Multiple species drive flexible lake food webs with warming

Climate change is rewiring the food webs that determine the fate of diverse ecosystems. Mobile generalist consumers are responding to climate change by rapidly shifting their behaviour and foraging, driving food webs to flex. Although these responsive generalists form a key stabilizing module in food web structure, the extent to which they are present throughout whole food webs is largely unknown. Here, we show that multiple species comprising key trophic roles drive flexible lake food webs with warming. By examining lakes that span a 7°C air temperature gradient, we found significant reductions in nearshore derived carbon and nearshore habitat use with increased temperature in three of four fish species. We also found evidence that the response of lake trout to increased temperatures may reduce their biomass and cascade to release their preferred prey, the pelagic forage fish cisco. Our results suggest that climate warming will shift lake food webs toward increased reliance on offshore habitats and resources. We argue that species across trophic levels broadly couple lake macrohabitats, suggesting that potentially stabilizing responsive consumers are present throughout food webs. However, climate change appears to limit their ability to responsively forage, critically undermining a repeated stabilizing mechanism in food webs.

Introduction data for these species was available for 40 or more lakes and the trophic level, thermal 156 classifications, and habitat preferences for these species are known (Coker et al. 2001;Hasnain 157 et al. 2013). Lake trout (Salvelinus namaycush) is the most common top predator of offshore this coldwater species (Dolson et al. 2009;Plumb & Blanchfield 2009;Tunney et al. 2014;160 Guzzo et al. 2017). Cisco (or lake herring, Coregonus artedi) is one of the most common cold-161 adapted planktivores and is a common prey item for lake trout and walleye. Walleye (Sander 162 vitreus) is a common cool-adapted piscivore and popular sport fish that is present in lakes across 163 the boreal shield. Yellow perch (Perca flavescens) is a widespread and abundant cool-adapted 164 mesopredator that is consumed by a wide variety of predatory fishes throughout its range, 165 including both lake trout and walleye. These four species comprise a large portion of the average 166 catch in the 59 lakes that we use here from the OMNRF's BsM surveys (see Supplementary 167 Information). The number of individual for each species sampled for stable isotope analysis in 168 each lake varied from 2 to 21 (mean 13.5) for lake trout, 3 to 32 (mean 17.0) for walleye, 1 to 20 169 (mean 10.8) for cisco, and 1 to 18 (mean 9.1) for yellow perch. For both lake trout and walleye, 170 only individuals greater than 250 mm were used for stable isotope analysis because these species 171 are known to show ontogenic diet shifts (Mittelbach & Persson 1998;Sherwood et al. 2002;172 Galarowicz et al. 2006).

174
Food Web Metric Calculations using Stable Isotopes 175 As in many previous studies, we used stable isotopic signatures from our four fish species 176 and baseline invertebrates to calculate both the nearshore carbon index (based on the proportion 177 nearshore carbon, e.g., Tunney et al. 2014) and trophic position (Vander Zanden et al. 1999b, a;178 Post 2002). Collection and processing methods for stable isotope data can be found in Dolson et 179 al. 2009;Tunney et al. 2014Tunney et al. , 2018. We used baseline invertebrates from the nearshore and 180 offshore zones to account for variability in isotopic signatures across lakes. We incorporated data 181 for multiple trophic groups into both our nearshore and offshore baseline isotopic signatures to reduce the number of estimated baseline values required for our analysis and to increase the 183 sample size for our baseline isotopic signatures. We corrected all δ 13 C signatures using C:N 184 ratios as 185 δ 13 C corr = δ 13 C raw + (−3.32 + 0.99 * CN) 186 where δ 13 C corr is the corrected δ 13 C signature, δ 13 C raw is the raw δ 13 C signature, and CN is the 187 C:N ratio of that tissue sample (Post et al. 2007). For lakes that were missing either nearshore or 188 offshore baseline isotopic signatures, we estimated the δ 13 C and δ 15 N signatures of the missing 189 baseline using the available baseline and simple linear regression between the baselines across 190 lakes (see Supplementary Information). 191 We used two source mixing models to estimate the nearshore carbon index and the 192 trophic position of each species in each lake based on their relative isotopic signatures (Post 193 2002). We calculated the nearshore carbon index for each fish species as 194 NCI fish = δ 13 C fish -δ 13 C osb δ 13 C nsb -δ 13 C osb

195
where NCI fish is the nearshore carbon index in the diet of a fish species, δ 13 C fish is the 196 average δ 13 C signature for that fish species, δ 13 C osb is the average or estimated δ 13 C signature 197 for all offshore baselines (i.e., mussels and/or zooplankton), and δ 13 C nsb is the average or 198 estimated δ 13 C signature for all nearshore baselines (i.e., snails and/or aquatic insect larvae). The 199 nearshore carbon index is similar to the proportion nearshore carbon used by others (e.g., Tunney

207
where TP fish is the trophic position of a fish species, NFI fish is the nearshore carbon index for that 208 fish species, δ 15 N fish is the average δ 15 N signature for that fish species, δ 15 N nsb is the average or 209 estimated δ 15 N signature for all nearshore baselines (i.e., snails and aquatic insect larvae), 210 δ 15 N osb is the average or estimated δ 15 N signature for all offshore baselines (i.e., mussels and  Behaviour Metrics and Biomass Index Using Catch-per-unit-effort 215 We used catch-per-unit-effort data for each depth stratum (Sandstrom et al. 2013 We found strong evidence that lake food web structure varied across an approximately 7° C 251 climate gradient. Three of the four species showed evidence of changes in nearshore feeding 252 with increased temperature. Both top predators (lake trout and walleye) showed a significant 253 decrease in the nearshore carbon index with increasing average recent air temperature, and the 254 cold-adapted cisco showed a similar marginally significant decrease (Figure 2a, c, and e, Table   255 3). The cool-adapted yellow perch showed no significant relationship between average recent air 256 temperature and the nearshore carbon index (Figure 2g, Table 3). Lake trout showed a significant  Information and Table S3).

262
Behaviour  In agreement with changes in nearshore carbon index, the same three species (lake trout, 264 walleye, and cisco) showed strong evidence of reduced nearshore habitat use. All three species 265 showed a significant increase in mean depth of capture with increasing average recent air temperature (Figure 3a, c, and e, Table 3). Consistent with these results, lake trout and cisco 267 showed a significant decrease in probability of nearshore presence and walleye showed a 268 marginally significant increase in probability of offshore presence with increasing average recent 269 air temperature (see Supplementary Information). In contrast, yellow perch showed no 270 relationship between either mean depth of capture or probability of offshore presence and 271 average recent air temperature (Figure 3g and inset, Table 3).

272
We found evidence of a reduction in biomass with increasing temperature for one 273 species, lake trout, which showed a significant decrease in CUE biomass index with increasing 274 average recent air temperature ( Figure 3b, the reason for the different responses of these species is unclear. Despite the lack of response in 307 yellow perch, the foraging responses of these fish species strongly indicate that boreal shield lake 308 food webs will collectively flex towards offshore resource and habitat use in response to 309 warming. 310 We found evidence that a reduction in nearshore coupling with warming may impact the  web architecture that has long interested ecologists (Sugihara et al. 1989).