AMPHIBIAN CONCENTRATIONS IN DESICCATING MUD MAY DETERMINE THE BREEDING SEASON OF THE WHITE-SHOULDERED IBIS (PSEUDIBIS DAVISONI)

Transcription

1 The Auk 130(4): , 2013 The American Ornithologists Union, Printed in USA. AMPHIBIAN CONCENTRATIONS IN DESICCATING MUD MAY DETERMINE THE BREEDING SEASON OF THE WHITE-SHOULDERED IBIS (PSEUDIBIS DAVISONI) Hugh L. Wright, 1,4 Nigel J. Collar, 2,3 Iain R. Lake, 1 and Paul M. Dolman 1 1 School of Environmental Sciences, University of East Anglia, Norwich Research Park, Norwich, NR4 7TJ, United Kingdom; 2 BirdLife International, Girton Road, Cambridge, CB3 0NA, United Kingdom; and 3 School of Biological Sciences, University of East Anglia, Norwich Research Park, Norwich, NR4 7TJ, United Kingdom Abstract. Many waterbirds reproduce seasonally in response to fluctuations in food supply. White-shouldered Ibises (Pseudibis davisoni) breed during the dry, water-drawdown season, but, unlike other waterbirds, they do not take advantage of prey concentrated in diminishing pools. To understand how this species successfully feeds and breeds at the driest time of year, we studied its habitat use and diet, and the habitat conditions that influence intake rates and prey density at waterholes. Foraging observations, prey sampling, and landscape-scale assessment of habitat availability were undertaken (at 7, 47, and 58 waterholes, respectively) over two breeding seasons. Although they rarely foraged in water, the birds used all exposed substrates, feeding on amphibians and small invertebrates. Amphibians were the most abundant prey in waterhole substrates and accounted for 81% of overall biomass intake. Both intake rates and density of amphibian prey biomass were greater in dry than in moist or saturated substrates. Intake rates and density of prey biomass changed little through the dry season, but exposed substrate extent increased by 74%. The White-shouldered Ibis s use of dry waterhole substrates to exploit seasonally concentrated prey is unusual among large waterbirds, and we suggest that its breeding season may be timed to coincide with receding water levels and increasing substrate exposure. Estimated prey requirements of a breeding pair over the nesting period were equivalent to nearly two-thirds of amphibian biomass found at large waterholes. Each pair may therefore require multiple waterholes to overcome prey depletion and breed successfully, which is consistent with the noncolonial dispersed distribution of nesting pairs. Received 13 December 2012, accepted 18 June Key words: Cambodia, foraging, intake rate, nesting time, prey density, prey depletion, Pseudibis davisoni, seasonal reproduction, White-shouldered Ibis. Las Concentraciones de Anfibios en el Lodo en Desecación Podrían Determinar la Temporada Reproductiva de Pseudibis davisoni Resumen. Muchas aves acuáticas se reproducen estacionalmente en respuesta a las fluctuaciones en la disponibilidad de alimento. Pseudibis davisoni se reproduce durante la temporada seca o de descenso del agua, pero a diferencia de otras aves acuáticas, no aprovecha las presas concentradas en los estanques en disminución. Para entender cómo esta especie se alimenta y reproduce exitosamente en el momento más seco del año estudiamos su uso del hábitat y dieta, y las condiciones del hábitat que afectan las tasas de obtención de las presas y su densidad en los cuerpos de agua. Hicimos observaciones de forrajeo, muestreo de presas y una evaluación del hábitat a escala del paisaje (en 7, 47 y 58 estanques, respectivamente) en dos temporadas reproductivas. Aunque rara vez forrajearon en el agua, las aves usaron todos los sustratos expuestos, alimentándose de anfibios y pequeños invertebrados. Los anfibios fueron la presa más abundante en los sustratos de los estanques y representaron el 81% de la ingesta de biomasa. La tasa de ingesta y la densidad de la biomasa de presas de anfibios cambiaron poco durante la temporada seca, pero la cantidad de sustrato expuesto aumentó un 74%. El uso de los sustratos de estanque por parte de P. davisoni para aprovechar las presas concentradas estacionalmente es inusual entre aves acuáticas grandes, y sugerimos que su temporada reproductiva podría estar sincronizada con la baja en los niveles del agua y el incremento en la exposición del sustrato. Los estimados del requerimiento de presas de una pareja en reproducción a lo largo del periodo de anidación fueron equivalentes a casi dos tercios de la biomasa de anfibios encontrada en estanques grandes. Cada pareja podría requerir de múltiples estanques para sobreponerse a la disminución de las presas y reproducirse exitosamente, lo que concuerda con la distribución dispersa de las parejas anidantes. 4 hughlewiswright@gmail.com The Auk, Vol. 130, Number 4, pages ISSN , electronic ISSN by The American Ornithologists Union. All rights reserved. Please direct all requests for permission to photocopy or reproduce article content through the University of California Press s Rights and Permissions website, com/reprintinfo.asp. DOI: /auk

