Proceedings of the Royal Society B: Biological Sciences, 284, 20171039. 29

711 Heino, J., Virkkala, R. & Toivonen, H. (2009) Climate change and freshwater biodiversity: detected 712 patterns, future trends and adaptations in northern regions. Biological Reviews, 84, 39-54. 713 Heino, M. (2014) Quantitative traits. Stock identification methods, pp. 59-76. Elsevier. 714 Heino, M., Dieckmann, U. & Godø, O.R. (2002) Measuring probabilistic reaction norms for age and size 715 at maturation. Evolution, 56, 669-678. 716 Helland, I.P., Vøllestad, L.A., Freyhof, J. & Mehner, T. (2009) Morphological differences between two 717 ecologically similar sympatric fishes. Journal of Fish Biology, 75, 2756-2767. 718 Hendry, A.P., Kinnison, M.T., Heino, M., Day, T., Smith, T.B., Fitt, G., Bergstrom, C.T., Oakeshott, J., 719 Jørgensen, P.S., Zalucki, M.P., Gilchrist, G., Southerton, S., Sih, A., Strauss, S., Denison, R.F. & 720 Carroll, S.P. (2011) Evolutionary principles and their practical application. Evolutionary 721 Applications, 4, 159-183. 722 Holmes, R.L. (1983) Computer-assisted quality control in tree-ring dating and measurement. 723 Hrabik, T.R., Jensen, O.P., Martell, S.J.D., Walters, C.J. & Kitchell, J.F. (2006) Diel vertical migration in 724 the Lake Superior pelagic community. I. Changes in vertical migration of coregonids in response 725 to varying perdation risk. Canadian Journal of Fisheries and Aquatic Sciences, 63, 2286-2295. 726 Hubbs, C.L. (1929) The fishes. The book of the Huron mountains, pp. 153-164. Huron Mountain Club, 727 Marquette, Michigan. 728 Huey, R. & Berrigan, D. (1996) Testing evolutionary hypotheses of acclimation. Animals and 729 temperature: Phenotypic and evolutionary adaptation, 59, 205-237. 730 Hunter, M.D. & Price, P.W. (1992) Playing Chutes and Ladders: Heterogeneity and the Relative Roles of 731 Bottom-Up and Top-Down Forces in Natural Communities. Ecology, 73, 724-732. 732 Jacobson, B., Grant, J.W.A. & Peres-Neto, P.R. (2015) The interaction between the spatial distribution of 733 resource patches and population density: consequences for intraspecific growth and morphology. 734 Journal of Animal Ecology, 84, 934-942. 735 Janhunen, M., Peuhkuri, N. & Piironen, J. (2009) Morphological variability among three geographically 736 distinct Arctic charr (Salvelinus alpinus L.) populations reared in a common hatchery 737 environment. Ecology of Freshwater Fish, 18, 106-116. 738 Japyassú, H.F. & Malange, J. (2014) Plasticity, stereotypy, intra-individual variability and personality: 739 Handle with care. Behavioural Processes, 109, 40-47. 740 Jeppesen, E., Meerhoff, M., Davidson, T.A., Trolle, D., SondergaarD, M., Lauridsen, T.L., Beklioglu, M., 741 Brucet Balmaña, S., Volta, P. & González-Bergonzoni, I. (2014) Climate change impacts on 742 lakes: an integrated ecological perspective based on a multi-faceted approach, with special focus 743 on shallow lakes. 744 Johnson, D.W., Grorud-Colvert, K., Sponaugle, S. & Semmens, B.X. (2014) Phenotypic variation and 745 selective mortality as major drivers of recruitment variability in fishes. Ecology Letters, 17, 743- 746 755. 747 Jonsson, B. & Jonsson, N. (2001) Polymorphism and speciation in Arctic charr. Journal of Fish Biology, 748 58, 605-638. 749 Keeley, E., Parkinson, E. & Taylor, E. (2007) The origins of ecotypic variation of rainbow trout: a test of 750 environmental vs. genetically based differences in morphology. Journal of Evolutionary Biology, 751 20, 725-736. 752 Kingsolver, J.G., Izem, R. & Ragland, G.J. (2004) Plasticity of Size and Growth in Fluctuating Thermal 753 Environments: Comparing Reaction Norms and Performance Curves1. Integrative and 754 Comparative Biology, 44, 450-460. 755 Kinnison, M.T. & Hairston Jr, N.G. (2007) Eco-evolutionary conservation biology: contemporary 756 evolution and the dynamics of persistence. Functional Ecology, 21, 444-454. 757 Kinnison, M.T., Hairston Jr, N.G. & Hendry, A.P. (2015) Cryptic eco‐evolutionary dynamics. Annals of 758 the New York Academy of Sciences, 1360, 120-144. 759 Klingenberg, C.P. (2010) Evolution and development of shape: integrating quantitative approaches. 760 Nature Reviews Genetics, 11, 623-635. 30

761 Klingenberg, C.P. (2014) Studying morphological integration and modularity at multiple levels: concepts 762 and analysis. Philosophical Transactions of the Royal Society B: Biological Sciences, 369, 763 20130249. 764 Lazić, M.M., Carretero, M.A., Crnobrnja-Isailović, J. & Kaliontzopoulou, A. (2015) Effects of 765 Environmental Disturbance on Phenotypic Variation: An Integrated Assessment of Canalization, 766 Developmental Stability, Modularity, and Allometry in Lizard Head Shape. The American 767 Naturalist, 185, 44-58. 768 Matta, E.M., Black, B.A. & Wilderbuer, T.K. (2010) Climate-driven synchrony in otolith growth- 769 increment chronologies for three Bering Sea flatfish species. Marine Ecology Progress Series, 770 413, 137-145. 771 McCann, K.S. (2012) Food Webs. Monographs in Population Biology. Princeton University Press, New 772 Jersey. 773 McCann, K.S., Rasmussen, J.B. & Umbanhowar, J. (2005) The dynamics of spatially coupled food webs. 774 Ecology Letters, 8, 513-523. 775 McMeans, B.C., McCann, K.S., Humphries, M., Rooney, N. & Fisk, A.T. (2015) Food Web Structure in 776 Temporally-Forced Ecosystems. Trends in Ecology & Evolution, 30, 662-672. 777 McPhee, M.V., Noakes, D.L.G. & Allendorf, F.W. (2015) Developmental rate: A unifying mechanism for 778 sympatric divergence in postglacial fishes? Current Zoology, 58, 21-34. 779 McShea, D.W. (2017) Evolution of Complexity. Evolutionary Developmental Biology: A Reference 780 Guide (eds L. Nuno de la Rosa & G. Müller), pp. 1-11. Springer International Publishing, Cham. 781 McShea, D.W. & Brandon, R.N. (2010) Biology's first law: the tendency for diversity and complexity to 782 increase in evolutionary systems. University of Chicago Press. 783 Merilä, J. & Hendry, A.P. (2014) Climate change, adaptation, and phenotypic plasticity: the problem and 784 the evidence. Evolutionary Applications, 7, 1-14. 785 Moczek, A.P. (2010) Phenotypic plasticity and diversity in insects. Philosophical Transactions of the 786 Royal Society B: Biological Sciences, 365, 593-603. 787 Montoya José, M. & Raffaelli, D. (2010) Climate change, biotic interactions and ecosystem services. 788 Philosophical Transactions of the Royal Society B: Biological Sciences, 365, 2013-2018. 789 Morris, M.R. (2014) Plasticity-mediated persistence in new and changing environments. International 790 journal of evolutionary biology, 2014. 791 Muir, A.M., Bronte, C.R., Zimmerman, M.S., Quinlan, H.R., Glase, J.D. & Krueger, C.C. (2014) 792 Ecomorphological diversity of Lake Charr Salvelinus namaycush at Isle Royale, Lake Superior. 793 Transactions of the American Fisheries Society, 143, 972-987. 794 Muir, A.M., Vecsei, P. & Krueger, C.C. (2012) A perspective on perspectives: a method toward reducing 795 variation in digital shape analysis. Transactions of the American Fisheries Society, 141, 1161- 796 1170. 797 Murdoch, A. & Power, M. (2013) The effect of lake morphometry on thermal habitat use and growth in 798 Arctic charr populations: implications for understanding climate-change impacts. Ecology of 799 Freshwater Fish, 22, 453-466. 800 Norberg, J., Urban, M.C., Vellend, M., Klausmeier, C.A. & Loeuille, N. (2012) Eco-evolutionary 801 responses of biodiversity to climate change. Nature Climate Change, 2, 747. 802 Nosil, P. (2012) Ecological speciation. Oxford University Press. 803 O'Dea, R.E., Lagisz, M., Hendry, A.P. & Nakagawa, S. (2019) Developmental temperature affects 804 phenotypic means and variability: A meta-analysis of fish data. Fish and Fisheries, 20, 1005- 805 1022. 806 O’Dea, R.E., Noble, D.W., Johnson, S.L., Hesselson, D. & Nakagawa, S. (2016) The role of non-genetic 807 inheritance in evolutionary rescue: epigenetic buffering, heritable bet hedging and epigenetic 808 traps. Environmental Epigenetics, 2. 809 Olsson, J., Svanbäck, R. & Eklöv, P. (2006) Growth rate constrain morphological divergence when driven 810 by competition. Oikos, 115, 15-22. 31

