Context-dependent costs and benefits of a heterospecific nesting association

Context-dependent costs and benefits of a heterospecific nesting association Abstract The costs and benefits of interactions among species can vary spatially or temporally, making them context-dependent. For example, benefits associated with nesting near species that deter predators may give way to costs if the association increases the risk of predation during other stages of reproduction. We examined the extent to which the costs and benefits of heterospecific aggregations between a declining shorebird, the Hudsonian Godwit (Limosa haemastica), and a potential protector and predator, the Mew Gull (Larus canus), varied with breeding stage. Specifically, we assessed the spatial distribution and fate of 43 godwit and 262 gull nests in Beluga, Alaska, from 2014 to 2016. We then evaluated the effect of habitat and proximity to gulls on daily survival rates of 120 godwit nests from 2009 to 2016. We also examined the relationship between the proximity to gulls and survival of godwit chicks to 5 days old, the period when they are vulnerable to gull predation. Nests of godwits and gulls were significantly clustered across the landscape, a pattern that habitat heterogeneity failed to explain. Hatching success of godwit nests improved with proximity to the gull colony and increasing numbers of gull nests within 200 m. In contrast, survival of godwit chicks to 5 days improved with increasing distance to the gull colony. The costs and benefits that godwits derived from associating with Mew Gulls were thus context-dependent, with benefits pre-hatch and costs post-hatch. Our results show how spatiotemporal variation in species interactions precludes simple generalizations about the nature of their outcomes. INTRODUCTION Heterospecific associations generally arise when participants benefit from living in groups but avoid the costs of competition (Farine et al. 2014). Benefits from aggregations, such as improved access to food, detection of predators, and/or nest defense, derive not only from a group-size effect, but also from the unique or complementary characteristics of each species (Harrison and Whitehouse 2011; Sridhar et al. 2012). Heterospecific associations are widely documented across taxa, including fish (Lukoschek and Mccormick 2000), amphibians (Phelps et al. 2007), mammals (Querouil et al. 2008), and birds (Sridhar et al. 2009). However, studies of these associations are often restricted to specific periods of the year, such as mixed species foraging flocks that form during the non-breeding season or protective associations occurring during nest incubation in the breeding season (Quinn and Ueta 2008; Sridhar et al. 2009). Understanding the costs and benefits to both the entire assemblage, as well as each species on its own, can inform how these interactions may shift throughout the duration of the association. One type of heterospecific association is a protective nesting association, which occurs when one or more species benefit directly from occupying nesting areas defended from predators by a protector species (Quinn and Ueta 2008). The protected species can derive a number of benefits from these associations including predator protection, information parasitism, reduced effort defending nests, and improved mate attraction. For example, Western Grebes (Aechmophorus occidentalis) react to the alarm calls of Forster’s Terns (Sterna forsteri) by covering eggs prior to departing the nest, and thereby increasing nest survival (Nuechterlein 1981). Of course, benefits for the protected species may vary among protector species. Yellow Warblers (Setophaga petechia) nesting near Gray Catbirds (Dumetella carolinensis), for instance, suffer less predation, whereas those nesting near Red-winged Blackbirds (Agelaius phoeniceus) are parasitized less frequently by Brown-headed Cowbirds (Molothrus ater; Clark and Robertson 1979). At the same time, however, protective associations can incur costs that individuals must try to behaviorally mitigate. For example, Red-breasted Geese (Branta ruficollis), which suffer from direct predation and harassment when nesting near Peregrine Falcons (Falco peregrinus), are able to optimize their fitness by nesting at intermediate distances from falcon nests and thereby minimizing the amount of harassment suffered (Quinn and Kokorev 2002; Quinn and Ueta 2008). Alternatively, some protector species may fail to protect from certain predators. For example, Spotted Sandpipers (Actitis macularia) nest within Common Tern (Sterna hirundo) colonies for protection from minks (Mustela vison), but experience higher egg predation by migrating Ruddy Turnstones (Arenaria interpres), which are attracted to the high density of tern eggs (Alberico et al. 1991). That said, if risk varies predictably with distance to a protector species—even if risk differs across life stages (e.g. adult, egg, chick), individuals still may be able to optimize their decisions (Mönkkönen et al. 2007). The extent to which interactions between species are positive or negative can be a function of the biotic or abiotic context (e.g. “context-dependent interactions”; Chamberlain et al. 2014). For instance, fluctuations in the population size of a predator’s primary prey can drive variation in the magnitude of the pressure predators place on alternative prey (McKinnon et al. 2014). Such scenarios have been reported for heterospecific breeding aggregations. The nesting association between Red Phalaropes (Phalaropus fulicarius) and Sabine’s Gulls (Xema sabini) improves nest success for phalaropes only in years when alternate prey are available for arctic foxes (Vulpes lagopus), one of the main predators of phalarope nests. Sabine’s Gulls are unable to defend against foxes; thus, nesting within the gull colony provides little protection for phalaropes when the abundance of foxes’ primary prey—collard lemmings (Dicrostonyx torquatus)—is low (Smith et al. 2007). Likewise, artificial nests near Long-tailed Skuas (Stercorarius longicaudus) gained no survival advantage because skuas depredated clutches in spite of defending their own nests (Larsen and Grundetjern 1997). In this way, context-dependent interactions can have important consequences for population demography and dynamics. Elucidating how protective associations may change over time can therefore be especially important for uncommon or declining species. The long-distance migratory shorebird, the Hudsonian Godwit (Limosa haemastica, hereafter: “godwits”) is one such species for which its conservation is limited by a poor understanding of the cues used to select breeding habitat (Senner 2010). Godwits breed in sedge bogs that are dominated by muskeg interspersed with small ponds, spruce tree islands, and drier upland areas (Walker et al. 2011; Swift et al. 2017a). Though the occurrence and density of breeding godwits varies widely within and across bogs, godwits appear to form semi-permanent clusters within a subset of suitable breeding areas (Swift et al. 2017b). Interestingly, nesting clusters are probably a result of social cues rather than underlying heterogeneity in vegetation or predation risk (Swift et al. 2017b). Our initial observations suggested that godwits may preferentially nest near Mew Gulls (Larus canus, hereafter: “gulls”), a semi-colonial breeder that forms loud, aggressive defensive flocks whenever predators enter the colony (Moskoff and Bevier 2002; Swift RJ, personal observation). Because godwits seldom defend nests (Walker et al. 2011), they potentially have much to gain from nesting near larids, which are common protector species (Quinn and Ueta 2008). At the same time, godwits may have to balance an important cost—gulls are the main predator of godwit chicks (Senner et al. 2017). In this study, we investigated the degree to which the nests of godwits and gulls were associated and the manner in which costs and benefits of this relationship might vary across different stages of the breeding season. METHODS Study area and species From 2009 to 2011 and 2014 to 2016, we monitored breeding godwits within an ~8 km2 area at Beluga River, Alaska (61.21°N, 151.03°W). The study area was divided into 2 study plots of uninterrupted muskeg bog—North (550 ha) and South (120 ha)—that were separated by ~7 km of unmonitored boreal forest and muskeg bog. From 2014 to 2016, each plot was censused for both godwit and gull nests, although gulls were only partially censused in 2014. Spatial aggregations of godwit nests are not explained by habitat heterogeneity (Swift et al. 2017b). At Beluga River, godwits breed at a density of 5 breeding pairs per square kilometer. Godwits seldom defend nests against gulls or other predators during incubation, instead relying on cryptic camouflage (Walker et al. 2011). Mew Gulls are a common, facultatively colonial breeder in both marine and freshwater habitats (Moskoff and Bevier 2002), with nest densities of 10–40 nests per km2 in our Beluga River study area. Mew Gulls are aggressive toward potential predators, engaging in loud calls and active mobbing. Because they communally defend nests, gull reproductive success correlates with the aggression of a colony (Moskoff and Bevier 2002). Additionally, godwits and gulls nest highly synchronously (nest initiation within 1 day; Swift RJ, unpublished data), despite gulls arriving on the breeding grounds several weeks prior to godwits (eBird 2017). The community of avian and mammalian predators active at Beluga River is diverse though only a small portion of godwit nests are depredated each year (Walker et al. 2011; Senner et al. 2017). The main nest predators are red foxes (Vulpes vulpes), Common Ravens (Corvus corax), and Sandhill Cranes (Grus canadensis). Godwit adults are also prone to Northern Harrier (Circus cyaneus) predation while incubating. Based on anecdotal observations as well as remains of young godwit chicks (i.e. legs, USFWS metal band, and/or radio) found near active gull nests, we believe that gulls are the main predator of young godwit chicks (Senner et al. 2017), though they rarely depredate eggs (Moskoff and Bevier 2002). In addition to gulls, Great Horned Owls (Bubo virginianus), Common Ravens, and red foxes commonly depredate godwit chicks. Nest distribution and fate Once found, nests were marked only with a GPS unit, as we did not physically mark nest locations to avoid associative learning of predator species (Reynolds 1985). For all godwit nests, we estimated hatch date using egg flotation and monitored nests every 2 to 3 days until signs of hatching, after which nests were monitored daily (Liebezeit et al. 2007). We typically checked nests by resighting incubating birds with binoculars from 20 to 30 m away. Adults were flushed weekly (at most) to minimize disturbances that might increase the probability of nest failure, and field teams did not visit nests directly when predators were observed nearby. Although we recorded the locations of all gull nests, only a subset of gull nests were monitored twice weekly. A nest was considered successful if ≥1 egg hatched and chicks successfully left the nest site. We used the presence of young at or near the nest as an indication of nest success. Nest failure was presumed when we found empty nests early in the incubation period and/or destroyed eggs. Due to low rates of nest abandonment and the strong influence of predators on nest survival in this system (Senner et al. 2017), we considered the failure rate of nests in our study to represent the depredation rate as well. Analyses of point patterns Point pattern analyses are the study of the spatial arrangements of points in space, where the datum of interest is the location of the point itself (Diggle 1979, 2003). Point pattern analyses assume a complete census of the study area, and most tests also assume that data are both stationary and isotropic (Fortin and Dale 2005). To comply with the assumption of a complete census, each plot was analyzed separately. Due to consistently small numbers of breeding godwits on the South plot (n = 5 each year), spatial analyses are only reported for the North plot. Multi-type spatial patterns were analyzed only for 2015 and 2016, as not all nests were located in 2014. To test the null hypothesis that godwit and gull nests were distributed randomly within our study plots, we used a combination of first- and second-order multi-type point pattern tests. We imported godwit and gull nest data into program R v.3.4.0 (R Development Core Team 2017) and used the SPATSTAT package for point pattern analysis (Baddeley and Turner 2005). Multi-type tests examine patterns of nest locations between species. Significant associations in first-order nearest neighbor interactions suggest potential local interactions between species from individual nests, which may be indicative of territoriality between species. Significant associations in second-order analyses provide an assessment of potential interactions associated with the total abundance of nests. Evaluations of protective associations among nesting species are more likely to be influenced by the overall abundance of birds rather than the proximity of nearest neighbors, and it is thought that they may be better examined with second-order analyses (Andersen 1992; Diggle 2003). We considered a second-order aggregation of godwit and gull nests as evidence of clustering between species. For godwit and gull nests in each year, we conducted a first-order multi-type G function analysis as a preliminary tool to assess spatial patterns between the 2 species’ nests. For multi-type point patterns, the G function estimated the distribution of the distance from a point of type i to the nearest point of type j, where i and j indicate the 2 species. The G function estimated the nearest neighbor distance distribution function G(r) from a point pattern within a defined window and compared it to the theoretical Poisson process. As our second-order test, we applied multi-type Ripley’s K (Ripley 1976, 1988) to detect spatial randomness at successively larger scales based upon the cumulative distribution function (i.e. the number of additional nests within a distance, r, of a random nest; Baddeley and Turner 2005). For a multi-type point pattern, the multi-type K function counted the expected numbers of points of type j within a given distance of a point of type i. We derived