Thermal preference influences depth use but not biomass of predatory fishes 1 in response to lake morphometry 2

19 Top predators’ responses to environmental conditions shape food web architecture and influence 20 ecosystem structure and stability. Yet the impacts of fundamental properties like ecosystem size 21 and morphometry on top predators’ behaviour are poorly understood. We examined how lake 22 morphometry impacts the behaviour (inferred by depth use) of three key fish top predators—the 23 cold-adapted lake trout, the cool-adapted walleye, and the warm-adapted smallmouth bass— 24 which can each strongly impact local food web structure. We used catch-per-unit-effort data 25 from nearly 500 boreal lakes of Ontario, Canada to evaluate the role of thermal preference in 26 dictating mean depth of capture and biomass index in response to lake morphometry. We found 27 evidence that thermal preferences influence how species’ depth use and biomass changed with 28 lake size, proportion of littoral area, and maximum lake depth, although we found no relationship 29 with lake shape. However, found no strong evidence that lake morphology influences these 30 species’ biomasses, despite theory that predicts such a relationship. Our results suggest that some 31 aspects of lake morphometry can alter habitat accessibility and productivity in ways that 32 influence the behaviour and biomass of these top predator species depending on their thermal 33 preferences. Our results have implications for how lake food webs expand and contract with lake 34 morphometry and other key abiotic factors. We argue that several key abiotic factors likely drive 35 top predator depth use in ways that may shape local food web structure and play an important 36 role in determining the ultimate fate of ecosystems with environmental change. 37

The degree to which predatory fish couple offshore and nearshore habitats is potentially 71 shaped by a few key factors. One possibility is that predators respond to changes in resource 72 abundance by shifting towards locations with abundant resources (i.e., the birdfeeder effect, see suggesting that thermal accessibility may be limiting for many species. Various morphometric 85 factors appear to influence the accessibility of the nearshore in lakes. During summer, large and 86 . CC-BY 4.0 International license not certified by peer review) is the author/funder. It is made available under a The copyright holder for this preprint (which was this version posted December 3, 2019. . https://doi.org/10.1101/572925 doi: bioRxiv preprint 5 reticulate lakes likely have larger nearshore zones, and predatory fish must spend more time and 87 travel greater distances through warm water to reach prey (Dolson et al. 2009). As a 88 consequence, increasing lake size and complexity are associated with reduced consumption less 89 from littoral resource pathways in lake trout (Dolson et al. 2009, Tunney et al. 2012. 90 Importantly, these foraging shifts are associated with behavioural shifts in the form of altered 91 habitat use (Guzzo et al. 2017, Bartley et al. 2018). However, cool-water species such as the 92 walleye are likely less thermally restricted than cold-water species, and so ought to show weak 93 behavioural response to changes in nearshore accessibility (Fig. 1). Furthermore, warm-water 94 species, such as smallmouth bass, may never be thermally restricted from accessing prey in the 95 nearshore and thus may show no behavioural responses to lake morphometry (Fig. 1 Eloranta et al. 2015). Thus, the differences in coupling responses to changes in lake 103 morphometry predicted here ought to produce a large reduction in energy availability for lake 104 trout but little reduction for walleye and no reduction for smallmouth bass. Thus, these three 105 species ought to show contrasting biomass responses to changes in lake size and complexity.

106
Such responses would strongly suggest that lake morphometry shapes whole food webs.

107
Here, we use an extensive catch-per-unit-effort database of fishes in over nearly 500 108 boreal lakes of Ontario, Canada to evaluate the role of thermal preference in driving the depth 109 . CC-BY 4.0 International license not certified by peer review) is the author/funder. It is made available under a The copyright holder for this preprint (which was this version posted December 3, 2019. . https://doi.org/10.1101/572925 doi: bioRxiv preprint 6 use (inferred from mean depth of capture) and biomass index responses to lake morphometry. 110 We predict that as lake size, complexity and proportion littoral area increase, lake trout will be 111 caught at much deeper depths (Fig. 1). We predict that walleye will show weak depth response to 112 lake morphometry, and that smallmouth bass will show no depth response to changes in lake 113 morphometry (Fig. 1). Because changes in depth use by fish are thought to influence species 114 biomass through changes in resource availability, we also predict that lake trout will show 115 decreasing biomass index with increasing lake size, complexity, and proportion littoral area. We 116 predict that walleye will show weak biomass index decreases with these lake morphometric 117 variables, and that smallmouth bass will show no change in biomass index. By evaluating the 118 depth use and biomass of these three key top predator species with changes in habitat 119 accessibility, we provide further insight into the response of mobile generalist consumers to key 120 ecosystem characteristics that can shape the ways that ecosystems are altered by global 121 environmental change.

144
For our study, we used data from Cycle 1 of BsM, which took place from 2008 to 2012. 145 We only included lakes with a maximum depth greater than 10 m to ensure that both nearshore 146 and offshore habitats were present for the lakes that we use in our analysis. This criterion left 556 147 lakes for our analyses. Because the BsM protocol uses a relatively small number of fixed depth 148 strata (especially in the littoral zone), it is limited in its ability to detect depth use change within 149 the nearshore habitat but is well suited to detect changes in depth use in terms of nearshore and 150 offshore habitat use.

152
Lake Attribute Selection 153 We selected seven lake attributes to use as predictor variables in our statistical analysis following 154 a two-step procedure. First, we retrieved 10 lake attributes available for the 556 lakes included in 155 . CC-BY 4.0 International license not certified by peer review) is the author/funder. It is made available under a The copyright holder for this preprint (which was this version posted December 3, 2019. . https://doi.org/10.1101/572925 doi: bioRxiv preprint 8 our analysis from OMNRF's data to evaluate lake size, lake complexity, and habitat availability, 156 but also variables to control for climatic conditions, and productivity. Lake size was measured by 157 the surface area of the lake and lake complexity by the shoreline development index (or SDI) 158 computed following Dolson et al. (2009) using lake area and lake shoreline. We used growing 159 degree days based on air temperature (yearly average cumulative number of degrees above 5°C 160 from 1981 to 2010) and latitude as proxies for climatic condition; Secchi depth, total phosphorus 161 and dissolved organic carbon to assess productivity; mean and maximum depth (based on the 162 bathymetry of the lake), and the proportion of littoral area (proportion of the lake less than 4.6 163 m) to assess habitat availability. In a second step, we transformed all these variables (see Table   164 1) and scaled them in order to perform a principal component analysis (PCA). Based on the 165 results (see Fig S1, Table S1 and S2), we narrowed down the control variables to two: growing    where SMD is the mean depth of capture of a fish species, CUEi is the CUE of that species for 204 depth stratum i, pi is the proportion of the lake in depth stratum i, and di is the middle depth of 205 the depth range for that depth stratum.

206
The stratified sampling protocol use for BsM means that the survey of each lake is 207 essentially a "snapshot" of the habitat use (or depth distribution) of each species. We are using 208 this snapshot to calculate the mean "behaviour" of a species-a value that typifies what that 209 species is doing on average in that lake. We interpret shifts in the averages of this metric of  in each lake, we transformed the mean fork length of each species in each lake into a mean 222 . CC-BY 4.0 International license not certified by peer review) is the author/funder. It is made available under a The copyright holder for this preprint (which was this version posted December 3, 2019. . https://doi.org/10.1101/572925 doi: bioRxiv preprint 11 weight using the following equations: For lake trout, weight = 10 -5.19 * (fork length) 3.1 ; for 223 walleye, weight = 10 -4.79 * (fork length) 2.9 ; for smallmouth bass, weight = 10 -4.7 * (fork length) 3 . 224 We then multiplied the CUE for each species in each lake by the estimated mean weight to get a 225 mean weight per unit effort in kg per 100 metres per night or WUE, hereafter referred to as 226 'biomass index'. Both fork length and CUE data as well as the equations we used to calculate  Prey fish distribution 230 We also calculated the proportion of prey fish captured in the offshore zone of each lake. 231 We used the total number of prey fish captured in offshore nets (>6 m depth) divided by the total 232 number of prey fish caught in that lake. We considered a species to be a prey fish if it was not 233 primarily piscivorous in inland lakes (Coker et al. 2001). Our calculations included data from 234 both small mesh and large mesh nets, but we only included species with at least 10 individuals 235 <250 mm in length for our calculation to remove rare species and individuals too large to be prey 236 for most predators. These criteria left 53 species as prey fish.   (Table 2). Scaled coefficients indicated that as lake surface area, lake maximum depth, and 266 growing degree days increased, lake trout were captured at deeper depths, with lake surface area 267 . CC-BY 4.0 International license not certified by peer review) is the author/funder. It is made available under a The copyright holder for this preprint (which was this version posted December 3, 2019. . https://doi.org/10.1101/572925 doi: bioRxiv preprint 13 having the strongest influence of these three variables (Table 2, Fig. 2, 3). Proportion littoral area 268 and shoreline development index each had relative weak influences on lake trout depth (Table 2).   Fig. 2, 3).  (Table 3). Scaled coefficients indicated that each of these variables had 288 moderate negative effects on lake trout biomass index (Table 3, Fig. 2, 4). For walleye, the top-ranked model accounted for 40% of the variation in biomass index 290 among lakes and contained the predictor variables surface area, lake maximum depth, growing 291 degree days, shoreline development index, Secchi depth, and proportion littoral area (Table 3). 292 Scaled coefficients indicated that growing degree days, and lake maximum depth and shoreline 293 development index had a strong and moderate negative effect on biomass index, while lake 294 surface area had a weak positive effect (Table 3, Fig. 2, 4). 295 For smallmouth bass, the top-ranked model accounted for 19% of the variation in 296 biomass index among lakes and contained the predictor variables lake maximum depth, shoreline 297 development index, and proportion littoral area (Table 3). Scaled coefficients indicated that lake 298 maximum depth and shoreline development index had weak negative effects, while the 299 proportion littoral area had a moderate positive effect (Table 3, Fig. 2, 4). The top-ranked model accounted for 39% of the variation in the proportion of the prey fish 303 captured in the offshore among lakes and contained the predictor variables lake surface area, lake 304 maximum depth, growing degree days, and proportion littoral area (Table 4). Scaled coefficients 305 indicated that each of these variables except surface area had a negative effect on the proportion 306 of prey fish located in the offshore. The strongest influences on the proportion of prey fish 307 located in the offshore were proportion of littoral area, which had a negative effect, and lake 308 surface area, which had a positive effect on the offshore prey fish (Table 4). Predator's responses to environmental conditions are now widely recognized as a key 312 part of food web architecture that influences ecosystem structure and stability (Bartley et al. 313 2019). In this study, we sought to understand how lake morphometry impacts the depth use and 314 biomass of three key fish predators that strongly influence food web structure while controlling 315 for other factors, such as climate and productivity. By examining three key predator species that 316 differ in thermal guild, we can better understand the drivers of species habitat use and related 317 consequences for populations and for food web dynamics. Such insights are likely to be key to 318 piecing together the fundamental factors that structure food webs, which form the backdrop for 319 how ecosystems respond to environmental change. 320 We found strong evidence that species' thermal preferences influence their responses to 321 lake size. Consistent with our predictions, the depth use of lake trout and walleye are strongly . 339 We found that the thermal preferences of these predator species influenced their depth-340 use responses to the proportion of littoral habitat in lakes. All species occupy shallower depths as 341 the proportion of littoral area in lakes increases, but this was stronger for walleye and 342 smallmouth bass than for lake trout. Because of the correlation between proportion littoral area, 343 mean lake depth, total phosphorous, and dissolved organic carbon, a larger proportion littoral 344 area likely generally indicates a shallower, more productive lake. Thus, it seems likely that 345 because walleye and smallmouth are less limited by thermal accessibility, these two species can 346 respond to this increased nearshore productivity by using this area of the lake more. In contrast, 347 the thermal limitations of lake trout (Evans 2007) limit their ability to take advantage of greater  We did not find strong evidence that lake shape influences the depth use of top predator 353 fishes in boreal lakes of Ontario. We found only some evidence for an influence of lake shape, 354 estimated by shoreline development index, on the depth use of smallmouth bass but not for lake 355 trout or walleye. This result is in contrast to previous work that shows a response of lake trout 356 diet to changes in lake shape (Dolson et al. 2009). This discrepancy is potentially due to the vast 357 . CC-BY 4.0 International license not certified by peer review) is the author/funder. It is made available under a The copyright holder for this preprint (which was this version posted December 3, 2019. . https://doi.org/10.1101/572925 doi: bioRxiv preprint differences in the ranges of lake area and other factors between our datasets since the lakes 358 studied by (Dolson et al. 2009) are relatively small and geographically constrained to Algonquin 359 Park in Ontario, Canada. It is possible that across a large geographic area, other abiotic factors 360 are more important in driving the depth use of these species. It is also possible that lake shape 361 does not present as much of a limitation to habitat accessibility as previously thought, especially 362 when controlling for other factors such as littoral area and lake size. The relatively high R 2 for 363 lake trout depth of capture suggests that this species' habitat use is better explained by the abiotic 364 factors we examined here, including lake morphometry. For walleye and smallmouth bass, a 365 significant portion of the overall variance is not explained by the models, suggesting that factors 366 not considered in the present study could more strongly influence these species' habitat use. 367 Interestingly, we found that maximum lake depth was an important factor for all species   investigation than what we can provide here.

419
Our examination of the prey fish catch data shows that the proportion of prey fish located 420 in the offshore habitat increases with lake size. This result can be interpreted in multiple ways.

421
One possibility is that this increased biomass of prey offshore indicates an increased availability 422 of resources in the offshore in larger lakes (i.e., a bottom-up response). Lake trout and walleye 423 may be responding to this change in resource density rather than habitat accessibility of the 424 nearshore. However, this would also suggest that smallmouth bass is not responding in this

595
. CC-BY 4.0 International license not certified by peer review) is the author/funder. It is made available under a The copyright holder for this preprint (which was this version posted December 3, 2019. . https://doi.org/10.1101/572925 doi: bioRxiv preprint