2 October 2013 Foraging White-shouldered Ibises 775 Food supply and foraging success have important influences on the seasonality of reproduction in birds, either through the condition of breeding adults or as cues for nesting to commence (Perrins 1970, Perrins and Birkhead 1983, O Brien and Hau 2005). Breeding in waterbirds is particularly shaped by seasonal changes in both food abundance and availability (Kushlan 1978, Gawlik 2002). Waterbird breeding in temperate and high-latitude zones (as for most birds in these regions) is determined primarily by the increased abundance of food in spring (Lack 1968). By contrast, seasonal or opportunistic breeding strategies are commonplace in the tropics and subtropics (Kushlan 1978, Kingsford and Norman 2002), following dynamic hydrological cycles that create temporal and spatial variation in food quantity, type, and accessibility (Kushlan 1986, Frederick and Collopy 1989). Many tropical and subtropical waterbirds time their breeding to coincide with either floods or the subsequent drawdown of water within wetlands (Kushlan 1978). Floods may dramatically increase food abundance, especially after periods of water scarcity, because an excess of water stimulates aquatic productivity (Crome 1988, Kingsford et al. 2004). Alternatively, food availability increases in the drawdown season as prey are concentrated in diminishing pools (Russell et al. 2002) or as profitable habitats are made more accessible by receding waters and/or vegetation dieback (Bildstein et al. 1990, Morrison et al. 2009). The coincidence of nesting with water drawdown has been documented in several regions, including the Florida Everglades (Kushlan et al. 1975), South African wetlands (Berruti 1983), and the Llanos of Venezuela (González 1997). The tropical dry deciduous forests of central Indochina contain a high density of shallow, seasonal pools, or trapaengs, that provide water and food for an assemblage of large waterbirds and wild ungulates (Critical Ecosystem Partnership Fund 2012). How these pools were formed remains unknown, but a monsoonal climate creates wet-season (May October) floods followed by water drawdown, desiccation, and vegetation senescence in the dry season (November April; Wright et al. 2010). Most large, dry-forest waterbirds adopt seasonal breeding strategies, but the timing varies. One crane and one ibis species nest in the wet season; one ibis, one adjutant, and one other stork nest in the dry season; and one adjutant nests from the late wet season to the middle of the dry season (Clements et al. 2013). Why these birds nest asynchronously and how their foraging ecologies differ is not well understood. The contrast between the sympatric Giant Ibis (Thaumatibis gigantea) and White-shouldered Ibis (Pseudibis davisoni), solitary breeders in the wet and dry seasons, respectively, is particularly intriguing because these critically endangered species use broadly similar habitats, including trapaengs (Wright et al. 2012, BirdLife International 2013). Foraging White-shouldered Ibises depend on trapaengs during the breeding season (Wright et al. 2012), but unlike many waterbirds that forage by wading in pools during the drawdown season (Kushlan 1978), preliminary observations suggest that they rarely feed in water (Wright 2008). Furthermore, Whiteshouldered Ibises provision their nestlings when water is scarcest (Fig. 1). We confirmed the timing and duration of nesting and investigated how foraging White-shouldered Ibises can profit from trapaengs at the driest time of year. To address this question, we measured White-shouldered Ibis habitat use, prey intake rate, and density of prey biomass. We expected that substrates around pool margins (exposed as water recedes) must provide access to abundant food and, given that many ibises probe for prey in wet habitats (Kushlan 1978), we expected that soft, saturated substrate would be most profitable. We also estimated prey depletion at trapaengs in the dry season, and we discuss the likely implications of the White-shouldered Ibis s foraging ecology for its seasonal reproductive strategy. Methods Study area. We studied the foraging ecology of Whiteshouldered Ibises within Western Siem Pang Important Bird Area (IBA), Stung Treng province, northern Cambodia (14 07 N, E; Fig. 2). This 138,000-ha site contains the largest known subpopulation of White-shouldered Ibises, a minimum of 346 birds (Wright et al. 2013b), and comprises deciduous dipterocarp forest interspersed with patches of agriculture, grassland, and other forest types. Trapaengs ( ha) occur frequently, and strongly seasonal rainfall (Fig. 1) creates spatiotemporal variation in trapaeng water levels and substrate moisture (Wright et al. 2010). Drying substrates often crack into polygonal blocks as the dry season progresses, and foraging Wild Boar (Sus scrofa) may also churn up exposed ground (H. L. Wright pers. obs.). Trapaeng habitat survey. To examine changes in the relative extent of habitat moisture types through the White-shouldered Ibis s breeding season, a random sample of 58 trapaengs (from a total of 290 known; Fig. 2) were mapped in the early dry season Fig. 1. Percentage of White-shouldered Ibis nests hatched per half-month period and mean monthly rainfall. Nest data were averaged over three breeding seasons (n = 9, 18, and 20 nests, respectively) in the period Rainfall was averaged over 2 years ( ). Error bars indicate 95% confidence intervals.