Ripley’s K from the multi-type nest dataset and compared it with the theoretical curve of the Poisson point pattern, which represented complete spatial randomness. We used the linearized form of K, L(r) = (K[r]) – πr2, to aid in interpretation and to stabilize the variance (Besag 1977; Haase 1995). Here, the expected number of nests in an area with radius r is subtracted from K[r], the observed number of nests in a circle with radius r. Under complete spatial randomness, the number of nests in a circle follows a Poisson distribution and L(r) = 0 for all distances. Though Ripley’s K function is widely recognized as a useful tool for detecting spatial aggregations, the cumulative character of this statistic often hampers the detection of scale-dependent patterns (Condit et al. 2000; Schurr et al. 2004). If clumping occurs on a relatively small scale, the point density at larger scales will be above average as well because the increasing circular scales are cumulative. Consequently, we also performed the pair-correlation function (PCF; Ripley 1981; Stoyan and Stoyan 1994), which tests for interactions between points (i.e. nests) separated by a distance r. Unlike Ripley’s K function, which counts all nests contained within a circle, the PCF can be thought of as a circle centered at a given nest, where the only nests counted are those that lie on the circle boundary (i.e. a ring). The PCF is the probability of observing a pair of nests separated by a distance r, divided by the corresponding probability for a Poisson process (Baddeley 2008). Interpretation of the PCF was similar to that of Ripley’s K in that values above the upper bounds of the confidence envelope indicate clustering and those below indicate inhibition. For a multi-type point pattern, the multi-type PCF function examines the probability of finding a point of type i at location x and a point of type j at location y. Lastly, we utilized multi-type Ripley’s K analyses to evaluate whether godwit nest fate was correlated with its spatial positioning relative to gull nests. For 2016 only, we evaluated successful and failed godwit nests separately relative to all gull nests found. We considered a second-order aggregation of successful godwit nests with all gull nests and second-order inhibition between failed godwit nests and all gull nests as evidence in support of the protective-association hypothesis. We compared the observed test statistic, Gij(r), Kij(r), or PCFij(r), against the distribution of Gij(r), Kij(r), or PCFij(r) from 199 permutations of point patterns based on a Poisson process model with the same density as the observed nests (Ripley 1976; Baddeley and Turner 2005). We graphed the confidence envelope to test for significant deviations from complete spatial randomness in each of our analyses. At each distance, observed Gij(r), Kij(r), or PCFij(r) below the confidence envelope indicated significant deviations from complete spatial randomness towards regularity or spatial inhibition. Observed Gij(r), Kij(r), or PCFij(r) above the confidence envelope indicated significant aggregation or clustering. Because variability in user-defined distances for this test can affect the outcome of Ripley’s K, we ran each test using the default range as prescribed by SPATSTAT. The recommended range for the distance lags (r) was 0–852 m for the North plot. We initially performed these tests separately by year to verify that the spatial pattern and location of clusters were comparable among years but then pooled across all 3 years given that our sample sizes were relatively small. Vegetation parameters After godwit nests were no longer active, we measured the habitat at each nest site and a suite of associated random points surrounding the nest. We defined the microhabitat (nest site) scale as the area within a 1-m diameter circle centered on the nest. In each godwit territory, we additionally placed 25 1-m diameter circular plots at randomly selected points. Points were selected from within a 200-m radius of the nest using a random number generator. All points were within the wet sedge dominated bog and study area boundaries. For each circular plot, we measured the distance to the nearest water body (≥ 2 cm deep) from the center of the circle, and within the plot itself, the percent cover for all plant species present. From this, we summarized the percentage of the circle covered by shrubs, sedges, and grasses, and forbs, as well as the percentage of bare ground (water, mud, or rocks). We also summed the number of plant species present in the circular plot as a metric of species richness (see Swift et al. 2017a, b for more information). Vegetation analyses We used Moran’s I test (Moran 1948) to examine if spatial patterning of godwit nest locations was correlated with an underlying spatial pattern in the habitat features used by godwits to choose their nest site. If certain vegetation characteristics drove settlement decisions, then clusters of nests should correspond to patches of especially favorable habitat. We selected focal vegetation parameters based on previous work (Swift et al. 2017a) showing that godwits selected areas with greater numbers of plant species; more sedge/grass; forb, and tall shrubby cover between 30-cm and 1-m tall; less bare ground; and were closer to shallow water than random sites. To reduce the number of variables and tests performed, we used the distribution of the results of a principal component analysis (PCA) using these 6 variables for our Moran’s I tests. To explore spatial autocorrelation, the principal components were tested at 3 different scales using a different number of distance classes (20, 50, 100) in the freely available software SAM (Rangel et al. 2010), with greater numbers of distance classes representing a finer-scale analysis. Each distance class was defined such that an approximately equal number of pairs of points were considered in each distance class. We determined the significance of Moran’s I for each distance class using a randomization procedure with 999 simulations (Fortin and Dale 2005). Vegetation data for nest locations and randomly selected points were analyzed in both a combined dataset and a nests-only dataset. To account for non-independence among distance classes, the significance for each class was assessed using a Bonferroni correction. Moran’s I values were then plotted as a correlogram against k distance classes to aid in interpretation (Fortin and Dale 2005). A significant positive Moran’s I value indicated a patch of similarly structured vegetation; a significant negative value indicated dissimilar vegetation characteristics and was interpreted as a space between patches (Amico et al. 2008). Godwit nest survival We examined the influence of the gull colony and habitat characteristics on godwit nest survival with mark-recapture analyses. Using all gull nests found from 2014 to 2016 combined, we created a minimum convex polygon for each plot that we defined as the gull colony. For each godwit nest, we calculated the minimum distance to the gull colony boundary, the number of gull nests within 200 m, and the minimum distance to the nearest gull nest using ArcGIS (ESRI 2015). We also selected 6 habitat variables known to be used by godwits when choosing their nest sites (Swift et al. 2017a): distance to the closest water body (≥2 cm), % tall shrubby cover (between 30-cm and 1-m tall), % bare ground (water, mud, or rocks), % sedge and grass cover, % herbaceous forb cover, and the number of species within the 1-m circle plot. We used program MARK to estimate daily survival rates (DSRs) of godwit nests in 6 separate analyses. First, we examined the effects of gull proximity and habitat characteristics on nest DSR for 43 nests monitored from 2014 to 2016 (Dinsmore et al. 2002; Rotella et al. 2004). We treated study plot and year as 2 subsets and initially modeled them separately. Within the subsets, we modeled each variable alone as well as in combined habitat and proximity to gulls models. Distance to the gull colony and the nearest gull nest were highly correlated (r2 = 0.86) and were therefore not included together in models. We evaluated models using Akaike’s information criterion corrected for small sample sizes (AICc; Burnham and Anderson 2002), and present beta estimates with standard errors and 95% confidence intervals (CIs). Second, we expanded our analysis to 120 godwit nests found from 2009 to 2011 and 2014 to 2016 and again examined the effects of gull proximity and habitat characteristics on nest DSR. However, because detailed data on gull nests were not collected from 2009 to 2011, our only gull-related metric was the distance to the gull colony boundary, which was presumed to be stable across years. We performed these tests on a combined dataset, by year, and by plot. Godwit chick survival To assess the influence of proximity to the gull colony on the survival of godwit chicks to 5 days-of-age, we radio-tracked a subset of godwit chicks from successfully hatching nests from 2014 to 2016. Generally, gulls are no longer predators of godwit chicks after day 5 when godwit chicks become too large a prey item for gulls and are highly mobile (Senner et al. 2017). We randomly selected 1 or 2 chicks from each brood to receive a small 0.62 g Holohill radio. We clipped the downy feathers from a small area on each chick’s back and attached radios above the uropygial gland with cyanoacrylate glue. We deployed up to 20 radios each year, but not all chicks were located alive within the first 5 days post-hatching. Each chick was located every 2–3 days until the chick had died or fledged. We randomly selected one location for each individual within the first 5 days post-hatch, leading to 29 observations from 25 broods over the 3 years. For each triangulated location, we calculated its distance to the gull colony, distance to the closest gull nest, number of gull nests within 200 m, and distance to the closest pond in ArcGIS (ESRI 2015). We also calculated the distance to the colony for the nest from which the chick hatched. We then used generalized linear mixed models with a logistic regression to examine the influence of gulls on chick survival to day 5, with brood and year as random effects. We evaluated each variable in separate univariate models using AICc scores (Burnham and Anderson 2002) in program R (R Development Core team 2017) with the “lme4” and “bbmle” packages (Bates et al. 2015; Bolker 2017). RESULTS Nest summary We found 43 godwit nests from 2014 to 2016, and 120 godwit nests in total from 2009 to 2016. Of these, 83 godwit nests were found within the gull colony (Figure 1). Daily nest survival was high in each year (>97%) for godwits. Apparent nest success (successful nests/total number of nests) averaged 83% for gull nests (n = 151 nests monitored of 252 located; Figure 1). Figure 1 View largeDownload slide Nest locations, gull colony boundary, and study plot for Hudsonian Godwits and Mew Gulls in Beluga River, Alaska. (a) Hudsonian Godwit nest locations on North plot from 2009 to 2016. (b) Hudsonian Godwit nest locations on South plot from 2009 to 2016. (c) Mew Gull nest locations on North plot from 2014 to 2016. (d) Mew Gull nest locations on South plot from 2014 to 2016. The dashed line shows the Mew Gull colony boundary, and the solid line shows the study plot boundary. Figure 1 View largeDownload slide Nest locations, gull colony boundary, and study plot for Hudsonian Godwits and Mew Gulls in Beluga River, Alaska. (a) Hudsonian Godwit nest locations on North plot from 2009 to 2016. (b) Hudsonian Godwit nest locations on South plot from 2009 to 2016. (c) Mew Gull nest locations on North plot from 2014 to 2016. (d) Mew Gull nest locations on South plot from 2014 to 2016. The dashed line shows the Mew Gull colony boundary, and the solid line shows the study plot boundary. Godwit and gull nests were spatially clustered on the North plot based on second-order tests (Figure 2; Supplementary Figures 2 and 4). Using a nearest neighbor G function, godwit and gull nests were randomly distributed in 2015, 2016, and the combined year dataset (Supplementary Figures 1 and 3). However, our second-order analyses suggested a strong aggregation in both 2015 and 2016, as well as the combined years, based on the comparison of Ripley’s K function with the Poisson point-process null model (Figure 2). Additionally, the PCF test showed similar clustering patterns for 2015, 2016, and the combined years (Supplementary Figures 2 and 4). In 2016, successful godwit nests clustered with all gull nests based on the Ripley’s K test (Supplementary Figure 5a). However, failed godwit nests also were clustered with all gull nests (Supplementary Figure 5b). Figure 2 View largeDownload slide Ripley’s K function (transformed to L(r)) for all Hudsonian Godwit and Mew Gull nests found on North plot in 2015 (a), 2016 (b), and combined year dataset (c). The solid black line represents values for the point pattern (observed), dashed black line represent the expectation under complete spatial randomness (theoretical) of the Poisson null model, and the gray lines represent the confidence envelope based on 199 randomizations of a Poisson point process. Values above the upper bounds of the confidence envelope indicate clustering at distance r, and value below the lower bounds indicated inhibition. Figure 2 View largeDownload slide Ripley’s K function (transformed to L(r)) for all Hudsonian Godwit and Mew Gull nests found on North plot in 2015 (a), 2016 (b), and combined year dataset (c). The solid black line represents values for the point pattern (observed), dashed black line represent the expectation under complete spatial randomness (theoretical) of the Poisson null model, and the gray lines represent the confidence envelope based on 199 randomizations of a Poisson point process. Values above the upper bounds of the confidence envelope indicate clustering at distance r, and value below the lower bounds indicated inhibition. Habitat We performed a principal components analysis on the microhabitat characteristics of godwit nests and associated random points from 2014 to 2016 to reduce habitat variables into a smaller set of principal components (PCs) and also to examine the combined effect of multiple habitat variables. At the microhabitat scale, the first 2 principal components were retained and explained about 55% of the variance. The first principal component (PC1; SD 1.45) described a gradient of vegetation from the number of plant species (positive values) to habitats dominated by sedges and grasses (negative values; Supplementary Table I); the second (PC2; SD 1.08) separated the distance to water (positive) from habitats characterized by forbs (negative). Vegetation attributes varied in the degree to which they were spatially autocorrelated (i.e. patchily distributed; Supplementary Table II). Of the 24 tests of spatial autocorrelation conducted (2 PC variables × 3 distance classes × 2 point subsets (nests and nests + random points) × 2 study plots), 75% (n = 18) yielded no significant autocorrelation (Supplementary Table II). No patchiness was detected within the nest-only dataset. Significant spatial autocorrelation was detected at distances ranging from 47 to 374 m for the North plot, depending on the number of classes and which PC variables were considered, and no spatial autocorrelation was detected for the South plot. Greater levels of patchiness were detected when all points were included than when restricted to nest locations, suggesting that areas surrounding godwit nests were similar in vegetation structure across the study area. Collectively, these results suggest that vegetation patchiness did not drive the spatial pattern of godwit nests. Godwit nest survival Models that included measures of proximity to gull nests better explained godwit nest survival from 2014 to 2016 than did the constant survival model, whereas models with habitat measures had the least explanatory power (Supplementary Table III). Godwit nests were more likely to succeed as distance to the gull colony decreased (β = −0.008; CI = −0.01, −0.003; Figure 3a), and the number of gull nests within 200 m increased (β = 0.29; CI = −0.007, 0.59; Figure 3b)—a pattern that persisted whether nests were grouped by year or plot. Figure 3 View largeDownload slide Daily survival rates of Hudsonian Godwit nests in Beluga River, Alaska. From 2014 to 2016, survival declined with increasing distance to the Mew Gull colony (a) and increased with increasing numbers of Mew Gull nests within 200 m (b). From 2009 to 2016, daily survival rates of Hudsonian Godwit nests declined with increasing distance to the Mew Gull colony (c). Ninety-five percent confidence intervals shown (gray lines). Figure 3 View largeDownload slide Daily survival rates of Hudsonian Godwit nests in Beluga River, Alaska. From 2014 to 2016, survival declined with increasing distance to the Mew Gull colony (a) and increased with increasing numbers of Mew Gull nests within 200 m (b). From 2009 to 2016, daily survival rates of Hudsonian Godwit nests declined with increasing distance to the Mew Gull colony (c). Ninety-five percent confidence intervals shown (gray lines). We tested the influence of microhabitat variables as well as the distance to the gull colony on godwit nest survival with a linear trend for the 2009 to 2016 dataset. The distance to the gull colony again was the top model (wi = 0.47), with most habitat measures falling below the null model (Supplementary Table IV). As the distance to the colony increased, godwit nests were more likely to fail (β = −0.003; CI = −0.006, −0.0009; Figure 3) regardless of whether nests were grouped by plot or year. Godwit chick survival Survival of godwit chicks to day 5 improved with increasing distance to the gull colony (β = 14.29; CI = 3.72, 24.85; Supplementary Table V). Results were similar whether we used our entire sample or randomly selected one chick from each brood. Eight of fifteen godwit chicks that survived the 5-day period moved out of the colony between locations. Seven of 22 (32%) godwit chicks born within the colony survived to day 5 compared to 8 of 13 (62%) godwit chicks born outside the colony. Only 8 of 19 chicks located within the gull colony at any point during the 5-day period survived through this period. Whereas, 7 of 9 chicks located outside the gull colony survived. Nevertheless, godwit chicks moved similar distances per day regardless of whether the chick was located inside or outside the colony (within: 263.6 m, SD = 201.82, n = 27; outside: 293.97 m, SD = 177.78, n = 18). However, godwit chicks that were born within the gull colony that survived moved farther per day then predated chicks (survived: 286.0 m, SD = 220.1, n = 15; predated: 235.6 m, SD = 181.9, n = 12). DISCUSSION Our results confirmed a heterospecific nesting association between Hudsonian Godwits and Mew Gulls in Beluga River, Alaska, but showed that benefits occurred only during the nesting stage when gulls played an indirect protective role. After hatch, the survival of godwit chicks was negatively associated with their proximity to gulls, which are an important chick predator. The association between gulls and godwits was thus context dependent, and godwits appear to optimize their fitness by adopting a strategy to nest within the gull colony but leave it after hatching. Thus, godwits seem to adaptively respond to a landscape where there is both spatial and temporal variation in suitable nesting habitat and predation risk throughout the breeding season (Mönkkönen et al. 2007; Seppänen et al. 2007). Although we found that nest survival of godwits was greater near gull colonies, a full demonstration that the nesting association is protective requires 3 conditions: 1) the ability to recognize potential protectors, 2) active selection of nest sites near protector species rather than simply in similar habitat, and 3) survival benefits exceed the effects of predator swamping (Quinn and Ueta 2008). The association between godwits and gulls meets each of these criteria. First, godwits nested near a species that exhibits loud, defensive behaviors that are easily detected by other species in the community. Protector species are chosen based on both quality and reliability (Larsen and Grundetjern 1997), and they must not affect resource availability for the protected species (Mönkkönen and Forsman 2002). Godwits seem to actively choose to nest near gulls, which are known to nest in association with shorebirds, waterfowl, and jaegers in Europe (Götmark and Andersson 1980; Moskoff and Bevier 2002). Combined, this suggests that godwits recognize gulls as a potential protector species. Second, the association occurs despite differences in microhabitat nesting preferences between the 2 species (Burger and Gochfeld 1988; Swift et al. 2017a). Mew Gull nests were most commonly found on islands in deep snowmelt ponds that had little vegetation (Swift RJ, unpublished data). Although we did not test the spatial distribution or availability of gull breeding habitat, in general, habitat features, such as the pond complexes used by gulls, are randomly distributed across the bog (Swift et al. 2017b). Furthermore, we found that habitat attributes of godwit nest sites (i.e. distance to water, tall shrubby cover) were not major determinants of godwit nest placement or survival; rather, the proximity to the gull colony and the density of nearby gull nests exhibited more influence on godwit nest survival. The aggregation of godwit and gull nests therefore probably has little to do with the spatial distribution of habitats for either gulls or godwits and, instead, is the result of social attraction. Hence, although vegetation characteristics explain some aspect of nesting associations, they do not fully account for the benefits derived from the association (Quinn et al. 2003; Kleindorfer et al. 2009). Third, the density of godwits nesting within gull colonies, though greater than that outside of the colony (average per year: 9.2 nests per km2 inside vs. 1.2 nests per km2 outside), was almost certainly too low to result in predator swamping given the abundance and diversity of predators in the system. However, given that other ground nesting birds (e.g. gulls, shorebirds, waterfowl, and passerines) also reach relatively high densities within gull colonies (average per year: 101.1 nests per km2 inside vs. 12.3 nests per km2 outside), predator swamping may still play a role. The extent to which these densities may result in swamping is thus unclear, as the higher densities may also attract predators to a prey-rich area. Several factors may promote a protective association between gulls and godwits. For protective associations to be effective, the breeding seasons of the 2 species must be synchronous, nest defense must continue throughout the active nesting period, and the protector species must be reliable (Quinn and Ueta 2008). Although we did not directly test whether godwits can differentiate among potential protectors based on their quality, their nesting distribution suggests that they can. One supporting observation is that godwits do not nest near two other defensive larid species, Arctic Terns (Sterna paradisaea) and Bonaparte’s Gulls (Chroicocephalus philadelphia), that arrive later to the breeding grounds and thus do not breed in synchrony with godwits (Swift RJ, unpublished data; eBird 2017). In Beluga River, godwits and gulls initiate their nests at approximately the same time (Swift RJ, unpublished data), but gulls arrive to the breeding grounds several weeks earlier and have established territories prior to godwit arrival (Swift RJ, unpublished data; eBird 2017). Because the gull colony is also relatively stable in size and location, and because gulls have high nest site fidelity (Moskoff and Bevier 2002), godwits may be able to use information from previous breeding seasons when choosing nest locations. Furthermore, from 2014 to 2016, the average initiation date for gull nests was within 1 day of that of godwits (Swift RJ, unpublished data). Highly synchronous nesting and the slightly shorter incubation period of godwits (22–23 days) compared with gulls (23–27 days) translates into earlier hatch dates for godwits—potentially minimizing the threat of gull predation during the vulnerable chick period. Godwits, therefore, may be initiating nests as early as possible after arrival to minimize the risk of nesting within the gull colony, while still actively choosing to aggregate with gulls as a potential protector. Godwits nesting in association with Mew Gulls had 27% higher nest success than did those nesting outside of gull colonies. This benefit is likely the result of active protection from predators by gulls, whereby the defensive behaviors of gulls protect godwit nests when mutual predators such as red foxes and Common Ravens are present in the gull colony. High nest survival is a common benefit of protective associations and has been documented in most known cases, but the mechanism for this protection is typically unknown (Quinn and Ueta 2008). Alternatively, godwits may use the defensive behavior of gulls simply as an early warning system for approaching predators. In this scenario, the gull colony could serve as an “information center,” with godwits acting as potential information parasites, gleaning information that alerts pairs to the presence of predators, and allowing them to engage in cryptic or defensive behaviors, similar to grebes nesting in tern colonies (Nuechterlein 1981; Burger 1984; Doligez et al. 2002). The effectiveness of protective aggregations, by way of deterring predators, may be more strongly driven by colony size than by species composition. Indeed, gulls experience greater nest success in larger colonies, presumably due to effective mobbing behaviors (Götmark and Andersson 1984). Colonies below a certain threshold density could attract predators, but not offer sufficient protection, and thus increase the likelihood of nest failure for both species, creating an ecological trap (Dwernychuk and Boag 1972; Schlaepfer et al. 2002). Accordingly, we detected the strongest effect on nest distributions with second-order tests, and larger numbers of gull nests increased godwit nest survival, suggesting that the gull colony size was important for godwit nest survival. Thus, further study is needed to identify the mechanisms that drive the protective association between nesting godwits and gulls at Beluga River. During the chick stage, the benefits of nesting near gulls gave way to costs, with only 42% of chicks located within the colony surviving to day 5 compared to 70% outside the colony. Although predation is a potential cost of nesting near a protector, few studies have shown that protectors can become predators during different breeding stages. For instance, Eurasian Kestrels (Falco tinnunculus) protect Eurasian Curlew (Numenius arquata) nests, but depredate a small percentage (5%) of curlew chicks each year (Norrdahl et al. 1995). However, curlew chicks are only an incidental prey item of kestrels. Furthermore, large colonies of gulls (over 500 pairs) have been known to completely eliminate cohorts of waterfowl chicks whose parents nested in association with the colony, creating an ecological trap for nesting waterfowl (Dwernychuk and Boag 1972). Therefore, the context-dependent relationship between godwits and gulls may not be unique. Despite the potential dual nature of heterospecific associations, the predictable spatial variation in predation risk from nesting and/or territorial predators can provide protected species with the opportunity to adaptively respond to the changing nature of these relationships (Thomson et al. 2006; Mönkkönen et al. 2007). For instance, individual godwits may be able to compensate for the trade-off between nest and chick predation risk by nesting at intermediate distances or on the edge of the gull colony (Mönkkönen et al. 2007). Alternatively, godwits could compensate for nesting within a stable, risky environment through brood movements, such as by leading their broods to safer habitats outside of dense gull breeding areas. Brood movements that avoid predator-rich or food-poor areas have been well-studied, including with Kentish plover (Charadrius alexandrinus) chicks that show increased survival and growth rates in a non-natal habitat that was food-rich and predator-poor (Kosztolányi et al. 2007). Accordingly, we relocated most godwit broods outside their natal territories and often in areas of the bog with few nesting gulls (Swift RJ, personal observation). Invertebrate prey biomass and habitat attributes vary little across the bog and are therefore unlikely to explain use of these areas (Senner et al. 2017; Swift et al. 2017b). In fact, individuals hatched from nests within and outside the gull colony moved similar distances each day, suggesting biological constraints on the distances moved. Rather, godwits with broods that survived to 5 days moved farther each day than those that were predated. Behavioral responses to nesting in risky environments may thus allow godwits to compensate and increase chick survival despite nesting in close proximity to gulls. Our study thus provides evidence that Hudsonian Godwits benefit from nesting inside the Mew Gull colony through increased hatching success, but bear a cost of lower chick survival due to gull depredation. Based on these findings, we suggest that the nature of interactions between godwits and gulls changes with breeding stage and is, therefore, context-dependent. Our study is among the first to examine the effects of protective associations beyond the nest stage and to document context-dependent interactions based on breeding stage. The costs and benefits of this association are clearly complex, and the lasting benefits (e.g. lifetime fitness) for nesting Hudsonian Godwits associating with Mew Gulls remain unclear and require further study. SUPPLEMENTARY MATERIAL Supplementary data are available at Behavioral Ecology online FUNDING This work was supported by the National Science Foundation (PCE-1110444 to N.R.S. and DGE-1144153 to R.J.S.); U.S. Fish and Wildlife Service (4074 and 5147) to N.R.S.; David and Lucile Packard Foundation to N.R.S.; Faucett Family Foundation to N.R.S. and R.J.S.; Arctic Audubon Society to N.R.S.; American Ornithological Society to N.R.S.; Cornell Lab of Ornithology to N.R.S. and R.J.S.; Athena Fund at the Cornell Lab of Ornithology to N.R.S. and R.J.S.; Arctic Audubon Society to N.R.S., and Cornell University to N.R.S. and R.J.S. Many thanks to numerous field assistants that assisted in data collection and the many colleagues that provided input along the way. All procedures performed in this study involving animals were in accordance with the ethical standards of Cornell University and as part of an approved animal use and care protocol. The authors declare that they have no conflict of interest. Data accessibility: Analyses reported in this article can be reproduced using the data provided by Swift et al. (2018). REFERENCES Alberico JAR , Reed JM , Oring LW . 1991 . Nesting near a Common Tern colony increases and decreases Spotted Sandpiper nest predation . Auk . 108 : 904 – 910 . Amico G , Garcia D , Rodriguez-Cabal MA . 2008 . Spatial structure and scale-dependent microhabitat use of endemic “tapaculos” (Rhinocryptidae) in a temperate forest of southern South America . Ecología Austral . 18 : 169 – 180 . Andersen M . 1992 . Spatial analysis of two-species interactions . Oecologia . 91 : 134 – 140 . Google Scholar CrossRef Search ADS PubMed Baddeley A . 2008 . Analysing spatial point patterns in R . In: Workshop Notes, version 3 . Australia : CSIRO . Available at www.csiro.au/resources/pf16h.html. Baddeley A , Turner R . 2005 . SPATSTAT: An R package for analyzing spatial point patterns . J Stat Softw . 12 : 1 – 42 . Google Scholar CrossRef Search ADS Bates D , Maechler M , Bolker B , Walker S . 2015 . Fitting linear mixed-effects models using lme4 . J Stat Softw . 67 : 1 – 48 . Google Scholar CrossRef Search ADS Besag J . 1977 . Contribution to the discussion of Dr. Ripley’s paper . J R Stat Soc Series B . 39 : 193 – 195 . Bolker B , Team RDC . 2017 . bbmle: Tools for general maximum likelihood estimation . R package version 1.0.20. Burger J . 1984 . Grebes nesting in gull colonies: Protective associations and early warning . Am Nat . 123 : 327 – 337 . Google Scholar CrossRef Search ADS Burger J , Gochfeld M . 1988 . Habitat selection in Mew Gulls: Small colonies and site plasticity . Wilson Bull . 100 : 395 – 410 . Burnham KP , Anderson DR . 2002 . Model selection and multimodel inference: a practical information-theoretic approach . New York: Springer Science & Business Media . Chamberlain SA , Bronstein JL , Rudgers JA . 2014 . How context dependent are species interactions ? Ecol Lett . 17 : 881 – 890 . Google Scholar CrossRef Search ADS PubMed Clark KL , Robertson RJ . 1979 . Spatial and temporal multi-species nesting aggregations in birds as anti-parasite and anti-predator defenses . Behav Ecol Sociobiol . 5 : 359 – 371 . Google Scholar CrossRef Search ADS Condit R , Ashton PS , Baker P , Bunyavejchewin S , Gunatilleke S , Gunatilleke N , Hubbell SP , Foster RB , Itoh A , LaFrankie JV , et al. 2000 . Spatial patterns in the distribution of tropical tree species . Science . 288 : 1414 – 1418 . Google Scholar CrossRef Search ADS PubMed Diggle PJ . 1979 . Statistical methods for spatial point patterns in ecology . Fairland, Maryland : International Cooperative Publishing House . Diggle PJ . 2003 . Statistical analysis of spatial point patterns , 2nd ed . Arnold, London . Dinsmore SJ , White GC , Knopf FL . 2002 . Advanced techniques for modeling avian nest survival . Ecology . 83 : 3476 – 3488 . Google Scholar CrossRef Search ADS Doligez B , Danchin E , Clobert J . 2002 . Public information and breeding habitat selection in a wild bird population . Science . 297 : 1168 – 1170 . Google Scholar CrossRef Search ADS PubMed Dwernychuk LW , Boag DA . 1972 . Ducks nesting in association with gulls—an ecological trap ? Can J Zool . 50 : 559 – 563 . Google Scholar CrossRef Search ADS eBird . 2017 . eBird: An online database of bird distribution and abundance . Ithaca, New York : eBird, Cornell Lab of Ornithology . Available: http://www.ebird.org. ESRI . 2015 . ArcView® 10.3.1 GIS . Redlands, California, USA : Environmental Systems Research Institute Inc .. Farine DR , Downing CP , Downing PA . 2014 . Mixed-species associations can arise without heterospecific attraction . Behav Ecol . 25 : 574 – 581 . Google Scholar CrossRef Search ADS Fortin MJ , Dale MRT . 2005 . Spatial analysis: a guide for ecologists . Cambridge, United Kingdom : Cambridge University Press . Götmark F , Andersson M . 1980 . Breeding association between Common Gull Larus canus and Arctic Skua Stercorarius parasiticus . Ornis Scandinavica . 11 : 121 – 124 . Google Scholar CrossRef Search ADS Götmark F , Andersson M . 1984 . Colonial breeding reduces nest predation in the common gull (Larus canus) . Anim Behav . 32 : 485 – 492 . Google Scholar CrossRef Search ADS Haase P . 1995 . Spatial pattern analysis in ecology based on Ripley’s K-function: Introduction and methods of edge correction . J Veg Sci . 6 : 575 – 582 . Google Scholar CrossRef Search ADS Harrison NM , Whitehouse MJ . 2011 . Mixed-species flocks: an example of niche construction ? Anim Behav . 81 : 675 – 682 . Google Scholar CrossRef Search ADS Kleindorfer S , Sulloway FJ , O’Connor JOD . 2009 . Mixed species nesting associations in Darwin’s tree finches: Nesting pattern predicts predation outcome . Biol J Linn Soc . 98 : 313 – 324 . Google Scholar CrossRef Search ADS Kosztolányi A , Székely T , Cuthill IC . 2007 . The function of habitat change during brood-rearing in the precocial Kentish plover Charadrius alexandrinus . Acta Ethol . 10 : 73 – 79 . Google Scholar CrossRef Search ADS Larsen T , Grundetjern S . 1997 . Optimal choice of neighbour: predator protection among tundra birds . J Avian Biol . 28 : 303 – 308 . Google Scholar CrossRef Search ADS Liebezeit JR , Smith PA , Lanctot RB , Schekkerman H , Tulp I , Kendall SJ , Tracy DM , Rodrigues RJ , Meltofte H , Robinson JA , et al. 2007 . Assessing the development of shorebird eggs using the flotation method: species specific and generalized regression models . Condor . 109 : 32 – 47 . Lukoschek V , Mccormick MI . 2000 . A review of multi-species foraging associations in fishes and their ecological significance . In: Proceeding of the 9th International Coral Reef Symposium; 2000 Oct 23–27; Bali, Indonesia. Vol. 1, 467–474 pp . McKinnon L , Berteaux D , Bêty J . 2014 . Predator-mediated interactions between lemmings and shorebirds: A test of the alternative prey hypothesis . Auk . 131 : 619 – 628 . Google Scholar CrossRef Search ADS Mönkkönen M , Forsman JT . 2002 . Heterospecific attraction among forest birds: a review . Ornithol Sci . 1 : 41 – 51 . Google Scholar CrossRef Search ADS Mönkkönen M , Husby M , Tornberg R , Helle P , Thomson RL . 2007 . Predation as a landscape effect: the trading off by prey species between predation risks and protection benefits . J Anim Ecol . 76 : 619 – 629 . Google Scholar CrossRef Search ADS PubMed Moran PAP . 1948 . The interpretation of statistical maps . J R Stat Soc Series B . 10 : 243 – 251 . Moskoff W , Bevier LR . 2002 . Mew Gull (Larus canus), The birds of North America . In: Rodewald PG , editor. Ithaca : Cornell Lab of Ornithology . Available from: Birds of North America: https://birdsna-org/Species-Account/bna/species/mewgul. Norrdahl K , Suhonen J , Hemminki O , Korpimäki E . 1995 . Predator presence may benefit: kestrels protect curlew nests against nest predators . Oecologia . 101 : 105 – 109 . Google Scholar CrossRef Search ADS PubMed Nuechterlein GL . 1981 . “Information parasitism” in mixed colonies of western grebes and Forster’s terns . Anim Behav . 29 : 985 – 989 . Google Scholar CrossRef Search ADS Phelps SM , Rand AS , Ryan MJ . 2007 . The mixed-species chorus as public information: tungara frogs eavesdrop on a heterospecific . Behav Ecol . 18 : 108 – 114 . Google Scholar CrossRef Search ADS Querouil S , Silva MA , Cascao I , Magalhaes S , Seabra MI , Machete MA , Santos RS . 2008 . Why do dolphins form mixed-species associations in the Azores ? Ethology . 114 : 1183 – 1194 . Google Scholar CrossRef Search ADS Quinn JL , Kokorev Y . 2002 . Trading-off risks from predators and from aggressive hosts . Behav Ecol Sociobiol . 51 : 455 – 460 . Google Scholar CrossRef Search ADS Quinn JL , Prop J , Kokorev Y , Black JM . 2003 . Predator protection or similar habitat selection in red-breasted goose nesting associations: Extremes along a continuum . Anim Behav . 65 : 297 – 307 . Google Scholar CrossRef Search ADS Quinn JL , Ueta M . 2008 . Protective nesting associations in birds . Ibis . 150 : 146 – 167 . Google Scholar CrossRef Search ADS R Development Core Team . 2017 . R: A Language and Environment for Statistical Computing . Vienna, Austria : R Foundation for Statistical Computing . Rangel TF , Diniz-Filho JAF , Bini LM . 2010 . SAM: a comprehensive application for spatial analysis in macroecology . Ecography . 33 : 46 – 50 . Google Scholar CrossRef Search ADS Reynolds JD . 1985 . Sandhill Crane use of nest markers as cues for predation . Wilson Bull . 97 : 106 – 108 . Ripley BD . 1976 . 2nd-order analysis of stationary point processes . J Appl Probab . 13 : 255 – 266 . Google Scholar CrossRef Search ADS Ripley BD . 1981 . Spatial Statistics . New York : Wiley . Google Scholar CrossRef Search ADS Ripley BD . 1988 . Statistical inference for spatial processes . Cambridge, United Kingdom : Cambridge University Press . Google Scholar CrossRef Search ADS Rotella JJ , Dinsmore SJ , Shaffer TL . 2004 . Modeling nest-survival data: A comparison of recently developed methods that can be implemented in MARK and SAS . Anim Biodivers Conserv . 27 : 187 – 205 . Schlaepfer MA , Runge MC , Sherman PW . 2002 . Ecological and evolutionary traps . Trends Ecol Evol . 17 : 474 – 480 . Google Scholar CrossRef Search ADS Schurr FM , Bossdorf O , Milton SJ , Schumacher J . 2004 . Spatial pattern formation in semi-arid shrubland: a priori predicted versus observed pattern characteristics . Plant Ecol . 173 : 271 – 282 . Google Scholar CrossRef Search ADS Senner NR . 2010 . Conservation Plan for the Hudsonian Godwit. Version 1.1 . Manomet, MA : Manomet Center for Conservation Science . Senner NR , Stager M , Sandercock BK . 2017 . Ecological mismatches are moderated by local conditions for two populations of a long-distance migratory bird . Oikos . 126 : 61 – 72 . Google Scholar CrossRef Search ADS Seppänen JT , Forsman JT , Mönkkönen M , Thomson RL . 2007 . Social information use is a process across time, space, and ecology, reaching heterospecifics . Ecology . 88 : 1622 – 1633 . Google Scholar CrossRef Search ADS PubMed Smith PA , Gilchrist HG , Smith JNM , Nol E . 2007 . Annual variation in the benefits of a nesting association between Red Phalaropes (Phalaropus fulicarius) and Sabine’s Gulls (Xema sabini) . Auk . 124 : 276 – 290 . Google Scholar CrossRef Search ADS Sridhar H , Beauchamp G , Shanker K . 2009 . Why do birds participate in mixed-species foraging flocks? A large-scale synthesis . Anim Behav . 78 : 337 – 347 . Google Scholar CrossRef Search ADS Sridhar H , Srinivasan U , Askins RA , Canales-Delgadillo JC , Chen CC , Ewert DN , Gale GA , Goodale E , Gram WK , Hart PJ , et al. 2012 . Positive relationships between association strength and phenotypic similarity characterize the assembly of mixed-species bird flocks worldwide . Am Nat . 180 : 777 – 790 . Google Scholar CrossRef Search ADS PubMed Stoyan D , Stoyan H . 1994 . Fractals, random shapes and point fields: methods of geometrical statistics . New York : Wiley . Swift RJ , Rodewald AD , Senner NR . 2017a . Breeding habitat of a declining shorebird in a changing environment . Polar Biol . 40 : 1777 – 1786 . Google Scholar CrossRef Search ADS Swift RJ , Rodewald AD , Senner NR . 2017b . Environmental heterogeneity and biotic interactions as potential drivers of spatial patterning of shorebird nests . Landsc Ecol . 32 : 1689 – 1703 . Google Scholar CrossRef Search ADS Swift RJ , Rodewald AD , Senner NR . 2018 . Data from: context-dependent costs and benefits of a heterospecific nesting association . Dryad Digital Repository . https://doi.org/10.5061/dryad.m8s2r36. Thomson RL , Forsman JT , Sardà-Palomera F , Mönkkönen M . 2006 . Fear factor: Prey habitat selection and its consequences in a predation risk landscape . Ecography . 29 : 507 – 514 . Google Scholar CrossRef Search ADS Walker BM , Senner NR , Elphick CS , Klima J . 2011 . Hudsonian Godwit (Limosa haemastica), The Birds of North America Online . In: Rodewald PG , editor. Ithaca : Cornell Lab of Ornithology . Available from: http://bna.birds.cornell.edu/bna/species/hudgod. © The Author(s) 2018. Published by Oxford University Press on behalf of the International Society for Behavioral Ecology. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/about_us/legal/notices) http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Behavioral Ecology Oxford University Press

