Cross-seasonal weather effects interact with breeding conditions to impact reproductive success in an alpine songbird

In alpine habitats, fluctuating early-season weather conditions and short breeding seasons limit reproductive opportunities, such that arriving and breeding earlier or later than the optimum may be particularly costly for migratory species. Given early-season energy limitations, the influence of environmental conditions across the annual cycle on breeding phenology may have pronounced fitness consequences, yet our understanding of cross-seasonal dynamics in alpine breeding organisms is severely limited. For an alpine-breeding, migratory population of horned lark (Eremophila alpestris) in northern British Columbia, Canada (54.8°N latitude) we assessed how spatially explicit weather conditions from across the annual cycle influenced clutch initiation date and offspring development. We also addressed how cross-seasonal effects on breeding parameters interact to influence reproductive fitness. With 12 years of intensive breeding data and 3 years of migration data from archival light-level geolocators, we used a sliding window approach to identify critical points during the annual cycle where weather events most influenced breeding phenology and offspring development. Consequences for reproductive success were assessed using nest survival simulations. Average clutch initiation varied up to 11 days among years but did not advance from 2003 to 2019. Colder temperatures with greater precipitation at wintering habitats, as well as colder temperatures upon arrival at the breeding site delayed clutch initiation, independent of arrival time. Extreme cold (sub-zero temperatures) within a staging area just prior to arrival at the breeding site carried over to prolong offspring development rate, potentially by influencing parental investment. Nest survival decreased with both later clutch initiation and prolonged offspring development, such that females that nested earlier and fledged offspring at a younger age were up to 45% more likely to reproduce successfully. We demonstrate pronounced carry-over effects acting through mechanisms that influence breeding phenology and offspring development independently. We also highlight the potential importance of staging areas for alpine songbirds, particularly given that environmental conditions are becoming increasingly decoupled across seasons. Understanding the cross-seasonal mechanisms shaping breeding decisions in stochastic environments like the alpine enables more accurate predictions of future individual- and population-level responses to climate change.


Introduction
years and, being just prior to when larks arrive in the alpine, is likely to influence breeding  202 We gathered temperature and precipitation data separately for the winter and staging 203 areas. Within each region, we laid out a grid of points spaced at 50 km intervals. We then 204 interpolated surface air temperature ('air.sig995') and precipitation rate ('prate.sfc') from the 205 NCEP R-1 database to each point four times daily (midnight, 0600, 1200, and 1800 hr) for all 12 at least 75% of the stage-specific location estimates were selected to differentiate the two regions 218 such that the areas did not overlap. The black triangle denotes the breeding site.

220
Statistical analysis 222 Weather variable selection and model fitting 223 We assessed the influence of spatially-explicit weather conditions on clutch initiation and age at variable importance. All analyses were performed using R version 3.6.3 (R Core Team 2020). 232 We constructed a sliding window that spanned different but overlapping time periods for  Figure S1). Therefore, we excluded models containing both 263 variables from consideration when conducting AICc model selection and averaging (see below).

264
For age at fledging, winter and stopover freeze days were strongly correlated (rp = 0.87; 265 Supplemental Appendix: Figure S2). These variables represented similar time windows (winter = pattern. Therefore, we chose to only include stopover freeze days because it had the greatest 268 effect size and lowest AICc relative to the null during sliding window selection (stopover ΔAICc 269 = -7.4; winter ΔAICc = -4.5). Following these steps, all predictor variables of both clutch 270 initiation and age at fledging had a subsequent VIF < 1.9, indicating no multicollinearity. 271 We built global models for both clutch initiation and age at fledging which included the 272 selected candidate weather variables and additional biologically relevant explanatory variables.

273
For clutch initiation date, we included snow depth as an explanatory variable. For age at 274 fledging, we included clutch initiation date and brood size to control for potential effects on 275 development rate. For both response variables, we also included year as a linear term to test for 276 an overall trend over the study period. All variables were standardized and centered to allow 277 comparison of relative effect sizes. We broke each global model into all possible subset models 278 (clutch initiation: n = 64 models; age at fledging: n = 32) and ranked each with AICc. As  Table S1 and S2 for top models). Bootstrapped means and 85% 284 credible intervals were used to evaluate parameter strength and significance. We chose an 85% 285 CI because it is more consistent with an information theoretic approach (Arnold 2010).

287
Fitness effects 288 We assessed the influence of advanced or delayed clutch initiation relative to the annual mean on

301
Model support was compared using AICc.    initiation, prolonged development times (13 days) had a predicted 14.6% lower probability of 398 success than rapid development (7 days; Figure 6). For early nests, the predicted cost of delayed 399 fledge was minimal (-6.9%), but the cost increased as the season progressed (-20.1%; Figure 6).      indicates weak to no correlation. A correlation greater than ±0.70 is considered a concern for 865 collinearity. weak to no correlation. A correlation greater than ±0.70 is considered a concern for collinearity. 881 882