Breeding populations of Marbled Godwits and Willets have high annual survival and strong site fidelity to managed wetlands

Abstract The Prairie Pothole Region of central Canada supports a diverse community of breeding waterbirds, but many species have declining populations and the demographic mechanisms driving the declines remain unknown. We conducted a 7‐year field study during 1995–2001 to investigate the demographic performance of Marbled Godwits (Limosa fedoa) and Willets (Tringa semipalmata) breeding in managed wetlands near Brooks, Alberta. Mark‐recapture analyses based on Cormack–Jolly–Seber models revealed that the annual rates of apparent survival for Marbled Godwits (ϕ^ = 0.953 ± 0.012SE) and Willets (ϕ^ = 0.861 ± 0.015SE) are among the highest rates of survivorship reported for any breeding or nonbreeding population of large‐bodied shorebirds. Our estimates of life expectancy for males were comparable to longevity records in godwits (17.3 years ±5.8SE vs. 25–29+ years) and willets (7.7 ± 1.5SE vs. 10+ years). The two species both showed strong breeding site fidelity but differed in rates of mate fidelity. Pairs that reunited and males that switched mates usually nested <300 m from their previous nests, whereas females that switched mates usually moved longer distances >1.1–1.5 km. Returning pairs usually reunited in godwits (85%) but not in willets (28%), possibly because of species differences in adult survival or patterns of migration. Baseline estimates of annual survival for banded‐only birds will be useful for evaluating the potential effects of new tracking tags or the environmental changes that have occurred during the past 20 years. Conservation strategies for large‐bodied shorebirds should be focused on reduction of exposure to anthropogenic mortality because low rates of natural mortality suggest that losses to collisions at breeding sites or harvest at nonbreeding areas are likely to cause additive mortality.


| INTRODUC TI ON
The Prairie Pothole Region of central Canada is a complex landscape of semipermanent wetlands that provide important breeding habitat for migratory waterbirds, including high densities of nesting waterfowl, shorebirds, wading birds, and marsh songbirds. Habitat availability has been reduced by losses to drainage and conversion to cultivation or pasture (Doherty et al., 2018;Gratto-Trevor, 2006), although conservation easements for waterfowl production have the potential to benefit a diversity of wildlife species (Koper & Schmiegelow, 2006;Tori et al., 2002). Habitat conditions in the Prairie Pothole Region are usually determined by annual variation in precipitation which affects soil moisture, water depth, and vegetative cover, but with predictions of drier conditions under future climate change (Niemuth et al., 2010). The local abundance and distribution of migratory waterbirds varies with annual conditions but many species are also showing long-term population declines (Forcey et al., 2007;Niemuth et al., 2008;Steen et al., 2014). Major threats to migratory waterbirds are thought to include habitat loss and degradation, climate change, and unsustainable harvest Pearce-Higgins et al., 2017;Watts et al., 2015), but the demographic mechanisms of ongoing population declines remain poorly understood. Reliable estimates of fecundity and survival are fundamental to conservation planning but remain unavailable for many widespread species (Méndez et al., 2018;Piersma et al., 1997;Sandercock, 2003). Baseline information on demographic rates for stable populations are particularly useful for identifying environmental perturbations that are likely to reduce future population viability (Burton et al., 2006;McDuffie et al., 2022;Piersma et al., 2016;Taylor & Dodd, 2013).

Marbled Godwits (Limosa fedoa) and Willets (Tringa semipalmata)
are two species of large-bodied shorebirds that breed in the Prairie Pothole Region of central Canada and the USA (Gratto-Trevor, 2006;Specht et al., 2020). The biogeography of these species in North America is complex with three separate breeding populations of Marbled Godwits and two disjunct breeding populations of Willets (Gibson & Kessel, 1989;Martínez-Curci et al., 2014;Oswald et al., 2016). The Prairie Pothole Region supports large breeding populations of the midcontinental subspecies (L. f. fedoa and T. s. inornata), and Canada has significant responsibility for stewardship of these birds. Breeding populations of godwits and willets have been declining at −0.8% to −2.1% per year in the Prairie Pothole Region of Canada during the last half century (Figure 1), although nonbreeding populations in coastal California appear to be more stable (Warnock et al., 2021). Both species are categorized as species of 'High Concern' in assessments of Shorebirds of Conservation Concern in Canada (Hope et al., 2019) and the United States (US Shorebird Conservation Plan, 2016).
Large-bodied shorebirds are often characterized by a "slow" lifehistory strategy with low reproductive rates, delayed maturity, and high survivorship (Myers et al., 1987;Watts et al., 2015). Shorebirds with socially monogamous mating systems also tend to have strong fidelity to breeding sites and to mates (Kwon et al., 2022;Oring & Lank, 1984). Thus, population growth rates are likely to be more sensitive to factors that affect adult survival and breeding site fidelity than components of reproduction (Hitchcock & Gratto-Trevor, 1997;Koivula et al., 2008;Ottvall & Härdling, 2005). Previous population studies of Marbled Godwits and Willets have been limited but have provided basic information on migratory connectivity (Gratto-Trevor, 2011;Haig et al., 2002;Olson et al., 2014), habitat preferences (Gratto-Trevor, 2006;Ryan & Renken, 1987;Specht et al., 2020), nesting success (Garvey et al., 2013;Kantrud & Higgins, 1992), and return rates or longevity (Colwell et al., 1995;Howe, 1982;Kelly & Cogswell, 1979). No previous analyses of adult demography exist for the prairie populations of Marbled Godwits or Willets. The three objectives of our field project were (i) to evaluate use of biometrics for sexing of adult birds, (ii) to conduct mark-recapture analyses to estimate apparent survival corrected for variation in the probability of encounter, and (iii) to use records of returning birds to compare patterns of mate and site fidelity between the two species.
Additional monitoring was conducted on two smaller patches to the west (2.7 km 2 ) and to the southeast (4.4 km 2 ) of the core study area.
The semi-arid habitats included native mixed-grass prairie interspersed with a network of ca. 50 small wetlands ranging in size from 16-120 ha. The study site was managed by Ducks Unlimited Canada for breeding waterfowl and cattle production, with flooding in spring F I G U R E 1 Population indexes and long-term trends from the Breeding Bird Survey program for Marbled Godwits and Willets in the Prairie Pothole Region of Canada (BCR 11), 1970. Source: Smith et al. (2020. and autumn via irrigation canals connected to the Bow River. Other infrastructure included pumpjacks and storage tanks for oil and gas extraction and transmission powerlines.
Annual variation in wetland conditions and small mammal numbers were assessed at a regional level (Table 1). For an index of wetland conditions, we used pond counts recorded in May for stratum 28 (ca. 40 K km 2 ) in the annual Waterfowl Breeding Population and Habitat Surveys coordinated by the Canadian Wildlife Service. High numbers of microtine rodents were observed during field work in 1997 but were low in other years. For an index of rodent activity and predator numbers, we compiled information on the relative abundance of Short-eared Owls (Asio flammeus) from June surveys in Alberta conducted under the Breeding Bird Survey program (Specht & Arnold, 2018

| Field methods
The study site at Kitsim was searched for nests twice per year during the first 6 years of the study. Nests were found by cable dragging with a 30 m cable or chain dragged slowly between a pair of allterrain vehicles. At first discovery, nest sites of shorebirds and waterfowl were marked with a pin flag. Nest locations were recorded on high-resolution aerial photos and later digitized with UTM coordinates for UTM zone 12 in eastern Alberta (±0.1 m). All adult godwits and about half of the adult willets were captured at nest sites during early incubation by lowering a mist net over the nest. The remaining willets were captured at their nests with passive walk-in traps made from chicken wire. At first capture, each adult was individually marked with a unique combination of three color bands, a white flag with two bands on each tibia, and a numbered stainless-steel band on the tarsus. The wrap-around color rings were not sealed, but only one Willet lost a color ring during the course of the study.
Willets were also temporarily marked with pink or yellow dyes on their white wing bars or rump for easier identification in flight and to reduce unnecessary disturbance from re-trapping. After capturing one parent, nests were revisited to capture any unbanded mates. In the case of godwits, females tended to incubate during the day and males at night so it was possible to capture mates by visiting the nest at different times of day. In the case of Willets, we attempted to trap the mate if a flushed bird did not have a temporary plumage mark.
No nests were deserted because of nest monitoring or capture of adults at the nest.
In the last six years of the project (1996)(1997)(1998)(1999)(2000)(2001), we conducted regular surveys during the prelaying and brood attendance periods to search for returning adults that had been banded in previous years. Godwits and willets encountered at foraging sites, nests, or attending broods were routinely scanned for color bands. We recorded color-band combinations, locations, and behavior for all banded individuals. Mated pairs were identified during the laying period by mate-guarding behavior and during the incubation and the brood-rearing periods by joint parental care. In addition, the larger wetland complexes surrounding the core study area were surveyed three times per year for banded birds by driving systematic survey routes with all-terrain vehicles (Gratto-Trevor, 2006).

| Morphometrics
Godwits and willets were sexed by morphometrics recorded in the field, including body mass, culmen (feathering to tip), maximum wing chord (flattened and straightened), tarsus, and vent length.
Preliminary sex assignments were confirmed in the field by behavioral observations of aerial displays and wing flashing by males, sex roles of males and females during mate guarding and copulations,

| Mark-recapture analyses
Resightings of birds during the breeding season were combined with recaptures of birds during incubation to create live encounter histories for each individual bird. Each year within the encounter history was coded as 1 = resighted or captured on the study area, or 0 = not detected during the breeding season. All birds included in the encounter histories were first captured and banded as adults of unknown age (after-hatch-year or AHY). Birds marked as flightless young were not included in the survival analysis because no godwits and only two willets were encountered again in the study area. In the case of known mortality events where an adult bird was found dead during the breeding season, we censored the individual as not released after the occasion with the dead recovery event. We used Cormack-Jolly-Seber (CJS) models to estimate annual probabilities of apparent survival (ϕ), corrected for the probability of encounter (p, Sandercock, 2020). Apparent survival is the joint probability of true survival (S) and site fidelity (F) to the study area between two consecutive years. The complement of apparent survival (1−ϕ) included losses to mortality and to permanent emigration caused by dispersal from the study site. The probability of encounter is the product of site propensity and the true detection rate, and the complement (1−p) included losses due to temporary emigration or incomplete detection.
Mark-recapture analyses were conducted in Program Mark (ver. 9.0) in an information theory framework (White & Burnham, 1999).
First, we selected factors to include in the global model for the probabilities of apparent survival (ϕ) and encounter (p). We modeled both parameters as a function of sex (sex) because males and females were expected to differ in sex roles during the breeding season and in timing of migration and use of wintering areas during the nonbreeding season. Males may have higher apparent survival and detection rates than females because the sexes differ in reproductive effort, males tend to have stronger site fidelity in male-territorial species, and males are more active in display, mate defense, and brood attendance (Sandercock et al., 2000). We did not consider time-since-marking models for apparent survival because few birds had encounter histories where they were detected on only a single occasion (godwits: 4 of 94, 4.3%; willets: 30 of 175, 17.1%). We modeled ϕ and p with time-dependence (year) to test for patterns of annual variation. Further, we modeled apparent survival as a function of number of ponds and Short-eared Owls as annual indices of wetland conditions and predator abundance ( Table 1). Number of ponds was bimodal so we grouped years into two bins and compared apparent survival between five years with typical water levels (1995 and 1998-2000: 56.6-67.2 K ponds) vs. two years with particularly wet conditions (1996-1997: 130.5-143.2 K ponds). Similarly, numbers of rodents and Short-eared Owls were high in only one year so we grouped years into two bins and compared apparent survival between the rodent peak (1997) vs. all other years (1995-1996 and 1998-2000). Last, we modeled encounter rates as a function of the number of days afield as an index of field effort ( Table 1, 1996-2001: 18-67 days afield). We combined factors in main effect models (+) and factorial models with interaction terms (×).
Our global model was the standard CJS model with the effects of sex and year in both parameters: ϕ(sex × year), p(sex × year). We used Program Release to evaluate the goodness-of-fit of the global model to the encounter histories, and calculated an overdispersion factor as the ratio of the component tests to their degrees of free- Moderate amounts of overdispersion are common in mark-recapture data and values of ĉ < 3 indicate that the global model is acceptable. The asymptotic value of ĉ is expected to be one in the case of no overdispersion (Sandercock, 2020).
We proceeded with model testing by fitting reduced models with fewer parameters. All models were constructed with design matrices and the logit-link function. Starting with our global model, we applied constraints to the probabilities in the following order: encounter rates (p) and then apparent survival (ϕ). Each probability was modeled by dropping interaction terms from factorial models (sex × year) to create additive models (sex + year) and then by dropping main effects to create single factor (sex or year) or constant models (con). Once we had identified a minimum AICc model, we added factors back in parameters to explore whether nearby models were a better fit. Parameter counts (K) were adjusted to match the structure of the model which was usually the number of columns in the design matrix. In the special case of models with annual variation in both apparent survival and the probability of encounter, we also adjusted K to account for inestimable parameters from the last transition of the CJS models (Sandercock, 2020). The parameter count (K) and the log-likelihood -2ln(L) were combined to calculate Akaike's Information Criterion (AICc) and models were ranked by the difference from the minimum AICc model (ΔQAICc). Akaike weights (w i ) were used to determine the relative likelihood of a model within the set of candidate models, and ratios of weights between groups of candidate models with or without explanatory factors ( were used to quantify the relative support for different effects. To examine patterns of variation between the sexes and annual conditions, we took estimates of apparent survival (̂ ) and encounter rates (p) from factorial models such as ϕ(sex × ponds), p(con), even if the unconstrained model was a relatively poor fit.
Models were considered equally parsimonious if ΔAICc ≤ 2, but we also examined the coefficients for the explanatory variable in cases where models differed by a single parameter (ΔK = 1, Arnold, 2010). To obtain overall parameter estimates, we reran the best-fit models with the MCMC sampling procedure of Program Mark and present estimates as means ±1SE with 95% credible intervals (95%CrI). We fitted MCMC models with a single chain and 40,000 tuning samples, 10,000 burn-in samples and 100,000 stored samples for the posterior distributions. We also used the variance components procedure of Program Mark to estimate process variance of the overall estimates of apparent survival.
Here, the method of moments approach was applied to the timedependent estimates of apparent survival from model ϕ(year), p(con). We used the analysis of deviance procedure in Program Mark to estimate the percentage of annual variation in apparent survival that was explained by the annual covariates for number of ponds and relative abundance of owls. The ANODEV procedure was used to compare three nested models, ϕ(year), ϕ(covar), and ϕ(con), and the calculations were based on differences in deviances and degrees of freedom. Life expectancy of adults (Ê) and 95% credible intervals were calculated as derived parameters from the mean estimates of apparent survival (̂ ) with 2.5% and 97.5% quartiles from the posterior distributions, where the SE(Ê) was calculated with a function based on the delta method (Powell, 2007):

| Mate fidelity and site fidelity
To determine patterns of mate fidelity and breeding site, we examined mating histories for pairs of marked birds where at least one partner returned to the study site in the same or a following year.
Pairs were considered to have reunited with the same mate if both members of the pair returned and were recorded together on a nest. one-way analysis of variance (aov).

| Morphometrics
Godwits were dimorphic in body size and coloration (F/M = 1.06-1.29), and the best character traits for distinguishing between the sexes were culmen length (Hedge's g = 4.32) and body mass (g = 2.55, Table 2). Females had longer bills with dark coloration (range = 108.9-130.9 mm, n = 49), whereas males had shorter bills with more orange at the base (89.3-108.5 mm, n = 46). The sexes overlapped in body mass, but females were 19% heavier than males on average. In contrast, Willets had greater overlap in body size (F/M = 1.02-1.29) and no sexual differences in coloration. The best characters for distinguishing the sexes in Willets were body mass (Hedge's g = 1.76) and length of the vent after egg laying (g = 1.44).
Females tended to be 11% heavier and had vents that were 34% longer than males.  was a good fit to the encounter histories for both Marbled Godwits

| Mark-recapture analyses
( 2 20 = 8.59, p = .987) and Willets ( 2 23 = 20.2, p = .628). Estimates of the overdispersion factor based on the ratio of the χ-statistic to the degrees of freedom were less than one, so we set ĉ = 1 and used AICc for model selection.
Models with a constant probability of encounter were a better fit than models where the parameter was constrained as a function of sex, year or annual variation in field effort (ratio of AICc weights, godwits: 0.97/0.07 = 13.9×, willets: 0.81/0.19 = 4.3×, Table 3).
Estimates of the probability of encounter were high for both species of shorebirds and were p = 0.919 ± 0.017SE in godwits and p = 0.887 ± 0.019SE in willets.
Models where apparent survival was modeled as a function of annual variation in water levels or relative abundance of Short-eared Owls were parsimonious in both species of shorebirds ( Table 3). Parameter estimates from an unconstrained model, ϕ(year), p(con), indicated that apparent survival tended to be higher in two wet years (1996)(1997)  analysis of deviance showed that annual variation in water conditions did not explain significant variation in apparent survival for either godwits (9.7%, F 1,4 = 0.43, p = .55) or willets (20.3%, F 1,4 = 1.02, p = .37). Estimates of apparent survival from a factorial model with effects of water levels and sex, ϕ(sex × ponds), p(con), showed that water levels had a greater effect on females than males (Table 4). In years of high water levels, the apparent survival of female godwits was 4 points higher, whereas the apparent survival of female willets was 5 points lower. The relative abundance of owls had poor performance as an explanatory factor compared to water levels. Apparent survival of godwits and willets in the year with high numbers of rodents and high relative abundance of owls (1997) was not unusual compared to the other study years (Figure 2, right panels). Coefficients for the effects of owl abundance were not significantly different from zero in either godwits (β = +0.45, 95%CI = −0.99 to +1.90) or willets (β = 0.005, 95%CI = −0.77 to +0.78). Owl abundance did not explain annual variation in apparent survival for either godwits (7.1%, F 1,4 = 0.31, p = .61) or willets (0.1%, F 1,4 < 0.001, p > .99).
Despite marking and resighting relatively large samples of birds over a 7-year study period, the minimum AICc models for both species had a relatively simple structure with either effects of sex or a constant probability of apparent survival ( Table 3). Parameter estimates based on MCMC sampling showed that overall apparent survival was high in Marbled Godwits (̂ = 0.934 ± 0.015SE) and tended to be higher among males (̂ = 0.944 ± 0.018SE) than females (̂ = 0.924 ± 0.021SE, Table 4). We applied the method of moments approach to decompose variance components among the parameter estimates from ϕ(year), p(con), and estimated apparent survival with process variance alone to be ̂ = 0.953 ± 0.012SE. In willets, models with an effect of sex received more than twice the support of a constant model (ratio of AICc weights: 0.229/0.103 = 2.2×), and apparent survival was again higher among males (̂ = 0.882 ± 0.022SE) than females (̂ = 0.816 ± 0.027SE, Table 4). Overall apparent survival from model phi(con), p(con) was also relatively high at ̂ = 0.849 ± 0.018SE and if estimated with the variance components procedure was ̂ = 0.861 ± 0.015SE. Based on our annual estimates of apparent survival, the expected life expectancy (Ê) of birds first captured as adults at the breeding grounds would be 14.6 years for Marbled Godwits and 6.1 years for willets (Table 5)

| Apparent survival of adults
Our  et al., 1992). Our annual estimates of apparent survival indicate that projected lifespan could be 17.3 years for male godwits and 7.7 years for male willets. Estimates of maximum longevity are known to be biased because they are a function of ringing and recovery effort (Sandercock, 2003). Nevertheless, our estimates of average lifespan are consistent with previous reports of maximum longevity for both species, including 25-29+ years for godwits (Colwell et al., 1995;Gratto-Trevor, 2020) and 10+ years for willets (Lowther et al., 2020).
Despite substantial variation in water levels from year to year and some differences in rodent numbers and predator abundance, we found little evidence for annual variation or effects of annual covariates on apparent survival of godwits or willets. Our ability to detect effects of environmental covariates was relatively low in a 7-year project. Nevertheless, a lack of annual variation is consistent with the overall high estimates of apparent survival that we found for both species in our study. Our work is consistent with previous population studies of shorebirds in showing that adult survival is often buffered against year-to-year variation in environmental conditions in long-lived species (Ottvall & Härdling, 2005;Weiser et al., 2018).
Species with a slow life-history strategy can be vulnerable to environmental change if anthropogenic factors cause additive mortality (Sandercock et al., 2011;Watts et al., 2015), but they may be less susceptible to environmental stochasticity if demographic buffering is important (Hilde et al., 2020).

| Breeding site fidelity and mate fidelity
We found small sexual differences in apparent survival with males having higher apparent survival than females in both species. Higher return rates or apparent survival among males are a common feature of shorebirds that are socially monogamous with a male-territorial social system (Kwon et al., 2022;Oring & Lank, 1984;Sandercock & Gratto-Trevor, 1997;Thompson & Hale, 1993), although no sex differences have been reported too (Groen & Hemerik, 2002;Ottvall, 2005). Mated pairs of shorebirds usually winter at separate nonbreeding sites but meet again at breeding sites (Gratto-Trevor, 2011;Gunnarsson et al., 2004). Our estimates of breeding dispersal indicated that males returned to the same territory, but females move farther than males when changing mates, similar to patterns previously reported in some Calidris sandpipers (Gratto et al., 1985;Sandercock et al., 2000;van Leeuwen & Jamieson, 2018).
Similarly, sex differences have been reported in homing rates to the same breeding territory in Eurasian Curlews (Numenius arquata, male 91% vs. females 78%, Berg, 1994) and Common Redshanks (Tringa totanus, males 84% vs. females 67%, Jackson, 1994) and to the same breeding population in Whimbrels (N. phaeopus, males 87% vs. females 68%, Grant, 1991). Our core study area was 21.6 km 2 in area and dispersal movements >2 km would have led to permanent F I G U R E 2 Estimates of apparent survival as a function of number of ponds and owl abundance for Marbled Godwits and Willets breeding near Brooks, Alberta, 1995Alberta, -2001. Breaks in the x-axis were used for clarity due to large differences among years in the two explanatory factors. Parameter estimates for apparent survival taken from models ϕ(year), p(con).
emigration from most of the area. The longest dispersal movements detected in our study were 1-6 km. Thus, sex differences in apparent survival can be due to differences in true survival, but here might also be explained by sex differences in breeding site fidelity.

TA B L E 4
Estimates of apparent survival (̂ ) and probability of encounter (p) for Marbled Godwits and Willets at the Kitsim wetland complex near Brooks, Alberta, 1995Alberta, -2001 breeding in a saltmarsh habitat (Howe, 1982). A range of hypotheses have been proposed to account for interspecific variation in matefidelity rates, including explanations that relate mate fidelity to habitat preferences, reproductive success, mortality rates, or aspects of migration (Choudhury, 1995;van Leeuwen & Jamieson, 2018). Mate fidelity is often lower among shorebirds that breed in ephemeral habitats (Friedrich et al., 2015), but seems unlikely to explain species differences here because godwits and willets bred sympatrically in the same managed wetlands. Similarly, variation in reproductive success cannot explain species differences in mate fidelity because nest success was similar for godwits (54%, n = 114 nests) and wil-  (Leach et al., 2020). Last, both species have wide nonbreeding ranges but Marbled Godwits are short-distance migrants that winter in southern California and Mexico (Gratto-Trevor, 2020;Warnock et al., 2021), whereas Willets migrate long distances to winter in South America (Martínez-Curci et al., 2014;Oswald et al., 2016). Pairs of Willets may reunite less frequently if they are more likely to arrive asynchronously at the breeding grounds after northbound migration.

TA B L E 6
Annual estimates of return rates (RR), apparent survival (ϕ), and true survival (S) from population studies of four genera of largebodied shorebirds (Limosa, Numenius, Bartramia, and Tringa Thompson and Hale (1993) Ukraine Breeding RR 0.724 Zhmud (1992) 0.78 following impoundment of an intertidal mudflat with a barrage for flood control (Burton et al., 2006), and apparent survival of wintering Eurasian Curlews dropped from 0.95 to 0.81 during 2 years with mechanized dredging of cockles in a coastal estuary (Taylor & Dodd, 2013). Reductions in apparent survival may be explained by increased mortality rates or by greater displacement from local areas. Estimates of vital rates can also provide insights into the demographic mechanisms that underly population dynamics. For example, high adult survival rates indicate that low reproductive rates are the likely driver of population declines among Eurasian Curlews and Black-tailed Godwits in western Europe (Franks et al., 2017;Robinson et al., 2020;Roodbergen et al., 2012).
High annual survival rates and strong site fidelity demonstrate that managed wetlands provide an important breeding habitat for godwits and willets in Alberta. Conservation easements for waterfowl production may not benefit all wildlife species (Koper & Schmiegelow, 2006), but did create suitable breeding conditions for prairie shorebirds (Gratto-Trevor, 2006). Additional effort to im-  (Cook et al., 2021). We conclude that reducing exposure to anthropogenic mortality will be important for conservation of large-bodied shorebirds because low natural mortality rates imply that any additional losses are likely to lead to additive mortality (Sandercock et al., 2011;Watts et al., 2015).

ACK N OWLED G M ENTS
We thank D. Pearson, H. Johnston, S. van Wilgenburg, G. Perrin, A.
Keeley, G. Thibault, and A. Neudorf for their assistance with field work. We greatly appreciated the help of Karla Guyn and her field assistants who located many godwit and willet nests in the course of a waterfowl study during the first 3 years of the project. Ducks Unlimited Canada (Alberta) provided logistical support, especially T.
Sadler and the Brooks office. We thank the participating landowners who gave us access to private lands, including The Eastern Irrigation