Free meals during breeding: increased resource access does not benefit Arcic-nesting shorebirds

Arctic ecosystems are facing intensifying impacts of climate change, notably an increase in air temperature that can boost local productivity. Yet whether such resource surge can benefit declining migratory birds is unclear. Here we experimentally increased prey abundance and measure its effect on the body condition, nesting patterns, and nesting success of the white-rumped sandpiper (Calidris fuscicollis). We captured females at the beginning and at the end of their incubation to assess body condition and we installed small temperature probes inside their nests to measure their incubation recess. To estimate nest survival, we regularly monitored experimental and control nests during two consecutive summers (2016–2017). For both experimental years with contrasting nest success (73% vs. 21%), we found no evidence of an effect of our supplementation experiment mimicking an increased abundance of arthropods in the Canadian Arctic. This suggests that in situ resources are not limiting during incubation. Breeding strategies and success in shorebirds seem to be driven by inter-individual traits related to body condition upon the initiation of incubation.

Incubation is a complex period as one or both parents must maintain the 41 temperature of their eggs above a threshold, deter nest predators, and feed. Generally, in 42 order for embryos to develop correctly, nest temperatures should not drop below a 43 temperature threshold of 35 °C (White and Kinney 1974). In High Arctic environments 44 where average air temperatures usually range from -10 to 10 °C in the summer, this 45 means that birds often cannot leave their nest unattended for long periods of time (Cantar 46 and Montgomerie 1985, Olson et al. 2006). However, income breeders also regularly 47 need to feed in order to maintain sufficient body stores reserves (Cresswell 2004, 48 Supplementation experiment 164 The experiment started as soon as possible with the random designation of control or 165 supplementation treatments to nests as they were discovered. We chose to use industrially 166 sold dried mealworms (Tenebrio molitor) as our supplementation because they have 167 previously been used as experimental resources for little stints (Tulp and Schekkerman,168 unpubl. data) and red knots (Vézina, unpubl. data) and are commonly used to feed captive 169 or rescued shorebirds (e.g., Gartell et al. 2014). Additionally, we observed white-rumped 170 sandpipers feeding on mealworms (video available along with data and code on figshare 171 taken during preliminary tests in 2015; see link at the end). 172 For all supplemented nests (n = 15 in 2016, n = 13 in 2017), we placed a plastic cup (~4 173 cm diameter) filled mealworms 5 m east of the nest (Fig. 2c). Approximately half of the 174 cup was buried in the ground but remained accessible to sandpipers. This 175 supplementation was available to the sandpipers for the rest of their incubation; since 176 refills were not required required, we considered the number of mealworms as ad libitum 177 resources for the breeding individual white-rumped sandpiper. We also used a cup 178 containing only a rock as a weight to maintain it in the ground and set the cup 5 m east of 179 each control nest (n = 14 in 2016, n = 15 in 2017) to account for any positive or negative 180 bias related to the installation of the cup. We later estimated that the average nest age was 181 8 days (SD = 4.7; n = 57) at the initiation of the experiment. 182 In 2016, we used bownets to capture females on their nest. We banded each female, 183 evaluated their body condition, and measured morphometric dimensions at the beginning 184 of the experiment. We measured mass using a hanging Pesola scale (±0.5 g precision), 185 the lengths of the head, culmen, and tarsus using a caliper (±0.1 mm), and wing length 186 using a ruler (±1 mm). Females were recaptured and weighed towards the end of the 187 incubation period: on average, 10 days after the first capture. Only in 2017, 11 females 188 were captured as part of other protocols. 189

Mass loss models (P1) 190
We calculated a daily rate of mass loss for each incubating female by subtracting their 191 initial weight from their final weight and dividing the result by the number of days 192 between the two captures. Head, culmen, and tarsus lengths were correlated (Pearson's r 193 = 0.82, 0.59, and 0.49) and were combined by extracting the principal axis from their 194 PCA to obtain a metric of bird size. In all analyses, nest cover was converted into a 195 categorical variable with two levels: 0-10% and 20-40%. To test for evidence in support 196 of P1, we identified 8 fixed-effects linear models with mass loss as their response 197 variable and which comprised the main predictor (categorical; supplementation and 198 control) and different covariates (Table S1). To limit the complexity and number of 199 models, we defined concurrent models based on categories of covariates. Three models 200 used environmental covariates: soil humidity (categorical; dry and wet), and nest cover 201 (categorical; 0-10% and 20-40%). Two models used biological covariates: nest initiation 202 date (numeric; Julian days), initial weight (numeric; grams), and size (numeric; principal 203 axis of PCA, see above). We expected the supplementation's effects to be modulated by 204 soil humidity or initiation date, so we included their interaction terms. Finally, two 205 models contained no covariates. 206 Incubation models (P2, P3, and P4) 207 For each dataset of nest temperatures recorded by the thermistors, we computed a moving 208 median and considered that a bird had vacated its nest if temperatures were 3 °C below 209 11 this threshold. We then calculated the average number of recesses per day per nest over 210 the study period (hereafter recess frequency) by dividing the total number of recesses by 211 the total number of minutes of data and multiplying the result by 1,440 minutes/day. We 212 divided the total number of recess minutes by the total number of recesses to obtain the 213 average duration of recesses (hereafter recess duration). Finally, we calculated the 214 proportion of time spent incubating (hereafter nest attendance) by dividing the total 215 number of incubation minutes by the total number of minutes of data. 216 We investigated the relationships between recess frequency and recess duration 217 using type 2 linear regressions. Using type 1 linear models, we examined whether nest 218 attendance could be accurately predicted by either or both previous metrics. To test for 219 evidence in support of P2, P3, and P4, we identified three sets of 6 fixed-effects models 220 with recess frequency, recess duration, and nest attendance, respectively, as their 221 response variable (Tables S1-S2-S3, respectively). Aside from the null model, these 222 models comprised combinations of the main predictor (supplementation) with categories 223 of covariates. Three models contained environmental variables, soil humidity 224 (categorical; dry and wet) and nest cover (categorical; 0-10% and 20-40%), and one 225 model contained the biological variable, nest initiation date (numeric; Julian days). We 226 expected the supplementation's effects to be modulated by soil humidity or initiation 227 date, so we included their interaction terms. Finally, two models contained no covariates. 228

Nesting success models (P5) 229
Nesting success was treated as a binomial variable indicating success (at least one 230 hatched chick) or failure (no hatched chicks) of the incubation. We evaluated nesting 231 success using the RMark package (Laake 2013) to estimate daily survival probabilities 232 over time and compensate for apparent nesting success by considering exposure days 233 (Mayfield 1961(Mayfield , 1975. To test for evidence in support of P5, we identified 10 fixed-234 effects models (Table S5)

Mass loss analyses (P1) 265
White-rumped sandpipers weighed 46.0g (SD = 3.5g, CV = 0.08) and 44.8g (SD = 2.7g, 266 CV = 0.06) on average at the beginning and end of the experiment, respectively. Changes 267 in mass were variable during the experiment: while some females lost up to 12.5 g, others 268 gained up to 6.0 g across the studied period. The top-ranked model explaining variation 269 in mass loss per day (R 2 adjusted = 0.44, Table S1) included the supplementation, initial 270 weight, and their interaction (Table 1) predicts that a 41g female will gain 0.29g/day while a 51g female will lose 0.41g/day. We 277 also found that an extreme value in our dataset (a 57g female that lost 1g/day) was not a 278 leverage point (Cook 1979) as it did not substantially change the outcome of the analysis. 279 Incubation analyses (P2, P3, and P4) 280 Over the course of our experiment, we found that female white-rumped sandpipers 281 incubated for 85. We carried out a supplementation experiment mimicking an increased abundance of 315 arthropods in the Canadian Arctic and found no evidence that white-rumped sandpipers 316 should be affected by additional resources while nesting. Contrary to our predictions, the 317 rate of mass loss per day, nest attendance, and nesting success of incubating females were 318 all unaffected by resource supplementation. This does not support our hypothesis that 319 increases in arthropod productivity during incubation could increase the nesting success 320 of female white-rumped sandpipers by improving their body condition and/or by allowing 321 them to spend more time incubating. However, we found that mass loss per day, nest 322 attendance, and nesting success were likely linked to the initial reproductive traits of 323 incubating females. This suggests that in situ resources may not be limiting for white-324 rumped sandpipers during incubation, yet the strategies that they employ during this 325 period, and the successfulness of the latter, seem to be driven by inter-individual traits 326 upon the initiation of incubation. We will explore how these traits impact breeding as 327 well as why our supplementation experiment had no detectable effects in the following 328 sections. 329

Mass loss (P1) 330
In an experiment carried out on the tropical wintering site of a migratory bird, Cooper et 331 al. (2015) observed that diminished abundances of arthropod prey had no effect on body 332 mass. However, birds that were exposed to food reduction deposited more fat and lost 333 more pectoral muscle mass, leading to delayed departure dates for spring migration 334 The positive linear relationship between weight loss and weight has already been 346 shown in birds (Monaghan et al. 1989) and suggests that mass loss of smaller birds is 347 more constrained physiologically than that of larger birds. Because weight loss increased 348 with initial weight and was unaffected by the size of birds, it is likely that heavier 349 females, who lost up to 1g/day, had more body reserves and did not need to feed as much. 350 On the other hand, leaner females gained weight (up to 0.7g/day) during the incubation 351 period, suggesting that they prioritized feeding. 2017). Our results suggest that nest abandonment became more prevalent as the season 389 progressed even though sandpipers had access to resource supplements. This implies that 390 the quantity of resources during incubation is unlikely a signal to leave the nest before 391 hatching. However, shorebirds face unpredictable abiotic (Tulp and Schekkerman 2006, 392 Smith and Wilson 2010) and biotic (Smith et al. 2012) conditions that may raise 393 incubation costs beyond their limit, forcing them to abandon. The fact that our 394 supplementation had no effect on nest survival indicates that white-rumped sandpipers 395 cannot compensate for these unexpected incubation costs with increased nutritional 396 intake. For the future scenarios of increased productivity in the Arctic, we suggest that an 397 increased peak of arthropods during incubation may not necessarily translate into higher 398 nesting success of these sandpipers. However, they could benefit before (increased 399 fueling) and after (prior to migration, chick survival) incubation.