A link between water availability and nesting success mediated by predator prey interactions in the Arctic

Transcription

1 Ecology, 90(2), 2009, pp Ó 2009 by the Ecological Society of America A link between water availability and nesting success mediated by predator prey interactions in the Arctic NICOLAS LECOMTE, 1,3 GILLES GAUTHIER, 1 AND JEAN-FRANC OIS GIROUX 2 1 De partement de Biologie and Centre d E tudes Nordiques, Universite Laval, Que bec, Canada 2 Groupe de Recherche en E cologie Comportementale et Animale, De partement des Sciences Biologiques, Universite du Que bec a` Montre al, Que bec, Canada Abstract. Although water availability is primarily seen as a factor affecting food availability (a bottom-up process), we examined its effect on predator prey interactions through an influence on prey behavior (a top-down process). We documented a link between water availability, predation risk, and reproductive success in a goose species (Chen caerulescens atlantica) inhabiting an Arctic environment where water is not considered a limited commodity. To reach water sources during incubation recesses, geese nesting in mesic tundra (low water availability) must move almost four times as far from their nest than those nesting in wetlands, which reduced their ability to defend their nest against predators and led to a higher predation rate. Nesting success was improved in high rainfall years due to increased water availability, and more so for geese nesting in the low water availability habitat. Likewise, nesting success was improved in years where the potential for evaporative water loss (measured by the atmospheric water vapor pressure) was low, presumably because females had to leave their nest less often to drink. Females from water-supplemented nests traveled a shorter distance to drink, and their nesting success was enhanced by 20% compared to the control. This shows that water availability and rainfall can have a strong effect on predator prey dynamics and that changes in precipitation brought by climate change could have an impact on some Arctic species through a top-down effect. Key words: Arctic-nesting Snow Geese; Bylot Island, Nunavut, Canada; Chen caerulescens; incubation; predator prey interactions; resource limitation; top-down control; water availability and limitation. INTRODUCTION Water can be a limiting factor for ecophysiological processes (e.g., Bartholomew and Cade 1963, Klaassen 2004, Williams and Tieleman 2005). However, water availability can also affect demographic processes such as survival and reproduction (Newton 1998, Tieleman et al. 2004). For instance, increased water levels and flooding can lead to a complete breeding failure and reduced survival in ground-breeding species (e.g., Myers et al. 1985, Thompson and Furness 1991, Ratcliffe et al. 2005). Reproductive success can also be positively linked with the amount of precipitation because of increased water availability at the onset of the breeding period (Coe and Rotenberry 2003, Ladia et al. 2007). This link is usually assumed to be mediated through an increase in food availability rather than water per se (bottom-up control; Blancher and Robertson 1987, Gibbs and Grant 1987, Ogutu et al. 2008). Water availability can also affect predator prey interactions (top-down control; Ostfeld and Keesing 2000). In arid ecosystems, where water limitations are Manuscript received 31 January 2008; revised 17 June 2008; accepted 23 June Corresponding Editor: E. G. Cooch. 3 Present address: Department of Biology, University of Tromsø, N-9037, Tromsø, Norway. nicolas.lecomte@ib.uit.no 465 ubiquitous, individuals are distributed according to availability of surface water, often trading off their nutritionnal requirements against water use (Redfern et al. 2003, Tielman et al. 2004). The concentration of individuals at water holes can considerably increase their vulnerability to predation (Ferns and Hinsley 1995, Doody et al. 2007, Valeix et al. 2007). However, the effect of water availability on predation prey interactions has been largely ignored in environments where water is not usually considered limiting (Morrison and Bolger 2002). Yet, recent evidence suggests that the spatial distribution of water or precipitation regimes can affect predator efficiency, with potentially strong consequences for predator prey dynamics and ecosystem processes (e.g., Hilton et al. 1999, Post et al. 1999, Lecomte et al. 2008a). Although prey behavior could also be affected by water availability and rainfall, we know very little about how it can influence predation risk, and hence predator prey dynamics, in non-arid environments. In birds, nest predation can have a considerable impact on reproductive success (Martin 1995). In species where only one sex incubates, incubating birds must balance the conflicting needs of foraging and protecting their nest because, when they leave to feed, predation risk of unattended eggs often increases (Thompson and Raveling 1987, Samelius and Alisauskas 2001). Change

2 466 NICOLAS LECOMTE ET AL. Ecology, Vol. 90, No. 2 FIG. 1. Constraints imposed by habitat on water supply in a Snow Goose colony. This figure is based on an aerial photograph of a portion of the colony showing hypothetical goose nests (white circles) and movements during incubation recesses (white arrows) toward permanent water sources (black) either in wetlands (dark gray) or in mesic tundra (light gray). in resource availability during incubation could then affect egg predation rate if this influences the time spent off the nest by the sole incubating parent (thereafter called recess). Much emphasis has been placed on food availability or feeding behavior during incubation (Martin and Ghalambor 1999), but constraints imposed by water requirements have been largely ignored, even though they can be significant in species that mostly fast during incubation (Le Maho et al. 1981, Boismenu et al. 1992, Reed et al. 1995). Conditions that favor evaporative water loss (e.g., low atmospheric water vapor pressure) or reduce water availability around the nest (e.g., low rainfall) may increase recess frequency or the distance traveled by the birds to drink, thereby exposing their nest to increased predation. Therefore, such a system is ideally suited for examining how