Feeding – Danger Trade-Offs Underlie Stopover Site Selection by Migrants Compromis alimentation-prédation sous-tendant la sélection d ’ une halte migratoire par des migrateurs

To migrate successfully, birds need to store adequate fat reserves to fuel each leg of the journey. Migrants acquire their fuel reserves at stopover sites; this often entails exposure to predators. Therefore, the safety attributes of sites may be as important as the feeding opportunities. Furthermore, site choice might depend on fuel load, with lean birds more willing to accept danger to obtain good feeding. Here, we evaluate the factors underlying stopover-site usage by migrant Western Sandpipers (Calidris mauri) on a landscape scale. We measured the food and danger attributes of 17 potential stopover sites in the Strait of Georgia and Puget Sound region. We used logistic regression models to test whether food, safety, or both were best able to predict usage of these sites by Western Sandpipers. Eight of the 17 sites were used by sandpipers on migration. Generally, sites that were high in food and safety were used, whereas sites that were low in food and safety were not. However, dangerous sites were used if there was ample food abundance, and sites with low food abundance were used if they were safe. The model including both food and safety best-predicted site usage by sandpipers. Furthermore, lean sandpipers used the most dangerous sites, whereas heavier birds (which do not need to risk feeding in dangerous locations) used safer sites. This study demonstrates that both food and danger attributes are considered by migrant birds when selecting stopover sites, thus both these attributes should be considered to prioritize and manage stopover sites for conservation. RÉSUMÉ. Afin de compléter leur migration, les oiseaux ont besoin d’emmagasiner les réserves énergétiques nécessaires pour franchir chaque étape de leur parcours. Les migrateurs obtiennent leurs réserves énergétiques sur les haltes migratoires, où sont souvent présents des prédateurs. Ainsi, les caractéristiques relatives à la sécurité d’un site peuvent être tout aussi importantes que les possibilités de s’y alimenter. De plus, le choix du site peut dépendre des réserves énergétiques de l’oiseau, un oiseau maigre étant plus disposé à accepter le danger afin d’obtenir une bonne alimentation. Dans la présente étude, nous avons évalué les facteurs sur lesquels repose l’utilisation des haltes migratoires par des Bécasseaux d’Alaska (Calidris mauri) migrateurs, à l’échelle du paysage. Nous avons mesuré les caractéristiques relatives à l’alimentation et à la sécurité de 17 haltes migratoires potentielles dans la région des détroits de Georgie et de Puget. Nous avons utilisé des modèles de régression logistique afin de tester si l’alimentation, la sécurité ou les deux, étaient les variables qui prédisaient le mieux l’utilisation de ces haltes par les Bécasseaux d’Alaska. Huit des 17 sites ont été utilisés par les bécasseaux en migration. En général, les sites sécuritaires et ayant un nombre élevé de proies étaient utilisés, tandis que ceux qui étaient non sécuritaires et dont le nombre de proies était faible ne l’étaient pas. Toutefois, les sites non sécuritaires étaient utilisés s’ils offraient un nombre élevé de proies, et ceux qui n’offraient que peu de proies étaient utilisés s’ils étaient sécuritaires. Le modèle qui incluait à la fois l’alimentation et la sécurité expliquait le mieux l’utilisation des haltes potentielles par les bécasseaux. En outre, les bécasseaux maigres utilisaient les haltes les moins sécuritaires, tandis que les oiseaux plus gras (qui n’ont pas besoin de prendre le risque de s’alimenter dans des haltes non sécuritaires) utilisaient des haltes plus sécuritaires. Cette étude démontre Centre for Wildlife Ecology, Simon Fraser University, Jacques Whitford AXYS, Pacific Wildlife Research Centre, Canadian Wildlife Service Avian Conservation and Ecology Écologie et conservation des oiseaux 3(1): 7 http://www.ace-eco.org/vol3/iss1/art7/ que les caractéristiques relatives à l’alimentation et à la sécurité sont prises en considération par les oiseaux migrateurs au moment de la sélection des haltes migratoires; ces caractéristiques devraient donc toutes deux être considérées afin de prioriser et de gérer les haltes migratoires dans une optique de conservation.

