2.4 Statistical analysis
All statistical analyses were performed in R v. 3.6.1 (R Core Team, 2019). For each season, we estimated differences in δ13C and δ15N values of western sandpiper plasma among demographic groups using general linear models (GLMs). To characterize the diet of western sandpipers in each season, we first estimated differences in δ13C and δ15N values among five groups of prey: biofilm, microphytobenthos, Bivalvia, Polychaeta, and other invertebrates (Clitellata, Gastropoda, and Malacostraca) using a linear discriminant analysis (LDA). Clitellata, Gastropoda, and Malacostraca were grouped into a single prey source because there was substantial overlap in their δ13C and δ15N values. Statistical significance of δ13C and δ15N values in the LDA was determined with forward stepwise selection using Wilks’ Lambda (λ) criterion in the klaR package (Weihs, Ligges, Luebke, & Raabe, 2005). Next, we estimated the proportional contribution of the five prey groups to the diets of western sandpipers using a stable isotope mixing model in the IsotopeR package (Hopkins & Ferguson, 2012). Model runs were conducted with three chains, a burn-in of 100,000 iterations, a run of 100,000 iterations, and a thinning rate of 100. We compared the results of mixing models using discrimination factors for plasma derived from dunlin from two previously published studies: Evans Ogden, Hobson, & Lank (Δ 13C ± standard deviation (SD) = 0.50 ± 0.42 , Δ 15N ± SD = 3.30 ± 0.32 ; 2004) and Lourenço et al. (Δ 13C ± SD = 0.32 ± 0.16 ,Δ 15N ± SD = 3.30 ± 0.20 ; 2015). Discrimination factors were added to the stable isotope values of all prey prior to model runs. Discrimination factors have been shown to vary among different environments, trophic levels, taxa, tissues, metabolic rates, modes of nitrogenous excretion, sample preparations, and food sources with different qualities and quantities of protein (Dalerum & Angerbjörn, 2005; Florin, Felicetti, & Robbins, 2011; McCutchan, Lewis, Kendall, & McGrath, 2003; Robbins, Felicetti, & Florin, 2010; Vanderklift & Ponsard, 2003). To help account for this variation, we incorporated uncertainty in discrimination factors by including standard deviation in our mixing models (Hopkins & Ferguson, 2012).
We compared the diet composition of western sandpipers obtained from the mixing models between seasons, among demographic groups, and among demographic groups within each season using a Permutational Analysis of Variance (PERMANOVA) with 9999 permutations based on Bray-Curtis dissimilarities in the Vegan package (Oksanen et al., 2019). We ensured homogeneity of multivariate dispersion using permutest in Vegan (Oksanen et al., 2019).
Finally, we examined the relationship between morphological traits of individual western sandpipers and diet composition using redundancy analysis (RDA). We estimated the proportion of variance in diet composition explained by two measures of body size, body mass and bill (culmen) length. The RDA was conducted using forward selection with 10,000 permutations using ordistep in Vegan (Oksanen et al., 2019). Differences in body mass and bill length among demographic groups were assessed using GLMs.
RESULTS
Western sandpiper plasma did not differ significantly among demographic groups in δ13C (all p > 0.05) or δ15N (all p > 0.05) in mid-winter. In spring, δ13C (all p > 0.05) values did not differ among demographic groups, but δ15N values were significantly lesser in juvenile males compared to adult females (p = 0.04). In each season, stable isotope signatures differed among the five prey groups (Fig. 1; Appendix Fig. A1). Only δ15N contributed significantly to the linear discriminant function that differentiated prey groups in winter (λ = 0.18, p < 0.001) and spring (λ = 0.17, p < 0.001).