water availability can affect a predator prey interaction through its influence on prey behavior. We investigated how water availability affected nesting success in an Arctic-nesting species, the Snow Goose (Chen caerulescens; see Plate 1). In this gyneparental incubating species, predation is the primary cause of nest failure and occurs mostly when females leave their nest to drink or feed during incubation (Cooke et al. 1995, Beˆty et al. 2001; and this study). Although males can limit egg losses by attacking and chasing off predators (Samelius and Alisauskas 2005), they most often follow their mate during recesses, leaving the nest unprotected. Unattended Snow Goose nests suffer a much higher predation rate than attended ones (Samelius and Alisauskas 2001) and this risk increases with distance from the nest by the parents (Beˆty et al. 2002). We tested the hypothesis of a link between water availability and nesting success mediated through predation. We first predicted that geese nesting in mesic tundra where water is scarce should be more vulnerable to predation than those in wetlands because they had to travel a greater distance to drink during recesses (Fig. 1). Distance to water is taken as an index of local water availability and predators should thus find more unattended nests in mesic tundra than in wetlands. We tested these predictions by monitoring the behavior of incubating females during recesses and the foraging behavior of predators in the two habitats during four breeding seasons. We also hypothesized that rainfall should increase water availability and thus decrease predation risk as distance to drinking sources is reduced after rainfall, especially in mesic tundra. We therefore predicted that difference in nesting success between wetland and mesic tundra should vary with rainfall, increasing in dry years and decreasing in wet years. We further hypothesized that atmospheric water vapor pressure should affect evaporative water loss by geese and influences their need to leave the nest to drink. Thus, we predicted that nesting success should increase in years where atmospheric water pressure was high. Finally, to eliminate the confounding factor of differences in food quality between wetlands and mesic tundra (i.e., wetlands have more food plants preferred by geese; Gauthier et al. 1996), we experimentally manipulated water availability by supplementing nesting pairs with water. We predicted that pairs that were provided with water should stay closer to their nest during recesses and achieve higher nesting success than controls. METHODS Study site Data were collected between 1995 and 2005 on Bylot Island, Nunavut, Canada ( N, W). This island is the most important breeding site for Greater Snow Geese (Chen caerulescens atlantica), with more than breeding pairs, the majority of which are nesting in the same colony (Reed et al. 2002). The habitat is a mosaic of wetlands and mesic tundra. Wetlands typically occur in small patches (average size: 3.4 ha) embedded into a landscape dominated by mesic tundra, which covers over 10 times more area than wetlands (Lecomte et al. 2008b). Wetlands are mostly found in polygon-patterned ground created by frost action in lowland tundra (Gauthier et al. 1996). Polygons often form a heterogeneous, intricate network of lakes, ponds, and narrow water channels that hamper the movement of walking predators (Fig. 1; Lecomte et al. 2008a). Wetlands have a rich cover of graminoids, the preferred food plants of geese (Gauthier et al. 1996). Mesic tundra is found in drier lowlands, gentle slopes of small hills, and low-altitude plateaus. Standing water is scarce in mesic tundra and is only found in isolated small ponds or streams (Fig. 1) and between hummocks (low mounds of earth formed by winter freezing) following a rainfall. This habitat is dominated by prostrate shrubs and sparse forbs and graminoids.

3 February 2009 WATER LIMITATION IN ARCTIC-NESTING GEESE 467 PLATE 1. Greater Snow Goose pair protecting their nest at the Bylot Island goose colony, Nunavut, Canada, in When they are present close or on the nest, the predation risk is at its lowest level for the eggs. Photo credit: N. Lecomte. The predator community of Bylot Island is relatively simple with only one terrestrial predator, the Arctic fox (Vulpex lagopus; hereafter fox), and several avian predators, such as Parasitic Jaegers (Stercorarius parasiticus), Glaucous Gulls (Larus hyperboreus), and Common Ravens (Corvus corax; Beˆty et al. 2001). Foxes are the most important predator, depredating 72% of all goose eggs (Beˆty et al. 2002). All predators used both habitats and easily moved between them because wetland patches are much smaller than their home range size (Lecomte et al. 2008a). Climatic data At our study site, an automated weather station has been recording hourly air temperature since 1995 and relative humidity since 1999 using a Vaisala HMP35CF probe (Vaisala, Blainville, Que bec, Canada). Mean daily temperature and relative humidity during the study was 3.78C (annual range: C) and 81.2% (annual range: %; typical value for High Arctic areas). Each year, we manually recorded daily rainfall (mm) from 1 June to 15 August with a calibrated rain gauge. We estimated the partial water vapor pressure of the air, p(h 2 O) in hectopascals (hpa), which drives evaporation rate in animals (Gates 1980), using the Goff-Gratch equation (Goff and Gratch 1946), which is based on temperature and relative humidity. Behavioral observations From 2002 to 2005, we carried out behavioral observations of foraging predators and incubating geese. We observed predators and geese throughout the 24-h daylight period of the Arctic during the incubation period (9 June to 6 July; 600 hours of observations during 120 days). From three blinds, we used binoculars (103) and spotting scopes (20 603) to cover an area of 2.8 km 2 where both nesting habitats (wetlands and mesic tundra) were present and where nest densities were representative of those encountered in the whole colony. The number of goose nests under observation ranged annually from 100 to 500. We continuously scanned incubating females, and every time a female was seen departing from a nest, we started a stopwatch and observed it continuously until she resumed incubation.

