Postbreeding elevational movements of western songbirds in Northern California and Southern Oregon

Abstract Migratory species employ a variety of strategies to meet energetic demands of postbreeding molt. As such, at least a few species of western Neotropical migrants are known to undergo short‐distance upslope movements to locations where adults molt body and flight feathers (altitudinal molt migration). Given inherent difficulties in measuring subtle movements of birds occurring in western mountains, we believe that altitudinal molt migration may be a common yet poorly documented phenomenon. To examine prevalence of altitudinal molt migration, we used 29 years of bird capture data in a series of linear mixed‐effect models for nine commonly captured species that breed in northern California and southern Oregon. Candidate models were formulated a priori to examine whether elevation and distance from the coast can be used to predict abundance of breeding and molting birds. Our results suggest that long‐distance migrants such as Orange‐crowned Warbler (Oreothlypis celata) moved higher in elevation and Audubon's Warbler (Setophaga coronata) moved farther inland to molt after breeding. Conversely, for resident and short‐distance migrants, we found evidence that birds either remained on the breeding grounds until they finished molting, such as Song Sparrow (Melospiza melodia) or made small downslope movements, such as American Robin (Turdus migratorius). We conclude that altitudinal molt migration may be a common, variable, and complex behavior among western songbird communities and is related to other aspects of a species’ natural history, such as migratory strategy.


| INTRODUCTION
Long-distance molt migration is a mechanism by which migratory species deal with the energetic demands of the postbreeding molt (definitive prebasic molt sensu Wolfe et al., 2014) by moving to seasonally food-rich environments to replace their body and flight feathers (Pyle et al., 2009). Even at smaller spatial scales, resident and facultative migratory birds must acquire the dispersed and seasonal food resources necessary for successful completion of postbreeding molt (Daan et al., 1988;Murphy & King, 1992). Seasonal food resources are particularly patchy in mountainous areas where insect and fruit abundance can vary dramatically across relatively short distances (Thomas, 2005). As such, altitudinal molt migration should be expected in many species that breed and molt in montane areas.
Molt is an energetically costly process necessary for the maintenance of feathers and plumage. As such, birds may suffer from a limited capacity to thermoregulate (Schieltz & Murphy, 1997) or sustain flight (Hedenström & Sunada, 1999) during molt. These limitations make birds more susceptible to the deleterious effects of inclement weather, or the inability to escape predators. The relatively early timing of Western Sandpiper (Calidris mauri) molt migration has been suggested as an adaptation to avoid a common migratory predator, the Peregrine Falcon (Falco peregrinus, Lank et al. 2003). The vulnerability of molting birds to predation and inclement weather may result in increased mortality, which in-turn can affect population viability (Swaddle & Witter, 1997). In addition to potential direct demographic effects, nonlethal events during the molting season may have indirect carry-over effects on other phases of the avian life cycle, such as breeding (Slagsvold & Dale, 1996). As such, the timing of molt is most likely highly adaptive and particularly susceptible to changes driven by natural selection. For example, western and eastern populations of Warbling Vireos (Vireo gilvus) molt on the winter and summer grounds, respectively (Pyle, 1997); such differences presumably reflect local adaptation and aid in the successful completion of molt across longitudes. To better understand selective pressures responsible for differences in molt strategies and the influence of lethal and nonlethal effects experienced during molt on population viability, we first need to determine when and where birds molt, and identify those landscape features associated with molt.
To date, few studies have endeavored to associate landscape features with altitudinal molt migration in the western United States.
However, the limited number of studies that examined altitudinal molt migration in the western United States suggests a general pattern of upslope movements after breeding to undergo molt. For example, Rowher, Rowher, and Barry (2008) used point counts conducted in the spring and fall and found that Cassin's Vireo (Vireo cassinii) were more abundant at higher elevations during the fall molting season than in the spring. Steele and McCormick (1995) captured birds at several different elevations in the Sierra Nevada Mountains of California and observed adults leaving the breeding grounds at lower elevations and moving to higher elevation sites, where they had not been captured during the breeding season, to undergo molt. Presumably, many birds move upslope to wet montane meadows during the molting season to take advantage of insect food resources (Van Dyke, 1919).
To examine relationships between landscape features and postbreeding movements prior to molt, we used data from a network of banding stations for species known to molt on their summer grounds.
Specifically, we examined relationships between abundances of breeding and molting birds and landscape features such as elevation and distance from coast. In total, we examined nine of the most commonly captured species in northern California and southern Oregon.
Studies suggest that higher elevation habitats retain more moisture, relative to lower elevations, during hot and dry late-summer and early fall periods throughout our study area (Patton & Judd, 1970;Robinson et al., 2013). We suspect that insect abundance, an important food resource for molting birds, is strongly correlated with moisture during these hot and dry periods (sensu Van Dyke, 1919). We formulated three a priori hypotheses regarding movements of western passerines between periods of breeding and molting. (i) We hypothesize that some species of western passerines move across elevations after the breeding season to complete molt. If our first hypothesis is correct, we expect that there will be higher abundances of molting birds at higher elevations and further from the coast when compared to abundances of breeding birds. (ii) Long-distance migrants are physiologically adapted to move great distances; therefore, we hypothesize that many long-distant migrant species have evolved to seek far-off habitats with greater resources during postbreeding molt. If this is true, we expect to find spatially disparate populations of breeding and molting long-distant migrant birds throughout our study area. (iii) Conversely, we hypothesize that resident birds are less equipped to make postbreeding movements, which would result in less-distant and adjacent populations of breeding and molting resident species throughout our study area.