3 776 Wright et al. auk, Vol. 130 Fig. 2. Study area in Western Siem Pang Important Bird Area (inset showing location within Cambodia), with the distribution of trapaengs sampled for White-shouldered Ibis intake, prey biomass, and habitat availability ( observation, prey-sampled, and habitat trapaengs, respectively). Observation trapaengs were also prey-sampled. (November 2009) and again in the subsequent middle to late dry season (early March 2010), coinciding with peak chickprovisioning time (Fig. 1). Trapaeng boundaries were defined by a border of terrestrial vegetation that marked maximal wetseason extent; recognized boundaries did not vary through the subsequent dry season. Within each trapaeng, habitat patches were sketch-mapped to distinguish major variation in moisture conditions and vegetation height and/or cover (Wright et al. 2010). Habitat patches (area = 861 ± 1,501 m 2 ; lower quartile = 120 m 2 ; and 4.0 ± 1.3 patches trapaeng 1 ; statistics are reported as means ± SD throughout) were mapped by one observer (H. L. Wright) using a hand-held GPS and laser rangefinder. Patch moisture was classified and recorded as (1) water (pools or flooded animal wallows), (2) saturated substrate (viscous, liquid mud at pool margins or in wallows), (3) moist substrate (damp, solid earth), and (4) dry substrate. Average vegetation height and cover per habitat patch were visually estimated by vegetation type (grass, sedge, reed, herb, and Sesbania spp.), following a training period using tape-rule measurements. We also recorded evidence that local people were harvesting amphibians and swamp eels (Synbranchidae); this food-gathering practice was indicated by levered-out blocks of earth on the trapaeng surface. Maps were georeferenced and digitized in a geographic information system (ArcGIS, version 9.3; ESRI, Redlands, California), and moisture classes were aggregated across multiple patches (following Wright et al. 2010) to calculate their percentage extent at each trapaeng. Preliminary analyses found that Whiteshouldered Ibises intake rate was unaffected by vegetation biomass when controlling for moisture (online Supplementary Material, Table S1; see Acknowledgments); therefore, analyses presented here focus on moisture effects, and vegetation is not considered further. Mean availability of moisture classes at the 58 repeat-sampled trapaengs was compared between the early and middle to late dry seasons using paired Wilcoxon tests (with Holm adjustment for Type I error rate). Prey sampling at trapaengs. We measured the density of prey biomass in exposed trapaeng substrates by moisture class in habitat patches at 47 trapaengs (55% overlap with habitatmapped trapaengs; Fig. 2) in the (n = 20) and (n = 27) breeding seasons. In each year, prey sampling took place evenly across the 4-month period of late November to early April. Trapaengs were randomly selected after stratification by their distance to nearest settlement (0 2.9 km, n = 11; km, n = 12; km, n = 11; 9 12 km, n = 13) because we anticipated that harvesting of amphibians and swamp eels would affect prey biomass (particularly at trapaengs close to villages). However, the trapaeng habitat survey (above) found that harvesting took place at only 11 of 58 trapaengs (19%) in the middle to late dry season and affected only 4.2 ± 3.3% of the surface areas of these 11 trapaengs. Harvesting at prey-sampled trapaengs proved even scarcer, and

4 October 2013 Foraging White-shouldered Ibises 777 although we cannot estimate the quantity of biomass removed, the small areas involved suggest that the effect on prey biomass was probably small. Prey samples were collected using 10 soil cores per trapaeng, taking at least three cores from each substrate moisture class (dry, moist, and saturated), or five from each when only two moisture classes were present. Waterborne prey and benthic substrates were not sampled, because White-shouldered Ibises rarely feed in water (Wright 2008). Within each moisture class, cores were taken from a range of vegetation conditions, representative of those mapped in trapaengs across the study area (above). Cores measured cm in surface area and 18 cm in depth (equivalent to an adult male White-shouldered Ibis s bill length), providing 5.29 m 3 total sampled volume across all trapaengs. Core dimensions were sufficient to capture the small prey routinely caught by Whiteshouldered Ibises; larger prey (such as eels >25 cm in length) may be underrepresented but were almost never caught by White-shouldered Ibises. Although cores were dug rapidly, some items could have escaped (e.g., amphibians withdrawing into deep cracks or swamp eels in wetter substrates withdrawing into burrows). For each core, the prey type (amphibians, small invertebrates [listed in online Supplementary Material, Table S2; see Acknowledgments], crabs, swamp eels, and snakes) and size (classified into body-length intervals of cm, cm, and 5 cm) of each item were counted; amphibian identification followed Neang and Holden (2008). Items <1 cm in size were rarely consumed and were excluded from analysis. Biomass was estimated from means of centigram ash-free dry mass (cg; AFDM) per prey type per size class (Piersma et al. 1994), determined from a sample of specimens collected at trapaengs (Table S2). Mapping of habitat patches followed procedures for the trapaeng habitat survey (above). Substrate microtopography (cracked holed vs. even uncracked ground) was also assessed; this varied at a finer scale than classes of substrate moisture and was recorded per soil core. Our data measure biomass density, but biomass availability to White-shouldered Ibises may differ among prey types and substrates, being lower, for example, in compacted, dry mud that cannot be probed. Prey data were summed per habitat patch; prey biomass density was not estimated per trapaeng, because placement of cores was not proportional to the area of moisture classes in each trapaeng. The proportionate contribution of each prey type to total biomass density (cg AFDM totaled across trapaengs) was estimated from its contribution within each substrate, multiplied by the substrate s total proportionate extent across the 58 trapaengs mapped in the middle to late dry season. Foraging observations. Intake rates, habitat use, diet, and activity of White-shouldered Ibises were measured at 7 of the 47 prey-sampled trapaengs (Fig. 2) during the and breeding seasons. Observations were spread across a 4-month period in each year, from early December to early April, corresponding with the timing of prey sampling. Following preliminary observations, we chose trapaengs that contained a range of moisture conditions and maximized the likelihood of visitation by White-shouldered Ibises and, thus, data collection (necessary to overcome the scarcity of encounters with this rare species). Five trapaengs were sampled in one year and two in both years, providing nine trapaeng year observation periods; trapaengs sampled in both years were observed in different months under different moisture conditions in each year. Observation trapaengs were larger, with a more even composition of substrate moisture types than those across the wider landscape (further details are presented in online Supplementary Material, Table S3; see Acknowledgments), but patterns in prey biomass density (Table S4) were similar to those across all prey-sampled trapaengs (Table 1). Observation trapaengs were situated along a gradient of distance to settlement (5.62 ± 2.36 