Context-dependent costs and benefits of a heterospecific nesting association

Loading next page...
 
/lp/ou_press/context-dependent-costs-and-benefits-of-a-heterospecific-nesting-tMhOjmjI1I
Publisher
Oxford University Press
Copyright
© The Author(s) 2018. Published by Oxford University Press on behalf of the International Society for Behavioral Ecology. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com
ISSN
1045-2249
eISSN
1465-7279
D.O.I.
10.1093/beheco/ary042
Publisher site
See Article on Publisher Site

Abstract

Abstract The costs and benefits of interactions among species can vary spatially or temporally, making them context-dependent. For example, benefits associated with nesting near species that deter predators may give way to costs if the association increases the risk of predation during other stages of reproduction. We examined the extent to which the costs and benefits of heterospecific aggregations between a declining shorebird, the Hudsonian Godwit (Limosa haemastica), and a potential protector and predator, the Mew Gull (Larus canus), varied with breeding stage. Specifically, we assessed the spatial distribution and fate of 43 godwit and 262 gull nests in Beluga, Alaska, from 2014 to 2016. We then evaluated the effect of habitat and proximity to gulls on daily survival rates of 120 godwit nests from 2009 to 2016. We also examined the relationship between the proximity to gulls and survival of godwit chicks to 5 days old, the period when they are vulnerable to gull predation. Nests of godwits and gulls were significantly clustered across the landscape, a pattern that habitat heterogeneity failed to explain. Hatching success of godwit nests improved with proximity to the gull colony and increasing numbers of gull nests within 200 m. In contrast, survival of godwit chicks to 5 days improved with increasing distance to the gull colony. The costs and benefits that godwits derived from associating with Mew Gulls were thus context-dependent, with benefits pre-hatch and costs post-hatch. Our results show how spatiotemporal variation in species interactions precludes simple generalizations about the nature of their outcomes. INTRODUCTION Heterospecific associations generally arise when participants benefit from living in groups but avoid the costs of competition (Farine et al. 2014). Benefits from aggregations, such as improved access to food, detection of predators, and/or nest defense, derive not only from a group-size effect, but also from the unique or complementary characteristics of each species (Harrison and Whitehouse 2011; Sridhar et al. 2012). Heterospecific associations are widely documented across taxa, including fish (Lukoschek and Mccormick 2000), amphibians (Phelps et al. 2007), mammals (Querouil et al. 2008), and birds (Sridhar et al. 2009). However, studies of these associations are often restricted to specific periods of the year, such as mixed species foraging flocks that form during the non-breeding season or protective associations occurring during nest incubation in the breeding season (Quinn and Ueta 2008; Sridhar et al. 2009). Understanding the costs and benefits to both the entire assemblage, as well as each species on its own, can inform how these interactions may shift throughout the duration of the association. One type of heterospecific association is a protective nesting association, which occurs when one or more species benefit directly from occupying nesting areas defended from predators by a protector species (Quinn and Ueta 2008). The protected species can derive a number of benefits from these associations including predator protection, information parasitism, reduced effort defending nests, and improved mate attraction. For example, Western Grebes (Aechmophorus occidentalis) react to the alarm calls of Forster’s Terns (Sterna forsteri) by covering eggs prior to departing the nest, and thereby increasing nest survival (Nuechterlein 1981). Of course, benefits for the protected species may vary among protector species. Yellow Warblers (Setophaga petechia) nesting near Gray Catbirds (Dumetella carolinensis), for instance, suffer less predation, whereas those nesting near Red-winged Blackbirds (Agelaius phoeniceus) are parasitized less frequently by Brown-headed Cowbirds (Molothrus ater; Clark and Robertson 1979). At the same time, however, protective associations can incur costs that individuals must try to behaviorally mitigate. For example, Red-breasted Geese (Branta ruficollis), which suffer from direct predation and harassment when nesting near Peregrine Falcons (Falco peregrinus), are able to optimize their fitness by nesting at intermediate distances from falcon nests and thereby minimizing the amount of harassment suffered (Quinn and Kokorev 2002; Quinn and Ueta 2008). Alternatively, some protector species may fail to protect from certain predators. For example, Spotted Sandpipers (Actitis macularia) nest within Common Tern (Sterna hirundo) colonies for protection from minks (Mustela vison), but experience higher egg predation by migrating Ruddy Turnstones (Arenaria interpres), which are attracted to the high density of tern eggs (Alberico et al. 1991). That said, if risk varies predictably with distance to a protector species—even if risk differs across life stages (e.g. adult, egg, chick), individuals still may be able to optimize their decisions (Mönkkönen et al. 2007). The extent to which interactions between species are positive or negative can be a function of the biotic or abiotic context (e.g. “context-dependent interactions”; Chamberlain et al. 2014). For instance, fluctuations in the population size of a predator’s primary prey can drive variation in the magnitude of the pressure predators place on alternative prey (McKinnon et al. 2014). Such scenarios have been reported for heterospecific breeding aggregations. The nesting association between Red Phalaropes (Phalaropus fulicarius) and Sabine’s Gulls (Xema sabini) improves nest success for phalaropes only in years when alternate prey are available for arctic foxes (Vulpes lagopus), one of the main predators of phalarope nests. Sabine’s Gulls are unable to defend against foxes; thus, nesting within the gull colony provides little protection for phalaropes when the abundance of foxes’ primary prey—collard lemmings (Dicrostonyx torquatus)—is low (Smith et al. 2007). Likewise, artificial nests near Long-tailed Skuas (Stercorarius longicaudus) gained no survival advantage because skuas depredated clutches in spite of defending their own nests (Larsen and Grundetjern 1997). In this way, context-dependent interactions can have important consequences for population demography and dynamics. Elucidating how protective associations may change over time can therefore be especially important for uncommon or declining species. The long-distance migratory shorebird, the Hudsonian Godwit (Limosa haemastica, hereafter: “godwits”) is one such species for which its conservation is limited by a poor understanding of the cues used to select breeding habitat (Senner 2010). Godwits breed in sedge bogs that are dominated by muskeg interspersed with small ponds, spruce tree islands, and drier upland areas (Walker et al. 2011; Swift et al. 2017a). Though the occurrence and density of breeding godwits varies widely within and across bogs, godwits appear to form semi-permanent clusters within a subset of suitable breeding areas (Swift et al. 2017b). Interestingly, nesting clusters are probably a result of social cues rather than underlying heterogeneity in vegetation or predation risk (Swift et al. 2017b). Our initial observations suggested that godwits may preferentially nest near Mew Gulls (Larus canus, hereafter: “gulls”), a semi-colonial breeder that forms loud, aggressive defensive flocks whenever predators enter the colony (Moskoff and Bevier 2002; Swift RJ, personal observation). Because godwits seldom defend nests (Walker et al. 2011), they potentially have much to gain from nesting near larids, which are common protector species (Quinn and Ueta 2008). At the same time, godwits may have to balance an important cost—gulls are the main predator of godwit chicks (Senner et al. 2017). In this study, we investigated the degree to which the nests of godwits and gulls were associated and the manner in which costs and benefits of this relationship might vary across different stages of the breeding season. METHODS Study area and species From 2009 to 2011 and 2014 to 2016, we monitored breeding godwits within an ~8 km2 area at Beluga River, Alaska (61.21°N, 151.03°W). The study area was divided into 2 study plots of uninterrupted muskeg bog—North (550 ha) and South (120 ha)—that were separated by ~7 km of unmonitored boreal forest and muskeg bog. From 2014 to 2016, each plot was censused for both godwit and gull nests, although gulls were only partially censused in 2014. Spatial aggregations of godwit nests are not explained by habitat heterogeneity (Swift et al. 2017b). At Beluga River, godwits breed at a density of 5 breeding pairs per square kilometer. Godwits seldom defend nests against gulls or other predators during incubation, instead relying on cryptic camouflage (Walker et al. 2011). Mew Gulls are a common, facultatively colonial breeder in both marine and freshwater habitats (Moskoff and Bevier 2002), with nest densities of 10–40 nests per km2 in our Beluga River study area. Mew Gulls are aggressive toward potential predators, engaging in loud calls and active mobbing. Because they communally defend nests, gull reproductive success correlates with the aggression of a colony (Moskoff and Bevier 2002). Additionally, godwits and gulls nest highly synchronously (nest initiation within 1 day; Swift RJ, unpublished data), despite gulls arriving on the breeding grounds several weeks prior to godwits (eBird 2017). The community of avian and mammalian predators active at Beluga River is diverse though only a small portion of godwit nests are depredated each year (Walker et al. 2011; Senner et al. 2017). The main nest predators are red foxes (Vulpes vulpes), Common Ravens (Corvus corax), and Sandhill Cranes (Grus canadensis). Godwit adults are also prone to Northern Harrier (Circus cyaneus) predation while incubating. Based on anecdotal observations as well as remains of young godwit chicks (i.e. legs, USFWS metal band, and/or radio) found near active gull nests, we believe that gulls are the main predator of young godwit chicks (Senner et al. 2017), though they rarely depredate eggs (Moskoff and Bevier 2002). In addition to gulls, Great Horned Owls (Bubo virginianus), Common Ravens, and red foxes commonly depredate godwit chicks. Nest distribution and fate Once found, nests were marked only with a GPS unit, as we did not physically mark nest locations to avoid associative learning of predator species (Reynolds 1985). For all godwit nests, we estimated hatch date using egg flotation and monitored nests every 2 to 3 days until signs of hatching, after which nests were monitored daily (Liebezeit et al. 2007). We typically checked nests by resighting incubating birds with binoculars from 20 to 30 m away. Adults were flushed weekly (at most) to minimize disturbances that might increase the probability of nest failure, and field teams did not visit nests directly when predators were observed nearby. Although we recorded the locations of all gull nests, only a subset of gull nests were monitored twice weekly. A nest was considered successful if ≥1 egg hatched and chicks successfully left the nest site. We used the presence of young at or near the nest as an indication of nest success. Nest failure was presumed when we found empty nests early in the incubation period and/or destroyed eggs. Due to low rates of nest abandonment and the strong influence of predators on nest survival in this system (Senner et al. 2017), we considered the failure rate of nests in our study to represent the depredation rate as well. Analyses of point patterns Point pattern analyses are the study of the spatial arrangements of points in space, where the datum of interest is the location of the point itself (Diggle 1979, 2003). Point pattern analyses assume a complete census of the study area, and most tests also assume that data are both stationary and isotropic (Fortin and Dale 2005). To comply with the assumption of a complete census, each plot was analyzed separately. Due to consistently small numbers of breeding godwits on the South plot (n = 5 each year), spatial analyses are only reported for the North plot. Multi-type spatial patterns were analyzed only for 2015 and 2016, as not all nests were located in 2014. To test the null hypothesis that godwit and gull nests were distributed randomly within our study plots, we used a combination of first- and second-order multi-type point pattern tests. We imported godwit and gull nest data into program R v.3.4.0 (R Development Core Team 2017) and used the SPATSTAT package for point pattern analysis (Baddeley and Turner 2005). Multi-type tests examine patterns of nest locations between species. Significant associations in first-order nearest neighbor interactions suggest potential local interactions between species from individual nests, which may be indicative of territoriality between species. Significant associations in second-order analyses provide an assessment of potential interactions associated with the total abundance of nests. Evaluations of protective associations among nesting species are more likely to be influenced by the overall abundance of birds rather than the proximity of nearest neighbors, and it is thought that they may be better examined with second-order analyses (Andersen 1992; Diggle 2003). We considered a second-order aggregation of godwit and gull nests as evidence of clustering between species. For godwit and gull nests in each year, we conducted a first-order multi-type G function analysis as a preliminary tool to assess spatial patterns between the 2 species’ nests. For multi-type point patterns, the G function estimated the distribution of the distance from a point of type i to the nearest point of type j, where i and j indicate the 2 species. The G function estimated the nearest neighbor distance distribution function G(r) from a point pattern within a defined window and compared it to the theoretical Poisson process. As our second-order test, we applied multi-type Ripley’s K (Ripley 1976, 1988) to detect spatial randomness at successively larger scales based upon the cumulative distribution function (i.e. the number of additional nests within a distance, r, of a random nest; Baddeley and Turner 2005). For a multi-type point pattern, the multi-type K function counted the expected numbers of points of type j within a given distance of a point of type i. We derived Ripley’s K from the multi-type nest dataset and compared it with the theoretical curve of the Poisson point pattern, which represented complete spatial randomness. We used the linearized form of K, L(r) = (K[r]) – πr2, to aid in interpretation and to stabilize the variance (Besag 1977; Haase 1995). Here, the expected number of nests in an area with radius r is subtracted from K[r], the observed number of nests in a circle with radius r. Under complete spatial randomness, the number of nests in a circle follows a Poisson distribution and L(r) = 0 for all distances. Though Ripley’s K function is widely recognized as a useful tool for detecting spatial aggregations, the cumulative character of this statistic often hampers the detection of scale-dependent patterns (Condit et al. 2000; Schurr et al. 2004). If clumping occurs on a relatively small scale, the point density at larger scales will be above average as well because the increasing circular scales are cumulative. Consequently, we also performed the pair-correlation function (PCF; Ripley 1981; Stoyan and Stoyan 1994), which tests for interactions between points (i.e. nests) separated by a distance r. Unlike Ripley’s K function, which counts all nests contained within a circle, the PCF can be thought of as a circle centered at a given nest, where the only nests counted are those that lie on the circle boundary (i.e. a ring). The PCF is the probability of observing a pair of nests separated by a distance r, divided by the corresponding probability for a Poisson process (Baddeley 2008). Interpretation of the PCF was similar to that of Ripley’s K in that values above the upper bounds of the confidence envelope indicate clustering and those below indicate inhibition. For a multi-type point pattern, the multi-type PCF function examines the probability of finding a point of type i at location x and a point of type j at location y. Lastly, we utilized multi-type Ripley’s K analyses to evaluate whether godwit nest fate was correlated with its spatial positioning relative to gull nests. For 2016 only, we evaluated successful and failed godwit nests separately relative to all gull nests found. We considered a second-order aggregation of successful godwit nests with all gull nests and second-order inhibition between failed godwit nests and all gull nests as evidence in support of the protective-association hypothesis. We compared the observed test statistic, Gij(r), Kij(r), or PCFij(r), against the distribution of Gij(r), Kij(r), or PCFij(r) from 199 permutations of point patterns based on a Poisson process model with the same density as the observed nests (Ripley 1976; Baddeley and Turner 2005). We graphed the confidence envelope to test for significant deviations from complete spatial randomness in each of our analyses. At each distance, observed Gij(r), Kij(r), or PCFij(r) below the confidence envelope indicated significant deviations from complete spatial randomness towards regularity or spatial inhibition. Observed Gij(r), Kij(r), or PCFij(r) above the confidence envelope indicated significant aggregation or clustering. Because variability in user-defined distances for this test can affect the outcome of Ripley’s K, we ran each test using the default range as prescribed by SPATSTAT. The recommended range for the distance lags (r) was 0–852 m for the North plot. We initially performed these tests separately by year to verify that the spatial pattern and location of clusters were comparable among years but then pooled across all 3 years given that our sample sizes were relatively small. Vegetation parameters After godwit nests were no longer active, we measured the habitat at each nest site and a suite of associated random points surrounding the nest. We defined the microhabitat (nest site) scale as the area within a 1-m diameter circle centered on the nest. In each godwit territory, we additionally placed 25 1-m diameter circular plots at randomly selected points. Points were selected from within a 200-m radius of the nest using a random number generator. All points were within the wet sedge dominated bog and study area boundaries. For each circular plot, we measured the distance to the nearest water body (≥ 2 cm deep) from the center of the circle, and within the plot itself, the percent cover for all plant species present. From this, we summarized the percentage of the circle covered by shrubs, sedges, and grasses, and forbs, as well as the percentage of bare ground (water, mud, or rocks). We also summed the number of plant species present in the circular plot as a metric of species richness (see Swift et al. 2017a, b for more information). Vegetation analyses We used Moran’s I test (Moran 1948) to examine if spatial patterning of godwit nest locations was correlated with an underlying spatial pattern in the habitat features used by godwits to choose their nest site. If certain vegetation characteristics drove settlement decisions, then clusters of nests should correspond to patches of especially favorable habitat. We selected focal vegetation parameters based on previous work (Swift et al. 2017a) showing that godwits selected areas with greater numbers of plant species; more sedge/grass; forb, and tall shrubby cover between 30-cm and 1-m tall; less bare ground; and were closer to shallow water than random sites. To reduce the number of variables and tests performed, we used the distribution of the results of a principal component analysis (PCA) using these 6 variables for our Moran’s I tests. To explore spatial autocorrelation, the principal components were tested at 3 different scales using a different number of distance classes (20, 50, 100) in the freely available software SAM (Rangel et al. 2010), with greater numbers of distance classes representing a finer-scale analysis. Each distance class was defined such that an approximately equal number of pairs of points were considered in each distance class. We determined the significance of Moran’s I for each distance class using a randomization procedure with 999 simulations (Fortin and Dale 2005). Vegetation data for nest locations and randomly selected points were analyzed in both a combined dataset and a nests-only dataset. To account for non-independence among distance classes, the significance for each class was assessed using a Bonferroni correction. Moran’s I values were then plotted as a correlogram against k distance classes to aid in interpretation (Fortin and Dale 2005). A significant positive Moran’s I value indicated a patch of similarly structured vegetation; a significant negative value indicated dissimilar vegetation characteristics and was interpreted as a space between patches (Amico et al. 2008). Godwit nest survival We examined the influence of the gull colony and habitat characteristics on godwit nest survival with mark-recapture analyses. Using all gull nests found from 2014 to 2016 combined, we created a minimum convex polygon for each plot that we defined as the gull colony. For each godwit nest, we calculated the minimum distance to the gull colony boundary, the number of gull nests within 200 m, and the minimum distance to the nearest gull nest using ArcGIS (ESRI 2015). We also selected 6 habitat variables known to be used by godwits when choosing their nest sites (Swift et al. 2017a): distance to the closest water body (≥2 cm), % tall shrubby cover (between 30-cm and 1-m tall), % bare ground (water, mud, or rocks), % sedge and grass cover, % herbaceous forb cover, and the number of species within the 1-m circle plot. We used program MARK to estimate daily survival rates (DSRs) of godwit nests in 6 separate analyses. First, we examined the effects of gull proximity and habitat characteristics on nest DSR for 43 nests monitored from 2014 to 2016 (Dinsmore et al. 2002; Rotella et al. 2004). We treated study plot and year as 2 subsets and initially modeled them separately. Within the subsets, we modeled each variable alone as well as in combined habitat and proximity to gulls models. Distance to the gull colony and the nearest gull nest were highly correlated (r2 = 0.86) and were therefore not included together in models. We evaluated models using Akaike’s information criterion corrected for small sample sizes (AICc; Burnham and Anderson 2002), and present beta estimates with standard errors and 95% confidence intervals (CIs). Second, we expanded our analysis to 120 godwit nests found from 2009 to 2011 and 2014 to 2016 and again examined the effects of gull proximity and habitat characteristics on nest DSR. However, because detailed data on gull nests were not collected from 2009 to 2011, our only gull-related metric was the distance to the gull colony boundary, which was presumed to be stable across years. We performed these tests on a combined dataset, by year, and by plot. Godwit chick survival To assess the influence of proximity to the gull colony on the survival of godwit chicks to 5 days-of-age, we radio-tracked a subset of godwit chicks from successfully hatching nests from 2014 to 2016. Generally, gulls are no longer predators of godwit chicks after day 5 when godwit chicks become too large a prey item for gulls and are highly mobile (Senner et al. 2017). We randomly selected 1 or 2 chicks from each brood to receive a small 0.62 g Holohill radio. We clipped the downy feathers from a small area on each chick’s back and attached radios above the uropygial gland with cyanoacrylate glue. We deployed up to 20 radios each year, but not all chicks were located alive within the first 5 days post-hatching. Each chick was located every 2–3 days until the chick had died or fledged. We randomly selected one location for each individual within the first 5 days post-hatch, leading to 29 observations from 25 broods over the 3 years. For each triangulated location, we calculated its distance to the gull colony, distance to the closest gull nest, number of gull nests within 200 m, and distance to the closest pond in ArcGIS (ESRI 2015). We also calculated the distance to the colony for the nest from which the chick hatched. We then used generalized linear mixed models with a logistic regression to examine the influence of gulls on chick survival to day 5, with brood and year as random effects. We evaluated each variable in separate univariate models using AICc scores (Burnham and Anderson 2002) in program R (R Development Core team 2017) with the “lme4” and “bbmle” packages (Bates et al. 2015; Bolker 2017). RESULTS Nest summary We found 43 godwit nests from 2014 to 2016, and 120 godwit nests in total from 2009 to 2016. Of these, 83 godwit nests were found within the gull colony (Figure 1). Daily nest survival was high in each year (>97%) for godwits. Apparent nest success (successful nests/total number of nests) averaged 83% for gull nests (n = 151 nests monitored of 252 located; Figure 1). Figure 1 View largeDownload slide Nest locations, gull colony boundary, and study plot for Hudsonian Godwits and Mew Gulls in Beluga River, Alaska. (a) Hudsonian Godwit nest locations on North plot from 2009 to 2016. (b) Hudsonian Godwit nest locations on South plot from 2009 to 2016. (c) Mew Gull nest locations on North plot from 2014 to 2016. (d) Mew Gull nest locations on South plot from 2014 to 2016. The dashed line shows the Mew Gull colony boundary, and the solid line shows the study plot boundary. Figure 1 View largeDownload slide Nest locations, gull colony boundary, and study plot for Hudsonian Godwits and Mew Gulls in Beluga River, Alaska. (a) Hudsonian Godwit nest locations on North plot from 2009 to 2016. (b) Hudsonian Godwit nest locations on South plot from 2009 to 2016. (c) Mew Gull nest locations on North plot from 2014 to 2016. (d) Mew Gull nest locations on South plot from 2014 to 2016. The dashed line shows the Mew Gull colony boundary, and the solid line shows the study plot boundary. Godwit and gull nests were spatially clustered on the North plot based on second-order tests (Figure 2; Supplementary Figures 2 and 4). Using a nearest neighbor G function, godwit and gull nests were randomly distributed in 2015, 2016, and the combined year dataset (Supplementary Figures 1 and 3). However, our second-order analyses suggested a strong aggregation in both 2015 and 2016, as well as the combined years, based on the comparison of Ripley’s K function with the Poisson point-process null model (Figure 2). Additionally, the PCF test showed similar clustering patterns for 2015, 2016, and the combined years (Supplementary Figures 2 and 4). In 2016, successful godwit nests clustered with all gull nests based on the Ripley’s K test (Supplementary Figure 5a). However, failed godwit nests also were clustered with all gull nests (Supplementary Figure 5b). Figure 2 View largeDownload slide Ripley’s K function (transformed to L(r)) for all Hudsonian Godwit and Mew Gull nests found on North plot in 2015 (a), 2016 (b), and combined year dataset (c). The solid black line represents values for the point pattern (observed), dashed black line represent the expectation under complete spatial randomness (theoretical) of the Poisson null model, and the gray lines represent the confidence envelope based on 199 randomizations of a Poisson point process. Values above the upper bounds of the confidence envelope indicate clustering at distance r, and value below the lower bounds indicated inhibition. Figure 2 View largeDownload slide Ripley’s K function (transformed to L(r)) for all Hudsonian Godwit and Mew Gull nests found on North plot in 2015 (a), 2016 (b), and combined year dataset (c). The solid black line represents values for the point pattern (observed), dashed black line represent the expectation under complete spatial randomness (theoretical) of the Poisson null model, and the gray lines represent the confidence envelope based on 199 randomizations of a Poisson point process. Values above the upper bounds of the confidence envelope indicate clustering at distance r, and value below the lower bounds indicated inhibition. Habitat We performed a principal components analysis on the microhabitat characteristics of godwit nests and associated random points from 2014 to 2016 to reduce habitat variables into a smaller set of principal components (PCs) and also to examine the combined effect of multiple habitat variables. At the microhabitat scale, the first 2 principal components were retained and explained about 55% of the variance. The first principal component (PC1; SD 1.45) described a gradient of vegetation from the number of plant species (positive values) to habitats dominated by sedges and grasses (negative values; Supplementary Table I); the second (PC2; SD 1.08) separated the distance to water (positive) from habitats characterized by forbs (negative). Vegetation attributes varied in the degree to which they were spatially autocorrelated (i.e. patchily distributed; Supplementary Table II). Of the 24 tests of spatial autocorrelation conducted (2 PC variables × 3 distance classes × 2 point subsets (nests and nests + random points) × 2 study plots), 75% (n = 18) yielded no significant autocorrelation (Supplementary Table II). No patchiness was detected within the nest-only dataset. Significant spatial autocorrelation was detected at distances ranging from 47 to 374 m for the North plot, depending on the number of classes and which PC variables were considered, and no spatial autocorrelation was detected for the South plot. Greater levels of patchiness were detected when all points were included than when restricted to nest locations, suggesting that areas surrounding godwit nests were similar in vegetation structure across the study area. Collectively, these results suggest that vegetation patchiness did not drive the spatial pattern of godwit nests. Godwit nest survival Models that included measures of proximity to gull nests better explained godwit nest survival from 2014 to 2016 than did the constant survival model, whereas models with habitat measures had the least explanatory power (Supplementary Table III). Godwit nests were more likely to succeed as distance to the gull colony decreased (β = −0.008; CI = −0.01, −0.003; Figure 3a), and the number of gull nests within 200 m increased (β = 0.29; CI = −0.007, 0.59; Figure 3b)—a pattern that persisted whether nests were grouped by year or plot. Figure 3 View largeDownload slide Daily survival rates of Hudsonian Godwit nests in Beluga River, Alaska. From 2014 to 2016, survival declined with increasing distance to the Mew Gull colony (a) and increased with increasing numbers of Mew Gull nests within 200 m (b). From 2009 to 2016, daily survival rates of Hudsonian Godwit nests declined with increasing distance to the Mew Gull colony (c). Ninety-five percent confidence intervals shown (gray lines). Figure 3 View largeDownload slide Daily survival rates of Hudsonian Godwit nests in Beluga River, Alaska. From 2014 to 2016, survival declined with increasing distance to the Mew Gull colony (a) and increased with increasing numbers of Mew Gull nests within 200 m (b). From 2009 to 2016, daily survival rates of Hudsonian Godwit nests declined with increasing distance to the Mew Gull colony (c). Ninety-five percent confidence intervals shown (gray lines). We tested the influence of microhabitat variables as well as the distance to the gull colony on godwit nest survival with a linear trend for the 2009 to 2016 dataset. The distance to the gull colony again was the top model (wi = 0.47), with most habitat measures falling below the null model (Supplementary Table IV). As the distance to the colony increased, godwit nests were more likely to fail (β = −0.003; CI = −0.006, −0.0009; Figure 3) regardless of whether nests were grouped by plot or year. Godwit chick survival Survival of godwit chicks to day 5 improved with increasing distance to the gull colony (β = 14.29; CI = 3.72, 24.85; Supplementary Table V). Results were similar whether we used our entire sample or randomly selected one chick from each brood. Eight of fifteen godwit chicks that survived the 5-day period moved out of the colony between locations. Seven of 22 (32%) godwit chicks born within the colony survived to day 5 compared to 8 of 13 (62%) godwit chicks born outside the colony. Only 8 of 19 chicks located within the gull colony at any point during the 5-day period survived through this period. Whereas, 7 of 9 chicks located outside the gull colony survived. Nevertheless, godwit chicks moved similar distances per day regardless of whether the chick was located inside or outside the colony (within: 263.6 m, SD = 201.82, n = 27; outside: 293.97 m, SD = 177.78, n = 18). However, godwit chicks that were born within the gull colony that survived moved farther per day then predated chicks (survived: 286.0 m, SD = 220.1, n = 15; predated: 235.6 m, SD = 181.9, n = 12). DISCUSSION Our results confirmed a heterospecific nesting association between Hudsonian Godwits and Mew Gulls in Beluga River, Alaska, but showed that benefits occurred only during the nesting stage when gulls played an indirect protective role. After hatch, the survival of godwit chicks was negatively associated with their proximity to gulls, which are an important chick predator. The association between gulls and godwits was thus context dependent, and godwits appear to optimize their fitness by adopting a strategy to nest within the gull colony but leave it after hatching. Thus, godwits seem to adaptively respond to a landscape where there is both spatial and temporal variation in suitable nesting habitat and predation risk throughout the breeding season (Mönkkönen et al. 2007; Seppänen et al. 2007). Although we found that nest survival of godwits was greater near gull colonies, a full demonstration that the nesting association is protective requires 3 conditions: 1) the ability to recognize potential protectors, 2) active selection of nest sites near protector species rather than simply in similar habitat, and 3) survival benefits exceed the effects of predator swamping (Quinn and Ueta 2008). The association between godwits and gulls meets each of these criteria. First, godwits nested near a species that exhibits loud, defensive behaviors that are easily detected by other species in the community. Protector species are chosen based on both quality and reliability (Larsen and Grundetjern 1997), and they must not affect resource availability for the protected species (Mönkkönen and Forsman 2002). Godwits seem to actively choose to nest near gulls, which are known to nest in association with shorebirds, waterfowl, and jaegers in Europe (Götmark and Andersson 1980; Moskoff and Bevier 2002). Combined, this suggests that godwits recognize gulls as a potential protector species. Second, the association occurs despite differences in microhabitat nesting preferences between the 2 species (Burger and Gochfeld 1988; Swift et al. 2017a). Mew Gull nests were most commonly found on islands in deep snowmelt ponds that had little vegetation (Swift RJ, unpublished data). Although we did not test the spatial distribution or availability of gull breeding habitat, in general, habitat features, such as the pond complexes used by gulls, are randomly distributed across the bog (Swift et al. 2017b). Furthermore, we found that habitat attributes of godwit nest sites (i.e. distance to water, tall shrubby cover) were not major determinants of godwit nest placement or survival; rather, the proximity to the gull colony and the density of nearby gull nests exhibited more influence on godwit nest survival. The aggregation of godwit and gull nests therefore probably has little to do with the spatial distribution of habitats for either gulls or godwits and, instead, is the result of social attraction. Hence, although vegetation characteristics explain some aspect of nesting associations, they do not fully account for the benefits derived from the association (Quinn et al. 2003; Kleindorfer et al. 2009). Third, the density of godwits nesting within gull colonies, though greater than that outside of the colony (average per year: 9.2 nests per km2 inside vs. 1.2 nests per km2 outside), was almost certainly too low to result in predator swamping given the abundance and diversity of predators in the system. However, given that other ground nesting birds (e.g. gulls, shorebirds, waterfowl, and passerines) also reach relatively high densities within gull colonies (average per year: 101.1 nests per km2 inside vs. 12.3 nests per km2 outside), predator swamping may still play a role. The extent to which these densities may result in swamping is thus unclear, as the higher densities may also attract predators to a prey-rich area. Several factors may promote a protective association between gulls and godwits. For protective associations to be effective, the breeding seasons of the 2 species must be synchronous, nest defense must continue throughout the active nesting period, and the protector species must be reliable (Quinn and Ueta 2008). Although we did not directly test whether godwits can differentiate among potential protectors based on their quality, their nesting distribution suggests that they can. One supporting observation is that godwits do not nest near two other defensive larid species, Arctic Terns (Sterna paradisaea) and Bonaparte’s Gulls (Chroicocephalus philadelphia), that arrive later to the breeding grounds and thus do not breed in synchrony with godwits (Swift RJ, unpublished data; eBird 2017). In Beluga River, godwits and gulls initiate their nests at approximately the same time (Swift RJ, unpublished data), but gulls arrive to the breeding grounds several weeks earlier and have established territories prior to godwit arrival (Swift RJ, unpublished data; eBird 2017). Because the gull colony is also relatively stable in size and location, and because gulls have high nest site fidelity (Moskoff and Bevier 2002), godwits may be able to use information from previous breeding seasons when choosing nest locations. Furthermore, from 2014 to 2016, the average initiation date for gull nests was within 1 day of that of godwits (Swift RJ, unpublished data). Highly synchronous nesting and the slightly shorter incubation period of godwits (22–23 days) compared with gulls (23–27 days) translates into earlier hatch dates for godwits—potentially minimizing the threat of gull predation during the vulnerable chick period. Godwits, therefore, may be initiating nests as early as possible after arrival to minimize the risk of nesting within the gull colony, while still actively choosing to aggregate with gulls as a potential protector. Godwits nesting in association with Mew Gulls had 27% higher nest success than did those nesting outside of gull colonies. This benefit is likely the result of active protection from predators by gulls, whereby the defensive behaviors of gulls protect godwit nests when mutual predators such as red foxes and Common Ravens are present in the gull colony. High nest survival is a common benefit of protective associations and has been documented in most known cases, but the mechanism for this protection is typically unknown (Quinn and Ueta 2008). Alternatively, godwits may use the defensive behavior of gulls simply as an early warning system for approaching predators. In this scenario, the gull colony could serve as an “information center,” with godwits acting as potential information parasites, gleaning information that alerts pairs to the presence of predators, and allowing them to engage in cryptic or defensive behaviors, similar to grebes nesting in tern colonies (Nuechterlein 1981; Burger 1984; Doligez et al. 2002). The effectiveness of protective aggregations, by way of deterring predators, may be more strongly driven by colony size than by species composition. Indeed, gulls experience greater nest success in larger colonies, presumably due to effective mobbing behaviors (Götmark and Andersson 1984). Colonies below a certain threshold density could attract predators, but not offer sufficient protection, and thus increase the likelihood of nest failure for both species, creating an ecological trap (Dwernychuk and Boag 1972; Schlaepfer et al. 2002). Accordingly, we detected the strongest effect on nest distributions with second-order tests, and larger numbers of gull nests increased godwit nest survival, suggesting that the gull colony size was important for godwit nest survival. Thus, further study is needed to identify the mechanisms that drive the protective association between nesting godwits and gulls at Beluga River. During the chick stage, the benefits of nesting near gulls gave way to costs, with only 42% of chicks located within the colony surviving to day 5 compared to 70% outside the colony. Although predation is a potential cost of nesting near a protector, few studies have shown that protectors can become predators during different breeding stages. For instance, Eurasian Kestrels (Falco tinnunculus) protect Eurasian Curlew (Numenius arquata) nests, but depredate a small percentage (5%) of curlew chicks each year (Norrdahl et al. 1995). However, curlew chicks are only an incidental prey item of kestrels. Furthermore, large colonies of gulls (over 500 pairs) have been known to completely eliminate cohorts of waterfowl chicks whose parents nested in association with the colony, creating an ecological trap for nesting waterfowl (Dwernychuk and Boag 1972). Therefore, the context-dependent relationship between godwits and gulls may not be unique. Despite the potential dual nature of heterospecific associations, the predictable spatial variation in predation risk from nesting and/or territorial predators can provide protected species with the opportunity to adaptively respond to the changing nature of these relationships (Thomson et al. 2006; Mönkkönen et al. 2007). For instance, individual godwits may be able to compensate for the trade-off between nest and chick predation risk by nesting at intermediate distances or on the edge of the gull colony (Mönkkönen et al. 2007). Alternatively, godwits could compensate for nesting within a stable, risky environment through brood movements, such as by leading their broods to safer habitats outside of dense gull breeding areas. Brood movements that avoid predator-rich or food-poor areas have been well-studied, including with Kentish plover (Charadrius alexandrinus) chicks that show increased survival and growth rates in a non-natal habitat that was food-rich and predator-poor (Kosztolányi et al. 2007). Accordingly, we relocated most godwit broods outside their natal territories and often in areas of the bog with few nesting gulls (Swift RJ, personal observation). Invertebrate prey biomass and habitat attributes vary little across the bog and are therefore unlikely to explain use of these areas (Senner et al. 2017; Swift et al. 2017b). In fact, individuals hatched from nests within and outside the gull colony moved similar distances each day, suggesting biological constraints on the distances moved. Rather, godwits with broods that survived to 5 days moved farther each day than those that were predated. Behavioral responses to nesting in risky environments may thus allow godwits to compensate and increase chick survival despite nesting in close proximity to gulls. Our study thus provides evidence that Hudsonian Godwits benefit from nesting inside the Mew Gull colony through increased hatching success, but bear a cost of lower chick survival due to gull depredation. Based on these findings, we suggest that the nature of interactions between godwits and gulls changes with breeding stage and is, therefore, context-dependent. Our study is among the first to examine the effects of protective associations beyond the nest stage and to document context-dependent interactions based on breeding stage. The costs and benefits of this association are clearly complex, and the lasting benefits (e.g. lifetime fitness) for nesting Hudsonian Godwits associating with Mew Gulls remain unclear and require further study. SUPPLEMENTARY MATERIAL Supplementary data are available at Behavioral Ecology online FUNDING This work was supported by the National Science Foundation (PCE-1110444 to N.R.S. and DGE-1144153 to R.J.S.); U.S. Fish and Wildlife Service (4074 and 5147) to N.R.S.; David and Lucile Packard Foundation to N.R.S.; Faucett Family Foundation to N.R.S. and R.J.S.; Arctic Audubon Society to N.R.S.; American Ornithological Society to N.R.S.; Cornell Lab of Ornithology to N.R.S. and R.J.S.; Athena Fund at the Cornell Lab of Ornithology to N.R.S. and R.J.S.; Arctic Audubon Society to N.R.S., and Cornell University to N.R.S. and R.J.S. Many thanks to numerous field assistants that assisted in data collection and the many colleagues that provided input along the way. All procedures performed in this study involving animals were in accordance with the ethical standards of Cornell University and as part of an approved animal use and care protocol. The authors declare that they have no conflict of interest. Data accessibility: Analyses reported in this article can be reproduced using the data provided by Swift et al. (2018). REFERENCES Alberico JAR , Reed JM , Oring LW . 1991 . Nesting near a Common Tern colony increases and decreases Spotted Sandpiper nest predation . Auk . 108 : 904 – 910 . Amico G , Garcia D , Rodriguez-Cabal MA . 2008 . Spatial structure and scale-dependent microhabitat use of endemic “tapaculos” (Rhinocryptidae) in a temperate forest of southern South America . Ecología Austral . 18 : 169 – 180 . Andersen M . 1992 . Spatial analysis of two-species interactions . Oecologia . 91 : 134 – 140 . Google Scholar CrossRef Search ADS PubMed Baddeley A . 2008 . Analysing spatial point patterns in R . In: Workshop Notes, version 3 . Australia : CSIRO . Available at www.csiro.au/resources/pf16h.html. Baddeley A , Turner R . 2005 . SPATSTAT: An R package for analyzing spatial point patterns . J Stat Softw . 12 : 1 – 42 . Google Scholar CrossRef Search ADS Bates D , Maechler M , Bolker B , Walker S . 2015 . Fitting linear mixed-effects models using lme4 . J Stat Softw . 67 : 1 – 48 . Google Scholar CrossRef Search ADS Besag J . 1977 . Contribution to the discussion of Dr. Ripley’s paper . J R Stat Soc Series B . 39 : 193 – 195 . Bolker B , Team RDC . 2017 . bbmle: Tools for general maximum likelihood estimation . R package version 1.0.20. Burger J . 1984 . Grebes nesting in gull colonies: Protective associations and early warning . Am Nat . 123 : 327 – 337 . Google Scholar CrossRef Search ADS Burger J , Gochfeld M . 1988 . Habitat selection in Mew Gulls: Small colonies and site plasticity . Wilson Bull . 100 : 395 – 410 . Burnham KP , Anderson DR . 2002 . Model selection and multimodel inference: a practical information-theoretic approach . New York: Springer Science & Business Media . Chamberlain SA , Bronstein JL , Rudgers JA . 2014 . How context dependent are species interactions ? Ecol Lett . 17 : 881 – 890 . Google Scholar CrossRef Search ADS PubMed Clark KL , Robertson RJ . 1979 . Spatial and temporal multi-species nesting aggregations in birds as anti-parasite and anti-predator defenses . Behav Ecol Sociobiol . 5 : 359 – 371 . Google Scholar CrossRef Search ADS Condit R , Ashton PS , Baker P , Bunyavejchewin S , Gunatilleke S , Gunatilleke N , Hubbell SP , Foster RB , Itoh A , LaFrankie JV , et al. 2000 . Spatial patterns in the distribution of tropical tree species . Science . 288 : 1414 – 1418 . Google Scholar CrossRef Search ADS PubMed Diggle PJ . 1979 . Statistical methods for spatial point patterns in ecology . Fairland, Maryland : International Cooperative Publishing House . Diggle PJ . 2003 . Statistical analysis of spatial point patterns , 2nd ed . Arnold, London . Dinsmore SJ , White GC , Knopf FL . 2002 . Advanced techniques for modeling avian nest survival . Ecology . 83 : 3476 – 3488 . Google Scholar CrossRef Search ADS Doligez B , Danchin E , Clobert J . 2002 . Public information and breeding habitat selection in a wild bird population . Science . 297 : 1168 – 1170 . Google Scholar CrossRef Search ADS PubMed Dwernychuk LW , Boag DA . 1972 . Ducks nesting in association with gulls—an ecological trap ? Can J Zool . 50 : 559 – 563 . Google Scholar CrossRef Search ADS eBird . 2017 . eBird: An online database of bird distribution and abundance . Ithaca, New York : eBird, Cornell Lab of Ornithology . Available: http://www.ebird.org. ESRI . 2015 . ArcView® 10.3.1 GIS . Redlands, California, USA : Environmental Systems Research Institute Inc .. Farine DR , Downing CP , Downing PA . 2014 . Mixed-species associations can arise without heterospecific attraction . Behav Ecol . 25 : 574 – 581 . Google Scholar CrossRef Search ADS Fortin MJ , Dale MRT . 2005 . Spatial analysis: a guide for ecologists . Cambridge, United Kingdom : Cambridge University Press . Götmark F , Andersson M . 1980 . Breeding association between Common Gull Larus canus and Arctic Skua Stercorarius parasiticus . Ornis Scandinavica . 11 : 121 – 124 . Google Scholar CrossRef Search ADS Götmark F , Andersson M . 1984 . Colonial breeding reduces nest predation in the common gull (Larus canus) . Anim Behav . 32 : 485 – 492 . Google Scholar CrossRef Search ADS Haase P . 1995 . Spatial pattern analysis in ecology based on Ripley’s K-function: Introduction and methods of edge correction . J Veg Sci . 6 : 575 – 582 . Google Scholar CrossRef Search ADS Harrison NM , Whitehouse MJ . 2011 . Mixed-species flocks: an example of niche construction ? Anim Behav . 81 : 675 – 682 . Google Scholar CrossRef Search ADS Kleindorfer S , Sulloway FJ , O’Connor JOD . 2009 . Mixed species nesting associations in Darwin’s tree finches: Nesting pattern predicts predation outcome . Biol J Linn Soc . 98 : 313 – 324 . Google Scholar CrossRef Search ADS Kosztolányi A , Székely T , Cuthill IC . 2007 . The function of habitat change during brood-rearing in the precocial Kentish plover Charadrius alexandrinus . Acta Ethol . 10 : 73 – 79 . Google Scholar CrossRef Search ADS Larsen T , Grundetjern S . 1997 . Optimal choice of neighbour: predator protection among tundra birds . J Avian Biol . 28 : 303 – 308 . Google Scholar CrossRef Search ADS Liebezeit JR , Smith PA , Lanctot RB , Schekkerman H , Tulp I , Kendall SJ , Tracy DM , Rodrigues RJ , Meltofte H , Robinson JA , et al. 2007 . Assessing the development of shorebird eggs using the flotation method: species specific and generalized regression models . Condor . 109 : 32 – 47 . Lukoschek V , Mccormick MI . 2000 . A review of multi-species foraging associations in fishes and their ecological significance . In: Proceeding of the 9th International Coral Reef Symposium; 2000 Oct 23–27; Bali, Indonesia. Vol. 1, 467–474 pp . McKinnon L , Berteaux D , Bêty J . 2014 . Predator-mediated interactions between lemmings and shorebirds: A test of the alternative prey hypothesis . Auk . 131 : 619 – 628 . Google Scholar CrossRef Search ADS Mönkkönen M , Forsman JT . 2002 . Heterospecific attraction among forest birds: a review . Ornithol Sci . 1 : 41 – 51 . Google Scholar CrossRef Search ADS Mönkkönen M , Husby M , Tornberg R , Helle P , Thomson RL . 2007 . Predation as a landscape effect: the trading off by prey species between predation risks and protection benefits . J Anim Ecol . 76 : 619 – 629 . Google Scholar CrossRef Search ADS PubMed Moran PAP . 1948 . The interpretation of statistical maps . J R Stat Soc Series B . 10 : 243 – 251 . Moskoff W , Bevier LR . 2002 . Mew Gull (Larus canus), The birds of North America . In: Rodewald PG , editor. Ithaca : Cornell Lab of Ornithology . Available from: Birds of North America: https://birdsna-org/Species-Account/bna/species/mewgul. Norrdahl K , Suhonen J , Hemminki O , Korpimäki E . 1995 . Predator presence may benefit: kestrels protect curlew nests against nest predators . Oecologia . 101 : 105 – 109 . Google Scholar CrossRef Search ADS PubMed Nuechterlein GL . 1981 . “Information parasitism” in mixed colonies of western grebes and Forster’s terns . Anim Behav . 29 : 985 – 989 . Google Scholar CrossRef Search ADS Phelps SM , Rand AS , Ryan MJ . 2007 . The mixed-species chorus as public information: tungara frogs eavesdrop on a heterospecific . Behav Ecol . 18 : 108 – 114 . Google Scholar CrossRef Search ADS Querouil S , Silva MA , Cascao I , Magalhaes S , Seabra MI , Machete MA , Santos RS . 2008 . Why do dolphins form mixed-species associations in the Azores ? Ethology . 114 : 1183 – 1194 . Google Scholar CrossRef Search ADS Quinn JL , Kokorev Y . 2002 . Trading-off risks from predators and from aggressive hosts . Behav Ecol Sociobiol . 51 : 455 – 460 . Google Scholar CrossRef Search ADS Quinn JL , Prop J , Kokorev Y , Black JM . 2003 . Predator protection or similar habitat selection in red-breasted goose nesting associations: Extremes along a continuum . Anim Behav . 65 : 297 – 307 . Google Scholar CrossRef Search ADS Quinn JL , Ueta M . 2008 . Protective nesting associations in birds . Ibis . 150 : 146 – 167 . Google Scholar CrossRef Search ADS R Development Core Team . 2017 . R: A Language and Environment for Statistical Computing . Vienna, Austria : R Foundation for Statistical Computing . Rangel TF , Diniz-Filho JAF , Bini LM . 2010 . SAM: a comprehensive application for spatial analysis in macroecology . Ecography . 33 : 46 – 50 . Google Scholar CrossRef Search ADS Reynolds JD . 1985 . Sandhill Crane use of nest markers as cues for predation . Wilson Bull . 97 : 106 – 108 . Ripley BD . 1976 . 2nd-order analysis of stationary point processes . J Appl Probab . 13 : 255 – 266 . Google Scholar CrossRef Search ADS Ripley BD . 1981 . Spatial Statistics . New York : Wiley . Google Scholar CrossRef Search ADS Ripley BD . 1988 . Statistical inference for spatial processes . Cambridge, United Kingdom : Cambridge University Press . Google Scholar CrossRef Search ADS Rotella JJ , Dinsmore SJ , Shaffer TL . 2004 . Modeling nest-survival data: A comparison of recently developed methods that can be implemented in MARK and SAS . Anim Biodivers Conserv . 27 : 187 – 205 . Schlaepfer MA , Runge MC , Sherman PW . 2002 . Ecological and evolutionary traps . Trends Ecol Evol . 17 : 474 – 480 . Google Scholar CrossRef Search ADS Schurr FM , Bossdorf O , Milton SJ , Schumacher J . 2004 . Spatial pattern formation in semi-arid shrubland: a priori predicted versus observed pattern characteristics . Plant Ecol . 173 : 271 – 282 . Google Scholar CrossRef Search ADS Senner NR . 2010 . Conservation Plan for the Hudsonian Godwit. Version 1.1 . Manomet, MA : Manomet Center for Conservation Science . Senner NR , Stager M , Sandercock BK . 2017 . Ecological mismatches are moderated by local conditions for two populations of a long-distance migratory bird . Oikos . 126 : 61 – 72 . Google Scholar CrossRef Search ADS Seppänen JT , Forsman JT , Mönkkönen M , Thomson RL . 2007 . Social information use is a process across time, space, and ecology, reaching heterospecifics . Ecology . 88 : 1622 – 1633 . Google Scholar CrossRef Search ADS PubMed Smith PA , Gilchrist HG , Smith JNM , Nol E . 2007 . Annual variation in the benefits of a nesting association between Red Phalaropes (Phalaropus fulicarius) and Sabine’s Gulls (Xema sabini) . Auk . 124 : 276 – 290 . Google Scholar CrossRef Search ADS Sridhar H , Beauchamp G , Shanker K . 2009 . Why do birds participate in mixed-species foraging flocks? A large-scale synthesis . Anim Behav . 78 : 337 – 347 . Google Scholar CrossRef Search ADS Sridhar H , Srinivasan U , Askins RA , Canales-Delgadillo JC , Chen CC , Ewert DN , Gale GA , Goodale E , Gram WK , Hart PJ , et al. 2012 . Positive relationships between association strength and phenotypic similarity characterize the assembly of mixed-species bird flocks worldwide . Am Nat . 180 : 777 – 790 . Google Scholar CrossRef Search ADS PubMed Stoyan D , Stoyan H . 1994 . Fractals, random shapes and point fields: methods of geometrical statistics . New York : Wiley . Swift RJ , Rodewald AD , Senner NR . 2017a . Breeding habitat of a declining shorebird in a changing environment . Polar Biol . 40 : 1777 – 1786 . Google Scholar CrossRef Search ADS Swift RJ , Rodewald AD , Senner NR . 2017b . Environmental heterogeneity and biotic interactions as potential drivers of spatial patterning of shorebird nests . Landsc Ecol . 32 : 1689 – 1703 . Google Scholar CrossRef Search ADS Swift RJ , Rodewald AD , Senner NR . 2018 . Data from: context-dependent costs and benefits of a heterospecific nesting association . Dryad Digital Repository . https://doi.org/10.5061/dryad.m8s2r36. Thomson RL , Forsman JT , Sardà-Palomera F , Mönkkönen M . 2006 . Fear factor: Prey habitat selection and its consequences in a predation risk landscape . Ecography . 29 : 507 – 514 . Google Scholar CrossRef Search ADS Walker BM , Senner NR , Elphick CS , Klima J . 2011 . Hudsonian Godwit (Limosa haemastica), The Birds of North America Online . In: Rodewald PG , editor. Ithaca : Cornell Lab of Ornithology . Available from: http://bna.birds.cornell.edu/bna/species/hudgod. © The Author(s) 2018. Published by Oxford University Press on behalf of the International Society for Behavioral Ecology. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/about_us/legal/notices)

Journal

Behavioral EcologyOxford University Press

Published: Apr 5, 2018

There are no references for this article.

You’re reading a free preview. Subscribe to read the entire article.


DeepDyve is your
personal research library

It’s your single place to instantly
discover and read the research
that matters to you.

Enjoy affordable access to
over 18 million articles from more than
15,000 peer-reviewed journals.

All for just $49/month

Explore the DeepDyve Library

Search

Query the DeepDyve database, plus search all of PubMed and Google Scholar seamlessly

Organize

Save any article or search result from DeepDyve, PubMed, and Google Scholar... all in one place.

Access

Get unlimited, online access to over 18 million full-text articles from more than 15,000 scientific journals.

Your journals are on DeepDyve

Read from thousands of the leading scholarly journals from SpringerNature, Elsevier, Wiley-Blackwell, Oxford University Press and more.

All the latest content is available, no embargo periods.

See the journals in your area

DeepDyve

Freelancer

DeepDyve

Pro

Price

FREE

$49/month
$360/year

Save searches from
Google Scholar,
PubMed

Create lists to
organize your research

Export lists, citations

Read DeepDyve articles

Abstract access only

Unlimited access to over
18 million full-text articles

Print

20 pages / month

PDF Discount

20% off