ABSTRACT.To migrate successfully, birds need to store adequate fat reserves to fuel each leg of the journey.Migrants acquire their fuel reserves at stopover sites; this often entails exposure to predators.Therefore, the safety attributes of sites may be as important as the feeding opportunities.Furthermore, site choice might depend on fuel load, with lean birds more willing to accept danger to obtain good feeding.Here, we evaluate the factors underlying stopover-site usage by migrant Western Sandpipers (Calidris mauri) on a landscape scale.We measured the food and danger attributes of 17 potential stopover sites in the Strait of Georgia and Puget Sound region.We used logistic regression models to test whether food, safety, or both were best able to predict usage of these sites by Western Sandpipers.Eight of the 17 sites were used by sandpipers on migration.Generally, sites that were high in food and safety were used, whereas sites that were low in food and safety were not.However, dangerous sites were used if there was ample food abundance, and sites with low food abundance were used if they were safe.The model including both food and safety best-predicted site usage by sandpipers.Furthermore, lean sandpipers used the most dangerous sites, whereas heavier birds (which do not need to risk feeding in dangerous locations) used safer sites.This study demonstrates that both food and danger attributes are considered by migrant birds when selecting stopover sites, thus both these attributes should be considered to prioritize and manage stopover sites for conservation.RÉSUMÉ.Afin de compléter leur migration, les oiseaux ont besoin d'emmagasiner les réserves énergétiques nécessaires pour franchir chaque étape de leur parcours.Les migrateurs obtiennent leurs réserves énergétiques sur les haltes migratoires, où sont souvent présents des prédateurs.Ainsi, les caractéristiques relatives à la sécurité d'un site peuvent être tout aussi importantes que les possibilités de s'y alimenter.De plus, le choix du site peut dépendre des réserves énergétiques de l'oiseau, un oiseau maigre étant plus disposé à accepter le danger afin d'obtenir une bonne alimentation.Dans la présente étude, nous avons évalué les facteurs sur lesquels repose l'utilisation des haltes migratoires par des Bécasseaux d'Alaska (Calidris mauri) migrateurs, à l'échelle du paysage.Nous avons mesuré les caractéristiques relatives à l'alimentation et à la sécurité de 17 haltes migratoires potentielles dans la région des détroits de Georgie et de Puget.Nous avons utilisé des modèles de régression logistique afin de tester si l'alimentation, la sécurité ou les deux, étaient les variables qui prédisaient le mieux l'utilisation de ces haltes par les Bécasseaux d'Alaska.Huit des 17 sites ont été utilisés par les bécasseaux en migration.En général, les sites sécuritaires et ayant un nombre élevé de proies étaient utilisés, tandis que ceux qui étaient non sécuritaires et dont le nombre de proies était faible ne l'étaient pas.Toutefois, les sites non sécuritaires étaient utilisés s'ils offraient un nombre élevé de proies, et ceux qui n'offraient que peu de proies étaient utilisés s'ils étaient sécuritaires.Le modèle qui incluait à la fois l'alimentation et la sécurité expliquait le mieux l'utilisation des haltes potentielles par les bécasseaux.En outre, les bécasseaux maigres utilisaient les haltes les moins sécuritaires, tandis que les oiseaux plus gras (qui n'ont pas besoin de prendre le risque de s'alimenter dans des haltes non sécuritaires) utilisaient des haltes plus sécuritaires.Cette étude démontre