| METHODS
To test the aforementioned hypotheses, we measured differences in the abundance of breeding and molting birds relative to migratory guild, elevation, and distance from coast using mist-netting data from 82 stations in the Klamath-Siskiyou Bioregion of northern California and southern Oregon from 1982 to 2011 (Figure 1, Appendix A). Each station was operated from 2 to 27 years, and the average length each station was operated is 9.2 years. We operated two banding stations year-round; all others were operated from April through October.
Each banding station was scheduled at least once every 10 days during months of operation. Each station had eight to 15 net sites that were opened 15 min prior to sunrise and operated for 5-6 hr during each sampling day. For complete information on the banding methodology, see Alexander, Ralph, Hollinger, and Hogoboom (2004). Birds were aged and sexed following Pyle (1997), and other morphometrics were obtained following Ralph, Geupel, Pyle, Martin, and Desante (1993). Study species were chosen based on their abundance and diversity of migratory behaviors. For some species, (e.g., Song Sparrow) it was difficult to determine whether populations in our study area are resident, short-distance, or altitudinal migrants. Therefore, we lumped study species into two distinct migratory groups: resident/ short-distance migrants and long-distance Neotropical migrants. To measure changes in the abundance of breeding birds and those in postbreeding molt, we included only adult birds aged as after hatching year according to Pyle (1997a). We classified individuals as breeding if they had vascularized or wrinkled brood patches or cloacal protuberances that were medium or large (following Ralph et al., 1993). We classified individuals as molting if they were captured growing flight feathers symmetrically. We removed recaptures of the same individual within a season to avoid pseudoreplication. We standardized abundance of breeding and molting birds by individual captures per 100 net hours per station.
For each species, we created two candidate model sets of 16 zeroinflated linear regression models with a Poisson distribution using R (R Core Team 2015) and the glmmADMB package (Fournier et al., 2012). The response variables for the two candidate model sets were either the standardized abundance of breeding birds of a particular species or the standardized abundance of molting birds of a species.
Due to the nature of the banding stations being operated some years and not others and the various durations each station was run, we felt that modeling species abundance by each sampling day coupled with a zero-inflated Poisson distribution was the best way to overcome unequal sampling effort while still maintaining our large sample size. Explanatory variables were elevation (meters), quadratic elevation (meters), distance from coast (kilometers), and quadratic distance from coast (kilometers). Julian date (day of the year) was included in every model as a nuisance parameter. Furthermore, we delineated the study area and aggregated banding stations into 10 regions based on similarities in geography (Table 1). We compared the inclusion of a random intercept effect associated with region to see whether it improved  These regions were used for the random intercept effects in candidate models. A complete list of the stations and associated region can be found in Appendix A.
We ranked all candidate models using Akaike Information Criterion (AIC) and interpreted results based on the inclusion of explanatory variables. We also examined explanatory variable beta estimates and associated 95% confidence intervals to assess effect sizes of covariates on abundance of breeding and molting study species. Ranking models based on their AIC scores allowed us to estimate a type I error (false positive) rate for each candidate model where the highest false positive rate among top models for the 18 candidate model sets was 0.035; the overall mean type I error rate for top models was 0.006. We found colinearity between elevation and distance from coast in our dataset; therefore we did not include both covariates in any candidate single model. For top models that included a quadratic covariate as an explanatory variable, we maximized the abundance predicted by the quadratic model to determine at what elevation or distance from coast breeding and molting birds were at their highest abundance.
Post hoc, we examined each study species' timing of the definitive prebasic molt to be certain that months of station operation (May to October) captured the breadth of breeding and molting activity. We have two banding stations (HOME and WIWI) that were run year around and would bias our results if study species were found breeding or molting before May or after October at these stations.

| RESULTS
During the course of the study, we captured 37,886 breeding and postbreeding molting adult birds of the nine study species (Table 2).
Of these, Song Sparrow (n = 12,877) and Orange-crowned Warbler (n = 1,552) were our most common and least common study species, respectively.

| DISCUSSION
Four of the nine study species exhibited greater abundances of molting birds at higher elevations and further from the coast when compared to breeding season abundances. These results support our first hypothesis that birds commonly move to higher elevations to molt following breeding. We also found support for our second and third hypotheses that long-distant migrants move greater distances to molt following breeding when compared to resident/short-distant migrant birds.
Candidate models that included quadratic terms ranked better  (Vanderhoff, Sallabanks, & James, 2014). Wheelwright (1986) found that American Robins consume an approximately even mix of invertebrates and fruits during the summer, but during the fall and winter their diets become more frugivorous. Blackberries and other fruit-bearing plants are thought to flower and fruit sooner at higher elevations and further from the coast (Sallabanks, 1993