4 468 NICOLAS LECOMTE ET AL. Ecology, Vol. 90, No. 2 Upon departure, we measured the distance traveled to the nearest water source to drink. Distance was either estimated visually using 250 georeferenced sticks located in the colony, or a posteriori with a measuring tape (median precision of 1.5 m). For each drinking event, we recorded whether the water source was permanent (ponds, rivers) or temporary (small depression at the surface of the tundra). We recorded all nest attacks by predators during the absence of the female, noted the predator involved, and whether it was successful (i.e., at least one egg taken) or not. We defined an attack as any attempt by a predator to take goose eggs. At least seven to eight foxes were identified individually by the distinctive patterns of their fur (shedding from winter to summer fur) each year (Lecomte et al. 2008a). Goose nest monitoring We systematically searched for goose nests in different areas of the colony during egg laying and early incubation. Nests located either in the central portion of the colony or in random plots scattered throughout the colony were found to differ little in terms of nesting parameters (Reed et al. 2005); thus, all nests were pooled and considered representative of the colony as a whole. For each nest, we determined the habitat (wetland or mesic tundra) and revisited them once during incubation and again at hatching to monitor their fate. We determined laying date of the first egg (nest initiation) following Beˆty et al. (2001). We defined total clutch size as the maximum number of eggs found in a nest after the start of incubation. At hatch, we counted the total number of goslings in the nest. Although partial predation occurs, only about 10% of nests surviving to hatch lost some eggs (Lepage et al. 2000) and most predation events usually resulted in the loss of a full clutch. We therefore focused mostly on whole-nest success in subsequent analyses and we defined a nest as successful if at least one egg hatched. Water supplementation experiment During the period when geese were incubating in 2005, we supplemented 40 randomly chosen nests in the colony with water, split equally between wetlands and mesic tundra. For each nest, we used two dark brown plastic containers with a volume of ;8 L (11.7 cm wide cm deep cm long). We buried them in the ground in opposite directions from the nest at distances ranging from three to six meters. This distance was chosen because Beˆty et al. (2002) showed that the success of predator attacks on goose nests increased with the distance between the parents and their nest, and was very high when parents were 10 m from the nest. This distance also ensured that water supplementation was restricted to parents of focus nests as geese are territorial in the vicinity of their nest (Cooke et al. 1995). We filled containers with water from neighboring ponds for the first time one or two days after the onset of the incubation period and we refilled all of them on the same day at mid-incubation. Since no containers were empty at the time of refilling, this suggests that females had access to supplementary water throughout their incubation period. We used as control a sample of nonmanipulated nests observed and monitored in the same year. The spatial distribution of control and experimental nests overlapped in the colony. We conducted behavioral observations at the 40 experimental nests using the same methods previously described, and we monitored their fate until hatch. We also recorded presence of males at the experimental nests and at 40 control nests randomly chosen in the colony. Male presence was defined as the percentage of total time spent close to the nest (10 m) during periods of observations (see also Samelius and Alisauskas [2001]). We detected no widowed females as males were seen at least once at all nests under observation. Data analysis Behavior of incubating females and predators. We first explored the effect of water on predation risk by comparing the distance traveled by females from their nest to a drinking site between years and habitats with two-way ANOVAs. Second, we compared this distance between days following significant rainfall (i.e., presence of.20 mm of rainfall in the preceding three days) and those that did not. We also used a v 2 statistic to compare the success of predators attacks in the two nesting habitats during recesses. Nesting success. We examined the effects of habitat and covariates such as rainfall and p(h 2 O) on goose nesting success with the daily nest survival (S) procedure of program MARK Version 4.2 (Dinsmore et al. 2002). We included the age of the nest at the time of discovery (date found minus estimated laying date) as an individual covariate in the model because predation rate can vary during the nesting period (Beˆty et al. 2002). We first considered the model with full time (i.e., years; t), habitat (h), day of the season (linear effect; d), and age effects with relevant interactions as the most general model. To select the most parsimonious model, we used the Akaike s Information Criterion modified for small sample size (AIC c ) and the relative weight of evidence in favor of a particular model (xaic c ; Burnham and Anderson 1998). In presence of uncertainty in model selection (DAIC c, 4), we estimated model-weighted average parameter values and their unconditional standard errors (Burnham and Anderson 1998). We estimated nesting success as the product of daily nest survival for the mean duration of laying and incubation (27 d). To examine the effect of rainfall and p(h 2 O), we substituted time (i.e., years) by these covariates. We parameterized rainfall and p(h 2 O) as binary variables to contrast years with higher (wet years ¼ 1) and lower (dry years ¼ 0) rainfall or p(h 2 O) than average. However, p(h 2 O) data were available for only seven years ( ) compared to 11 years ( ) for rainfall. We

5 February 2009 WATER LIMITATION IN ARCTIC-NESTING GEESE 469 FIG. 2. Annual rainfall abundance on Bylot Island during the Snow Goose nesting season (12 June 25 July) from 1995 to The dotted line shows the long-term average. thus ran two separate analyses with covariates, a first one using only rainfall for and a second using p(h 2 O) and rainfall for We determined the proportion of variation in nest survival explained by these time-dependent covariates by calculating the ratio in difference in deviance between the constant model and the model with covariates vs. the difference between the constant and the full time-dependent (i.e., year) model (analogous to an r 2 ; Agresti 2002). Finally, we formally tested the effect of rainfall or p(h 2 O) on nest survival using the analysis of deviance (ANODEV; Agresti 2002), which contrasts the amount of deviance accounted for by a model with covariate vs. the total amount of deviance accounted by the full timedependent model. Water supplementation experiment. We determined the effect of habitat on the use of water containers vs. natural sources of water with a generalized estimating equation (GEE) model and a logit link function (Hardin and Hilbe 2003). This procedure accounts for the nonindependence of multiple observations on the same individuals (see Beˆty et al. [2002] for details on the procedure). The same procedure was used to determine the effects of habitat and experiment on male presence at the nest. We used a generalized linear mixed regression model to examine the effects of habitat and experimental treatment (fixed effects) on distance traveled to drink by females during recesses, adding female identity as a random effect. We compared nesting success between experimental and control nests following the same procedure presented earlier (Dinsmore et al. 2002). We compared the mean number of hatchlings per successful nest between the two habitats and between experimental vs. control nests with a two-way ANOVA. Because the dependant variable was not normally distributed, we ranked transformed data and we tested interactions using the aligned rank test procedure as recommended by Salter and Fawcett (1993). Finally, to test for possible confounding factors affecting nesting success, we compared laying date and clutch size between habitats and experimental and control nests with two-way ANOVAs, again using rank-transformed variables. General statistics procedures. Statistical analyses other than nest survival were performed using SAS software, Version 9 (SAS Institute 2002). We considered each goose nest as independent because (1) individual geese can defend their nests against predators, (2) the fate of a nest is independent of the fate of its nearest neighbors (Beˆty et al. 2001), and (3) several foxes (7 8 per year) were known to forage within the study area (Lecomte et al. 2008a). For all data, normality was checked with the Kolmogorov-Smirnov test (Lilliefors option). All probabilities are two-tailed and the level was set to a ¼ Unless mentioned otherwise, values are reported as mean 6 SE. RESULTS Rainfall and water availability Between 1995 and 2005, the mean total precipitation during the incubation period was 51.6 mm, with an almost a fourfold interannual variation (range mm; Fig. 2). During incubation, most precipitation fell as rain, although snow occasionally occurred. After significant precipitation events, rainwater persisted in small pools and/or in shallow depressions between hummocks for one to three days (N. Lecomte, unpublished data), thereby increasing the availability of drinking water for geese in the landscape. Between 1999 and 2005, the mean p(h 2 O) during the incubation period was 6.84 hpa, with small interannual variation (range, hpa). Annual rainfall and p(h 2 O) were not correlated (r ¼ 0.42; df ¼ 6; P ¼ 0.34).