INTRODUCTION
One challenge to understanding the causes underlying population declines of Neotropical migrants is that declines may be caused by a suite of factors operating at various stages of a migrant's life cycle.Most research on factors limiting these populations (i.e., reproduction and mortality) has focused on breeding and non-breeding stages (Robinson et al. 1995, Burke and Nol 1998, Zanette et al. 2000, Rodewald and Yahner 2001, Gill et al. 2001, Latta and Faaborg 2002, Whitfield 2003a, Stillman et al. 2005b), but comparatively less work has been carried out on limiting factors during migration (Hutto 1985, Moore et al. 1995) even though migration appears to be a time of high mortality (Sillett andHolmes 2002, Baker et al. 2004).In light of this, biologists have advocated that the protection of high-quality stopover habitat is essential to conserve migratory bird species (Myers et al. 1987, Skagen and Knopf 1993, Donovan et al. 2002, Mehlman et al. 2005).
Conserving stopover sites requires understanding what makes a site important.Mehlman et al. (2005) propose a conceptual framework to prioritize stopover sites based on their capacity to meet migrants' needs at a given point in space and time and facilitate their successful migration.The authors suggest that criteria to classify and prioritize stopover sites for conservation should include: 1) ecological context (e.g., proximity to ecological barriers, degree of spatial isolation, etc.); 2) intrinsic characteristics (e.g., food abundance, safety from predators); and 3) migrant use (e.g., relative abundance, including frequency and consistency of use as a stopover site).Simply put, to identify appropriate stopover sites for conservation, an understanding of the factors underlying their usage on a broad scale is required.
To migrate successfully, sufficient energy reserves are needed to fuel long-distance flights.In light of the rapid, impressive fuelling by long-distance migrants, high-quality stopover sites are generally considered as those that can provide the required energy.Therefore, much research on stopover-site usage has focused on the abundance and distribution of food (Schneider and Harrington 1981, Hicklin and Smith 1984, Loria and Moore 1990, Colwell and Landrum 1993, Piersma et al. 1994, Baker et al. 2004, Placyk and Harrington 2004).
More recently, investigators have begun to recognize a role for predation danger in shaping where, when, and how quickly migrants travel, and how they select and use stopover sites (Moore 1994, Weber et al. 1998, Clark and Butler 1999, Lank et al. 2003).For example, migrants may adjust their fat loads (Ydenberg et al. 2002, 2004, Schmaljohann and Dierschke 2005) and alter the amount of time spent being vigilant (Cimprich et al. 2005, Pomeroy 2006) with changing levels of predation danger, or may avoid feeding in dangerous places (Sapir et al. 2004, Cimprich et al. 2005, Pomeroy 2006, Pomeroy et al. 2006).Feeding-danger trade-offs are central to decisions made by foragers; in fact, this fundamental trade-off can affect processes ranging from individual fitness to population-and community-level processes (i.e., Cresswell 2008, and references therein).At stopover sites, greater danger makes foragers behave more cautiously, whereas higher food availability makes foragers more accepting of danger.
Although these studies inform us of the attributes that influence within-site habitat selection by migrants, they do not necessarily convey which attributes migrants use to select a site at which to stopover (Hutto 1985).Migrants likely use sitelevel attributes to determine if the site is suitable, and later adjust within-site usage to optimize fuel deposition.Large-scale studies are needed to understand how food and danger affect selection and usage of stopover sites by migrants.Furthermore, the physical attributes or current state http://www.ace-eco.org/vol3/iss1/art7/(fuel load) of individuals might influence their stopover-site selection.For example, lean migrants may be more willing to risk using dangerous sites that offer greater feeding benefits, whereas birds with ample fat reserves might use sites that offer greater safety (e.g., Ydenberg et al. 2002).
Here we designed a landscape-scale field study to assess the food and danger attributes of a variety of potential stopover sites to evaluate the factors underlying their usage by Western Sandpipers (Calidris mauri).We included stopover sites known to be used by Western Sandpipers, as well as sites that appeared suitable but were not used (criteria to determine "used" and "unused" sites are given below).We predicted that if usage of stopover sites depends on either food abundance or safety from predators alone, then sandpipers will use sites with high food abundance or safety and avoid sites with low food abundance or safety.If usage of stopover sites by sandpipers reflects a trade-off between food abundance and predation danger, then we predict that migrant sandpipers will use stopover sites high in both food and safety and not use sites that are low in both food and safety.Furthermore, birds should use dangerous sites only if they offer high food abundance and sites low in food abundance only if they are safe from predators.If the individual state (fuel load) of migrant sandpipers affects stopoversite usage, then we predict that lean birds should be willing to use dangerous sites whereas heavy birds with ample fuel reserves will use only sites that offer safety from predators.Understanding how these factors affect stopover-site usage allows us to classify the importance of sites to prioritize for conservation (cf.Mehlman et al. 2005).

Study System
Western Sandpipers are small (25-30 g on migration) calidrine shorebirds that migrate along the Pacific Coast from non-breeding grounds between northern Oregon and Peru, to breeding grounds in western and northern Alaska and eastern Siberia (Wilson 1994).En route, Western Sandpipers stopover on mud-and sandflats in the Georgia Basin (British Columbia, Canada) and Puget Sound (Washington, USA) regions (Butler et al. 1987, Buchanan 1988, Iverson et al. 1996, Evenson and Buchanan 1997) during mid-April to mid-May on northward migration, and during July (adults) and August (juveniles) on southward migration (Butler et al. 1987).We collected food abundance, safety, and site-use data from 17 mudand sandflats in the Georgia Basin/Puget Sound (GB/PS) region in 2002 (Fig. 1).Three sites, Boundary Bay (BB), Roberts Bank (RB), and Skagit Bay (SB), were sampled in more than one migratory stage; all other sites were only sampled once during either the northward or southward migratory stage (Table 1).

Food Abundance
Western Sandpipers feed on a range of epibenthic and infaunal invertebrates (Wilson 1994, Elner and Seaman 2003, Mathot and Elner 2004).To measure food abundance for Western Sandpipers, we quantified benthic invertebrate abundance in sediment cores, according to the methods described by Sutherland et al. (2000).We collected samples within 500 m of the shoreline using a 60-mL syringe (26 mm inner diameter) to a depth of 40 mm.Samples were collected within 1.5 h of mud exposure by the receding tide and frozen within 1 h of collection.Cores were later thawed and rinsed through a 0.5-mm mesh sieve.Material on the sieve was preserved in vials with 85% ethanol.We identified and counted invertebrates in each vial to the lowest taxon possible under a 40X dissecting microscope.Each taxon sampled has been shown to be consumed by Western Sandpipers (Wolf 2001 and references therein).Wolf (2001) shows that the size range of invertebrates ingested by the Western Sandpiper are effectively sampled by this method, as the sizes of the amphipod Corophium recovered from sandpiper feces were within the size range of those in the core samples.
We collected and measured 15-146 cores at each site during at least one migratory stage (Table 1).Three sites were sampled in multiple migratory stages; invertebrate abundance did not differ among three migratory stages at Boundary Bay (ANOVA: F 2, 143 = 1.80,P = 0.17) or among two stages at Roberts Bank (ANOVA: F 1, 52 = 2.89, P = 0.10).At Skagit Bay, invertebrate abundance differed significantly between northward (mean invertebrate abundance ± 95% confidence interval (C.I.) = 76.5 ± 19.0, N = 30) and southward adult (mean invertebrate abundance ± 95% C.I. = 26.33 ± 5.8, N = 30) migratory stages (F 1, 58 = 25.4,P < 0.0001).1975, Whitfield 1985, Buchanan et al. 1988, Dekker 1988, Cresswell 1994, 1996), and achieve greatest hunting success using cover to conceal their approach (Whitfield 1985, Cresswell 1994, Whitfield 2003b, Dekker and Ydenberg 2004).As the shoreline poses a great deal of danger for foraging sandpipers, stopover sites that are small and have a large proportion of available foraging habitat close to cover are more dangerous than stopover sites that are large where sandpipers can spend large amounts of time feeding on the open mudflat far from the danger lurking along the shoreline (Pomeroy et al. 2006).
As an index of safety, we used the distance tool in ArcMap v.9.

Site Usage
Sites were classified as "used" or "unused" based on the intensity and frequency of use, using as sources published literature, unpublished reports, the Bird Studies Canada Coastal Waterbirds Survey (Bird Studies Canada, unpublished data, 2005) database, and local knowledge (Table 1).Sites were classified as "used" if hundreds or thousands of sandpipers on stopover are recorded there on a regular and annual basis (Butler et al. 1987, 1992, Buchanan 1988, Iverson et al. 1996, Evenson and Buchanan 1997, Warnock and Bishop 1998, Acevedo Seaman et al. 2006).Sites were considered "unused" if there are no records of Western Sandpipers there.A few sites (see Table 1) had single-day records up to 75 birds, but the sites are in fact visited only rarely by western sandpipers, and were therefore classified as "unused." At sites that were used during migration, we used mist nets to capture sandpipers to measure the body mass of individuals using the site (for sample sizes see Table 1).Western Sandpipers were removed from the mist nets immediately after capture and weighed within 10 min.Tarsae were measured using callipers (to the nearest 0.1 mm).

Statistical Analysis
We tested for differences between sites in food abundance and safety using analysis of variance (ANOVA).T-tests were used to test if food abundance and safety differed between used and unused sites.Separate logistic regressions were used to test whether food abundance, safety, or food abundance and safety affected stopover-site usage by sandpipers.
For analysis of body mass, we used analysis of covariance (ANCOVA) including tarsus length as a covariate to account for body size differences between individuals.As northbound migrants were significantly heavier (0.8 g) than southbound migrants (P = 0.05), we also included migratory stage (northward, southward adult, southward juvenile) in the analysis.Body-mass values at each site are reported as the least-squares mean.To investigate the relationship between site safety and state of individuals using a site, we then used these least-squares means of body mass at each site.
Means and 95% C.I. are used throughout.JMPIN V. 4.04 (SAS 2001) was used for all statistical analyses.

DISCUSSION
Our analysis shows that including both food abundance and safety as factors improved our ability to predict which stopover sites were used by migrating sandpipers and which were not.Either factor on its own was also able to predict stopoversite usage, but migrant sandpipers used sites that were dangerous only if food abundance there was high, and used sites with low food abundance only if they were very safe.To our knowledge, no other study has investigated the usage of stopover sites by migrants or compared multiple site attributes at both used and unused sites on a regional scale.Our results also confirm that, on this scale, the usage of stopover sites depends on the fuel load of individual sandpipers, previously shown in a comparison of just two sites by Ydenberg et al. (2002).Sandpipers on dangerous stopover sites carry on average less fuel than those on safer sites.We hypothesize that this occurs because sandpipers with adequate fuel loads do not need to risk feeding in dangerous places.
Our best model incorrectly predicted usage of three of the 17 sites, and an examination of their characteristics can help to understand these assignment errors.Two sites (Doug Banks and Totten Inlet) were classified by the model as "unused," but are in fact used regularly by sandpipers on stopover.The cover surrounding these two sites differs from that lining the shoreline of the other six "used" sites, which consists of low marsh grasses whereas Doug Banks and Totten Inlet are surrounded by tall coniferous trees.From the perspective of a migrant sandpiper, these sites may be functionally safe, as falcons do not appear to use these trees for attack cover, but hunt by attacking over the open mudflat, much like the attack strategy they employ at larger sites (J.Buchanan, pers. comm.).This difference suggests that cover type might be an important factor influencing stopoversite usage by migrants and may be used, in addition to distance from cover, to remotely asses the danger of a potential site.
The third incorrectly classified site was Kye Bay, which, according to the model, should be used by sandpipers due to its high food abundance; however, there is no record of usage by Western Sandpipers.
A noticeable difference between this and all other used stopover sites is the relatively high percentage (60%) of nematodes (Phylum Nematoda, Class Adenophorea) among the potential prey items, compared with a maximum of 46% and an average of 25% nematodes among used sites (Appendix 1).Although all prey taxa presented here have been documented as Western Sandpiper prey (Wolf 2001), there is evidence that nematodes are less preferred by Western Sandpipers than other prey items such as polychaetes (Sutherland et al. 2000).It is possible that prey type may also be an important factor influencing stopover-site selection by migrants and may be used, in addition to prey abundance, to remotely assess the food availability of a potential site.
One possible source of bias in this study is from our measure of food at each site.We measured invertebrate abundance from core samples extracted within 500 m of the shoreline at each site; however, this methodology might not accurately represent food abundance across entire sites.At Boundary Bay, for example, there is a strong declining gradient in food abundance as distance from the shoreline increases (Pomeroy 2006).For large sites where these gradients are likely to occur (Swennen et al. 1982, Reise 1985, Kennish 1986), we might be overestimating food abundances, whereas our measures of food abundance at smaller sites are likely more accurate.Despite this, our analyses remain predictive.
This study indicates that both food and danger attributes of a site influence stopover-site usage by migrants.Stopover-site selection by migrants is thought to be hierarchical (Hutto 1985).To avoid the costs of sampling potential stopover sites, migrants likely use features of the environment to approximate levels of food abundance and predation danger at the site level to decide whether or not to use the site.For example, high densities of invertebrates and biofilm, on which sandpipers feed (Kuwae et al. 2008), are often associated with fine-grained sediments (Kennish 1990, Yates et al. 1993).It is possible that migratory shorebirds can visually assess food abundance of potential stopover sites from the air, e.g., by the reflectance or sediment properties of a beach or mudflat (Rainey et al. 2003, Pomeroy andButler 2005).To remotely assess predation danger at a site, migrants may use proximity to cover as a cue.Large mudflats with a vast expanse of foraging habitats distant from the shoreline are indicative of a relatively safe site for a migrant shorebird, whereas a small mudflat surrounded by marsh grass would be deemed a potentially dangerous place to stop.Upon selecting a stopover site, migrants can more accurately assess the levels of food abundance, and gauge the level of danger there based on their encounters with predators, and they can make appropriate adjustments to anti-predator behaviors to carefully balance food and danger over short temporal and small spatial scales.This work suggests that migrants assess food and danger attributes of stopover sites remotely and that these attributes affect which sites they will use.
Assessment of the conservation value of stopover sites for migrants should evaluate measures of both food and danger.As migration is energetically expensive, stopover-site quality is often associated with the quantity or quality of food resources (Yong et al. 1998, Baker et al. 2004, Placyk and Harrington 2004, Battley et al. 2005, Stillman et al. 2005a, van Gils et al. 2005, Smith et al. 2007).However, migration is also dangerous.As danger from predators increases, migrant birds have been shown to alter their behavior at stopover sites by decreasing length of stay at the site (Ydenberg et al. 2004), allocating more feeding time to vigilance (Cimprich et al. 2005, Pomeroy 2006), and carrying lower fuel loads (Ydenberg et al. 2002, 2004, Schmaljohann and Dierschke 2005).Predation danger influences habitat usage by migrants within and between stopover sites.Assessments of stopover-site quality then must include both food and danger attributes of sites.Furthermore, stopover-site quality is condition dependent.Whereas heavy individuals with ample fuel reserves may prefer safe sites with high bird densities despite few food opportunities, lean birds would be more likely to consider a highquality site as one where it could fatten quickly, despite the added danger.Our results suggest that a variety of site types (i.e., both high food and high safety) need to be conserved.
According to this study, usage of stopover sites by migrants will change if levels of food and/or danger change depending on a) the magnitude of the change, and b) the level of the other attribute.For example, if food abundance declines at a safe site, usage by migrants might not change, whereas if the site is dangerous it might no longer be used.At the landscape scale, increases in predator abundance would likely cause migrants to shift usage from small, dangerous sites to larger, safer ones (e.g., Ydenberg et al. 2004, Taylor et al. 2007).Moreover, changes in stopover-site usage will be condition dependent (Ydenberg et al. 2002(Ydenberg et al. , 2004)).A decrease in food abundance at a dangerous site will have a greater effect on lean birds, whereas increasing danger at safe sites will have a greater effect on heavy birds.If we neglect the role of predators and the effect they have on stopover-site usage, the behavioral adaptations used by migrants to avoid mortality by predation, including increasing vigilance, decreasing length of stay, and adjusting habitat usage, could instead be attributed to declines in food abundance at a site or to population decline (Ydenberg et al. 2004).Efforts to conserve habitat for migrants must consider the role of food abundance and predation danger in stopover-site usage.http://www.ace-eco.org/vol3/iss1/art7/Stopover-site conservation is essential to maintain populations of migratory birds.The importance of identifying, prioritizing, and protecting stopover habitat for birds is gaining increasing recognition (Myers et al. 1987, Donovan et al. 2002, Mehlman et al. 2005, Skagen 2006).Many bird conservation programs identify and prioritize sites of importance based on the number of individuals (or proportion of a population) using a particular site (e.g., Western Hemispher Shorebird Reserve Network (WHSRN), Important Bird Area (IBA)).However, this approach may overemphasize the importance of large (safe) sites that can support significant numbers of birds.Mehlman et al. (2005) assert that efforts to conserve stopover habitat for landbirds should target a mosaic of different types of stopover habitats.The authors suggest three categories of stopover sites representing points on a continuum of their capacity to meet the needs of migrants.These categories range from fire escapes (infrequently used with few available resources, but vital in emergencies) to convenience stores (midquality stopover sites that are used for rest, and replenishment of some fuel, but where food resources might be low and/or predation danger high), to full-service hotels (high-quality sites with abundant resources where migrants can safely rest and refuel for the next leg of migration).In our study system, the sites with ample food and safety would be considered full-service hotels, and sites that are high in either food or safety could be thought of as convenience stores, whereas sites that are low in food and safety may be potential fire escapes.Within each of these categories, the sites can then easily be ranked based on their food and safety attributes, and prioritized accordingly for conservation.
We show here evidence that migrant Western Sandpipers select stopover sites according to tradeoffs between food abundance and predation danger.Furthermore, usage of these sites depends on the state of an individual.This study suggests that migrants use habitat features, such as cover, to assess predation danger, and that they mediate their probability of mortality from predation by adjusting habitat usage on a landscape scale.Results such as these can be applied to predict the behavior of migrants at stopover sites, and usage of stopovers if food and/or danger attributes at a site change.Furthermore, stopover-site usage by migrants depends on state.As danger from predators changes on the landscape, the state of the birds that use those sites might also change (Ydenberg et al. 2004).This study indicates that to identify migration stopover sites for conservation, both food and danger attributes must be considered.

Fig. 1 .
Fig. 1.The Georgia Basin/Puget Sound region of British Columbia, Canada and Washington, USA.
1 (ESRI 2005)  to measure the furthest distance from the shoreline (as indicated by the upper water mark on marine charts (Nautical Data International 2005, National Oceanic and Atmospheric Administration (NOAA) 2005)) in the intertidal zone at each of the sites.For a large open mudflat or beach, this measure is the distance (m) between the shoreline (upper water mark) and the waterline (low water mark) at the widest point.For an enclosed bay surrounded by shoreline, the index of safety is measured as the distance from the shoreline to the midpoint of the widest distance across the bay.

Fig. 2 .
Fig. 2. The mean invertebrate density (± 95% C.I.) and safety index for sites that are used and not used by Western Sandpipers on migration.

Fig. 3 .
Fig. 3.The relationship between the safety index measured at stopover sites and the least-squares mean body mass (± 95% C.I.) of sandpipers captured.The line is that predicted from the nonlinear regression (LS mean body mass = 26.6 * (1-e(-0.0061*safetyindex))).

Table 1 .
The name and location of stopover sites sampled in the Georgia Basin/Puget Sound region, sample sizes for invertebrate cores and body-mass samples, and usage data for each migratory stage (N = northward, SA = southward adult, SJ = southward juvenile, n = number).

Table 2 .
Actual usage at sample sites on migration by sandpipers, and usage as predicted by each logistic regression model.(Y = yes; N = no.)