Discrimination factors from Evans Ogden et al. (2004) and Lourenço et al. (2015) were similar, yielding only minor differences in mixing model results, especially in spring (Fig. 2; Appendix Fig. A2). In winter, results from the mixing model that used the discrimination factor of Evans Ogden et al. (2004) indicated that western sandpipers consumed greater proportions of biofilm, Bivalvia, and Polychaeta compared to the results of the mixing model that used the discrimination factor of Lourenço et al. (2015; Fig. 2; Appendix Fig. A2). We present the results of subsequent analyses using the mixing model with the discrimination factor of Evans Ogden et al. (2004); results using the mixing model with the discrimination factor of Lourenço et al. (2015) are presented in Appendix A.
The diet composition of western sandpipers estimated by the mixing model differed significantly between seasons (pseudo-F = 3040.86,R2 = 0.95, p < 0.001), among demographic groups (pseudo-F = 37.04, R2= 0.01, p < 0.001), and among demographic groups within each season (pseudo-F = 26.15, R2 = 0.01, p < 0.001; Appendix Table A2). The difference between seasons explained 95% of the variation in western sandpiper diets, whereas the differences among demographic groups and among demographic groups within each season each explained 1% of the variation in diet. The proportions of biofilm and Bivalvia in western sandpiper diets were similar between seasons (Fig. 2). Western sandpipers from all demographic groups consumed a large proportion of other invertebrates in winter, whereas they consumed greater proportions of Polychaeta and microphytobenthos in spring (Fig. 2). However, the shift from diets dominated by other invertebrates in winter to diets dominated by Polychaeta in spring observed in all demographic groups should be interpreted with caution; this shift could be an artifact of the substantial overlap in stable isotope values between other invertebrates and Polychaeta in spring (Fig. 1). In contrast, the stable isotope values of microphytobenthos had minimal overlap with other prey groups during both seasons (Fig. 1). In winter, the average consumption of biofilm and microphytobenthos combined was approximately 2% of the diet of western sandpipers regardless of demographic group, whereas in spring, biofilm and microphytobenthos comprised 9%, 13%, 19%, and 24% of the diets of adult females, adult males, juvenile females, and juvenile males, respectively. Some juveniles consumed a large proportion of microphytobenthos in spring (Figs. 2-3). The mixing model indicated that the combined contribution of biofilm and microphytobenthos made up greater than 25% of the diet in 13 of 70 (19%) western sandpipers captured in spring; of these 13 individuals, all were juveniles, and 9 (69%) were juvenile males (Fig. 2).
In winter, there was little variation in prey consumption among demographic groups (Fig. 2); therefore, we did not analyze these data with an RDA. In contrast, diet composition differed among demographic groups in spring (Fig. 3; Appendix Fig. A3). Bill length and body mass were both selected as significant variables that explained differences in spring diet composition among western sandpipers (Fig. 3). Together, these variables explained 31% of the total variation (R2adj = 0.31). Bill length and body mass were significantly greater in female western sandpipers compared to males (all p < 0.05), but, within each sex, morphometric measurements did not differ significantly between age classes (all p > 0.05). Males with lesser body mass consumed a greater proportion of other invertebrates compared to females (Fig. 3). Consumption of Polychaeta increased with increasing bill length and was greatest in adult females (Fig. 3). In contrast, consumption of microphytobenthos increased with decreasing bill length and was greater in juveniles, particularly juvenile males (Fig. 3).
DISCUSSION
Our results provide the first evidence of seasonal differences in diet composition among demographic groups of western sandpipers, strongly supporting the conclusion that age- and sex-related dietary specialization facilitate seasonal resource partitioning in this species. In spring, juveniles, particularly juvenile males, consumed more biofilm and microphytobenthos than adults supporting our hypothesis that sandpipers with shorter bills, lesser mass, and less prey handling experience would consume the greatest amount of biofilm and microphytobenthos. Further, differences in diet composition among demographic groups were more pronounced at the onset of migration in spring compared to mid-winter following the expectation that the consumption of biofilm and microphytobenthos provides energy for sandpipers during their breeding migration and reduces competition when high densities of birds occupy foraging sites.