6 470 NICOLAS LECOMTE ET AL. Ecology, Vol. 90, No. 2 Model selection for the effect of habitat type (wetland vs. mesic tundra) and annual rainfall on daily nest survival (S )of Greater Snow Geese on Bylot Island, (n ¼ 3024). TABLE 1. Type of model Model DAIC c xaic c np Deviance With time variation 1) S (h3t)þd ) S (h3t)þdþage ) S hþtþd ) S hþtþdþage ) S tþdþage With time-dependent covariates 6) S (h3p)þd ) S hþpþd Without time variation 8) S hþd Notes: The most relevant models are ranked by their AIC c value. For each model, the difference in AIC c values is given in relation to the most parsimonious model (DAIC c ), the model s AIC c weight (xaic c ), the number of estimable parameters (np), and the deviance. Model subscripts are: age ¼ age of the nest when found (date found minus the nest initiation date), d ¼ linear effect of day of the season, h ¼ habitat (wetlands vs. mesic), P ¼ precipitation (dry vs. wet years), t ¼ year ( þ indicates additive effect; 3 indicates interactive effect. Use of water by geese and predator attacks We observed 102 incubation recesses for 96 nests. After departure from the nest, drinking was the first activity performed by females in 89.3% of the cases (n ¼ 91) whereas feeding behavior was the first activity in 8.8% of recesses (n ¼ 9), and preening was the first activity in 1.9% (n ¼ 2). Only 10.8% (n ¼ 11) of all recesses involved no drinking by females and occurred when the birds had to precipitately return to their nest due to presence of a predator. Globally, drinking, feeding, and preening were the most important activities during these recesses. Moreover, several drinking bouts were required by females to restore their hydric balance since they drank on average seven times per recess (n ¼ 15). During recesses, females nesting in mesic tundra traveled 3.8 times farther from their nest to drink than those nesting in wetlands (mesic, m; wetland, 9 6 3m;F 1,98 ¼ 57.7; P, 0.001), with no difference between years (F 1,98 ¼ 0.001; P ¼ 0.97). During the three days following a significant rainfall, distance traveled to drink decreased in both habitats (mesic tundra before rain, m; after rain, m; wetlands before rain, m; after rain 6 6 1m;F 1,74 ¼ 13.5; P, 0.001). This occurred because females drank in temporary water depressions rather than in permanent water sources (ponds, rivers). Compared to wetlands, females nesting in mesic tundra were 13.9 times more likely to exceed the 10-m threshold distance where the success of predator attacks increases considerably during incubation recesses (Bêty et al. 2002). As a result, both avian and terrestrial predators were more successful in robbing nests in mesic tundra (89% of attacks were successful; n ¼ 200 predator attacks) than in wetlands (79% of attacks were successful; n ¼ 77 predator attacks; v 2 ¼ 4.3; df ¼ 1; P ¼ 0.038) during incubation recesses. Water availability and nesting success Between 1995 and 2005, we monitored the fate of 1331 and 1693 nests in wetlands and mesic tundra, respectively. During these years, daily nest survival differed among habitats and years, increased with the day of the season, and possibly with the age when a nest was found, as suggested by the four top models in Table 1. When time was substituted by rainfall in the model, this variable in interaction with habitat types (Model 6, Table 1) explained 34% (rdev 2 ) of the annual variation in nest survival, a significant result (ANODEV of Model 6 vs. Models 1 and 8 in Table 1; F 2,18 ¼ 4.54; P ¼ 0.025). Overall, nesting success of geese was higher in wet than in dry years, and the difference in success between habitats was higher during dry years (11%, on average) than during wet ones (3%; Fig. 3A). This suggests that rainwater was relatively more beneficial for pairs nesting in mesic tundra than for those nesting in wetlands where success was higher. When we considered simultaneously the effect of rainfall and p(h 2 O) for the subset of years when both measurements were available, we found that both variables in interaction explained 81% of the annual variation in nest survival, a significant result (ANODEV of Model 2 vs. Models 1 and 7 in Table 2; F 3,10 ¼ 14.3; P, 0.001). Overall, nesting success of geese was high when either rainfall or p(h 2 O) was high, but success was reduced by almost 39% when both of these variables were low (Fig. 3B). The latter case corresponds to situations where potential evaporative water loss for geese was high and drinking water availability around the nest was low. Water supplementation and goose behavior In 2005, we observed 44 incubation recesses by females for the 40 nests equipped with water containers, and 30 recesses for 40 control nests. Both females and males used water containers, either for drinking or bathing. Females in wetlands used nearly four times more natural sources of water to drink (n ¼ 15) than the water containers (n ¼ 4), whereas geese in mesic tundra used containers more often (11 times in natural sources vs. 14 in containers; GEE (generalized estimating equation) v 2 ¼ 5.1; df ¼ 1; P ¼ 0.02).

7 February 2009 WATER LIMITATION IN ARCTIC-NESTING GEESE 471 FIG. 3. Annual nesting success (mean þ SE) of Snow Geese on Bylot Island as a function of (A) the habitat and annual precipitation (dry vs. wet years), and (B) the annual precipitation [P] and the partial water vapor pressure of the air [p(h 2 O)]. Nesting success corresponds to the daily nest survival, S, over the 27-day laying and incubation period. Wet and dry years correspond to years with rainfall abundance higher and lower than the average, respectively. The þ and symbols indicate years with conditions above and below the average, respectively. Numbers over bars indicate sample size. When we considered only the use of natural sources of water, distances traveled from the nest to those sources still differed between habitats in experimental nests (mesic, m; wetlands, 9 6 7m;F 1,26 ¼ 18.7; P, 0.001). However, distances traveled to natural sources did not differ between experimental ( m) and control nests (23 6 6m;F 1,54 ¼ 0.93; P ¼ 0.33), with no interaction between habitats and the experimental treatment (F 1,54 ¼ 0.94; P ¼ 0.34). When considering the use of all water sources (containers and natural), females with water containers in mesic tundra were 2.2 times less likely to exceed the 10-m threshold distance when they left the nest to drink (38% of cases; n ¼ 21) than females without water containers (84% of cases; n ¼ 13). In comparison, females with water containers in wetlands were only 1.2 times less likely to exceed the 10- m threshold than those without containers (26% vs. 35%; n ¼ 15 and 17, respectively). Overall, presence of males at the nest was higher for experimental nests (93%) than for control nests (70%; n ¼ 89; GEE v 2 ¼ 4.73; df ¼ 1; P ¼ 0.03), with no effect of habitat (v 2 ¼ 0.01; df ¼ 1; P ¼ 0.90). The daily nest survival analysis provided strong support for a treatment effect (Table 3). Overall, nesting success (i.e., probability that at least one egg survived to hatching) increased by almost 20% in nests with water containers (89%) compared to control nests (68%) in both habitats (Fig. 4). We found no difference in the mean number of hatched goslings per successful nest between habitats or between experimental and control nests ( ; F 3, 117 ¼ 0.10, P ¼ 0.95). Similarly, there was no difference in laying date and clutch size between experimental and control nests or between habitats (P for all tests). DISCUSSION Our study demonstrates that water availability had an effect on the nesting productivity of Snow Geese, and that this effect was due to a change in nest predation during incubation recesses. The presence of abundant and permanent water sources close to nests enhanced the quality of wetland habitat for nesting geese, especially in dry summers with limited rainfall. In addition, nest predation also increased in years where the potential for evaporative water loss of geese was high. An alternative explanation for our results could be that plant growth is enhanced in wet years, thereby improving the condition of incubating females and allowing them to increase incubation constancy. However, we reject this explanation for two reasons. First, our experimental manipulation of water availability (independent of food supply) Model selection for the effect of habitat type (wetland vs. mesic tundra), annual rainfall, and partial water vapor pressure of the air on daily nest survival (S) of Greater Snow Geese on Bylot Island, (n ¼ 2053). TABLE 2. Type of model Model DAIC c xaic c np Deviance With time variation 1) S (h3t)þd With time-dependent covariates 2Þ S hþ½pðh2oþ 3 PŠþd Þ S hþpðh2oþþpþd Þ S hþpðh2oþþd Þ S ½h 3 pðh2oþšþd ) S (h3p)þd Without time variation 7) S hþd Notes: The most relevant models are ranked by their AIC c value; AIC c components are as in Table 1. Model subscripts: p(h 2 O) ¼ partial water vapor pressure; all others are as shown in Table 1.

8 472 NICOLAS LECOMTE ET AL. Ecology, Vol. 90, No. 2 TABLE 3. Model selection for the effect of experimental increase in water availability on daily nest survival (S) of Greater Snow Geese nesting in two habitats (wetlands vs. mesic tundra) on Bylot Island, 2005 (n ¼ 300). Model DAIC c xaic c np Deviance 1) S cþage ) S c ) S cþhþage ) S cþd ) S cþh ) S cþhþdþage ) S hþage Notes: The most relevant models are ranked by their AIC c value; AIC c components are as in Table 1. Model subscripts: c ¼ experimental treatment (i.e., presence of water container); all others are as shown in Table 1. clearly showed the effect of water availability on nesting success, and second, Graham-Sauve (2008) showed that plant growth was actually reduced in wet years at our study site. Most studies showing ecological effects of water availability have dealt with species inhabiting arid environments (e.g., Morrison and Bolger 2002, Tieleman et al. 2004, Williams and Tieleman 2005, Loveridge et al. 2006). Here, we show that water availability can also have demographic consequences in a species inhabiting the Arctic environment where water is not usually considered a limited commodity. Predation-mediated water effect Predation has a large influence on breeding tactics and success of birds (Fontaine and Martin 2006). Any abiotic factors imposing constraints on prey behavior have the potential to indirectly affect susceptibility to predation, although empirical evidence is lacking (Schmidt 1999). Our study provides such evidence as both environmental conditions and life history of the prey interacted to influence top-down (i.e., predation) processes. The reproductive output of many species is often limited by food resources, a bottom-up control (e.g., Martin 1987, Nagy and Holmes 2005). This is supported by many food supplementation experiments that showed an increase in breeding success (Boutin 1990). However, to our knowledge, the 20% increase in nesting success found in our study is among the strongest ever reported. Moreover, even though we manipulated a resource (i.e., water), this increase resulted from a top-down effect. Several studies showing a strong effect of food supplementation on reproductive success were conducted during years of low natural food availability (e.g., Siikama ki 1998). Our experiment, however, was conducted in 2005, the year with the highest amount of rainfall recorded during our study period. This suggests that drinking water is a strong limiting resource and that incubating females benefit from any sources of water close to their nest, such as those appearing after a rainfall. Overall, nesting success of geese at the study colony was 66% in 2005 (compared to a long-term average of 63%, ; G. Gauthier, unpublished data). A key factor affecting predation rate on goose nests is lemming abundance, the alternative prey of predators (Beˆty et al. 2002), and in 2005, lemming abundance was intermediate (Lecomte et al. 2008a). Therefore, despite the high rainfall in 2005, nesting conditions for geese were otherwise near average and thus our experimental manipulation was not conducted under unusual conditions. Moreover, a recent investigation of the effect of several climatic factors on the reproduction of Greater Snow Geese confirmed the positive influence of rainfall on nesting success reported here (Dickey et al. 2008). The sudden appearance of temporary water holes for nesting females after rainfall can be seen as a sudden pulse of resources because individuals have little impact on the depletion of these resources (Burger 1981, Schmidt and Ostfeld 2008, Yang et al. 2008). Females apparently showed an immediate response to this unpredictable and ephemeral resource, and this enhanced their reproductive success. In addition, by creating numerous water sources in the landscape, rainfall provides a density-independent resource that changes the quality of many territories, especially those depleted in permanent water sources. Our results thus show a novel way by which an ephemeral resource pulse can affect the predator prey dynamic of a system. It is somewhat surprising that the treatment also increased nesting success in wetlands where water is abundant relatively close to the nest. It is highly unlikely that the presence of water containers deterred predators like foxes from experimental nests as those were small and inconspicuous (buried in the ground and the same color as the tundra soil). A possible explanation is that females using water containers in both habitats had an unobstructed view of the surroundings while drinking. FIG. 4. Nesting success (mean þ SE) of experimental (i.e., with water containers) and control Snow Goose nests in two habitats on Bylot Island in Nesting success corresponds to the daily nest survival, S, over the 27-day laying and incubation period. Numbers over bars indicate sample size.

9 February 2009 WATER LIMITATION IN ARCTIC-NESTING GEESE 473 In contrast, small ponds or water channels in wetlands are often located in depressions surrounded by polygon rims, which hinder the vision of drinking females, thereby limiting their detection of approaching egg predators (e.g., Kahlert 2003). Finally, higher water availability near the nest not only reduced the distance traveled by females during incubation recesses but also increased the presence of males at the nest. However, male presence was enhanced only when females were already at the nest, because when females leave the nest to drink, males almost invariably follow them closely; presumably because it is more important to protect the female than the eggs from attacks by predators or neighboring territorial males. Nonetheless, because nest defense against foxes is more efficient by both membears of the pair than by a single mate (Samelius and Alisauskas 2001), the increased presence of males at the nest may have contributed to the higher success of manipulated nests. Given the benefits accrued to birds having access to drinking water close to their nest, we would expect that proximity to predictable water resources should be an abiotic factor influencing nest site selection. Not only is drinking water more available in wetlands than in mesic tundra but the intricate network of water channels in wetlands also impedes movements of terrestrial predators like foxes (Lecomte et al. 2008a). In accordance with these observations, we have found that female Snow Geese prefer to nest in wetlands (Lecomte et al. 2008b). Although rainfall events are unpredictable, they are not rare as in arid ecosystems and could allow individuals nesting in lower quality territories (i.e., in mesic tundra) to attenuate, to some extent, predation risk. Water economy in the Arctic Comparative studies show a link between incubation constancy and body mass in waterfowl (Thompson and Raveling 1987). For example, large goose species generally take fewer recesses than smaller ones due to their relatively lower mass-specific metabolic rate and greater amount of nutrient reserves (Afton and Paulus 1992). Although these authors have typically emphasized energetic constraints to explain interspecific variations in incubation patterns, we argue that water requirement may be an additional constraint on recess frequency. Although we did not have data on recess frequency per se, we found a positive link between partial water vapor pressure in the air and nesting success. We suggest that when the potential for evaporative water loss of geese is high, females have to leave their nest more often to drink, thereby increasing exposure to predation. Geese are likely to face high water needs during reproduction. First, animals suffer an increased tendency to dehydrate in the Arctic because cold temperatures result in low atmospheric water vapor pressure (Willmer et al. 2000). Strong winds and high solar radiation, conditions that are frequent in the Arctic, should further enhance the potential for evaporative water loss. Second, birds face major physiological challenges to regulate their water balance while fasting during incubation (Le Maho et al. 1981). In Eider Ducks (Somateria molissima L.), Criscuolo et al. (2000) suggested that birds adjusted their drinking behavior during incubation recesses to regulate their body water while fasting. Anecdotal evidence also suggests that Snow Geese have high evaporative water loss, and hence high water requirements. J. Larochelle ( personal communication) and Ratte (1998) reported high water loss by gosling Greater Snow Geese during experiments in metabolic chambers. In addition, geese feeding in farmlands typically return to roosting ponds in midday to drink under sunny conditions, though not under overcast skies or during rain (Be chet et al. 2003). We therefore contend that the water economy of birds, which has received little attention in comparison to food habits even after the seminal paper of Bartholomew and Cade (1963), has been hitherto underrated in reproductive studies of geese and should be investigated in other bird species as well. In most biomes, change in rainfall abundance resulting from climate warming could be at least as great as change in air temperature, but future trends in precipitations are especially difficult to predict (Allen and Ingram 2002). Simulations predict a mean increase of 10% in the next 20 years (6 mm/month) around the Canadian Arctic, a larger increase than for the world as a whole (IPCC 2001, ACIA 2005). Given that goose nesting success is positively linked to rainfall, we could expect some changes in goose productivity in the future under such a scenario. The importance of water in the functioning of tundra ecosystems is only beginning to emerge in the literature (Hodkinson et al. 1999) and our study provides a novel illustration of it. How much Arctic birds will balance their water requirements with nest predation risk in face of changing climatic conditions remains an open question for the moment. Conclusion Our study provides a clear demonstration that abiotic factors such as water availability and rainfall can have a strong effect on predator prey dynamics through an influence on prey behavior, even in a non-arid environment such as the tundra. Such relations may have significant consequences in tundra ecosystems where it is becoming increasingly evident that predator prey interactions play a key role (Bêty et al. 2002, Gilg et al. 2003, Gauthier et al. 2004). Our results also highlight the complexity of these relations due to the fact that rainfall creates unpredictible pulses of resources (water holes), and that the benefit of this resource for the prey is heterogeneous at the habitat scale. This is an emerging perspective in the literature (Holt 2008, Schmidt and Ostfeld 2008), where the spatial component of ephemeral resources is predicted to fine tune predator prey interactions. When variability in prey productivity is

10 474 NICOLAS LECOMTE ET AL. Ecology, Vol. 90, No. 2 linked to the dynamics of the pulse, this could allow persistance of the prey in lower quality habitats and its exploitation by predators during interpulse intervals. ACKNOWLEDGMENTS Funding was provided by grants from the Natural Sciences and Engineering Research Council of Canada to G. Gauthier, the Arctic Goose Joint Venture (Canadian Wildlife Service), the Fonds Québécois pour la Nature et les Technologies, ArcticNet, and the Northern Ecosystem Initiative (Environment Canada). Université Laval, the Centre d Études Nordiques, the Société Provancher, Le Fonds, Richard Bernard, and the Department of Indian Affairs and Northern Development provided additional financial assistance to N. Lecomte. Logistic support was generously provided by the Polar Continental Shelf Project (PCSP). We thank all the people who participated in the fieldwork. We are grateful to S. D. Coˆte, M.-A. Giroux, M. Humphries, J. Larochelle, and J. Williams for their comments on this manuscript. We are indebted to the Hunters and Trappers Association of Pond Inlet and to Parks Canada for allowing us to work on Bylot Island. Finally, we thank M. Lambrechts for encouraging us to manipulate water availability. All methods described in the paper were approved by the Animal Care Committee of Université Laval following guidelines of the Canadian Animal Care Council. This is PCSP contribution No LITERATURE CITED ACIA Impacts of a warming Arctic: Arctic climate impact assessment. Cambridge University Press, New York, New York, USA. Afton, A. D., and S. L. Paulus Incubation and brood care. Pages in B. D. J. Batt, A. D. Afton, M. G. Anderson, C. D. Ankney, D. H. Johnson, J. A. Kadlec, and G. L. Krapu, editors. Ecology and management of breeding waterfowl. University of Minnesota Press, Minneapolis, Minnesota, USA. Agresti, A Categorial data analysis. Second edition. John Wiley, New York, New York, USA. Allen, M. R., and W. J. Ingram Constraints on future changes in climate and the hydrologic cycle. Nature 419: Bartholomew, G. A., and T. J. Cade The water economy of land birds. Auk 80: Béchet, A., J.-F. Giroux, G. Gauthier, J. D. Nichols, and J. E. Hines Spring hunting changes the regional movements of migrating Greater Snow Geese. Journal of Applied Ecology 40: Bêty, J., G. Gauthier, J.-F. Giroux, and E. Korpimaki Are goose nesting success and lemming cycles linked? Interplay between nest density and predators. Oikos 93: Bêty, J., G. Gauthier, E. Korpimaki, and J.-F. Giroux Shared predators and indirect trophic interactions: lemming cycles and Arctic-nesting geese. Journal of Animal Ecology 71: Blancher, P. J., and R. J. Robertson Effect of food supply on the breeding biology of Western Kingbirds. Ecology 68: Boismenu, C., G. Gauthier, and J. Larochelle Physiology of prolonged fasting in Greater Snow Geese (Chen caerulescens atlantica). Auk 109: Boutin, S Food supplementation experiments with terrestrial vertebrates: patterns, problems, and the future. Canadian Journal of Zoology 68: Burger, J A model for the evolution of mixed-species colonies of Ciconiiformes. Quarterly Review of Biology 56: Burnham, K. P., and D. R. Anderson Model selection and inference: a practical information-theoretic approach. Springer-Verlag, New York, New York, USA. Coe, S. J., and J. T. Rotenberry Water availability affects clutch size in a desert sparrow. Ecology 84: Cooke, F., R. F. Rockwell, and D. B. Lank The Snow Geese of La Pérouse Bay, natural selection in the wild. Oxford University Press, New York, New York, USA. Criscuolo, F., M. Gauthier-Clerc, G. W. Gabrielsen, and Y. Le Maho Recess behaviour of the incubating Common Eider Somateria mollissima. Polar Biology 23: Dickey, M.-H., G. Gauthier, and M.-C. Cadieux Climatic effects on the breeding phenology and reproductive success of an Arctic-nesting goose species. Global Change Biology, in press. Dinsmore, S. J., G. C. White, and F. L. Knopf Advanced techniques for modeling avian nest survival. Ecology 83: Doody, J. S., R. A. Sims, and M. Letnic Environmental manipulation to avoid a unique predator: drinking hole excavation in the agile wallaby, Macropus agilis. Ethology 113: Ferns, P. N., and S. A. Hinsley Importance of topography in the selection of drinking sites by sandgrouse. Functional Ecology 9: Fontaine, J. J., and T. E. Martin Parent birds assess nest predation risk and adjust their reproductive strategies. Ecology Letters 9: Gates, D. M Biophysical ecology. Springer-Verlag, New York, New York, USA. Gauthier, G., J. Bêty, J.-F. Giroux, and L. Rochefort Trophic interactions in a High Arctic Snow Goose colony. Integrative and Comparative Biology 44: Gauthier, G., L. Rochefort, and A. Reed The exploitation of wetland ecosystems by herbivores on Bylot Island. Geoscience Canada 23: Gibbs, H. L., and P. R. Grant Ecological consequences of an exceptionally strong El-Nin o event on Darwin s finches. Ecology 68: Gilg, O., I. Hanski, and B. Sittler Cyclic dynamics in a simple vertebrate predator prey community. Science 302: Goff, J. A., and S. Gratch Low-pressure properties of water from 160 to 2128 F. Pages nd annual meeting of the American Society of Heating and Ventilating Engineers, New York, New York, USA. Graham-Sauvé, M Effets en cascade du climat et des interactions trophiques indirectes sur les plantes de la toundra par l oie des neiges. M.S. Thesis. Université du Québec, Rimouski, Canada. Hardin, J. W., and J. M. Hilbe Generalized estimating equations. Chapman and Hall/CRC, Boca Raton, Florida, USA. Hilton, G. M., G. D. Ruxton, and W. Cresswell Choice of foraging area with respect to predation risk in redshanks: the effects of weather and predator activity. Oikos 87: Hodkinson, I. D., N. R. Webb, J. S. Bale, and W. Block Hydrology, water availability and tundra ecosystem function in a changing climate: the need for a closer integration of ideas? Global Change Biology 5: Holt, R. D Theoretical perspectives on resource pulses. Ecology 89: IPCC Climate change 2001: synthesis report. A contribution of working groups I, II, and III to the third assessment report of the intergovernmental panel on climate change. Cambridge University Press, Cambridge, UK. Kahlert, J The constraint on habitat use in wingmoulting Greylag Geese Anser anser caused by anti-predator displacements. Ibis 145:E45 E52.

11 February 2009 WATER LIMITATION IN ARCTIC-NESTING GEESE 475 Klaassen, M May dehydration risk govern long-distance migratory behaviour? Journal of Avian Biology 35:4 6. Lada, H., J. R. Thomson, R. Mac Nally, G. Horrocks, and A. C. Taylor Evaluating simultaneous impacts of three anthropogenic effects on a floodplain-dwelling marsupial Antechinus flavipes. Biological Conservation 134: Le Maho, Y., H. Vu Van Kha, H. Koubi, G. Dewasmes, J. Girard, P. Ferre, and M. Cagnard Body composition, energy expenditure, and plasma metabolites in long-term fasting geese. American Journal of Physiology 241:E342 E354. Lecomte, N., V. Careau, G. Gauthier, and J.-F. Giroux. 2008a. Predator behaviour and predation risk in the heterogeneous Arctic environment. Journal of Animal Ecology 77: Lecomte, N., G. Gauthier, and J.-F. Giroux. 2008b. Breeding dispersal in a heterogeneous landscape: the influence of habitat and nesting success in Greater Snow Geese. Oecologia 155: Lepage, D., G. Gauthier, and S. Menu Reproductive consequences of egg-laying decisions in Snow Geese. Journal of Animal Ecology 69: Loveridge, A. J., J. E. Hunt, F. Murindagomo, and D. W. Macdonald Influence of drought on predation of elephant (Loxodonta africana) calves by lions (Panthera leo) in an African wooded savannah. Journal of Zoology 270: Martin, T. E Food as a limit on breeding birds a lifehistory perspective. Annual Review of Ecology and Systematics 18: Martin, T. E Avian life history evolution in relation to nest sites, nest predation, and food. Ecological Monographs 65: Martin, T. E., and C. K. Ghalambor Males feeding females during incubation. I. Required by microclimate or constrained by nest predation? American Naturalist 153: Morrison, S. A., and D. T. Bolger Variation in a sparrow s reproductive success with rainfall: food and predator-mediated processes. Oecologia 133: Myers, P., L. L. Master, and R. A. Garrett Ambient temperature and rainfall: an effect on sex ratio and litter size in deer mice. Journal of Mammalogy 66: Nagy, L. R., and R. T. Holmes Food limits annual fecundity of a migratory songbird: an experimental study. Ecology 86: Newton, I Population limitation in birds. Academic Press, San Diego, California, USA. Ogutu, J. O., H. P. Piepho, H. T. Dublin, N. Bhola, and R. S. Reid Rainfall influences on ungulate population abundance in the Mara-Serengeti ecosystem. Journal of Animal Ecology 77: Ostfeld, R. S., and F. Keesing Pulsed resources and community dynamics of consumers in terrestrial ecosystems. Trends in Ecology and Evolution 15: Post, E., R. O. Peterson, N. C. Stenseth, and B. E. McLaren Ecosystem consequences of wolf behavioural response to climate. Nature 401: Ratcliffe, N., S. Schmitt, and M. Whiffin Sink or swim? Viability of a Black-tailed Godwit population in relation to flooding. Journal of Applied Ecology 42: Ratte, J Thermorégulation et croissance chez les oisons de la grande oie des neiges, Anser caerulescens atlantica. M.S. Thesis. Université Laval, Québec, Canada. Redfern, J. V., R. Grant, H. Biggs, and W. M. Getz Surface-water constraints on herbivore foraging in the Kruger National Park, South Africa. Ecology 84: Reed, A., R. J. Hughes, and H. Boyd Patterns of distribution and abundance of Greater Snow Geese on Bylot Island, Nunavut, Canada Wildfowl 53: Reed, A., R. J. Hughes, and G. Gauthier Incubation behavior and body-mass of female Greater Snow Geese. Condor 97: Reed, E. T., G. Gauthier, and R. Pradel Effects of neck bands on reproduction and survival of female Greater Snow Geese. Journal of Wildlife Management 69: Salter, K. C., and R. Fawcett The ART test of interaction: a robust and powerful rank test of interaction in factorial models. Communications in Statistics: Simulation and Computation 22: Samelius, G., and R. T. Alisauskas Deterring Arctic fox predation: the role of parental nest attendance by Lesser Snow Geese. Canadian Journal of Zoology 79: Samelius, G., and R. T. Alisauskas Sex-biased costs in nest defence behaviours by Lesser Snow Geese (Chen caerulescens): consequences of parental roles? Behavioral Ecology and Sociobiology 59: SAS Institute SAS guide for personal computers. Version 9. SAS Institute, Cary, North Carolina, USA. Schmidt, K. A Foraging theory as a conceptual framework for studying nest predation. Oikos 85: Schmidt, K. A., and R. S. Ostfeld Numerical and behavioral effects within a pulse-driven system: consequences for shared prey. Ecology 89: Siikamäki, P Limitation of reproductive success by food availability and breeding time in Pied Flycatchers. Ecology 79: Thompson, K. R., and R. W. Furness The influence of rainfall and nest-site quality on the population-dynamics of the Manx Shearwater Puffinus puffinus on Rhum. Journal of Zoology 225: Thompson, S. C., and D. G. Raveling Incubation behavior of Emperor Geese compared with other geese: interactions of predation, body size, and energetics. Auk 104: Tieleman, B. I., J. B. Williams, and G. H. Visser Energy and water budgets of larks in a life history perspective: parental effort varies with aridity. Ecology 85: Valeix, M., H. Fritz, R. Matsika, F. Matsvimbo, and H. Madzikanda The role of water abundance, thermoregulation, perceived predation risk and interference. African Journal of Ecology 46: Williams, J. B., and B. I. Tieleman Physiological adaptation in desert birds. Bioscience 55: Willmer, P., G. Stone, and I. A. Johnston Environmental physiology of animals. Blackwell Scientific Publications, Oxford, UK. Yang, L. H., J. L. Bastow, K. O. Spence, and A. N. Wright What can we learn from resource pulses? Ecology 89: