Flexibility in the face of climate change? A rapid and dramatic shift towards later spring migration in Hudsonian godwits (Limosa haemastica)

Abstract

With rapid environmental change, shifts in migration timing are vitally important for population stability in migratory species and have been widely documented. However, little remains known about how migrants make these shifts and what factors influence the utilization of these strategies, limiting assessments of their vulnerability to climate change. Hudsonian godwits (Limosa haemastica) are extreme long-distance migratory shorebirds that (i) have previously advanced their population-level migration timing and (ii) are sexually dimorphic. We combined over a decade of tracking data from one breeding population with a historical predictive model to assess ongoing shifts in migration timing and investigate potential sex-specific migration strategies. We found that irrespective of sex, godwit departure and arrival timing shifted 6 days later from 2010 to 2023. The population maintained an average migratory duration of 24 days and drove shifts in arrival timing entirely by changing their non-breeding-ground departure. Yet, we also found godwits arrived later than predicted by the historical model, indicating that conditions on the non-breeding grounds may constrain their ability to respond to changes on the breeding grounds. These results emphasize the need for a more holistic approach to assessing the vulnerability of migratory species and the adaptiveness of changes in migration timing.

1. Introduction

As climate change alters the phenology of the resources on which migratory species and their offspring rely [1], the ability to adjust migratory timing is vital for maintaining individual fitness and stable populations [2]. As a result, large-scale changes in migratory timing—particularly advancements during spring—have been widely documented and the degree to which populations have adjusted their timing remains a critical measure when assessing current and future vulnerabilities to climate change [3,4]. The rate at which migratory timing is shifting, however, remains highly variable at the population and species levels [5]. This suggests the need to not only document shifts in migratory timing but also investigate the factors that influence an individual or population’s ability to utilize different time-shifting strategies—an often-overlooked component of large-scale migration phenology studies [6].

Long-distance migrants, in particular, face an array of climatic and non-climatic factors that are likely to influence their ability to adequately respond to phenological changes. For instance, long-distance migrants are thought to inherently rely on invariant photoperiodic cues to initiate their non-breeding-ground departure, which could make this an unlikely point in the annual cycle to catalyse a shift in migratory timing [7,8]. Recent studies have shown, though, that some long-distance migrants can use information gathered during previous migrations to adjust their departure timing [9], but it remains unclear whether these shifts can lead to advancements in breeding-ground arrival [10]. In general, the success of this strategy depends on the interannual variation in conditions along the migration route experienced by a population and the anomalousness of the conditions within a given year [11].

Utilizing en route decisions may be more viable for long-distance migrants, as they may be able to encounter increasingly reliable cues about breeding-ground conditions along their migration route [12,13]. Yet, decisions made en route, such as spending less time at stopover sites or skipping stopover sites altogether, may reduce the time needed for refuelling—potentially affecting males and females differently because of divergent, sex-specific goals during migration. For instance, males are likely under strong time selection to secure breeding territories [14] and avoid the fitness consequences of divorce [15,16], whereas females have the added pressure of accruing the resources needed for reproduction [17]. Thus, despite leading to earlier breeding-ground arrival, a reduction in refuelling time can result in reduced female energetic condition upon arrival, delaying reproduction and failing to alleviate climate-induced phenological mismatches [18]. The potential for females to face such tradeoffs between arrival timing and condition emphasizes that sex-specific goals can affect the utilization and effectiveness of time-shifting strategies for each sex, as well as constrain an entire population’s response to phenological shifts [19]. Ultimately, an increased understanding of sex-specific differences in timing and time-shifting strategies could further elucidate the potential for climate change-induced reversible state effects to affect reproduction [20], a critical gap in our understanding that is inhibiting comprehensive vulnerability assessments of migratory species.

Hudsonian godwits (Limosa haemastica; hereafter ‘godwits’) are extreme long-distance migrants that spend the non-breeding season in southern South America and breed in arctic and sub-arctic North America [21]. Despite their predicted vulnerability to climatic change, the godwit population breeding in Beluga (Alaska) appears to be using information from previous migrations to effectively track climate-driven shifts in phenology [11]. As a result, on average, Beluga-breeding godwits advanced both their spring migratory and reproductive timing by 9 days from 1974 to 2010 [11,22]. Nonetheless, in anomalously early and warm years, godwits experience phenological mismatches that significantly reduce their reproductive success [23]. Godwits also exhibit high levels of mate fidelity and, in some cases, rapid clutch initiation following migration [21], suggesting that sex-specific goals may affect the strategies each sex can use to adjust their migratory timing. Yet, it remains unknown whether godwits have been continuing to effectively track more recent, increasingly variable, climatic trends, and how individuals might be achieving these shifts.

Here, we used a long-term geolocation tracking dataset to assess ongoing changes in godwit migratory timing by investigating (i) non-breeding-ground departure timing, (ii) breeding-ground arrival timing and (iii) migration duration, while investigating potential sex-specific differences at each stage. With parameters from a historical model [11], we then predicted the arrival date of the first godwit(s) to the breeding grounds using eBird and environmental data to compare with first arrival dates from our tracking data and determine if godwits have continued to track recent shifts in the climate regime. We predicted that godwit breeding-ground arrival timing would continue to trend earlier as we had found previously. Additionally, because we expected males to be under strong time selection, we predicted that males would utilize en route time-minimizing strategies, reducing migration duration over time and driving population-level trends in breeding-ground arrival timing. However, while we still expected females to shift their spring migratory timing earlier to some degree, we predicted that they would exhibit little reduction, if any, in migration duration because of the added pressure to accrue the resources needed for egg production en route. Our study illuminates the time-shifting strategies used by individuals to alter migratory timing and investigates if demographic factors, such as sex, can influence the utilization and effectiveness of such strategies. Ultimately, our work adds to the growing body of knowledge regarding the ability of migratory species to respond to a changing climate.

2. Methods

(a). Godwit capture and tracking

We monitored godwits in Beluga (61°21′ N, 151°03′ W) during the breeding season (May–July) from 2009−2023 and tracked the migrations of these breeding birds using light-level geolocators during two separate periods: 2009−2012 (British Antarctic Survey, 1.4 g Mk14 geolocators in 2009 and 1.1 g Mk10 geolocators in 2010−2011) [24] and 2020−2023 (Migrate Technology, 1.1 g Intigeo geolocators). Tracked individuals were captured on the nest using mist nets during incubation and their sex was determined by plumage and morphology [21]. Each individual was fit with a unique USGS metal band on the right tibia, in addition to a plastic cohort colour band and an individually specific green leg flag with an alphanumeric code to which a geolocator was attached on the left tibia. In subsequent breeding seasons, geolocators were deployed on previously tagged individuals that were recaptured and had their old geolocator replaced, as well as previously unmarked individuals. In total, we retrieved 81 geolocators, 52 from 2009 to 2012 and 29 from 2020 to 2023; 19 individuals were tracked in multiple years (hereafter, ‘multi-year individuals’), meaning we tracked 20 females and 30 males for at least one migration.

(b). Data filtering

Raw light data for each individual were extracted using BASTrack (2009−2012) or IntigeoIF (2020−2023), then analysed in the R Programing Environment (v. 4.2.2) [25] using the packages ‘GeoLight’ (v. 2.0.1) [26] and ‘FLightR’ (v. 0.5.3) [27] following Lisovski et al. [28]. All filter settings were kept the same between the two types of tags except for those used with the preprocessLight function within GeoLight, where lmax (i.e. maximum light level in lumens) was set to 64 for geolocators from 2009 to 2012 and 75 for those from 2020 to 2023 to account for the more limited light spectrum measured by the earlier models. Twilights (i.e. sunrises and sunsets) were annotated and outliers that were incorrectly classified (e.g. periods where the geolocator was shaded because of incubation) were identified by visual inspection of the raw light data and either adjusted manually or deleted. Filtered twilights were then calibrated in ‘FLightR’ using the known deployment location coordinates, with individually distinct calibration periods chosen based on when an individual’s geolocator was deployed or through an assessment of the raw light data. We extracted migration schedules using a maximum flight distance between two consecutive twilight periods of 1900 km and a probability cutoff of 0.2. Mean latitudes and longitudes were used to designate stopover locations. The estimated non-breeding-ground departure and breeding-ground arrival dates were selected from the median quartile and transformed into Julian dates for analysis. In some cases, geolocators failed to capture data during the migration window, raw light data exhibited timing drift, or location estimations produced in ‘FLightR’ were poor (e.g. location estimates falling outside of the species’ known range) despite using the above parameters. This resulted in an exclusion of 16 geolocators from subsequent analyses. Of the 65 remaining geolocators, five had data from two migrations, giving us a total of 70 tracks from round-trip migrations.

Accurately identifying arrival and departure dates with geolocators is notoriously difficult [27]. We therefore also used the wet/dry data collected by the geolocators to corroborate our estimates of godwit departure and arrival timing throughout their migrations. British Antarctic Survey geolocators recorded conductivity measurements every 10 min, with a scale ranging from 0 to 200, where measurements of 0 represent complete dryness. Migrate Technology Ltd. geolocators recorded how many times the logger was wet every 30 s over each 4 h period (measures of 0−480), as well as the maximum conductivity value recorded during that 4 h period (scale of 0−127). For species that take multi-day non-stop migratory flights like godwits, distinct periods of dryness (i.e. conductivity or wetness measures of 0) can be used to accurately gauge when an individual departed from one location and arrived at another [29]. The accuracy of this method is further strengthened by godwit foraging behaviour, as they feed primarily on coastlines or in wetlands, ensuring submersion of the geolocator when an individual has ceased flying.

When timing estimates varied between ‘FLightR’ and our wet/dry data, we utilized the first day of a recorded dry block as the ‘true’ departure date and the first day of a recorded wet block following a dry block as the ‘true’ arrival date. Nonetheless, timing estimates never varied by more than 3 days between our methods. For five tracks used in the final analysis, wet/dry data were not obtained—in most cases, this was a result of tags that were retrieved with 2 years of movement data and for which wet/dry data were only captured during the first year the geolocator was deployed. In these instances, ‘FLightR’ location and timing estimates were used without wet/dry corroboration. To account for the potential reduction in accuracy in departure timing estimation for these individuals, we chose the latest possible departure date identified by ‘FlightR’ as the ‘true’ departure and the earliest possible arrival date as the ‘true’ arrival to the breeding grounds.

(c). Analysis

All models were run in the R Programming Environment using the package ‘lme4’ [30], and their statistical significance was evaluated using the package ‘lmertest’ [31].

(i). Population- and individual-level analysis

First, we investigated shifts in breeding-ground arrival and non-breeding ground departure within the two distinct tracking periods (2009−2012, tracks = 40; 2020−2023, tracks = 30). To do this, we used linear mixed-effects models with year as the predictor variable and individual as a random effect to account for individuals that were tracked for more than 1 year. We then combined both tracking periods and used the same model to explore if there were linear, population-level shifts in breeding-ground arrival, non-breeding ground departure and migration duration across all years. In this way, we could test whether migratory timing had changed both within and across our two periods of tracking despite the gap in time between them. To assess sex-specific differences at each stage, we re-ran each model with an interaction term between sex and year included among the predictor variables.

Next, we wanted to examine whether individuals tracked for multiple years exhibited trends that were reflective of population-level patterns in breeding-ground arrival and non-breeding ground departure timing. Following Conklin et al. [10], we used subject-centring [32] to determine if there were differences in within- and between-individual trends. To do this, we used linear models with two variables, one for within-individual variation, to ask if individuals with repeated measures were significantly shifting their migratory timing during their lifetimes, and the other for between-individual variation, to ask if, over time, cohorts of individuals were shifting their migratory timing. These variables were then included as additive terms and arrival/departure date was used as the dependent variable. For these models, we only included tracks from multi-year individuals (n = 19, tracks = 46).

(ii). Historic model of godwit arrival date

Previously, we developed a statistical model that related weather variables from across the godwit migratory corridor with the arrival date of the first godwit each year to nearby Anchorage (Alaska), as reported by community scientists from 1974 to 2010 [11]. That model identified six factors as influencing godwit migratory timing: (i) the previous 5 year mean May temperature from the breeding grounds, (ii) the previous year’s first godwit arrival date at the breeding grounds, (iii) the previous year’s mean May temperature from the breeding grounds, (iv) the mean April wind speed from a northbound stopover site (Houston, TX, USA), (v) the amount of precipitation during the non-breeding season (Puerto Montt, Chile), and (vi) the March value of the Southern Oscillation Index. The model explained 74% of the variation in arrival date. Here, we gathered data from the period since that paper was published (2011−2023) from the same data sources and used the same methodology in order to predict the timing of the first godwit arrival based on these environmental factors. This included godwit first arrival dates to Anchorage derived from eBird (www.ebird.org), weather data from the US National Climate Data Center (www.ncei.noaa.gov) and the online global weather station repository TuTiempo (www.tutiempo.net), and Southern Oscillation Index Values from the Australian Bureau of Meteorology (www.bom.gov.au/climate/enso/soi/).

Using the predict function in ‘lme4’ with the previous historical model results and the weather and eBird data from 2011 to 2023, we predicted the first arrival dates for godwits over the same period for which we had geolocator data. We then ran a simple linear regression, with the predicted first arrival date as the dependent variable and year as the predictor variable, to derive an expected rate of change in godwit arrival dates since 2010.

3. Results

(a). Population-level shifts in migratory timing

We recovered full migration tracks from an average of 12 ± 2 individuals per year (n = 6 years), representing 36 ± 6% of the entire breeding population in Beluga. There was no significant trend in either non-breeding ground departure (β = −0.03, s.e. = 0.49, p = 0.96) or breeding-ground arrival timing (β = −0.36, s.e. = 0.57, p = 0.53) over time for individuals tracked from 2010 to 2012. For individuals tracked within the second tracking period (2021−2023), there was no significant trend in non-breeding ground departure timing (β = −0.95, s.e. = 0.51, p = 0.08), but breeding-ground arrival advanced by approximately 2.5 days over the 3 year period (β = −0.85, s.e. = 0.30, p = 0.01).

When the two tracking periods were combined, we found significant, directional, population-level shifts in both non-breeding ground departure and breeding-ground arrival timing across the 13 year period (n = 70). Godwits shifted both their non-breeding ground departure (β = 0.36, s.e. = 0.07, p < 0.01; figure 1C) and breeding-ground arrival timing (β = 0.33, s.e. = 0.05; p < 0.01; figure 1A) later by approximately 6 days after 2010. Despite changes in departure and arrival timing, migration duration stayed the same over this period (β = −0.02, s.e. = 0.06, p = 0.72; figure 1B), with individuals completing their northward migration in 24 ± 3 days.

Figure 1.

Changes in migratory timing for Hudsonian godwits (Limosa haemastica) breeding in Beluga (Alaska, USA) over a 13-year period (2010−2023). (A) Change in godwit population-level arrival timing to the breeding grounds in Beluga. (B) Change in godwit population-level migration duration. (C) Change in godwit population-level departure timing from the non-breeding grounds in southern Chile. Points are scaled by colour to represent the number of individuals with the same migration arrival, duration and departure.

(b). Sex-specific shifts in migratory timing

We found no significant interaction between sex and year in population-level non-breeding-ground departure (interaction term: β = −1.69, s.e. = 1.52, p = 0.27), breeding-ground arrival (interaction term: β = −2.23, s.e. = 1.22, p = 0.07) or migration duration (interaction term: β = −0.40, s.e. = 1.47, p = 0.79).

(c). Shifts at the individual level

For individuals that were tracked for multiple years, we found no significant difference in within-individual trends for either non-breeding-ground departure (β = 0.69, s.e. = 1.21, p = 0.57; figure 2A,E) or breeding-ground arrival timing (β = 2.15, s.e. = 1.14, p = 0.07; figure 2B,F). However, between-individual trends were significantly different—both non-breeding-ground departure (β = 1.63, s.e. = 0.36, p < 0.01) and breeding-ground arrival (β = 1.08, s.e. = 0.34, p < 0.01) shifted later between individuals. Only one individual, a female godwit (hereafter, ‘1PJ’), was tracked during both tracking periods. That individual shifted both its non-breeding ground departure and breeding-ground arrival timing later over time (figure 2C,D), aligning with population-level trends (figure 1A,C).

Figure 2.

Variation in individual-level trends in migratory timing of Hudsonian godwits (Limosa haemastica) breeding in Beluga (Alaska, USA). (A) Non-breeding-ground departure and (D) breeding-ground arrival timing from individuals tracked for either 2 or 3 years during the first tracking period, 2010−2012. (B) Non-breeding-ground departure and (E) breeding-ground arrival timing for individual ‘1PJ’, which was tracked during both tracking periods. (C) Non-breeding-ground departure and (F) breeding-ground arrival timing from individuals tracked for either 2 or 3 years during the second tracking period, 2021−2023. Colours represent unique individuals.

(d). Predictive model

The historical model predicted no consistent change in population-level breeding-ground arrival dates between 2011 and 2023 (β = −0.28, s.e. = 0.15, p = 0.08; figure 3). However, actual first arrival dates derived from our geolocator data got significantly later during this period (β = 0.44, s.e. = 0.11, p = 0.02; figure 3).

Figure 3.

Trends in first arrival dates to Anchorage (Alaska, USA) of Hudsonian godwits (Limosa haemastica) predicted with a historical model (blue squares and line) and actual first arrival dates to Anchorage derived from community scientist observations (1974−2010; orange triangles and line) and Beluga (Alaska, USA) using geolocators (2010−2023; red dots and line).

4. Discussion

Contrary to the findings of many large-scale studies of long-distance migratory birds [3–5], we found that a population of Hudsonian godwits, an extreme long-distance migratory shorebird, can (i) rapidly alter its migratory timing, (ii) use non-breeding ground departure timing to drive changes in breeding-ground arrival timing, and (iii) dramatically reverse the direction in which its migratory timing has been changing within only a few years. Our results thus point to the need to better link individual- and species-level studies to understand both the patterns and drivers of migratory timing in the face of global change.

(a). Rapid change in population-level migratory timing

Most large-scale studies that have assessed adjustments in the migratory timing of long-distance migratory birds have found relatively slow rates of change over time—usually only 1−3 days per decade [4,33,34]. Previously, godwits had shifted their migratory timing by 9 days on average over a period of nearly 40 years, including two separate decades during which they had shifted their arrival by approximately 1 day per year [11]. Here, too, we found that godwits have shifted their migratory timing by approximately 6 days over a 13 year period. Previous large-scale studies have also frequently noted that short-distance migrants outpace long-distance migrants in the rate at which they have shifted their migratory timing [33]. Yet, our results demonstrate that a species with one of the most extreme long-distance migrations of any bird species can alter its migratory timing at rates just as fast, if not faster, than short-distance migrants.

These rapid shifts in godwit migratory timing suggest that individuals are capable of flexibly responding to environmental changes within their lifetime, something which has rarely, and only recently, been documented in other long-distance migrants [10]. That said, in this study, we can only conclude that between-individual (i.e. intergenerational) and not within-individual (i.e. individual phenotypic flexibility) trends are changing and contributing to population-level shifts in migratory timing. Many studies, like ours, do not have enough repeated individual measures over a sufficiently long period to differentiate the mechanistic role of between- and within-individual trends [35,36]. As a result, deciphering whether developmental plasticity, phenotypic flexibility or evolutionary processes—or all of them simultaneously—are driving trends in migration timing remains a need in studies of responses to climate change [35]. For example, developmental plasticity has been found to account for only a fraction of responses to changing climatic trends, suggesting microevolution is taking place at the same time in at least some species [37]. Nevertheless, 1PJ, the one individual that we successfully tracked across our entire study period, attests to the possibility that godwits can exhibit dramatic phenotypic flexibility in response to changes in environmental conditions during their lifetimes (see also [10]). Therefore, with longer-term monitoring of repeatedly tracked individuals, we might expect to observe that both within- and between-individual trends are contributing to timing shifts at the population level.

(b). Migratory timing shift driven by changes in departure timing

Historically, long-distance migrants have been predicted to have a limited capability of responding to climatic changes occurring on their breeding grounds because of the inherent constraints associated with their long-distance trips. For instance, the absence of cues predictive of breeding-ground conditions on the non-breeding grounds is thought to force individuals to predominantly rely on photoperiodic cues when initiating migration, making shifts in non-breeding-ground departure seem unlikely, if not, impossible [7,8]. Furthermore, the repeatability of non-breeding-ground departure timing in a plethora of avian species is relatively high [38]. For these reasons, we predicted that godwits would achieve shifts in migratory timing primarily by utilizing en route strategies and that the two sexes would exploit strategies differently because of sex-specific goals.

