Flexible migration and habitat use strategies of an endangered waterbird during hydrological drought

Wildlife species confront threats from climate and land use change, exacerbating the influence of extreme climatic events on populations and biodiversity. Migratory waterbirds are especially vulnerable to hydrological drought via reduced availability of surface water habitats. We assessed how whooping cranes (Grus americana) modified habitat use and migration strategies during drought to evaluate their resilience to changing conditions and adaptive capacity. We categorized >8000 night‐roost sites used by 146 cranes from 2010 to 2022 and examined relative use during non‐drought, moderate drought, and extreme drought conditions. We found cultivated and uncultivated palustrine and lacustrine wetlands were generally used less during droughts than non‐drought conditions. Conversely, impounded palustrine and lacustrine systems and rivers served more frequently as drought refugia (i.e., used more during drought than non‐drought conditions). Night roosts occurred primarily on private lands (86% overall); public land use decreased with latitude and increased with drought severity, with greatest use (56%) occurring during severe autumn drought in the southern Great Plains. Quantifying use of identified critical habitats in the United States indicated that Cheyenne Bottoms State Waterfowl Management Area and Quivira National Wildlife Refuge were used less during drought, and the Central Platte River and Salt Plains National Wildlife Refuge received similar use during drought compared to non‐drought conditions. Our findings provide insights into compensatory use of habitats, where impounded surface water may function in a complementary fashion with natural wetlands. Collectively, these and other types of wetlands distributed across the migration corridor provided a reliable network of habitat available across the Great Plains. A diversity of wetlands available during variable environmental conditions would be useful in supporting continued recovery of whooping cranes and likely have benefits for a wide array of migratory birds.

Extreme weather conditions shape natural communities and can be as or more influential drivers of ecosystem change than incremental trends or average conditions (Maxwell et al., 2018).Drought is a natural process to which species have adapted and has shaped natural communities (Basara et al., 2013;Clark et al., 2002).However, on-going climate and land use change can exacerbate effects of drought and other extreme weather events on biodiversity and population persistence.Climate change projections indicate potential for more frequent, severe, and longer lasting droughts (Strzepek et al., 2010;Trenberth et al., 2013;Zhao et al., 2020), which may create environmental conditions outside of some species' tolerances (Rom an-Palacios & Wiens, 2020) or drive habitats into undesired states (Ansley et al., 2023).Land conversion and habitat degradation have reduced the resilience of many ecosystems, decreasing their capacity to withstand intense perturbations such as extreme drought and return to their preceding equilibrium, putting species at greater risk of decline than they may have been historically (Chambers et al., 2019;Wada et al., 2013).Resilience also can be conceptualized at the species level regarding the ability of a species to tolerate significant stochastic disturbances and rebound to a prior and sustainable demographic equilibrium (Capdevila et al., 2020).
Drought influences the distribution of species, biodiversity, and net primary productivity of both terrestrial and aquatic ecosystems (Clark et al., 2016;Tilman & Haddi, 1992;Walls et al., 2013).For avian communities, specifically, drought can influence individual body condition and behaviors (Anderson et al., 2021;Stanek et al., 2022), population abundance and demography (Chan, 1999;Mooij et al., 2002), and avian species richness (Albright et al., 2010).Waterbirds, bird species reliant on surface water to meet all or most of their lifecycle events, are especially vulnerable to hydrological drought (i.e., a reduction in surface and groundwater; Van Loon, 2015).Hydrological drought influences population distribution, status, and persistence of waterbird species by creating climactic-induced resource bottlenecks and modifying wetland habitats (Barbaree et al., 2020;Maron et al., 2015;Wen et al., 2016).The conservation status of North American waterbirds varies from least concern to endangered, with decades-long decreases in abundance for most non-waterfowl species (International Union for the Conservation of Nature, 2022; Rosenberg et al., 2019).Moreover, migratory birds present additional conservation challenges in identifying and maintaining connected stopover habitats across multiple locations and potentially thousands of kilometers to successfully complete migration events (Runge et al., 2014).Thus, drought is often identified as a regional concern, but because of the habitat connectivity migratory waterbirds require, regional droughts can affect populations across vast ranges (Donnelly et al., 2022;Xu et al., 2019).
Considerable research exists on how climate change affects phenology of migration events (Ill an et al., 2014;Wood & Kellermann, 2015), but climate and land use change also affect availability of stopover habitat needed for waterbirds to successfully complete migration under variable and changing conditions (Haig et al., 2019).Determining how waterbirds and their habitats respond to conditions caused by drought and other extreme events during migration can improve conservation planning for ecosystem restoration across migratory networks considering climate change projections (Selwood et al., 2015;Uden et al., 2015;Wen et al., 2020).Behavioral plasticity to varying conditions can improve individual fitness and population persistence as long as the conditions do not persist outside of the species' tolerance (Hadfield & Strathmann, 1996).Quantifying how individuals modify behaviors, including habitat use, diet and foraging patterns, and migration strategies, when experiencing varying environmental conditions allows for greater understanding of adaptive capacity and ability of a species to persist under threats related to climate change (Beever et al., 2017;Teitelbaum et al., 2021;Thurman et al., 2021).Knowledge of how individuals alter behaviors in response to droughts allows for identification of specific locations or types of habitats that serve as refugia during conditions brought upon by drought (Selwood et al., 2015).If drought refugia differ in type, location, or configuration from habitats used during nondrought conditions, conservation plans could be modified to protect drought refugia.Because of the potential for more frequent and severe drought in the future (Fudickar & Ketterson, 2018;Maron et al., 2015), it will be essential to engage different conservation partners and stakeholders capable of taking conservation actions to mitigate effects of drought.
We used a population of whooping cranes (Grus americana) to assess how a migratory waterbird evaluates and modifies migration and habitat use strategies during drought conditions.Whooping cranes of the Aransas-Wood Buffalo population provide an interesting case study because of their public interest, conservation status as an endangered species, and requirement to migrate across the Great Plains of North America twice annually between breeding and nonbreeding areas (Canadian Wildlife Service and U.S. Fish and Wildlife Service, 2007).This population has experienced decadeslong modest growth, but climate and land use changes are identified as threats to the species' continued recovery as they affect habitat availability and quality at breeding and nonbreeding sites, with hydrological drought identified as an exacerbating stressor (Chavez-Ramirez & Wehtje, 2012).Our objectives were to quantify types of nighttime-roosting habitats used by migrating whooping cranes and determine the extent to which whooping cranes modified migration and habitat use behaviors in response to varying drought conditions across the Great Plains and Canadian Prairies.Continued recovery can be supported by identifying critical drought refugia, enhancing existing resources, and considering other anthropogenic stressors that may exacerbate effects of drought.Quantifying the response of whooping cranes to varying intensities of drought will increase understanding of the species' and the ecosystem's resilience to changing conditions throughout the migration corridor, providing insight into the adaptive capacity of this species and the habitats on which it depends.Low landscape resilience would be punctuated by an overall lack of wetland habitat that can alter normal movement patterns and a limited diversity of wetland types that may constrain quality habitat options under varying hydrological conditions.For the whooping crane, resilience can be defined by the bird's ability to adapt to and exploit whatever habitats are available through altered habitat selection and movement patterns.Our findings could be used to inform conservation plans for whooping cranes and other waterbirds that account for predicted increases in severe drought conditions and other climate changes that are likely to be seen in the future across the Great Plains (Bradford et al., 2020;McIntyre et al., 2014).

| Study area
For this study, we used locations of migrating whooping cranes from the Aransas-Wood Buffalo population across the Central Flyway of North America (Figure 1; Austin et al., 2019;Pearse et al., 2018).We focused our analyses on the grassland biome, which spans a large portion of the central United States and Canada.These lands are primarily in private ownership and have experienced land use changes that have resulted in a mosaic of cultivated lands used for agricultural production and grasslands used in grazing systems (Hartman et al., 2011).The study area includes complexes of wetlands, lakes, reservoirs, and rivers that attract millions of migratory waterbirds (Laubhan & Fredrickson, 1997).Whooping cranes preferentially select migratory stopover sites with greater wetland density and use different types of surface water as nightly roosts and diurnal foraging sites during migration (Baasch et al., 2019;Ellis et al., 2022;Pearse et al., 2017).We identified 4 regions within our study area for analyses, based on similarity in wetland density, level of land use change, and latitudinal gradient which influences baseline climactic conditions (La Sorte et al., 2014; Figure 1): (1) the Boreal Transition Zone (BTZ); (2) Northern Great Plains (NGP); (3) Central Great Plains (CGP); and (4) Southern Great Plains (SGP).The BTZ is an identified ecoregion in Canada (Ecological Stratification Working Group, 1995).We designated the NGP region as the Prairies Ecozone in Canada and Level 1 Ecological Regions of North America designated as Great Plains in the U.S. States of North Dakota and South Dakota, Montana, and Minnesota (U.S. Environmental Protection Agency, 2010).The CGP region included the Great Plains of Nebraska, Kansas, Colorado, Iowa, and Missouri.Finally, the SGP region consisted of Oklahoma and Texas.Within our study area, we explored specific drought responses at locations designated by the U.S. Fish and Wildlife Service as critical habitat for migrating whooping cranes: Salt Plains National Wildlife Refuge (NWR) in northern Oklahoma, Quivira NWR and Cheyenne Bottoms State Waterfowl Management Area (WMA), both in central Kansas, and a reach of the Platte River in southcentral Nebraska (Canadian Wildlife Service and U.S. Fish and Wildlife Service, 2007).For the Platte River, we included a larger area from what was identified initially as critical habitat, because this expanded area has been targeted for long-term ecosystem restoration efforts to support recovery of whooping cranes and other listed species (Farnsworth et al., 2011).

| Data sources and processing
During 2009-2022, we captured 166 whooping cranes at sites along coastal Texas and in and around Wood Buffalo National Park, Canada and marked them with leg-mounted transmitters capable of collecting locations using global position system capabilities and transmitting data using platform transmitting terminals (North Star Science and Technology LLC, Baltimore, MD, United States; Geotrak, Inc., Apex, NC, United States) or Global System for Mobiles and Global Packet Radio Service third and fourth generation cellular networks (Ornitela, Vilnius, Lithuania).We used helicopters to locate pre-fledged juveniles at breeding sites in Canada and captured individuals by hand (Kuyt, 1979).At nonbreeding sites in Texas, we captured adult and juvenile cranes with leg snares (Folk et al., 2005) Our analyses used location data from 145 individuals, collected in the study area during spring (inter-quartile range [IQR] = 4 Apr-23 Apr) and autumn (IQR = 8 Oct-5 Nov) migration seasons, 2010-2022.For each crane, we extracted 1 roost location per night (location closest to 00:00 Central Standard Time Zone) that was classified as occurring on the ground (instantaneous velocity reading <2.6 m/s; Byrne et al., 2017).Nightly roost locations were further collapsed across all sequential locations occurring on the same surface water body or reach of a river.We identified unique stopovers sites as those with ≥1 collapsed nightly location within 15 km of each other (Pearse et al., 2020).Using stopover site designations, we calculated residency time and distance between stopover sites within the area of study.
We characterized locations into seven categories of roosting habitat.We used wetland classification systems to assist where available (U.S. Fish and Wildlife Service, 2020) and satellite and aerial imagery (Lisle, 2006).Aerial imagery was accessed across several decades via Google Earth Pro version 7.3 using the "time slider" tool, which aided in confirming wetland presence and classification (Google, 2018).Palustrine wetlands included shallow non-tidal wetlands which can have emergent vegetation or open-water forms and are saturated or inundated for a significant portion of the growing season annually (Cowardin et al., 1979;Tiner, 2016).These wetlands tend to fluctuate in water level seasonally, vary in hydroperiod length, and dry out to varying degrees and frequencies.We divided palustrine wetlands into 3 categories for this study.Palustrine wetlands (PAL) included shallow emergent and open water wetlands.Palustrine agricultural wetlands (PAL-AG) were those within cultivated fields, with limited hydrophytes, and a history of cultivation across the entirety of the wetland basin, which we assessed using aerial imagery.Any palustrine wetland that was modified to increase wetland hydroperiod (i.e., impounded, excavated, or diked) was categorized separately (PAL-IMP).Lacustrine wetlands (LAC) were primarily deep-water wetlands >8 ha in size or deeper than 2.5 m at low water with <30% coverage of vegetation (Cowardin et al., 1979;Tiner, 2016); impounded lacustrine wetlands, including reservoirs, were categorized separately (LAC-IMP).Riverine wetlands (RIV) were lotic systems of all flooding frequencies, but generally included intermittently and permanently flooded designations.Finally, we identified uplands (UPL) as non-wetland locations generally located on grasslands or cultivated lands.We also classified each wetland into four size categories (small = <100 m; medium = 100-400 m; large = 400-1000 m; extra-large = >1000 m), using the shortest distance across the wetland or water body center (i.e., wetland width; Caven et al., 2022).We identified ownership of locations using the Protected Areas Database or the Canadian Protected and Conserved Areas Database (Environment and Climate Change Canada, 2021; U.S. Geological Survey Gap Analysis Project, 2022).Finally, we noted if locations occurred within areas designated as critical habitat for migrating whooping cranes.
We compiled information on status of drought within the study area from the U.S. and Canada drought monitor systems, which use multiple assessment indices to characterize soil moisture, streamflow, and precipitation (Svoboda et al., 2002).We downloaded spatial vector data from the month of April to represent spring migration and October to represent autumn migration (https:// droughtmonitor.unl.edu/,https://agriculture.canada.ca/en/agricultural-production/weather/canadian-droughtmonitor).In the United States, drought data were available each week; we used information from the middle of each month to characterize status of seasonal drought.Both drought classifications used a five-category system to characterize drought status across each nation.We collapsed these categories to form three levels of drought severity.Locations with no drought designation or identified as "D0-abnormally dry" were identified as nondrought condition.Locations identified as "D1-moderate" or "D2-severe" drought were identified as moderate drought conditions.Finally, we identified locations in "D3-extreme" or "D4-exceptional" drought as experiencing extreme drought conditions.Year-, season-, and location-specific drought information was extracted for whooping crane locations.We developed spatially explicit drought probability maps of the migration corridor for the spring and autumn based on 20 years of existing drought data from 2002 to 2022 to explore the variation in drought potential and severity by region.Maps were created by converting spatial data from vector to raster (100-m cell size) and calculating the proportion of years in which moderate or extreme drought occurred within each cell.

| Data analyses
We summed the number of nights that marked whooping cranes used the seven roost habitat categories and derived proportional use, standard errors, and 95% confidence intervals (CIs) by region, season, and drought severity.As our main goal was to compare how roost habitat use changed within a roost habitat type with respect to drought severity, we used a ratio of proportional use during moderate or extreme drought divided  by proportional use during non-drought conditions to identify situations where use during moderate or extreme drought conditions was greater, equal to, or less than use during non-drought conditions.This use ratio is equivalent to a resource selection ratio, except we compared habitat use during different conditions rather than use versus availability as with a traditional habitat selection analysis (Manly et al., 2002).We estimated use ratios for each habitat and combination of drought severity, region, and season.Standard errors were derived under the assumption that both proportions were estimates of used units (Manly et al., 2002;formula 4.25).We used 95% CIs of use ratios to infer if roost habitat types were used less (ratio <1.0), equally (ratio = 1.0), or more (ratio >1.0) during drought.We repeated this process for ownership and size categories.
We explored how use of critical habitat areas changed with drought severity by first identifying migrations where we had complete nightly data in the CGP and SGP.We calculated the proportion of migrations where birds stopped over at individual critical habitat areas overall and for each drought severity.As with habitat categories, we calculated use ratios for each critical habitat area and combined seasons because of low sample size.
To determine how whooping cranes modified daily migration distances in relation to drought conditions, we first filtered data to include only observations where there were no missing daily location data between stopover sites.We also removed migration flights initiating from the BTZ, as these were fewer in frequency compared to other regions and we had no observations of extreme drought during spring (Table 1).For each autumn and spring migration separately, we modeled variation in movement distances (km) in relation to the region of origin and drought severity.We first compared two models to determine if there was an interaction of region and drought severity; we used parameter estimates from the model with the lowest Akaike information criterion (AIC) for inference.We fit models with the glmer function in the lme4 package and R version 4.2.2 (Bates et al., 2014; R Core Team, 2022), using a gamma distribution, identify link function, and individual bird as a random effect to account for multiple observations for each bird.We derived inference regarding differences in movement lengths during normal compared to moderate and extreme drought severities using predicted model coefficients and 95% CIs.
To determine if whooping cranes modified residency time at stopover sites during drought, we first removed observations where we were unable to account for the location of individuals before or after the stopover.We also removed observations from the BTZ region for reasons stated above.By season, we modeled variation in time at stopover sites (nights) in relation to region and drought severity.As with movement distance, we compared models with and without an interaction of main effects using the model with the lowest AIC value for inference.We fit models with the glmer function in the lme4 package and R version 4.2.2 (Bates et al., 2014;R Core Team, 2022), using a gamma distribution, log link function, and including individual bird as a random effect.We derived inference regarding differences in stopover residency time during non-drought compared with drought conditions using predicted model coefficients and 95% CIs.The data generated in this study are provided in a U.S. Geological Survey data release (Pearse et al., 2024).

| RESULTS
We collected location data from 145 migrating birds from spring 2010 to autumn 2022.We monitored an average of 23.1 (2-55) birds for an average of 722.8 (56-1903) roost nights during each of 26 migration seasons.The Aransas-Wood Buffalo population had an estimated 283-543 birds during our study period; thus, our sample represented $4-10% of the population each year (Butler et al., 2014(Butler et al., , 2022)).Birds were under observation for an average of 4.1 migration seasons (1-17) and 129.6 (5-529) nights.Nightly roost locations were more prevalent during autumn (57%) compared to spring migration seasons (43%).By region, we observed 6% of roost nights in the BTZ, 69% in the NGP, 16% in the CGP, and 9% in the SGP.
Drought conditions occurred most often in SGP, became less common with increasing latitude, and were less common in the eastern than western portion of the corridor (Figure 1).Non-drought conditions were most common at roost sites (67%).Moderate drought conditions occurred at 21% and extreme drought at 12% of roosts.During migration events, whooping cranes experienced only 1 drought severity category during 32% of the time, 2 drought severity categories 34% of the time, and 3 drought categories 34% of the time.Most cranes (93%) experienced non-drought conditions during at least some portion of their migrations, with moderate drought conditions also common (69% of migrations) and extreme drought conditions least common (41%).
Across the NGP, >50% of roost nights were spent in PAL in both seasons during non-drought conditions, with shifts to greater use of PAL-AG during spring and LAC during autumn (Table 1).During autumn, PAL were used less during moderate and extreme drought conditions, whereas during spring, this roost category served as drought refugia (Figure 2).PAL-AG were consistently used less during drought across seasons.LAC had mixed relationships with drought, serving as refugia during extreme autumn droughts and moderate spring droughts, used less during moderate autumn droughts, and used consistently during extreme spring droughts compared to non-drought conditions.LAC-IMP systems offered cranes refugia during all conditions except extreme spring droughts.
Whooping cranes used roost categories most evenly across the CGP; however, use of PAL was more common during spring than autumn and of RIV and LAC-IMP more common in autumn than spring (Table 1).PAL and PAL-AG tended to be used less during moderate and extreme drought conditions, especially during spring (Figure 2).PAL-IMP wetlands were used similarly during moderate and extreme autumn droughts, used more during moderate spring droughts, and used less during extreme spring droughts compared to non-drought conditions.LAC-IMP systems were consistently used or refuge during droughts.RIV systems were generally drought refugia during both seasons.
PAL-IMP and LAC-IMP wetlands made up approximately 50% of use in the SGP, and RIV use was similar between seasons (18-19%; Table 1).PAL were consistently used less during droughts, and PAL-IMP use was similar during droughts compared to non-drought conditions (Figure 2).PAL-AG were used less during spring droughts and used similarly during autumn droughts compared to non-drought conditions.LAC-IMP systems served as drought refugia in this region, becoming the most-used roost habitat during extreme droughts.RIV use was similar during moderate drought conditions in both seasons, used less during extreme autumn droughts, and used more during extreme spring droughts compared to non-drought conditions.
We observed considerable differences in use and drought response of wetlands by size, which we categorized by the shortest width, by region and migration season.Small wetlands (<100 m width) comprised ≥33% of use across regions during spring but were used less during autumn (≤21%; Table 1).During droughts, proportional use of small wetlands was reduced compared to non-drought conditions across regions and seasons.Medium-sized wetlands (100-400 m width) were generally the second-most proportionally used size category during spring and the most-used size category during autumn.Medium wetlands were largely but not exclusively used in greater proportion during drought conditions compared to non-drought conditions across regions and seasons.Large-sized wetlands (400-1000 m width) were not commonly used among regions and seasons (≤21%; Table 1) with mixed drought responses for all regions except the SGP, where large wetlands were consistently used less during droughts.Finally, extralarge wetlands (>1000 m width) were uncommonly used during spring and commonly used during autumn during non-drought periods across seasons (≥32%; Table 1).As with large wetlands, whooping crane use of extra-large wetlands varied for all regions except the SGP, where these wetlands were used as drought refugia.
Across migrations and regions, whooping cranes used private lands for 86% of nightly roosts and public lands for 14% of roosts.Public ownership was dominated by federally (United States and Canada) owned lands (58%); other public lands included Canadian Provincial (22%), U.S. State (13%), and local government (7%).In the BTZ, public lands were used less during moderate autumn drought and used similarly during moderate spring drought (Figure 3).Whooping cranes used public lands in the NGP as drought refugia except during extreme autumn droughts.Public lands in the CGP were used less during moderate autumn drought, used similarly during extreme autumn drought and moderate spring drought, and used more during extreme spring droughts compared to non-drought conditions.Public lands in the SGP were refugia during autumn droughts and were used similarly during spring droughts.In circumstances where public lands served as drought refugia in the NGP, the U.S. Army Corps of Engineers owned 49% of sites, U.S. Fish and Wildlife Service 17%, Canadian Ministry of Environment 12%, and other owners <10% each (Table S1).In the CGP, public drought refugia were owned by the U.S. Bureau of Reclamation (35%), U.S. Army Corps of Engineers (25%), U.S. State Fish and Wildlife Agencies (23%), and the U.S. Fish and Wildlife Service (17%).Public drought refugia in the SGP were owned by the U.S. Army Corps of Engineers (59%), local municipalities (20%), and other owners (<10% each).

| Critical habitat
When quantifying overall percentage of marked cranes using identified critical habitat locations, we found Salt Plains NWR was used by an average of 27% of birds with active transmitters each migration season, 16% for Quivira NWR, 8% for the Platte River, and 5% at Cheyenne Bottoms WMA.Across sites, proportional use was consistently greater during autumn than spring migration, varying from 1.4 times more often at Quivira NWR and Cheyenne Bottoms WMA to 1.9 times more often at Salt Plains NWR.Most sites experienced similar use during moderate and extreme drought compared to non-drought conditions, although CIs of use ratios were wide in many cases (Figure 4).Cheyenne Bottoms WMA was used less during extreme drought conditions.The Platte River and Quivira NWR also showed indications of less use during moderate droughts, whereas Salt Plains NWR indicated greater potential to serve as refugia during extreme drought.

| Movement distances
After removing movement distances where location information was missing between subsequent days, 117 cranes moved an average of 279.5 km (n = 2401, SD = 176.6)between stopover sites during spring, and 128 cranes moved an average of 326.1 km (n = 1836, SD = 219.2) during autumn.In spring, the model that included region as an additive factor was preferred to one where region interacted with drought status (ΔAIC = 5.5).The difference in migration movements occurring after a whooping crane had been at a stopover site during non-drought conditions compared to moderate drought conditions was not different from zero (9.4 km; 95% CI: À7.2-25.9),whereas movements after extreme drought were 52.6 km (95% CI: 36.0-68.7;18%) longer compared to non-drought conditions.In autumn, a model describing region as an additive factor similarly described the data (ΔAIC = 0.7).The difference in migration movements occurring after a whooping crane had been at a stopover site during non-drought conditions compared to moderate drought conditions was not different from zero (13.1 km; 95% CI: À2.1-29.1),whereas movements after extreme drought were 27.1 km (95% CI: 11.5-42.8;9%) shorter compared to non-drought conditions.

| Stopover duration
Including site use where we did not have missing data before or after a stopover, 117 cranes resided an average of 2.9 days at spring stopover sites (n = 2371, SD = 4.5) and 126 birds resided an average of 4.7 days at autumn stopover sites (n = 1761, SD = 9.3).During spring, a model including an interaction between drought severity and region was better supported by the data than additive effects (ΔAIC = 16.6).Stopover duration did not differ between non-drought and moderate or extreme drought conditions in the CGP and SGP.In the NGP, cranes spent 0.2 days (95% CI: 0.1-0.3;15%) longer at sites during extreme drought compared to non-drought conditions.During autumn, the interactive model also was better supported by the data (ΔAIC = 32.1).We found little support for differences in stopover duration in association with drought severity in the CGP and SGP.In the NGP, stopover duration decreased 0.7 days (95% CI: 0.6-0.8;34%) during moderate drought compared to non-drought conditions.

| DISCUSSION
Whooping cranes modified habitat use and migration strategies to varying degrees depending on severity and location of drought.These broad patterns indicate a level of behavioral plasticity that has served this species in the past and will be needed for its continued recovery with ongoing climate and land use change.Quantifying the extent to which individuals can modify behaviors to extreme conditions like drought which are expected to intensify in the future, represents a knowledge gap for many species (Beever et al., 2017).Whooping cranes experienced moderate and extreme drought regularly during portions of most migration events.Whooping cranes readily modified habitat use of different types of wetlands in response to drought conditions, aligning with previous observations that this species used wetlands with a wide variety of broad-scale characteristics (Austin & Richert, 2005).Responses to drought varied spatially, likely because of regional differences in availability of wetlands across this large area.Whooping cranes have been characterized as being faithful to a migration corridor but generally selecting stopover sites opportunistically within it (Pearse et al., 2020).We suggest this opportunism arises partially by necessity, being driven by regular encounters with drought and other conditions that influence availability of quality stopover sites.For instance, the reconstruction of drought periods from tree ring growth rates indicates that extended droughts surpassing the severity of the 1930s have periodically occurred in the NGPs during the last several hundred years (Edmondson et al., 2014;Shapley et al., 2005).Whooping cranes persisted through these events and likely evolved under similar pressures.Though short duration droughts occurred during our study, it is unclear what the influence of a prolonged extreme drought severely limiting surface water availability would be on whooping crane habitat use patterns.
Changes in land use can exacerbate effects of climate change (Northrup et al., 2019).We observed mixed responses to how land use influenced drought resilience of wetlands depending on how wetland modifications influenced hydroperiod.Cultivated palustrine wetlands were commonly used (11-40%) among regions, especially during spring, when and where these shallow and ephemeral wetlands were available during non-drought conditions.Use of palustrine wetlands generally decreased during drought conditions, but use of cultivated palustrine wetlands was more affected by drought than uncultivated palustrine wetlands.Decreased proportional use by whooping cranes likely reflects reduced wetland function and indicates lower resilience of cultivated analogs to increasing intensities of drought.Cultivation in and around wetlands increases sedimentation, drainage, and loss of function (i.e., ponding) in palustrine wetlands (Gleason & Euliss, 1998;Johnson et al., 2012;McCauley et al., 2015), which are potential mechanisms for observed drought vulnerability.Programs to conserve wetlands subject to cultivation may mitigate effects of this practice on reducing hydroperiod of wetlands used by migrating and breeding waterbirds (Niemuth et al., 2006).
Palustrine wetlands modified to increase hydroperiod (e.g., impoundment, excavation), by contrast, were largely used similarly during drought or served as refugia.Impounding a site to modify or create a wetland increases water persistence under a range of hydrological conditions (Euliss & Mushet, 2004).Impounded palustrine wetlands were commonly used by roosting cranes during non-drought periods in the CGP and SGP (15-24%) where standing water can be scarce for migrating waterbirds (Barbaree et al., 2020;Stahlecker, 1992).These modified ponds are widespread across the central and western United States, created for drinking water for livestock and other uses (Natural Resources Conservation Service, 1997).Impoundments can be managed for a diversity of wildlife, serving as habitat for breeding and nonbreeding birds (Heitmeyer & Vohs, 1984;Lokemoen, 1973;Nelms et al., 2012).Whooping cranes generally use wetlands with low bank slopes, shallow depth profiles, and wide viewscapes (Austin & Richert, 2005;Baasch et al., 2022;Pearse et al., 2017).Impoundment can alter these preferred ecological conditions by increasing bank slopes and, by association, height of visual obstructions (e.g., vegetation) surrounding the wetland basin as well as influencing the food web (Euliss & Mushet, 2004).Moreover, although artificial or impounded wetlands are used by waterbirds, they do not support as diverse or abundant biological communities (Bellio et al., 2009).Our results indicate these roost habitats were useful locations as stopover sites for whooping cranes where natural wetlands were either not present or dry and will continue to serve as habitat if site characteristics match needs of migratory waterbirds like whooping cranes.If impoundments are intentionally created to meet the needs of waterbird communities, including whooping cranes, where wetland habitat is limited and droughts frequent, maintaining relatively natural topographical forms (e.g., low bank slopes) could bolster their value across hydrological contexts (Austin & Richert, 2005;Caven et al., 2022;Pearse et al., 2017).
Whooping cranes frequently used impounded lacustrine systems, or reservoirs, during non-drought periods in the SGP during both seasons and the CGP during autumn.Waterbird use of artificial or managed wetlands and impoundments can be extensive, especially when natural wetlands are less available because of land conversion or drought (Da Silva et al., 2020;Li et al., 2013;Navedo et al., 2011).Moreover, reservoirs provided whooping cranes reliable refugia during drought conditions across the Great Plains and this was most evident in the SGP, where autumn use during extreme droughts represented two-thirds of stopover nights.Thus, reservoirs partially ameliorated effects of drought by increasing probability of surface water persistence and providing refugia.Stopover stay length was generally reduced at larger lacustrine systems that included greater maximum water depths and wetted widths (Caven et al., 2022).However, whooping cranes have had extended stays at reservoirs, including during nonbreeding periods (Jung et al., 2022), indicating these sites have potential to provide quality habitat for whooping cranes.Further investigations regarding specific sites used and water conditions during extended stopovers may provide insight to appropriate land management and restoration techniques that could be incorporated when compatible with the diverse array of primary objectives of reservoirs in the Great Plains (McConnell, 2018;Vanausdall & Dinsmore, 2021).Nonetheless, impoundment of rivers across the Great Plains has greatly affected hydroperiods and vegetation communities of riparian areas (Costigan & Daniels, 2012), reducing the quality of riverine stopover sites in regulated rivers for waterbirds (Caven et al., 2019;Krapu et al., 2014).
Promoting conservation of a diverse network of stopover site complexes across the migration corridor aligns well with how whooping cranes modified habitat use in response to drought as compared to focusing attention only on specific sites or types of wetlands.We found considerable variability in how whooping cranes used different wetland types, sizes, and ownerships under different drought severity, regional, and seasonal contexts.In addition, although critical habitat areas were generally used in similar intensity regardless of drought severity, they were able to serve this function only in a portion of the migration corridor from central Nebraska to northern Oklahoma.Thus, whooping cranes did not consistently rely upon a specific subset of roost characteristics or locations to serve as drought refugia.In contrast, dynamic shifts in use were evident for many seasons and regions.For example, in the CGP during spring migration, palustrine and cultivated palustrine wetlands were most used during non-drought condition, impounded palustrine and lacustrine wetlands dominated use during moderate drought, and finally, river use was greatest during extreme drought conditions.Therefore, each of these wetland categories served as useful stopover sites for whooping cranes when they were most available and needed, depending on hydrological conditions.Complementary use of different wetland habitats, depending on their availability and quality under varying conditions, provides waterbirds with more stable and reliable stopover areas in variable environments than isolated and less diverse types of habitats (Barbaree et al., 2020;Kloskowski et al., 2009;Li et al., 2013).Conserving a functionally connected, diverse, and complementary network of wetland habitats to facilitate migration, which may benefit a wider range of migratory waterbirds, will entail transforming these concepts into identifiable areas where such habitats exist and may be currently insufficient or could be so in the future (Donnelly et al., 2022).
Whooping cranes expressed more subtle migration adjustments, primarily in movement distances when encountering extreme drought compared to moderate or non-drought conditions.The magnitude and direction of these modifications differed between spring and autumn migrations, which may provide insight to the relative stress birds are under during each migration season.
Duration of stopovers has been positively associated with subsequent migration bout distances (Pearse et al., 2020), and we found that same pattern for birds using the NGP during both seasons.During spring droughts, birds resided 15% longer at stopover sites and flew 18% longer distances, whereas during autumn birds responded to drought with 34% shorter stays in the NGP and 9% shorter migration flights.These seasonally different responses to drought indicate that whooping cranes may have a more limited ability to respond to extreme droughts in autumn compared to spring, putting greater emphasis on providing quality habitat during autumn.
Effective conservation plans for migratory birds emulate their extensive yearly range by being wide-ranging and accounting for the needs of species under varying environmental conditions, especially when that variation leads to resource bottlenecks (Maron et al., 2015).Our findings inform resource managers that diverse and complementary wetland habitats that are distributed across the migration corridor are beneficial for whooping cranes.In addition, certain portions of the corridor with regionally limited availability of quality stopover habitat may benefit from a conservation focus on wetland types used most frequently by whooping cranes.For example, we identified reservoirs as drought refugia across most of the migration corridor and as a high use migratory habitat in the SGP.These reservoirs are managed by many different federal, state, and local governments (Table S1), many of which may not have been engaged by the whooping crane conservation community or migratory bird managers in general.Therefore, opportunity exists to identify new partners in whooping crane recovery by initially recognizing their role in providing habitat for migrating cranes.Opening lines of communication with new partners may also alleviate potential issues related to disturbance, as use of non-traditional sites can lead to disturbance that the cranes may not normally be exposed to during non-drought conditions (Luo et al., 2012).Challenges faced by reservoir managers resulting in loss of water capacity may have collateral benefits to whooping cranes and other waterbirds, as sedimentation at inflow areas likely creates extensive shallowly flooded areas that provide safe and open roost sites (Rahmani et al., 2018).Similarly, drawn down reservoirs generally support a greater spatial extent of shallow water and mudflats, which provide improved habitat for shorebirds and wading birds such as cranes (Colwell & Taft, 2000).Nonetheless, other threats such as reduction in groundwater levels and climate change could be common to reservoir managers and whooping cranes (Brikowski, 2008).Ultimately, conservation programs that enhance a diverse array of wetland habitat within the migration corridor will increase the resilience of the whooping crane to drought and climate change.
We explored the concept of resilience at multiple ecological scales relating to whooping crane migration and hydrological drought.First, we explored resilience of whooping cranes to different levels of drought across their migration corridor and documented capacity to shift habitat use across hydrological contexts and regions.Secondly, we examined resilience of different wetland types and sizes to gradients of drought severity using the whooping crane as an ecological indicator, finding that small wetlands were generally less resilient than larger size categories as they mostly demonstrated reduced use during extreme droughts.Different wetland types were resilient to extreme drought in distinct regions, but impounding waterbodies generally improved resilience for individual palustrine and lacustrine wetlands.Nonetheless, research demonstrates that numerous impoundments or excavations can have the opposite effect and decrease resilience to drought at the broader watershed scale (Perkin et al., 2015;Tang et al., 2018).Pumping to inundate wetlands for agricultural purposes (e.g., cattle ponds, rice fields) can similarly have positive local impacts on whooping crane habitat availability (Johnsgard, 2015;Jorgensen & Dinan, 2016), but cumulative negative impacts on watershed functionality (Perkin et al., 2017).This paradox highlights how resilience varies across scales and is dependent upon interrelated ecological relationships at multiple spatial and temporal extents (Gladstone-Gallagher et al., 2019;Peterson et al., 1998).Applying habitat conservation and restoration approaches that consider and balance ecological functionality at multiple levels of organization will benefit migratory waterbird conservation.

F
I G U R E 1 Regions within the Canadian Prairies and United States Great Plains where we studied use of nightly roost sites by migrating whooping cranes, 2010-2022, including critical habitat areas within the whooping crane migration corridor (Central Platte River, Cheyenne Bottoms Waterfowl Management Area, Quivira National Wildlife Refuge, Salt Plains National Wildlife Refuge).Moderate and extreme drought conditions were spatially variable across the migration corridor of whooping cranes during spring and autumn migration seasons, 2002-2022.
T A B L E 1 Nightly proportional use of sites by migrating whooping cranes across the Canadian Prairies and United States Great Plains, 2010-2022.

F
I G U R E 2 Ratio of proportion of roost habitat categories used by whooping cranes during moderate or extreme drought to proportion used during non-drought conditions across four regions and two migration seasons (autumn and spring) within the Canadian Prairies and United States Great Plains, 2010-2022 (±95% confidence intervals).Roost habitat categories are included by region when use was >0.10.Values >1.0 indicate greater use during drought, values = 1.0 indicate similar use during drought, and values <1.0 indicate reduced use during drought compared to non-drought conditions.

F
I G U R E 3 Ratio of proportion of publicly owned lands used by whooping cranes during moderate or extreme drought to proportion used during non-drought conditions across 4 regions and 2 migration seasons (autumn and spring) within the Canadian Prairies and United States Great Plains, 2010-2022.BTZ = Boreal Transition Zone; NGP = Northern Great Plains; CGP = Central Great Plains; SGP = Southern Great Plains (±95% confidence intervals).Values >1.0 indicate greater use during drought, values = 1.0 indicate similar use during drought, and values <1.0 indicate reduced use during drought compared to non-drought conditions.F I G U R E 4 Ratio of proportion of locations identified as critical habitat used by whooping cranes used during moderate or extreme drought to proportion used during non-drought conditions (±95% confidence intervals).Values >1.0 indicate greater use during drought, values = 1.0 indicate similar use during drought, and values <1.0 indicate reduced use during drought compared to non-drought conditions.NWR = National Wildlife Refuge.