811 Olsson, J., Svanbäck, R. & Eklöv, P. (2007) Effects of resource level and habitat type on behavioral and 812 morphological plasticity in Eurasian perch. Oecologia, 152, 48-56. 813 Pacifici, M., Foden, W.B., Visconti, P., Watson, J.E.M., Butchart, S.H.M., Kovacs, K.M., Scheffers, B.R., 814 Hole, D.G., Martin, T.G., Akçakaya, H.R., Corlett, R.T., Huntley, B., Bickford, D., Carr, J.A., 815 Hoffmann, A.A., Midgley, G.F., Pearce-Kelly, P., Pearson, R.G., Williams, S.E., Willis, S.G., 816 Young, B. & Rondinini, C. (2015) Assessing species vulnerability to climate change. Nature 817 Climate Change, 5, 215. 818 Parsons, K.J., Sheets, H.D., SkÚLason, S. & Ferguson, M.M. (2011) Phenotypic plasticity, heterochrony 819 and ontogenetic repatterning during juvenile development of divergent Arctic charr (Salvelinus 820 alpinus). Journal of Evolutionary Biology, 24, 1640-1652. 821 Pereira, D.L. (1995) Construction of a 110-year biochronology from sagittae of freshwater drum 822 (Aplodinotus grunniens). Recent Developments in Fish Otolith Research, 177-196. 823 Persson, L. & De Roos, A.M. (2003) ADAPTIVE HABITAT USE IN SIZE-STRUCTURED 824 POPULATIONS: LINKING INDIVIDUAL BEHAVIOR TO POPULATION PROCESSES. 825 Ecology, 84, 1129-1139. 826 Pinheiro, J.C. & Bates, D.M. (2000) Linear mixed-effects models: basic concepts and examples. Mixed- 827 effects models in S and S-Plus, 3-56. 828 Poesch, M.S., Chavarie, L., Chu, C., Pandit, S.N. & Tonn, W. (2016) Climate Change Impacts on 829 Freshwater Fishes: A Canadian Perspective. Fisheries, 41, 385-391. 830 Proulx, R. & Magnan, P. (2004) Contribution of phenotypic plasticity and heredity to the trophic 831 polymorphism of lacustrine brook charr (Salvelinus fontinalis M.). Evolutionary Ecology 832 Research, 6, 503-522. 833 Quinn, T.J., II & Deriso, R.B. (1999) Quantitative fish dynamics. Oxford University Press, New York. 834 R Core Team (2016) R: A language and environment for statistical computing. R Foundation for 835 Statistical Computing, Vienna, Austria. 836 Ramler, D., Mitteroecker, P., Shama, L.N.S., Wegner, K.M. & Ahnelt, H. (2014) Nonlinear effects of 837 temperature on body form and developmental canalization in the threespine stickleback. Journal 838 of Evolutionary Biology, 27, 497-507. 839 Reed, T.E., Waples, R.S., Schindler, D.E., Hard, J.J. & Kinnison, M.T. (2010) Phenotypic plasticity and 840 population viability: the importance of environmental predictability. Proceedings of the Royal 841 Society B: Biological Sciences, 277, 3391-3400. 842 Reist, J.D., Wrona, F.J., Prowse, T.D., Power, M., Dempson, J.B., King, J.R. & Beamish, R.J. (2006) An 843 Overview of Effects of Climate Change on Selected Arctic Freshwater and Anadromous Fishes. 844 AMBIO: A Journal of the Human Environment, 35, 381-387, 387. 845 Schindler, D.E. & Scheuerell, M.D. (2002) Habitat coupling in lake ecosystems. Oikos, 98, 177-189. 846 Schluter, D. (2000) The ecology of adaptive radiation. Oxford University Press, Oxford. 847 Schoener, T.W. (2011) The Newest Synthesis: Understanding the Interplay of Evolutionary and 848 Ecological Dynamics. Science, 331, 426. 849 Sheridan, J.A. & Bickford, D. (2011) Shrinking body size as an ecological response to climate change. 850 Nature Climate Change, 1, 401-406. 851 Shuter, B.J., Finstad, A.G., Helland, I.P., Zweimüller, I. & Hölker, F. (2012) The role of winter 852 phenology in shaping the ecology of freshwater fish and their sensitivities to climate change. 853 Aquatic Sciences, 74, 637-657. 854 Simonsen, M.K., Siwertsson, A., Adams, C.E., Amundsen, P.-A., Præbel, K. & Knudsen, R. (2017) 855 Allometric trajectories of body and head morphology in three sympatric Arctic charr (Salvelinus 856 alpinus (L.)) morphs. Ecology and Evolution, 7, 7277-7289. 857 Skúlason, S., Parsons, K.J., Svanbäck, R., Räsänen, K., Ferguson, M.M., Adams, C.E., Amundsen, P.-A., 858 Bartels, P., Bean, C.W., Boughman, J.W., Englund, G., Guðbrandsson, J., Hooker, O.E., Hudson, 859 A.G., Kahilainen, K.K., Knudsen, R., Kristjánsson, B.K., Leblanc, C.A.L., Jónsson, Z., Öhlund,

860 G., Smith, C. & Snorrason, S.S. (2019) A way forward with eco evo devo: an extended theory of 861 resource polymorphism with postglacial fishes as model systems. Biological Reviews, 0. 862 Smith, N.G., Krueger, C.C. & Casselman, J.M. (2008) Growth chronologies of white sucker, 863 Catostomuscommersoni, and lake trout, Salvelinusnamaycush: a comparison among lakes and 864 between trophic levels. Environmental Biology of Fishes, 81, 375-386. 865 Stearns, S.C. (1989) The Evolutionary Significance of Phenotypic Plasticity. Bioscience, 39, 436-445. 866 Svanbäck, R., Mario Pineda‐Krch, a.M.D., Krch, M., Doebeli, M., Associate Editor: Claire de, M. & 867 Editor: Donald, L.D. (2009) Fluctuating Population Dynamics Promotes the Evolution of 868 Phenotypic Plasticity. American Naturalist, 174, 176-189. 869 Svanbäck, R., Zha, Y., Brönmark, C. & Johansson, F. (2017) The interaction between predation risk and 870 food ration on behavior and morphology of Eurasian perch. Ecology and Evolution, 7, 8567- 871 8577. 872 Thresher, R.E., Koslow, J.A., Morison, A.K. & Smith, D.C. (2007) Depth-mediated reversal of the effects 873 of climate change on long-term growth rates of exploited marine fish. Proceedings of the 874 National Academy of Sciences, 104, 7461. 875 Tonn, W.M., Holopainen, I.J. & Paszkowski, C.A. (1994) Density-Dependent Effects and the Regulation 876 of Crucian Carp Populations in Single-Species Ponds. Ecology, 75, 824-834. 877 van der Have, T.M. & de Jong, G. (1996) Adult Size in Ectotherms: Temperature Effects on Growth and 878 Differentiation. Journal of Theoretical Biology, 183, 329-340. 879 Vigliola, L. & Meekan, M.G. (2009) The back-calculation of fish growth from otoliths. Tropical fish 880 otoliths: information for assessment, management and ecology, pp. 174-211. Springer. 881 Voelker, S.L., Wang, S.Y.S., Dawson, T.E., Roden, J.S., Still, C.J., Longstaffe, F.J. & Ayalon, A. (2019) 882 Tree-ring isotopes adjacent to Lake Superior reveal cold winter anomalies for the Great Lakes 883 region of North America. Scientific Reports, 9, 4412. 884 Wagner, G.P. & Schwenk, K. (2000) Evolutionarily stable configurations: functional integration and the 885 evolution of phenotypic stability. Evolutionary biology, pp. 155-217. Springer. 886 Wang, T., Hamann, A., Spittlehouse, D. & Carroll, C. (2016) Locally Downscaled and Spatially 887 Customizable Climate Data for Historical and Future Periods for North America. PloS one, 11, 888 e0156720. 889 Ware, I.M., Fitzpatrick, C.R., Senthilnathan, A., Bayliss, S.L.J., Beals, K.K., Mueller, L.O., Summers, 890 J.L., Wooliver, R.C., Van Nuland, M.E., Kinnison, M.T., Palkovacs, E.P., Schweitzer, J.A. & 891 Bailey, J.K. (2019) Feedbacks link ecosystem ecology and evolution across spatial and temporal 892 scales: Empirical evidence and future directions. Functional Ecology, 33, 31-42. 893 Weisberg, S., Spangler, G. & Richmond, L.S. (2010) Mixed effects models for fish growth. Canadian 894 Journal of Fisheries and Aquatic Sciences, 67, 269-277. 895 West-Eberhard, M.J. (2003) Developmental plasticity and evolution. Oxford University Press. 896 Westneat, D.F., Wright, J. & Dingemanse, N.J. (2015) The biology hidden inside residual within- 897 individual phenotypic variation. Biological Reviews, 90, 729-743. 898 Willmore, K.E., Zelditch, M.L., Young, N., Ah-Seng, A., Lozanoff, S. & Hallgrímsson, B. (2006) 899 Canalization and developmental stability in the Brachyrrhine mouse. Journal of Anatomy, 208, 900 361-372. 901 Wood, Z.T., Fryxell, D.C., Moffett, E.R., Kinnison, M.T., Simon, K.S. & Palkovacs, E.P. (2020) Prey 902 adaptation along a competition-defense tradeoff cryptically shifts trophic cascades from density- 903 to trait-mediated. Oecologia. 904 Yamaguchi, K. (1991) Event history analysis. Sage. 905 Zelditch, M.L., Swiderski, D.L. & Sheets, H.D. (2012) Geometric morphometrics for biologists: a 906 primer. Academic Press. 907

908 33

909 Table 1. Tests of fixed effects for differences between lake charr ecotypes (lean, huronicus), 910 sample years (2007, 2018), and sexes (male, female) from a nonlinear mixed-effects model of 911 back-calculated length at sagittal otolith age, with random fish effects (fish-to-fish variation in 912 growth) sampled in Rush Lake.

(-0.5*Δi) Models df logLik AIC Δi e wi Ecotypes + Years 19 -7672.0 15382.0 0.0 1 1 Ecotypes 13 -7702.4 15430.8 48.8 0 0 Years 13 -7705.7 15437.4 55.4 0 0 Sexes 13 -7729.0 15484.0 102.0 0 0 Null 10 -7732.9 15485.8 103.8 0 0

913

914

915 Table 2. Tests of fixed effects for differences between lake charr morphs (huronicus or lean), 916 sample years (2007 or 2018), and sexes (males or females) from a linear mixed-effects model of 917 annual sagittal otolith growth increments as a function of a fixed age effect (age of increment 918 formation), random year effects (year of increment formation), and random fish effects (fish-to- 919 fish variation in growth) sampled in Rush Lake.

Df Df Morph Effect Numerator Denominator F-Ratio P-Value Both Age 34 1,927 718.5 ≤ 0.01 Ecotype 1 1,927 16.9 ≤ 0.01

Sex 1 1,927 0.4 0.5

Year 1 1,927 0.08 0.8

Huronicus Age 30 1,182 515.4 ≤ 0.01 Sex 1 1,182 0.05 0.8

Year 1 1,182 0.9 0.3

Lean Age 34 682 255.9 ≤ 0.01 Sex 1 682 1.4 0.2

Year 1 682 0.6 0.4

920

921 922

923 Fig. 1. a) PCA of lake charr body shape with percentage representing the variation explained by 924 that component and b) CVA of lake charr body shape with 95% confidence ellipses delineating 925 groups. Deformation grid plots represent the body and head shape variation of lake charr 926 sampled in Rush Lake in 2007 and 2018, in a positive direction, along each axis. 35

927

928

929 930

931 Fig. 2. a) PCA of lake charr head shape with percentage representing the variation explained by 932 that component and b) CVA of lake charr head shape with 95% confidence ellipses delineating 933 groups. Deformation grid plots represent the body and head shape variation of lake charr 934 sampled in Rush Lake in 2007 and 2018, in a positive direction, along each axis 36

935 936 Fig. 3. Body and head shape changes through growth of lean and huronicus lake charr ecotypes 937 from 2007 and 2018. PC1 scores of body shape (a) and head shape (b) are plotted against 938 centroid size. Only lean 2018 regressions were significantly different from 0, for both body and 939 head shape (body shape: R2 = 0.56, p≤0.01; head shape: R2= 0.36, p≤0.01). Lean are represented 940 by squares and huronicus by circles, whereas filled symbols are individual sampled in 2007 and 941 empty symbols are lake charr caught in 2018.

942

943

944

945

946

947 948 949 Fig. 4. Relative body condition of huronicus and lean lake charr sampled from Rush Lake, in 950 2007 and 2018. 951

952

953

954

955

956

957 958

959 960 Fig. 5. Asymptotic length (mm) and early growth rate (mm/year; first year) calculated from 961 sections sagittal otoliths of huronicus and lean lake charr sampled from Rush Lake, in 2007 and 962 2018. 963 964 965

966 967 Fig. 6. Annual growth increments (mm; random year effects from a linear mixed-effects model 968 that also included fixed age effects and random fish effects; Weisberg et al. 2010) by calendar 969 year for huronicus and lean lake charr ecotypes sampled from Rush Lake, in 2007 and 2018. 970 971 972 973

974 975 Fig. 7. Pearson correlation coefficients between ARSTAN-detrended otolith growth increments 976 and selected seasonal climate variables for each ecotype. Results include data combined across 977 fish sampled in 2007 and 2018. Correlation values for each ecotype that used a proportional 978 weighting scheme for seasonal variables across the current and two previous years is 979 represented by dark bars, whereas light bars indicate correlation values with no weighting. 980 Panels including proportional weight values (i.e., sum of weights equal to one) whereby the 981 combination of weights was optimized to maximize correlation values shown in dark bars 982 within the panel immediately above. Year lags correspond to the current year = 0, one previous 983 year = -1 and two previous years = -2).