km; range: km); harvesting of amphibians and swamp eels had occurred in 3 of the 9 trapaeng years, but again only small areas were affected (3.9 ± 3.5 % of the surface of these 3 trapaengs), and anthropogenic prey depletion probably had only a minor influence on foraging White-shouldered Ibises. Habitat patches were mapped and recorded following protocols used for the trapaeng habitat survey. One person (H.L.W.) undertook observations from dawn to dusk, for a mean 4.0 ± 0.7 continuous days per site, using a telescope (32 magnification) from blinds at trapaeng perimeters. The observer was typically 2 40 m from foraging birds, allowing prey Table 1. Influence of substrate moisture on White-shouldered Ibis intake rate and biomass density of amphibians and small invertebrates. Dry substrate and year 1 ( ) were reference classes ( ), and comparisons of saturated versus moist substrate (given by reordering classes in the model) are shown in italics. Date is the number of days since the sampling (dry) season began. Amphibians Small invertebrates Ibis intake rate model Prey biomass density model Ibis intake rate model Prey biomass density model Term β ± 95% CI Z P β ± 95% CI Z P β ± 95% CI Z P β ± 95% CI Z P Intercept 1.98 ± ± ± < ± Substrate moisture Dry Moist 0.46 ± < ± ± ± < Saturated 1.73± < ± ± < ± Saturated 1.27 ± < ± ± < ± vs. moist Year ± ± ± < ± Date 0.01 ± ± ± ±

5 778 Wright et al. auk, Vol. 130 captures to be seen clearly. Individual birds and breeding pairs could not be individually identified, so we cannot estimate the number of birds observed. However, the broad spacing of observation trapaengs (range: km; Fig. 2) improved the likelihood of observing multiple birds, given that White-shouldered Ibises disperse widely to breed (H. L. Wright unpubl. data). Within trapaengs, repeat observation of single individuals was reduced at three sites by the regular presence of birds (suspected to be flocks of failed, later-breeding, and/or nonbreeding birds), but other trapaengs typically hosted 1 3 birds at a time. Although we acknowledge that repeat observations of the same individuals are non-independent, approximately different individuals are likely to have been observed overall (estimation described in online Supplementary Material, Table S5; see Acknowledgments). Intake rates were obtained from replicate six-minute focal samples of adult birds (totaling h); recently fledged juveniles were excluded. For each capture, the type and size of the prey item were recorded (prey body length visually estimated in relation to Whiteshouldered Ibis bill length) using the categories applied for prey sampling and biomass measurement (above). Within focal observations, White-shouldered Ibises use of habitat patches was timed (assisted by markers placed on habitat patch boundaries), enabling habitatspecific intake rates to be calculated. Focal observations rotated or alternated between individuals when more than one bird was present; however, some individuals were observed repeatedly when no other birds were active or present. Because none of the habitats used by White-shouldered Ibises was densely vegetated, there was no bias toward more easily observed birds or habitats. We recorded White-shouldered Ibis habitat use by instantaneous scan-samples, scanning all visible individuals at 6-min minimum intervals to record their location (habitat patch identification). Habitat use was assessed by comparing the proportionate use of habitat moisture classes (from scansample records) to proportionate availability (from digitized habitat maps) across trapaeng years. We also recorded Whiteshouldered Ibis activity (feeding yes no) during each scan-sample to estimate the portion of the day spent foraging. Repeat observation of individuals and resulting non-independence of data can be problematic in both scan- and focal sampling, but can at least partly be mitigated by allowing time between observations (Sutherland 2004). Scan- and focal samples were typically taken alternately, thereby providing time for White-shouldered Ibises to move to different parts of trapaengs between samples. In biological terms, non-independence could potentially result in the oversampling of individuals with particular foraging efficiencies or specializations, making observation data unrepresentative of the overall population. This potential limitation cannot be ruled out and, therefore, results of our statistical analyses should be treated with caution. Nevertheless, given the large number of individuals observed, these findings remain a valuable starting point for understanding the White-shouldered Ibis s ecology. Nest and rainfall observation. White-shouldered Ibis nests were monitored to determine the timing and duration of breeding. Nests were located using a locally advertised reward scheme, giving villagers a small cash incentive for reliable nest reports (Wright et al. 2013a). Field staff undertook additional active searching at previous nesting locations, and at sites where White-shouldered Ibis pairs were regularly encountered. Nests were monitored every 5 7 days using ground-based observations. Each monitoring visit lasted min, sufficient to determine the nest s status. Rainfall was measured daily from April 2008 to May 2010 using a rain gauge in Siem Pang town (Fig. 2). Analyses. White-shouldered Ibis dietary composition was estimated from the proportionate contribution of each prey type to overall intake rate in each substrate type (from focal sample data across prey types and trapaeng years, AFDM per minute), multiplied by the proportionate use of that substrate type by White-shouldered Ibises (from scan-samples). Differences in overall Whiteshouldered Ibis intake rate between substrate moisture classes (dry, moist, and saturated) were compared with Mann-Whitney tests with Holm adjustment. The effects of substrate moisture class on White-shouldered Ibis intake rates and prey biomass density were then modeled separately for amphibians and small invertebrates, the White-shouldered Ibises main prey items. Intake rates and prey biomass densities were non-normal and overdispersed, with frequent zeros. Therefore, to allow Poisson errors to be fitted, both were converted from AFDM to count data, accounting for differences in biomass by standardizing counts to the equivalent number of prey of the smallest size class (for amphibians, cm; for small invertebrates, insect larvae of size cm; Table S2). Intake rate (count of smallest-item equivalents, per habitat patch per focal observation, n = 1,927), and biomass density (count of smallest-item equivalents summed across soil cores per habitat patch, n = 159) were modeled in generalized linear mixed models (GLMMs) with Laplace approximation in R, version 2.15 (Bolker et al. 2009, R Development Core Team 2012). Poisson and negative binomial error distributions available in the glmmadmb package (Skaug et al. 2012) were initially compared (with and without zeroinflation) for each model, and that which resulted in the lowest model Akaike s information criterion (AIC) value was selected for final modeling (following Bolker et al. 2012). Final models were run in the lme4 package (Bates et al. 2011) and used log-normal Poisson error containing an observation-level random effect (the identification number of each observation) to model extra-poisson variation (Maindonald and Braun 2010). Results were checked for consistency by remodeling with alternative error distributions and, where possible, using both lme4 and glmmadmb. Models of White-shouldered Ibis intake rate included the log number of minutes per habitat patch per focal observation as an offset, and models of prey biomass density included the log number of soil cores per habitat patch as an offset. Sampling date (days since the sampling season began; fixed effect) was also included to assess temporal changes over the 4-month sampling periods. Year and site (trapaeng ID) were included in models as fixed and random effects, respectively, to account for grouping of prey-sampled habitat patches by trapaeng (models of biomass density) or repeat visits to observation trapaengs (models of intake). Results revealed that large amphibians ( 5 cm in length) provided a substantial part of White-shouldered Ibis intake; however, we could only approximately assess their contribution to consumed and available prey biomass, because AFDM varied widely among specimens of this size class (more so than for other amphibian size classes; Table S2). To compare the relative importance of substrates to captures of large amphibians (and corroborate results from the model of amphibian intake), we undertook a supplementary chi-square test to compare the

6 October 2013 Foraging White-shouldered Ibises 779 frequency of White-shouldered Ibis foraging observations with or without captures of these prey. Biomass density data included only four large amphibians. During data collection, we also observed White-shouldered Ibises frequently catching amphibians from cracks and holes in dry substrate. Microtopography was not recorded in foraging observations, but we undertook a supplementary analysis to assess its effect on amphibian biomass density in dry substrate (n = 207 soil cores). Microtopography, date, and year (fixed effects) were included in a GLMM with lognormal Poisson error and site as a random effect. The potential for White-shouldered Ibises to deplete available prey was estimated to examine their susceptibility to intraspecific competition. The predicted prey consumption by one pair feeding for 69 days (duration of the nesting period, determined by nest observations) was compared to the mean prey biomass available in the middle to late dry season at a large trapaeng (defined by the upper quartile area of habitat-surveyed trapaengs). Only large trapaengs were considered to match the positive size bias in observation trapaengs (Table S3). If estimated prey consumption exceeds available biomass, nominal depletion rates of >100% are predicted, which indicates that multiple trapaengs would be required by White-shouldered Ibis pairs to sustain a breeding attempt. On the basis of nest monitoring and nest camera observations (Wright et al. 2013b) that showed nest attendance by adults, we assumed that only one bird of the pair could feed at a time during incubation and brooding (44 days), after which both birds fed and provisioned chicks simultaneously. Foraging duration per day (assuming 11.8 daylight hours) was estimated from the percentage of scan-samples that recorded foraging by White-shouldered Ibises in four time brackets ( , , , and hours), accounting for varying activity patterns and numbers of observed birds with time of day. Time spent traveling to trapaengs or visiting nests for chick provisioning could not be gauged; however, having already accounted for time spent incubating and brooding, provisioning commutes probably occupied only a small proportion of daily activity and produced only slight overestimation of foraging time. Depletion of prey biomass was estimated for all consumed prey types (cg AFDM, excluding crabs and swamp eels, which were rarely consumed) and for that part of the diet comprising amphibians alone (count of smallest-item equivalents). Average intake rates per substrate moisture type were scaled up to trapaeng level on the basis of average proportionate use of these substrates. Prey biomass density averages were scaled up using average proportionate extent of dry, moist, and saturated substrates per trapaeng in early March, when large areas of substrate are exposed. We calculated boundaries of uncertainty by simulating depletion 1,000 times (randomly sampling parameter values for each iteration from a normal distribution defined by the mean and standard error of estimates) and identifying the 95% confidence intervals (CI) within the distribution of predicted depletion estimates. combined took 44 ± 3 days (n = 17). Rainfall during the breeding season averaged 28 ± 56 mm month 1. Habitat change at trapaengs. The extent of water and dry substrate at trapaengs changed dramatically from the early to the middle and late dry seasons (November early March). Mean water cover dropped from 79.7% to 5.6% over the 4 months (Wilcoxon test V = 1,711, df = 58, P < 0.001), while mean dry substrate increased from 4.3% to 87.3% (V = 1,711, df = 58, P < 0.001), becoming by far the most abundant substrate type. Moist substrate extent did not differ between the early and middle to late dry seasons (V = 662, df = 52, P = 0.468), but saturated substrate cover decreased significantly, from 7.9% to 2.2% (V = 1,236, df = 52, P < 0.001). Foraging activity and habitat use. Observations at trapaengs provided 5,122 records of White-shouldered Ibis activity and habitat use (including repeat-observation of individuals) from 1,477 scan-samples (mean = ± 80.5 scans trapaeng year 1 ; range: ). The percentage of records that involved foraging individuals was similar across periods of the day (F = 0.84, df = 3 and 32, P = 0.484), averaging 80.0 ± 8.6% overall. Foraging White-shouldered Ibises made negligible use of water in relation to exposed substrates, with a mean of only 0.2 ± 0.3% foraging records per trapaeng in pools of water, considering 7 trapaeng years with both aquatic and substrate habitats available. Whiteshouldered Ibises fed in all substrate moisture conditions (Fig. 3): dry (commonest), moist, and saturated (scarcest). The proportion of foraging records for each moisture type was variable across trapaeng years, but overall, mean proportionate use of substrates was similar to their mean proportionate availability (Fig. 3). Composition of diet and prey biomass density. Amphibians formed the majority of the White-shouldered Ibis s diet at trapaengs (Fig. 4A), providing an estimated 80.6% of overall biomass intake. Most amphibians caught by White-shouldered Ibises were <5 cm in body length (99% of individual captures), but larger amphibians contributed to 19.5% of overall biomass intake, despite being caught much less frequently. Amphibians were also the most abundant prey type in trapaeng substrates (Fig. 4B), contributing 53.3% of estimated prey biomass density (accounting for average extent of substrate types). Small invertebrates accounted for just 9.7% of overall intake, compared with 20.2% of Results Timing of breeding. White-shouldered Ibises began breeding in the early dry season (Fig. 1), with the earliest incubation record on 19 December in Incubation lasted 31 ± 3 days (n = 17), with the nestling period (39 ± 7 days, n = 22) taking place from January to early May. Incubation and brooding of young nestlings Fig. 3. Mean, proportionate use of substrate moisture conditions by White-shouldered Ibises (dark gray) and substrate proportionate availability (pale gray) over 9 trapaeng year observations. Proportionate use was determined from scan-sampled foraging records, and availability is in relation to total substrate area per trapaeng year. Bars indicate 95% confidence interval (CI).

7 780 Wright et al. auk, Vol. 130 Fig. 4. White-shouldered Ibis intake rate and biomass density of all prey types (A and B, respectively) and of amphibians (C and D) by substrate moisture type. Column totals show (A) overall mean intake rate and (B) mean prey biomass density in each substrate, across prey types. Column subdivisions indicate proportionate prey type composition of (A) overall prey biomass intake rate and (B) overall prey biomass density (cg ash-free dry mass [AFDM] min 1 and g AFDM/m 3, respectively). Small invert. is small invertebrate; Eel is swamp eel. Amphibian intake rate (min 1 observation 1 ) and biomass density (soil core 1 habitat patch 1 ) are the square roots of counts of smallest item equivalents. Boxes indicate interquartile ranges (IQR; C and D), thick bars indicate medians and circles are outliers (1 IQR beyond the upper quartile). Intake rate data comprised 676 focal observations in dry, 623 in moist, and 628 in saturated substrates. Biomass density data comprised 207 soil cores and 76 habitat patches in dry, 129 cores and 45 patches in moist, and 123 cores and 38 patches in saturated substrates. prey biomass density. No crabs and only one small swamp eel were caught by White-shouldered Ibises, despite together accounting for 21.6% of prey biomass density. Unidentified prey items, probably small invertebrates or parts of amphibians, comprised an estimated 8.8% of overall White-shouldered Ibis intake. Influence of substrate moisture on intake rate and prey biomass density. Overall White-shouldered Ibis intake rate (pooled across trapaeng years and combining all prey types) varied among substrate moisture conditions. The overall intake rate in dry substrate was variable and, after accounting for Type I error rate, was not significantly different from that in moist (W = 197,102.5, df = 676 and 623, P = 0.063) or saturated (W = 226,232.5, df = 676 and 628, P = 0.063) substrates (Fig. 4A). However, intake was greater in moist substrate than in saturated substrate (W = 228,497.5, df = 623 and 628, P < 0.001; Fig. 4A), where intake rate was low, given that this held a combined biomass density of amphibians and small invertebrates similar to or higher than that in dry or moist substrates (Fig. 4B). Intake rate of amphibians differed among all substrate moisture types, being significantly greater in dry than in moist, and in moist than in saturated substrates (Table 1 and Fig. 4A, C). The contrast between dry and moist substrates may be related to differences in intake of large amphibians ( 5 cm in length), which contributed to 38% of outliers in intake rate data (Fig. 4C). The frequency at which White-shouldered Ibises captured large amphibians varied among substrate types (χ 2 = 15.8, df = 2, P < 0.001), with dry substrate providing more and moist substrate providing fewer captures than expected by chance. Amphibian biomass density was also greater in dry than in saturated substrate (Table 1 and Fig. 4B, D) but did not differ between dry and moist substrates. Supplementary analysis in relation to substrate microtopography showed that amphibian biomass density in soil cores was significantly greater in cracked holed than in even uncracked dry substrate (online Supplementary Material, Table S6; see Acknowledgments). The intake rate of small invertebrates was significantly greater in saturated (89.5% of intake of small invertebrates) than in dry and moist substrates (Table 1 and Fig. 4A), whereas biomass

8 October 2013 Foraging White-shouldered Ibises 781 density of small invertebrates was greater in moist and saturated than in dry substrates (Table 1 and Fig. 4B). Date had nonsignificant effects on intake and biomass density of both prey types during the dry season. Prey depletion. The scenario of a White-shouldered Ibis pair feeding at a large trapaeng over one nesting period (69 days) predicted 61.6% (95% CI: %) depletion of amphibian biomass and 43.5% (95% CI: %) depletion of biomass from all prey types combined. Greater depletion rates for amphibians than for all prey combined reflect the Whiteshouldered Ibis s apparent selection of frogs and more limited use of small invertebrates, in relation to biomass densities. Discussion Dry-season foraging ecology. As preliminary observations suggested (Wright 2008), White-shouldered Ibises virtually never fed in aquatic habitats at trapaengs, instead using substrates exposed around drying pools. This habitat preference resembles that of terrestrial-feeding ibises, including the Buff-necked Ibis (Theristicus caudatus) in Venezuelan savanna (Frederick and Bildstein 1992) and the Northern Bald Ibis (Geronticus eremita) in arid environments (Serra et al. 2008). White-shouldered Ibises foraged in a range of substrate moisture conditions (dry, moist, and saturated), but their relative use varied among trapaeng years. Assessing preference for these conditions was complicated by their close proximity within trapaengs and the birds use and movement between them all. Furthermore, the results derived from foraging observations require cautious interpretation because of the uneven contribution of different individuals to repeated observations. Overall intake rate (combining all prey types) differed only between dry and saturated substrates, but further effects were found when we treated major prey types (amphibians and small invertebrates) separately. Contrary to our expectation, dry rather than saturated substrate was the most profitable, providing the highest intake rate and containing high biomass densities of amphibian prey. Higher intake rates in dry than in moist substrate are most likely to be explained by captures of large amphibians (low frequency overall, but caught most often in dry substrate) rather than routinely higher intake rates of all amphibian prey. Nevertheless, given that amphibians contributed 8 more biomass to White-shouldered Ibis intake than any other prey type, dry and also moist substrates were clearly key microhabitats. High intake and biomass density of amphibians in dry substrate were probably both related to deep cracks (caused by desiccation) and holes (created by boars), where amphibian biomass density was higher than in even uncracked ground. With their decurved bills, White-shouldered Ibises are well adapted to probing into cracks and crevices (Cunningham et al. 2010), and dry substrate generally contained more crevices than moist or saturated ground (H. L. Wright pers. obs.), presumably owing to greater desiccation and longer exposure to boars. Amphibian habitat use in Indochinese dry forests is not well understood, but cracks and holes may be important refuges, providing access to moist, cool conditions away from hot ground surfaces. Microhyla frogs, and secondarily Paddy Frogs (Fejervarya limnocharis) dominated the amphibians consumed by White-shouldered Ibises and those found in trapaeng substrates. Our findings demonstrate that foraging White-shouldered Ibises profit from dry-season wetlands by exploiting seasonal prey concentrations within exposed substrates, especially in dry ground. This is an unusual strategy among large waterbirds. Although many large egrets, herons, ibises, and storks (including several dry-forest species co-occurring with the White-shouldered Ibis) also benefit from increasingly concentrated and/or accessible prey in drying wetlands, these typically select wet or saturated rather than dry habitats (Kushlan 1978, Russell et al. 2002, H. L. Wright pers. obs.). Other storks and many cranes forage in a wide range of moisture conditions (del Hoyo et al. 1992, Meine and Archibald 1996) but rarely dig or probe deeply for prey in dry ground. The sympatric Giant Ibis adopts a more aquatic foraging strategy than White-shouldered Ibis, frequently probing in saturated mud and shallow water (Wright et al. 2012). Foraging White-shouldered Ibises appeared to select amphibian prey, which formed a larger proportion of intake than of prey biomass density. However, this disparity is partly related to the absence of swamp eels and crabs in the White-shouldered Ibis diet, in that these organisms contributed one-fifth of prey biomass density. Swamp eels and crabs were found in saturated substrate and may have escaped predation by withdrawing into burrows (Hill and Watson 2007), burying themselves, or moving to nearby water (where White-shouldered Ibises did not feed). Similar avoidance techniques may explain the disproportionately low intake, in relation to their biomass, of amphibians in saturated substrate. Intake of small invertebrate biomass was half that suggested by their biomass density and contributed little to White-shouldered Ibises diet. Calorific prey contents in relation to the costs of different feeding strategies are not yet known, but invertebrates may provide a lower net energy gain compared to other prey. White-shouldered Ibises consumed most small invertebrates in saturated substrates, although more than half of small invertebrate biomass occurred in moist and dry substrates. Given that substrate penetrability influences foraging success and energy costs (Mouritsen and Jensen 1992), White-shouldered Ibises may have found it more profitable to catch invertebrates in soft, saturated mud. Foraging for small invertebrates may have provided White-shouldered Ibises with an alternative feeding strategy, possibly adopted when amphibians in dry and moist substrates became depleted. Food supply and White-shouldered Ibis reproduction. White-shouldered Ibises forage almost exclusively at trapaengs in the breeding season (Wright et al. 2012), and so the seasonality of reproduction is likely to be linked to food availability there. Trapaengs in the early dry season contained a small proportion of exposed substrate, mostly comprising less profitable moist and saturated ground. Although intake rates and prey biomass densities showed no clear differences across the dry season, the extent of exposed substrate increased substantially (20-fold for profitable dry substrate, from November to March), presumably making trapaengs increasingly attractive to White-shouldered Ibises. The species breeding may, therefore, be timed to coincide with water drawdown, a strategy documented in other wetland birds (Kushlan et al. 1975, Kushlan 1979, Morrison et al. 2009). To confirm this, further research should compare prey availability and foraging success in the breeding and nonbreeding seasons, and examine the potential links among rainfall, habitat extent, nesting date,

9 782 Wright et al. auk, Vol. 130 and breeding productivity. This may also help to determine the mechanisms that initiate breeding activity, because food can influence tropical birds both as an ultimate and as a proximate factor in the timing of reproduction (Fogden 1972, O Brien and Hau 2005). We estimated that a single breeding pair of White-shouldered Ibises foraging at one large trapaeng would cause considerable prey depletion, although uncertainty surrounding estimates was large. Without knowing the functional response of White-shouldered Ibis intake to prey density, we cannot accurately quantify the extent of habitat needed; however, with amphibians predicted to become depleted by nearly two-thirds, each White-shouldered Ibis pair plausibly requires more than one trapaeng to breed successfully, an inference supported by the White-shouldered Ibis s dispersed, territorial, and solitary nesting behavior (H. L. Wright unpubl. data) and consistent with intraspecific competition in a prey-scarce environment (Lack 1968). Our estimates of prey depletion are conservative in that we used the near-maximal extent of substrate found in the dry season, a 0.36-ha trapaeng area (larger than 75% of trapaengs in the landscape), and did not consider provisioning of fledged offspring. Prey replenishment was not considered but is unlikely to be significant, given that frogs in the genera Microhyla and Fejervarya spawn mainly in the wet season (Heyer 1973), making congregative breeding movements to trapaengs less likely in the middle to late dry season. Wetland resilience and conservation. Wetlands in Indochinese dry forests are threatened by changing land use and climate (Critical Ecosystem Partnership Fund 2012). Habitat conversion is the largest threat to the White-shouldered Ibis (Wright et al. 2013a), driven by dam construction, expansion of agro-industrial plantations, and growth of local agriculture. If each breeding pair requires multiple trapaengs, and intraspecific competition underpins a dispersed distribution within breeding populations, wetland habitat needs protection at the landscape scale. The effects of anthropogenic climate change are not yet clearly understood, but altered rainfall and evaporation would probably affect trapaeng hydrology, especially during the Whiteshouldered Ibis s breeding season when water stress is already high (Timmins 2011). Further study of trapaeng formation and optimal configuration will assist conservation efforts to improve waterhole resilience. However, attempts to manage trapaengs must balance the requirements of a suite of dry-forest waterbirds, with contrasting foraging and breeding strategies, as well as the herbivore populations that form the prey base for emblematic carnivores. Acknowledgments Supplemental material for this article is available at dx.doi. org/ /auk We thank M. Mem, B. P. Lourn, and P. Sum for assisting with data collection; BirdLife International in Indochina, particularly V. Bou and N. Net, for field-work assistance; T. Clements, J. Eames, T. Evans, and O. Keo for thoughts on ibis foraging ecology; P. Men and H. E. K. S. Cheng of Forestry Administration, and H. E. S. Chay of GDANCP, Ministry of Environment of Cambodia, for support; and D. Baker for comments on the manuscript. Funding was provided by Critical Ecosystem Partnership Fund (a joint initiative of l Agence Française de Développement, Conservation International, Global Environment Facility, Government of Japan, MacArthur Foundation and World Bank), Natural Environment Research Council, and Economic and Social Research Council. Literature Cited Bates, D., M. Maechler, and B. Bolker lme4: Linear mixed-effects models using S4 classes. [Online.] Available at R-Forge.R-project.org/projects/lme4. Berruti, A The biomass, energy consumption and breeding of waterbirds relative to hydrological conditions at Laicx St Lucia. Ostrich 54: Bildstein, K. L., W. Post, J. Johnston, and P. Frederick Freshwater wetlands, rainfall, and the breeding ecology of White Ibises in coastal South Carolina. Wilson Bulletin 102: BirdLife International IUCN Red List for birds. BirdLife International, Cambridge, United Kingdom. [Online.] Available at Bolker, B. M., M. E. Brooks, C. J. Clark, S. W. Geange, J. R. Poulsen, M. H. H. Stevens, and J.-S. S. White Generalized linear mixed models: A practical guide for ecology and evolution. Trends in Ecology & Evolution 24: Bolker, B., H. Skaug, A. Magnusson, and A. Nielsen Getting started with the glmmadmb package. [Online.] Available at glmmadmb.r-forge.r-project.org/glmmadmb.pdf. Clements, T., H. J. Rainey, D. An, V. Rours, S. Tan, S. Thong, W. J. Sutherland, and E. J. Milner-Gulland An evaluation of the effectiveness of a direct payment for biodiversity conservation: The Bird Nest Protection Program in the Northern Plains of Cambodia. Biological Conservation 157: Critical Ecosystem Partnership Fund Indo-Burma Biodiversity Hotspot, Indochina Region: Ecosystem Profile 2011 Update. Critical Ecosystem Partnership Fund, Arlington, Virginia. Crome, F. H. J To drain or not to drain? Intermittent swamp drainage and waterbird breeding. Emu 88: Cunningham, S. J., M. R. Alley, I. Castro, M. A. Potter, M. Cunningham, and M. J. Pyne Bill morphology of ibises suggests a remote-tactile sensory system for prey detection. Auk 127: del Hoyo, J., A. Elliott, and J. Sargatal Handbook of the Birds of the World, vol. 1: Ostrich to Ducks. Lynx Edicions, Barcelona, Spain. Fogden, M. P. L The seasonality and population dynamics of equatorial forest birds in Sarawak. Ibis 114: Frederick, P. C., and K. L. Bildstein Foraging ecology of seven species of Neotropical ibises (Threskiornithidae) during the dry season in the Llanos of Venezuela. Wilson Bulletin 104:1 21. Frederick, P. C., and M. W. Collopy Nesting success of five Ciconiiform species in relation to water conditions in the Florida Everglades. Auk 106: Gawlik, D. E The effects of prey availability on the numerical response of wading birds. Ecological Monographs 72: González, J. A Seasonal variation in the foraging ecology of the Wood Stork in the southern Ilanos of Venezuela. Condor 99: