Sixty-years of community-science data suggest earlier fall migration and short-stopping of waterfowl in North America

ABSTRACT Worldwide, migratory phenology and movement of many bird species is shifting in response to anthropogenic climate and habitat changes. However, due to variation among species and a shortage of analyses, changes in waterfowl migration, particularly in the fall, are not well understood. Fall migration phenology and movement patterns dictate waterfowl hunting success and satisfaction, with cascading implications on economies and support for habitat management and securement. Using 60 years of band recovery data for waterfowl banded in the Canadian Prairie Pothole Region (PPR), we evaluated whether fall migration timing and/or distribution changed in Mallard (Anas platyrhynchos), Northern Pintail (A. acuta), and Blue-winged Teal (Spatula discors) between 1960 and 2019. We found that in the Midcontinent Flyways, Mallards and Blue-winged Teal migrated faster in more recent time periods, whereas Northern Pintail began fall migration earlier. In the Pacific Flyway, Mallards began fall migration earlier. Both Mallards and Northern Pintails showed evidence of short-stopping in the Midcontinent Flyways. Indeed, the Mallard and Northern Pintail distribution of band recovery data shifted 180 and 226 km north, respectively, from 1960 to 2019. Conversely, Blue-winged Teal recovery distributions were consistent across years. Mallards and Northern Pintails also exhibited an increased proportion of band recoveries in the Pacific Flyway in recent decades. We provide clear evidence that the timing and routes of fall migration have shifted over the past 6 decades, but these phenological and spatial shifts differ among species. We suggest that using community-science data collected by hunters themselves to explain one of the group's major concerns (changes in duck abundance at traditional hunting grounds), within the environmental lens of climate change, may help lead to further engagement and two-way dialogue to support effective waterfowl management for these culturally and ecologically important species. How to Cite Cox, A. R., B. Frei, S. E. Gutowsky, F. B. Baldwin, K. Bianchini, and C. Roy (2023). Sixty-years of community-science data suggest earlier fall migration and short-stopping of waterfowl in North America. Ornithological Applications 125:duad041. LAY SUMMARY Understanding how waterfowl migration may shift in time and space is important to guide waterfowl hunting policy, species management, hunter satisfaction, and ecosystem function. We use 60 years of community-science banding and recovery data for Mallard, Northern Pintail, and Blue-wingedTeal banded in Prairie Canada to assess long-term changes in fall migration over time and space. Migration begins earlier for Mallard and Northern Pintail in recent decades, while Blue-winged Teal appears to migrate more quickly. Mallard and Northern Pintail complete their migration farther north in recent decades, with fewer birds reaching the Louisiana Bayous, and are more likely to migrate to the Pacific coast. RÉSUMÉ Dans le monde entier, la phénologie migratoire et les déplacements de plusieurs espèces d'oiseaux se modifient en réponse aux changements anthropogéniques du climat et de l'habitat. Cependant, en raison des variations entre les espèces et du manque d'analyses, les changements dans la migration de la sauvagine, particulièrement à l'automne, ne sont pas bien compris. La phénologie de la migration automnale et les patrons de déplacements dictent le succès et la satisfaction des chasseurs de sauvagine, avec des implications successives sur les finances et le soutien de la gestion et la préservation des habitats. En utilisant 60 ans de données de retour de bagues pour des canards baguée dans la région des fondrières des Prairies du Canada, nous avons évalué si le moment de la migration automnale et la répartition ont changé chez le canard colvert (Anas platyrhynchos), le canard pilet (A. acuta) et la sarcelle à ailes bleues (Spatula discors) entre 1960 et 2019. Nous avons constaté que dans les voies migratoires du centre du continent, les canards colverts et les sarcelles à ailes bleues migraient plus rapidement au cours des périodes les plus récentes, tandis que le canard pilet commençait sa migration d'automne plus tôt. Dans la voie migratoire du Pacifique, les canard colverts ont commencé leur migration d'automne plus tôt. Le canard colvert et le canard pilet A. acuta présentaient des signes de raccourcissement de leur distance de migration dans les voies migratoires du centre du continent. En effet, selon les données sur les retours de bagues, la répartition du canard colverts et du canard pilet s'est déplacée respectivement de 180 km et de 226 km vers le nord entre 1960 et 2019. À l'inverse, la répartition des retours de bagues de sarcelle à aile bleues était constante d'une année à l'autre. Le canard colvert et le canard pilet ont également présenté une proportion accrue de retours de bagues dans la voie migratoire du Pacifique au cours des dernières décennies. Nous démontrons clairement que le moment et les routes de la migration automnale ont changé au cours des six dernières décennies, mais que ces changements phénologiques et spatiaux diffèrent d'une espèce à l'autre. Nous suggérons que l'utilisation de données scientifiques communautaires collectées par les chasseurs pour expliquer l'une des principales préoccupations de ce groupe dans le contexte environnemental des changements climatiques (c.-à-d. les changements dans l'abondance des canards sur les lieux de chasse traditionnels) peut aider à renforcer l'engagement et le dialogue afin de soutenir une gestion efficace de la sauvagine pour ces espèces culturellement et écologiquement importantes.


INTRODUCTION
Climate change and widespread habitat transformation are impacting bird distributions and migration phenology worldwide (Chen et al. 2011, Hurlbert and Liang 2012, Romano et al. 2022, Saunders et al. 2022).Climate-related changes in the timing of spring migration in the Northern Hemisphere are well documented (Gallinat et al. 2015), with many bird species exhibiting advanced pre-breeding migration and earlier breeding (Gallinat et al. 2015, Lehikoinen et al. 2019).Likewise, many species have shifted their distributions in response to changing climatic conditions (e.g., precipitation, temperature) and land-use practices (e.g., agriculture, forestry) (Chen et al. 2011, Regos et al. 2018, Saunders et al. 2022).To date, most research has focused on songbird spring migration, with relatively few studies investigating large-scale phenological and distributional shifts of waterfowl or fall migration (Gallinat et al. 2015; but see Thurber et al. 2020).
While available research suggests that migration of birds in the Northern Hemisphere is shifting in both the spring (i.e.northward on return to breeding areas) and fall (i.e., southward to wintering areas) in response to climate and habitat changes, fall phenology changes are less ubiquitous than during the spring season and vary by species.Previous research has reported delayed, advanced, and unchanged postbreeding migration timing in different avian species during fall migration (Gallinat et al. 2015, Romano et al. 2022).Climaterelated changes in fall migration timing appear to be related to a species' migration distance, with short-distance migrants tending to delay fall migration and long-distance migrants tending to advance fall migration (Jenni andKéry 2003, Van Buskirk et al. 2009).Long-distance migrants generally exhibit lower phenotypic flexibility (Both et al. 2010, Moussus et al. 2011, Kullberg et al. 2015; but see Brown et al. 2021) as their migration schedule is typically more strictly controlled by photoperiod and endogenous circannual cycles (Berthold 1996).Although relatively few studies exist in waterfowl, research to date suggests that fall migration is occurring later in most species, both in North America (Andersson et al. 2022) and Europe (Lehikoinen and Jaatinen 2012).Delays are primarily linked to warming temperatures, reduced snow cover, and large-scale weather patterns (Schummer et al. 2010, Xu and Si 2019, Thurber et al. 2020, Andersson et al. 2022), which prolong post-breeding food availability, causing individuals to postpone migration departures (Schummer et al. 2010, Xu andSi 2019).Land-use changes may also affect the suitability of stopover and staging sites (Guillemain et al. 2015b), causing birds to stay longer where more resources are available.Many factors may impact migration timing, including precipitation (O'Neal et al. 2018), wind (O'Neal et al. 2018, Xu and Si 2019, Haest et al. 2019), intraspecific competition (Eichhorn et al. 2009, Stirnemann et al. 2012), predation pressure (Jonker et al. 2012), and human disturbance (Väänänen 2001), which can complicate identifying and understanding changes in fall migration phenology.Moreover, unlike spring, where migration timing is driven by a desire to migrate quickly and arrive on the breeding grounds early, fall migration is more flexible (Yohannes et al. 2009, Karlsson et al. 2012, Nilsson et al. 2013), which may make shifts in fall migration timing more difficult to decipher than in the spring.
Climate and habitat alterations may not only drive changes in bird migration phenology, but lead to overall shifts in avian species distributions.In the northern hemisphere, ranges of numerous avian species are moving northward in response to climate warming (Hitch and Leberg 2007, Devictor et al. 2008, Saunders et al. 2022, Bosco et al. 2022).The extent of northward range shifts can vary according to several species' traits, including body size, diet type and breadth, habitat, migratory status, and clutch size (reviewed by Rushing et al. 2020).In short-distance migrants, climate-related range shifts are predicted to expand distributions and shorten migration distances (Curley et al. 2020, Howard et al. 2020, Rushing et al. 2020); however, climate change may produce longer migration distances and range contractions in long-distance migrants (Curley et al. 2020, Rushing et al. 2020).During fall migration, many species, particularly geese, show evidence of short-stopping (Lehikoinen et al. 2013, Elmberg et al. 2014), whereby populations migrate shorter distances, leading to northward wintering range shifts (Elmberg et al. 2014).For geese, short-stopping is primarily due to changes in agricultural practices (Jefferies et al. 2003, Gauthier et al. 2005), but climate change is an important driver in many other species (La Sorte andThompson III 2007, Teitelbaum et al. 2016).Reduced snow cover and ice and warming temperatures are similarly predicted to cause northward wintering range shifts in dabbling ducks (Musil et al. 2011, Notaro et al. 2016, Reese and Skagen 2017).To date, more northern wintering ranges have been observed in several species, including Mallard (Anas platyrhynchos), American Black Duck (A. rubripes), and Northern Pintail (A. acuta; Brook et al. 2009, Meehan et al. 2021).
Understanding changing waterfowl migration patterns is critical for ensuring effective management and conservation planning.Accurate estimates of waterfowl phenology and distribution throughout the annual cycle are critical to ensuring that key resources (e.g., food, habitat) are present when and where they will be most useful (reviewed by Andersson et al. 2022).This is important from a conservation perspective, as asynchrony between resource requirements and availability can negatively impact survival and reproduction (Newton 2007, Kirby et al. 2008, Stafford et al. 2014) and have been linked to declines in numerous migratory bird species (Baker et al. 2004, van Bemmelen et al. 2021).Moreover, shifting fall migration could decouple monitoring and management efforts from true population numbers (Devers et al. 2021, Roberts et al. 2022) and could substantially alter foraging pressures in wetland and agricultural habitats (Schummer et al. 2014).Understanding changes to the timing and distribution of fall migration could also help direct waterfowl harvest policies by ensuring that hunting seasons coincide with waterfowl occurrence (Andersson et al. 2022).Furthermore, waterfowl hunting and birdwatching provide substantial economic benefits, with both hunters and bird watchers spending billions of dollars annually (reviewed by Notaro et al. 2016).Changes in the distribution and timing of fall migration could have repercussions on local economies and affect hunter harvest and satisfaction.
Because factors affecting fall migration occur on the breeding range and at different points along the migratory corridor, fall migration is best studied across large spatiotemporal scales (Elmberg et al. 2014).In particular, broad spatial and temporal datasets are necessary for disentangling whether species migrate through an area at a different time, whether short-stopping is occurring, or whether species are using alternative migration routes.To date, delayed fall migration in waterfowl has largely been documented at small scales, over relatively short timeframes, or for single species (Reese and Weterings 2018, Thurber et al. 2020, Andersson et al. 2022, Masto et al. 2022, Weller et al. 2022 but see Meehan et al. 2021) highlighting the need for broader-scale analyses.Due to limited availability of long-term data, most studies of changing phenology do not include time before climate change effects accelerated in the 1980s and thus many omit crucial periods of adaptation (Van Buskirk et al. 2009, Brisson-Curadeau et al. 2020, Thurber et al. 2020).Previous work has largely focused on smaller spatial scales given the challenges associated with monitoring species across large geographic areas (Bauer et al. 2008, Faaborg et al. 2010, Bridge et al. 2011 but see Reese andSkagen 2017, Meehan et al. 2021).
Programs that gather and facilitate the use of data collected by non-scientists (hereafter, community science) offer an opportunity to collect data across spatial and temporal scales that would be impossible for traditional survey methods (Dickinson et al. 2010, Theobald et al. 2015).In North America, large-scale waterfowl banding programs have been in operation since the 1920s (Alisauskas et al. 2014, Anderson et al. 2018).Waterfowl hunters are requested to report recov-eries of banded birds they harvested to the U.S. Geological Survey (USGS) Bird Banding Laboratory.Reporting rates have increased over time due to the introduction of new technologies for reporting, and current reporting rates are as high as 95%, indicating that waterfowl hunters are actively engaged in this long-running community science program (Alisauskas et al. 2014, Vrtiska 2021).The band reporting program was originally intended to inform species distribution and migratory patterns (Anderson et al. 2018), and band recoveries have since been used to identify the migratory flyways (Lincoln 1935, Roberts et al. 2022).However, the use of the banding program quickly evolved, and by 1930 Lincoln suggested that bands reported by hunters could also be used to estimate waterfowl population size.The banding program is now a mainstay of the waterfowl management program in North America and reported bands are used to estimate harvest rates, survival rates, and population sizes (Alisauskas et al. 2009(Alisauskas et al. , 2014;;Bartzen and Dufour 2017).Waterfowl hunting seasons are timed to coincide with fall migration, across the North American continent.As approximately 22,000 ducks were banded in Canadian prairie provinces in 2022 alone (Yates 2022), community-science band recovery data represents an exceptional opportunity to assess how waterfowl fall migration timing and distribution are changing.
We investigated long-term changes in waterfowl migration using 60 years of hunter band recoveries for ducks banded in the Canadian Prairie Pothole Region (PPR).The importance of the PPR to breeding waterfowl has been extensively documented, and it is estimated that the PPR produces over 50% of North America's duck population (Johnson and Grier 1988, Batt et al. 1989, Doherty et al. 2018).Our objective was to identify whether waterfowl fall (southward) migration phenology and/or distribution changed over time between 1960 and 2019.We focused on 3 species: Mallard, Northern Pintail, and Blue-winged Teal (Spatula discors).Band recoveries for these species provided sufficient data for quantitative spatiotemporal analyses, and all 3 species are of high economical and ecological importance.

Data Acquisition
Duck banding in prairie Canada has been carried out by a variety of agencies, but predominantly by personnel with the U.S. Fish and Wildlife Service and the Canadian Wildlife Service.The target species have historically been Mallards and Northern Pintails, but other species are captured and banded incidentally.A variety of capture methods have been used, but the vast majority of birds are captured using baited swim-in traps (North American Banding Council 2017).The objective of the program is to provide data in support of harvest management decisions (e.g., age-specific annual survival and harvest rates), and individuals are considered representative of the overall population, although sample sizes for cohorts are unbalanced.Evaluations of potential bias in condition of birds captured using bait during fall support the representativeness of sampling using baited traps (Reinecke and Shaiffer 1988), and the large sample size and distribution of effort by a large number of personnel minimize risk of biased marking.
We requested banding and recovery records from the Bird Banding Office (Bird Banding Biology, CWS, bbo@ec.gc.ca), which works in collaboration with the USGS to handle Canadian banding and data requests, for 6 waterfowl species banded in Manitoba (MB), Saskatchewan (SK), or Alberta (AB) available as of 3 December 2020, when we downloaded the data.Our initial 6 focal species all commonly breed in the Canadian prairies: Mallard, Northern Pintail, Bluewinged Teal, American Green-winged Teal (A. carolinensis), Gadwall (Mareca strepera), and American Wigeon (M.americana).Digitization of banding records was incomplete prior to 1960, so we restricted analyses to 1960-2019.To assess spatiotemporal changes in waterfowl fall migration, we restricted analyses to recoveries that occurred during the first hunting season after banding and those for which the precision of recovery locations was known within 10 min (~18.5 km).The "Migratory Bird Convention" stipulates that hunting seasons in the US and Canada must occur between September 1 and March 10 and be no more than 107 days in length, though additional restrictions may be set by either party.While the hunting season may extend beyond December in some portions of the United States, we excluded records beyond this point in our analysis to avoid geographic bias between areas that still had active hunting seasons later in the winter vs. those where hunting no longer occurred.By the end of December most if not all individuals of our focal species will have reached terminal points.Blue-winged Teal and Northern Pintails are some of the first ducks migrating southward in the fall, with peak migration for both species occurring in September and October (Clark et al. 2020, Rohwer et al. 2020).Mallards migrate later in the season, with migration peaks between late-October to mid-December, with some of the most southern migrants overwintering in Mexico beginning to arrive by November (Drilling et al. 2020).We included both hunter recoveries (i.e., birds harvested by a hunter) and birds that were found dead, restricted to Canada and the United States.Relying on hunter recoveries will bias our results towards areas with more hunting activity; this bias is difficult to disentangle as hunting activity is tied to waterfowl densities (Schroeder et al. 2019).Wintering distributions of all 3 species extend into Mexico, and Blue-winged Teal and Northern Pintail wintering distributions extend into Central and South America (Clark et al. 2020, Drilling et al. 2020, Rohwer et al. 2020).However, waterfowl bands are only irregularly reported outside of Canada and the U.S. and the analytic methods we used require large numbers of band recoveries with unbiased reporting rates.We did not filter by age or sex of the bird or time of banding.

Evaluating Spatiotemporal Change in Waterfowl Recovery Distribution
To assess spatiotemporal changes in the phenology and distribution of fall migration, we divided the banding and recovery records into three 20-year periods (1960-1979, 1980-1999, and 2000-2019) and eight 2-week intervals from September through December.Smaller time-steps (e.g., 10-yr periods) resulted in insufficient and unequal sample sizes, while larger time-steps (e.g., 30-yr periods) amalgamated too many years to capture changes.Two-week intervals provided the shortest time interval to capture a "snapshot" of the distribution of recoveries through time that enabled sufficient sample sizes.The first 2 weeks of September had far fewer recovery records; Northern Pintail recovery distributions could not be assessed during this interval.We restricted analysis of spatiotemporal trends in fall migration to Mallards, Northern Pintails and Blue-winged Teal due to sample size limitations for other species (i.e., Gadwalls, American Widgeons, and American Green-winged Teal; Supplementary Material Table 1).
We conducted a kernel density estimator (KDE) analysis using the R package adehabitatHR (Calenge 2006), with an ad hoc smoothing parameter with the default grid value of 60 for each species, time period, and 2-week interval.KDE is well suited to identifying areas of high probability density from point locations, and has successfully identified spatio-temporal distributions of waterfowl recoveries across the entire hunting season (Green andKrementz 2008, Calenge et al. 2010).Expanding on the approaches taken by Calenge et al. (2010) and Green and Krementz (2008), we tracked spatio-temporal distributions of waterfowl band recoveries during migration by 2-week intervals and across decades.We extracted utilization distribution (UD) contours to produce spatial polygons for each of the 50%, 70%, and 90% probability density contours using the function getverticeshr.We assessed spatial relationships between the 50% UD KDE contours for each 2-week interval and 20-yr period using approaches from STAMP (Spatio-Temporal Analysis of Moving Polygons, R package stampr) to generate metrics describing change events on spatial relationships (Long et al. 2018).For each time interval, we describe stepwise change between the 20-yr periods in relation to measures of distance and direction between polygons, and changes in polygon shape.We quantified 3 distinct types of distribution change events between subsequent periods (e.g., comparing 1960-1979 to 1980-1999, comparing 1980-1999 to 2000-2019).To do this, we quantified areas of stability (i.e., areas of complete overlap between the time periods), areas of expansion (areas only included in the later time period), and areas of contraction (areas only included in the earlier time period).Then, we quantified the absolute area (km 2 ) of each periods' KDE and the proportion of that KDE that was stable, contracting, or expanding compared to the previous time period for each 2-week interval (Supplementary Material Figure S8).

Accounting for Changes in Banding Effort through Time and by Province
There are waterfowl banding stations distributed across the prairie provinces (Figure 1).There have been some changes in banding station activity in the PPR since the 1960s, with a greater proportion of Mallards banded in Alberta and a greater proportion of Blue-winged Teal and Northern Pintail banded in Saskatchewan through time (Supplementary Material Figure 1).Only Mallard recoveries showed strong spatial segregation by province of banding (Supplementary Material Figure 3).Early data exploration indicated Mallards banded in Alberta were substantially more likely to be recovered following the Pacific Flyway than Mallards banded in Saskatchewan or Manitoba, where migration typically followed the Mississippi and Central Flyways (collectively known as the Midcontinent).Mallards have been identified as having low to moderate migration connectivity (Roberts et al. 2022), with most of the mixing between flyways occurring in the Mississippi Flyway (Buhnerkempe et al. 2016).Because the relative proportion of birds banded in Alberta had changed over time, we chose to repeat the entire analysis to evaluate spatiotemporal changes in recovery distribution separately for (1) Mallards banded in Alberta and (2) Mallards banded in Saskatchewan or Manitoba.This allowed us to assess whether patterns of change were due simply to changes in relative provincial banding effort.However, addressing within province redistribution of banding effort was beyond the scope of this analysis.We did not detect spatial patterns in recovery based on the province of banding for the Northern Pintails or the Blue-winged Teal so we combined all 3 provinces in a single analysis.There were no directional biases in latitude.

RESULTS
Of the more than 2.5 million Mallard, Northern Pintail, and Blue-winged Teal banded in the Canadian Prairie provinces from 1960 to 2019, 185,000 were shot or found dead during the hunting season.The number of waterfowl banded peaked in the 1980-1999s at an average of 59,000 annually, but was similar in 1960-1970 and 2000-2019 (32,000 and 36,000, respectively).Most recoveries were dead recoveries from hunters (99%).Due to higher rates of banding, Mallards dominate the recovery datasets with 150,172 recovered, compared to 11,766 Northern Pintail, and 17,421 Blue-winged Teal (Table 1).71% of banded Mallards were male, compared to 51% of banded Northern Pintail and 62% of banded Bluewinged Teal.Males made up 84% of recoveries for Mallards, but only 62% for Northern Pintail and Blue-winged Teal.

Mallard
Mallards banded in the Canadian Prairies primarily migrated along the Pacific or Midcontinent Flyways.Prairie Mallards

Species
Alpha code Banded Recoveries migrating along the Pacific Flyway wintered in Washington and Oregon, while those migrating along the Midcontinent Flyway wintered farther south in Louisiana, Arkansas, and Mississippi.The 50%, 70%, and 90% UD KDE contours of Mallard dead recoveries suggest that Mallards following the Midcontinent Flyway in more recent time periods (1980-1999 and 2000-2019) migrated south more quickly during early migration (September to October) than they did in 1960-1979 (Figure 2; Supplementary Material Figure 8).By December, Mallards exhibited short-stopping, and settled into their wintering ranges farther north in 1980-1999 and 2000-2019 than they did in 1960-1979, with fewer birds reaching the Louisiana Bayous in the contemporary periods than 50-60 years ago (Figure 2).Based on centroids of 50% KDE, during the final two weeks of December, Mallards shifted 55 km north between 1960-1979 and 1980-1999, and another 125 km north between 1980-1999 and 2000-2019.Our initial analysis of the 50%, 70%, and 90% UD KDE contours of dead recoveries showed that more Mallards began migrating to the Pacific Flyway between the 1960-1979 and 1980-1999 time-periods (Figure 2; Supplementary Material Figure 8).

Mallard
In total, 76% of Mallards migrating through the Pacific Flyway were banded in Alberta.In contrast, most birds banded in Saskatchewan or Manitoba were recovered in Midcontinent Flyway (78% and 91%, respectively).However, Alberta banded Mallards accounted for a larger proportion of recoveries (Supplementary Material Table 2) due the combination of increasing banding effort in Alberta, and decreased relative banding effort in Saskatchewan and Manitoba (Supplementary Material Figure 1).A secondary analysis, restricted only to birds banded in Alberta, showed that Mallards have always moved through the Pacific Flyway to settle in Washington and Oregon, but began migrating earlier in the 1980-1999 and 2000-2019 periods than they did in the 1960-1979 period (Supplementary Material Figure 7).Separate analysis of Mallards banded in Saskatchewan and Manitoba aligned with the combined analysis of all prairie banded Mallards (Supplementary Material Figure 6).

Northern Pintail
Northern Pintail banded in the Canadian Prairies migrated along both the Pacific and Mississippi Flyways (Supplementary Material Figure 4).Pacific Flyway migrators wintered in California, while Midcontinent Flyway migrators wintered primarily in eastern Texas and Louisiana.A total of 59% of Pintail recovered in the Pacific Flyway were banded in Alberta.
Northern Pintail recovery distribution along the Pacific Flyway was similar across all three time-periods (i.e., migration followed a similar path, started at the same time, and ended in the same area).However, in 2000-2019, there was some indication that Northern Pintails remained farther west for longer during late October and November (Figure 3).In the Midcontinent Flyway, similar to Mallards, Northern Pintails appeared to exhibit shortstopping, and reached their terminal U.S. range as evident by no continued movement past November 16th farther north in 2000-2019 than they did in the 1960-1979 and 1980-1999

Blue-winged Teal
Blue-winged Teal banded in the Canadian Prairies migrated along the Mississippi Flyway and wintered along the Gulf Coast (Supplementary Material Figure 5).Recovery distributions were very similar across all years.However, in mid-October through to early November, Blue-winged Teal in the 1960-1979 time-period migrated more slowly than in the later periods, resulting in teal arriving slightly earlier on the Gulf coast in the later time periods (Figure 4).Blue-winged Teal also showed a small westward range contraction from the eastern part of their wintering grounds in the south-eastern United States in the most recent time periods (Figure 4).

DISCUSSION
Using 60 years of community-science data, we found evidence of spatiotemporal shifts of harvest banding recoveries for 3 culturally and ecologically important waterfowl species.This included earlier southward movements in the first part of the fall migration period (Northern Pintail) and evidence of short-stopping for wintering (Mallard and Northern Pintail), as inferred using STAMP methods with reported recoveries of dead birds.These empirical findings match community reports from hunters and biologists over the last decades that fewer waterfowl appear to be reaching their wintering grounds in the southern United States (Green and Krementz 2008, Slagle and Dietsch 2018, Whitaker et al. 2019, Moorman 2020).Despite challenges and possible biases inherent with communityscience databases like the waterfowl band recovery database, the sample size (>185,000 reported bands used in this study) allows for unprecedented spatiotemporal insight of changes in bird movements and timing.
Our analysis of band recoveries found changes in fall migration and wintering movements of duck species which supports the growing body of literature finding large effects of milder falls and winters on waterfowl migration in North America (Notaro et al. 2016, O'Neal et al. 2018, Xu and Si 2019, Thurber et al. 2020, Weller et al. 2022), and elsewhere in the Northern Hemisphere (Lehikoinen et al. 2013, Guillemain et al. 2013).Interestingly, previous work on Mallard harvest distributions in the Mississippi and Central Flyways using recovery data from 1980-2003 showed no consistent shifts northwards, leading to a rejection of the short-stopping hypothesis, and suggested that waterfowl biologists should offer concerned hunters alternative hypotheses to perceived declines of Mallard numbers (Green and Krementz 2008).Our study differed from the Green and Krementz (2008) study in several key ways: we included more years of data (60 vs. 23 years), a larger scope (3 flyways instead of 2), and created KDE in 2-week vs. 5-month periods.These differences, together or in part, may explain the differences in our analysis of banding recoveries from harvested birds, and our results demonstrating short-stopping by Mallards in the most recent time periods.While it appears that there was contraction from the southernmost region of the wintering ground for Mallards when comparing the first and middle 20-year periods of the dataset (1960-1979 vs. 1980-1999), there was not as much northward expansion in these periods as compared to the middle and last 20-yr periods (1980-1999 vs. 2000-2019).
Mallard and Northern Pintail recoveries were found at two distinct wintering areas, one along the Pacific coast in northern California and the other in south-central United States.Mallard and especially Northern Pintail exhibit mixing between flyways, whereby individuals banded in one flyway are commonly recovered in another flyway (Buhnerkempe et al. 2016, Roberts et al. 2022).This mixing is thought to be due to movement and connection on the breeding grounds (Buhnerkempe et al. 2016, Roberts et al. 2022).We found a similar mixing of recoveries across flyways in our analysis; in particular, Northern Pintail banded in the central prairie province of Saskatchewan had recoveries spread across all 3 flyways.Movement changes associated with these wintering grounds differed; there was little change with the Pacificwintering population, but the central population appeared to migrate earlier and exhibit short-stopping in the most recent time periods.Consistent with this finding, recent analyses of 50 yr of Christmas Bird Count (CBC) data for the Atlantic and Mississippi Flyways found that declines of wintering Northern Pintail were most pronounced in the southernmost regions, such as Louisiana, where the species was historically most abundant, and that increases were most pronounced in the north (Meehan et al. 2021).The CBC also points to declines in wintering Mallards in Louisiana (Meehan et al. 2021), despite increases in continental Mallard populations.Mallard abundance in Louisiana coastal wildlife management areas and refuges also does not reflect national increases and instead has been stable white wintering Northern Pintail populations declined precipitously in Louisiana refuges (Whitaker et al. 2019).In both cases, the authors suggested that short-stopping due to climate changes could explain the patterns found in their respective studies.Comparatively, our finding that there was little change in the Pacific-wintering population of Northern Pintails differs from work that documents a northward shift in wintering populations due to habitat changes (Fleskes et al. 2018), which may be explained by differences in wintering locations of pintail that migrate from the Canadian prairies vs. pintail originating in the Pacific Flyway.
The wintering ranges of many species of waterfowl, including Mallard and Northern Pintails, are predicted to keep shifting toward colder regions and northern latitudes under a global warming scenario of a 3°C increase in global annual temperature at the end of the century (Bateman et al. 2020).These observations and their related predictions are consistent with the frost-wave hypothesis, whereby successive waves of frost drive migration of waterfowl farther and farther south along their fall migration (Schummer et al. 2014, Xu andSi 2019).Short-stopping in migratory waterfowl has potentially serious consequences for ecosystem function and processes.Waterfowl migration connects ecosystems across the continent through nutrient cycling, causing nutrient transfer both locally and continentally between aquatic and terrestrial habitats (e.g., Manny et al. 1994, Post et al. 1998, Olson et al. 2005), and driving seasonal ecosystem cycles (e.g., Olson et al. 2005, Nakamura et al. 2010).Waterfowl are also an important long-distance dispersal mechanism for aquatic organisms (Figuerola and Green 2002, Frisch et al. 2007, Hessen et al. 2019).For most waterfowl species, food is thought to be one of the most limiting factors outside the breeding season and that winter habitat limitations can carry-over during the next breeding period (Sedinger andAlisauskas 2014, Williams et al. 2014).
Land cover change can work independently or cumulatively with climate change to drive shifts in fall migration and wintering in waterfowl.Wetlands provide crucial food resources to support the energetic requirements of migrating waterfowl and wetland losses along migratory pathways can reduce survival and population sizes, particularly for species with inflexible migratory routes (Bellrose andTrudeau 1988, Xu et al. 2019a, b).At the same time, agricultural landscape alteration can affect waterfowl distribution during the fall and winter by offering alternative food sources (Jefferies et al. 2003, Gauthier et al. 2005, Guillemain et al. 2015b).Positive benefits of specific agricultural practices and other forms of habitat management have also been used to provide refuges from the negative impacts of climate change on migrating waterfowl (Gaget et al. 2023).Agricultural changes have been the major drivers of dramatic short-stopping in goose species, like the now overabundant Greater (Anser caerulescens atlanticus) and Lesser Snow Geese (A. c. caerulescens; Jefferies et al. 2003, Gauthier et al. 2005).Human-induced habitat changes in the Camargue region of France affected the distribution and phenology of Common Teal (Anas crecca) and possibly Mallard (Guillemain et al. 2015a, b).Similarly, the distribution of waterfowl wintering within the Central Valley of California shifted northward in response to shifts in the distribution of agricultural fields flooded after harvest in the region (Fleskes et al. 2018), and may be related to the increased proportion of Northern Pintail and Mallard migrating to the Pacific Flyway.
In addition to short-stopping, Northern Pintail, but not Mallard, are beginning fall migration earlier in the Midcontinent.Earlier initiation of migration may be due to either changes in the demography of the population, or an advancing annual cycle.Both nonbreeding or unsuccessful adult waterfowl tend to begin migration earlier than successful breeders (Reed et al. 2003), so reduced breeding success or an increased proportion of nonbreeders could explain the earlier migration.Northern Pintails have experienced low productivity and nest failure in agricultural habitats, particularly in drought years when natural wetlands are unavailable, resulting in population declines since the 1960s, which may contribute to their earlier departure (Zhao et al. 2019(Zhao et al. , 2020)).Productivity for Mallards is more variable, with some studies concluding no reduced productivity for the species (Osnas et al. 2016), and others suggesting reduced productivity may occur in the northern parts of the Mallard's breeding population (Specht and Arnold 2018).Alternatively, juveniles tend to be harvested earlier than adults in the hunting season (Alisauskas et al. 2014, Roy et al. 2015), so a more productive population might appear to migrate earlier due to a larger proportion of juveniles.However, there are no directional shifts in age ratios of harvested Mallards or Northern Pintail that would suggest that increased juvenile harvest is responsible for this pattern (Osnas et al. 2016, Smith et al. 2021).Mallard and Northern Pintail spring migration has also moved earlier (Murphy-Klassen et al. 2005, Ward et al. 2016, Andersson et al. 2022), so an alternative, non-exclusive explanation is that the whole annual cycle has shifted earlier in response and earlier completion of breeding allows for earlier migration.
Mallards and Northern Pintails banded in Prairie Canada also show a larger portion of recoveries in the Pacific Flyway in recent decades, for which there are several, non-exclusive explanations.First, hunting effort allocation may have shifted and caused a pattern that does not reflect the birds' true movements.However, the proportion of Mallards and Northern Pintails harvested in the Pacific Flyway has remained relatively stable, hovering at 14.3% (11.5-16.9)and 31.7% (23.3-42.0),respectively (Supplementary Material Figure S2).Second, changes in band recoveries may reflect shifts in banding effort allocation.This is plausible for Mallards as the proportion of Mallards banded in Alberta has increased and 76% of prairie banded Mallards that were recovered in the Pacific Flyway were banded in Alberta.This proportional shift in banding effort likely contributed to the observed increase in prairie Mallards migrating to the Pacific Flyway.Northern Pintails do not show the same patterns in proportional banding effort.Finally, there may have been a true change in flyway affinity.A recent analysis of migratory connectivity based on banding/recovery records supports this hypothesis: Mallards were more likely to move from the Central Flyway to the Pacific Flyway in the 2000s compared to the 1960s, and Northern Pintails were less likely to stay in the Central Flyway (Roberts et al. 2022).Unlike in the Midcontinent, there is no evidence of short-stopping occurring for either species in the Pacific Flyway.We hypothesize this is a result of Mallards and Northern Pintails both relying heavily on irrigated agricultural landscapes in the Pacific Flyway, particularly flooded rice fields (Fleskes et al. 2005), and therefore food availability is less impacted by climate change than wetlands in the Midcontinent.
Blue-winged Teal breeding in the Canadian Prairies consistently migrate along the Midcontinent Flyway, with lower migratory connectivity to the coastal flyways than Mallards (Roberts et al. 2022).Blue-winged Teal is a long-distance migrant, which imposes additional constraints on migration phenology (Bellrose 1980, Notaro et al. 2016).Teal are also wetland obligate foragers and as such are not able to take advantage of agricultural food sources (Rohwer et al. 2020), unlike the other two species.In contrast to Mallard and Northern Pintail, photo-period, rather than weather conditions, is thought to govern Blue-winged Teal migration phenology (Van Den Elsen 2016).Accordingly, we found little evidence of consistent shifts in migratory phenology of Blue-winged Teal.These results are consistent with local-scale studies of Blue-winged Teal fall migration in the Great Lakes (Van Den Elsen 2016, Thurber et al. 2020) and Southern Quebec (B.Frei et al. personal observation), and CBC trends for Blue-winged Teal in the Atlantic and Mississippi Flyways, which are independent of regional temperature (Meehan et al. 2021).Likewise, Blue-winged Teal wintering range is not expected to change appreciably under a 3°C increase in global annual temperature (Bateman et al. 2020).Blue-winged Teal are also considerably smaller bodied than Mallard or Northern Pintail, and thus may be more likely to time migration to avoid energetic constraints, as is the case for the Common Teal in Europe (Dalby et al. 2013).

Management Implications
Mallards are often used as a reference species for dabbling ducks in harvest management and habitat conservation plans, in part because of the large amount of data available for this species, but also because the single-species approach also allows for simplicity and cost benefits (Herbert et al. 2021, Roberts et al. 2022).However, patterns in fall migration are not consistent across species, as exemplified by the Blue-winged Teal here.Similar divergent responses to climate change during the fall migration season within the waterfowl community have also been documented (Thurber et al. 2020, B. Frei et al. personal observation).The value of the Mallard as a representative of all dabbling ducks in management plans could therefore be expected to decline in the future and it has been suggested that harvest management and habitat conservations plans incorporate multispecies approaches in the future (Petrie et al. 2011, Williams et al. 2014, Humburg et al. 2018).Identifying species-specific shifts in the breeding and wintering distributions of migratory species is a high-priority knowledge gap, with critical implications for evidence-based conservation action and avoiding costly mistakes in wildlife management (Pacifici et al. 2014, Rushing et al. 2020).Following multi-stock management, as recently adopted by the Atlantic Flyway (Johnson et al. 2019), provides a more flexible and resilient approach to monitoring and managing harvested species.However, it would require reinvesting banding effort currently focused on Mallards towards other species, for whom banding numbers are currently too low to be modelled annually.Indeed, there will be trade-offs if habitat planning conservation actions are targeted toward the species that face large climatic stressors but have little migratory plasticity and are thus most vulnerable to climate change, as these species will have specific needs that will not be representative of the needs of all other species (Summers et al. 2012, Pacifici et al. 2014).
Overall, waterfowl conservation in North America owes its success to a history of government, stakeholders, private individual cooperation, and partnerships over small and large scales to coordinate management and conservation efforts (Brasher et al. 2019).Hunters have often been described as the bedrock of those conservation efforts, but climate change is expected to disrupt the traditional conservation model of waterfowl (Anderson et al. 2018, Vrtiska 2021).Expenditures of migratory bird hunters in southeastern states represents roughly a third of all waterfowl hunter expenditure in the U.S. and waterfowl hunting represent an important contribution to the economy of those states (Grado et al. 2001(Grado et al. , 2011;;Poudel et al. 2016).The success of the North American Waterfowl Management Plan hinges in part on the ability to protect and manage wetlands in USA, Canada, and Mexico though funds acquired under the North American Wetlands Conservation Act (NAWCA; USFWS 2018).States provide an important source of matching funds required by NAWCA and with the ongoing shifts northward driven by climate change, there is a resulting fear that the decoupling of southern wintering habitats with northern breeding grounds will result in decreased funding for these ecologically valuable northern wetlands.There is little hope of increasing hunters upstream in the flyway to compensate for the potential loss of Louisiana hunters, as upstream states have been losing waterfowl hunters more rapidly than Louisiana since 2000 (Supplementary Material Figure 11).Short stopping will be a particularly difficult hurdle to address, as waterfowl hunters tend to hunt locally and avoid long trips (Patton 2018).Therefore, in order to adapt to climate change, the waterfowl management community should not limit itself to the development of new harvest management and habitat conservation plans, but it will also need to invest time and resources in addressing social impacts.With contraction of wintering range, social impacts will include increased competition between hunters and costs to hunt, as well as decreased hunting access and hunter satisfaction; all of which will cumulatively lead to decreased recruitment and retention of hunters, who are the bedrock of waterfowl conservation.One of the critical components of the life-cycle of community-science is the scientific analysis of collected data and the communication of those results back to community stakeholders (Rüfenacht et al. 2021).Two-way dialogue with waterfowl hunters about the impact that climate change and agricultural practices have on waterfowl migration and hunting opportunities remains crucial for adaptive waterfowl harvest management in a changing world.

FIGURE 1 .
FIGURE 1. Banding stations in the Canadian Prairies are widely distributed for the three focal species.Point size represents the number of recovered waterfowl banded at the station.

FIGURE 2 .
FIGURE 2. Midcontinent migration of Mallards begins earlier and exhibits short stopping in recent time periods with a smaller proportion of Mallards reaching the Gulf Coast based on the STAMP comparison of the 1960-1979 and 2000-2019 50% UD KDE contours of Mallard band recoveries during 2-week intervals through fall migration (September-December).Note that the 50% kernel density estimation has been restricted to continental North America.An animated GIF of Figure 2 showing all three-time periods is available in the Supplementary Material.
time periods.In contrast to Mallards, Northern Pintails in the Midcontinent Flyway began fall migration earlier in 1980-1999 and 2000-2019 relative to 1960-1979.Also, unlike the Mallards, there was no indication of a smaller proportion of migrating Northern Pintails reaching the southern edge of Midcontinent Flyway wintering distribution.Based on centroids of the 50% KDE (restricted to overland), during the final two weeks of December, Northern Pintail shifted 113 km north between1960-1979 and 1980-1999, and a farther 113 km north  between 1980-1999 and 2000-2019.

FIGURE 3 .
FIGURE 3. Northern Pintail begin fall migration earlier and a larger proportion stay further north based on a STAMP comparison of the 1960-1979 and 2000-2019 50% UD KDE contours of Northern Pintail band recoveries during 2-week intervals.Note that the 50% kernel density estimation has been restricted to continental North America.An animated GIF of Figure 3 showing all three-time periods is available in the Supplementary Material.

FIGURE 4 .
FIGURE 4.There has been little change in the winter distribution of in Blue-winged Teal winter distribution in the U.S. based on a STAMP comparison of the 1960-1979 and 2000-2019 50% UD KDE contours of Blue-winged Teal band recoveries during 2-week intervals through fall migration (September-December).Note that the 50% kernel density estimation has been restricted to continental North America.An animated GIF of Figure4showing all three-time periods is available in the Supplementary Material.