984

985

986 987 Figure 8. Otolith increment growth variation among ecotypes and collection periods for ages 1- 988 3 plotted versus regional cloud cover from July to September for the same year without 989 weighting or lag-effects included in a) and b). Data were detrended using regional chronology 990 standardization (see Methods and table A1). Within a given year, ecotype and collection period, 991 otolith growth data for early stage growth were averaged across ages 1-3. In c), reductions in 992 regional cloud cover by month from 1983 to 2017. Shown here are the predicted values for 993 1983 and 2017 using linear regressions fitted to interannual cloud cover data by month across 994 the same time period. These two years represent endpoints for which otolith data were 995 replicated enough to allow chronology construction. Cloud cover data were from airports 996 within 7 km of Lake Superior since the lake itself strongly controls nearby weather (see Voelker 997 et al. 2019).

998

999 1000 42

1001 Appendix

1002 Table A1. Listing and comparison of analytical methods used to analyze otolith growth data for 1003 different purposes and associated results. Analytical method Advantages Disadvantages Results for otolith growth Non-linear mixed- Uses all data, so it can be used Cross-dating not confirmed, Table 1, effects modeling with all fish morphology more environmental noise Table 2, Fig. measurements to back- not related to climate drivers 5, Fig. 6, calculate fish growth and Table A1 asymptotic length Cross-dated and Conventional approach for Short age span of fish Fig. 7, Table ARSTAN- identifying climate drivers of excludes identification of A2, Fig A3. detrended inter-annual variation in long- decadal or greater lived organisms such as trees environmental drivers (i.e., "segment length curse"). Some loss of data when otolith time seres do not cross-date Cross-dated and Alternative approach for Not as efficient at accurately Fig. 8, Fig A3 RCS-detrended identifying climate drivers extracting inter-annual that preserves decadal and variation compared to longer signals ARSTAN-detrending

1004 1005

1006 Table A2. Difference in annual growth increments (mm; random year effects from a linear 1007 mixed-effects model that also included fixed age effects and random fish effects; Weisberg et 1008 al. 2010) by calendar year for huronicus and lean lake charr ecotypes sampled from Rush Lake, 1009 in 2007 and 2018 (Figure 6). Metric Huronicus Lean %

SD 0.0049 0.0088 44 Slope -0.002 -0.0026 20 Decline 0.018 0.024 24 Mean 0.0004 0.0018 76

1010

1011 Table A3. Forward selection multiple regression modeling of detrended otolith increment 1012 growth chronologies for lake charr ecotypes (huronicus or lean) sampled in Rush Lake in 2007 1013 and 2018, as predicted by weighted seasonal climate variables. Environmental variables are 1014 represented as follow: Twin = winter temperature, SNOwin = winter snow, Tsum = summer 1015 temperature, and Pfall = fall precipitation.

1016 Models R2 P-value AIC

Huronicus Twin 0.23 0.01 -90.3

SNwin 0.26 <0.01 -91.5

SNwin + Tsum 0.35 <0.01 -93.1

Lean Twin 0.31 <0.01 -59.6

Pfal 0.37 <0.01 -61.5

Twin + Pfal 0.56 <0.01 -67.7

1017

1018

1019

1020

1021

1022

1023

1024

1025

1026

1027

1028

1029

1030

1031 1032 a)

1033 1034 b)

1035 Fig. A1. Lean-like (a) and huronicus (b) lake charr Salvelinus namaycush ecotypes of Rush Lake. 1036 Illustration by P.Vecsei, after Chavarie et al. (2016)

1037

1038

1039 1040 Fig. A2. Otolith growth increment plotted by age between lake charr ecotypes (lines) and 1041 detailed across the first three years for lake charr morphs (huronicus or lean) sampled in 2007 1042 and 2018.

1043

1044

1045

1046

1047

1048

1049 1050

1051 Fig. A3. Standard deviations (SD) of ARSTAN-detrended growth increment versus previous 1052 winter temperatures and precipitation as snow. Growth SD between lake charr ecotypes 1053 sampled in Rush Lake in 2007 and 2018, plotted versus weighted winter climate variables (A, B), 1054 SD of the difference between ecotype growth versus weighted winter variables (C, D) and SD of 1055 the difference between ecotype growth plotted versus a “winter index” that averaged the SD of 1056 sign-reversed temperature and precipitation as snow.

1057

1058

1059

1060

1061

1062

1063

1064

1065 1066