Despite this, we found that godwits did not utilize en route strategies, as we observed no change in migration duration over time and no difference in migratory timing between the sexes. These findings support our prediction that females are limited in their ability to effectively utilize en route strategies because of the condition-dependent pressures of egg production [18]. However, they contradict our prediction that males would utilize en route strategies, risking optimal body condition on arrival to achieve optimal arrival timing. This suggests that selection pressures may not be sex-specific but, rather, acting uniformly across the population. Nevertheless, sub-arctic and arctic breeding migrants are tasked with migrating vast distances in a matter of weeks and, in general, all individuals depart and arrive within a matter of days [24]. Thus, if differences are present between the sexes, they may be small and difficult to distinguish with (relatively) small sample sizes.

Instead, contrary to past predictions [7,8] we observed a proportional shift, whereby non-breeding-ground departure changed in directional synchrony with arrival on the breeding grounds. To our knowledge, shifts of this magnitude have not been previously documented. For instance, bar-tailed godwits (L. lapponica) have been documented flexibly advancing their non-breeding ground departure timing, but nonetheless failing to exhibit earlier arrival to their breeding grounds [10].

One possible explanation for this pattern in Hudsonian godwits is that individuals could be using knowledge of the previous year’s breeding-ground conditions to inform their departure timing in subsequent years. While this concept is not novel [9,11], the predictive power of knowledge from previous years is likely low in highly stochastic climate regimes and only beneficial when breeding-ground conditions are either unchanging or changing slowly, and linearly, over time. For instance, when godwits experienced mostly linear changes in climate in the past, information from previous breeding-ground conditions strongly predicted breeding-ground arrival timing [11]. The Arctic and sub-Arctic, however, are currently experiencing some of the most rapidly changing climatic regimes [39,40]—causing godwits to experience increasingly variable conditions upon arrival in recent years [23]. Thus, knowledge gathered about previous years’ breeding-ground conditions may be increasingly unreliable for godwits, limiting their ability to predict conditions within a given year and inform their subsequent migratory timing accordingly.

(c). Why are godwits changing their migration timing?

Most studies have concluded that the shifts in migratory timing in recent decades have been in response to changes in temperature ([3] but see [41]). Indeed, the historical model we employed here indicated a strong relationship between godwit first arrival dates and climatic conditions—including temperature—occurring along their migratory corridor and accurately predicted previous rapid changes in migration timing (e.g. 1997−2004) [11]. Since 2010, however, the discrepancy between observed and predicted godwit arrival dates using this model has increased and been the highest of any 13-year period since 1974, averaging 3−4 days. What is now driving godwit arrival timing in Beluga?

Given our result that godwits are synchronously shifting their migratory departure and arrival timing, it follows that whatever has changed since 2010 has likely occurred on their non-breeding grounds. Certainly, climatic factors not included in the historical model could now be influencing godwit migration timing. Wind, for instance, can strongly influence migration timing [42] and both advancements and delays in the timing of non-breeding-ground departure in long-distance migrants have been tied to the availability of beneficial tailwinds [43]. Godwits that depart from southern Chile are known to utilize prevailing tailwinds during their trans-Pacific flights [44]. Thus, capitalizing on beneficial winds, or avoiding disadvantageous winds, is likely an important determinant of godwit migratory timing. It is possible that these winds are changing, with favourable wind conditions becoming increasingly unpredictable due to changes in the Pacific subtropical anticyclone [45,46]. In some years, individual godwits may therefore be energetically ready to depart, but remain grounded, delaying their departure until wind conditions are favourable; godwits could also be capable of making use of beneficial tailwinds even when they might otherwise spend more time fuelling, allowing them to depart earlier than otherwise expected in some years [43].

It is also possible that non-climatic factors could now be influencing godwit migratory timing [47,48]. In southern Chile, the burgeoning aquaculture industry has changed many coastal areas, making coastal habitats increasingly dynamic. Godwits and other shorebirds exclusively forage in the intertidal mudflats that are also used for the cultivation and harvest of fish, shellfish and algae [49–51]. The effects of such practices on shorebird communities are complex. On the one hand, they may be mutually beneficial—areas with algae cultivation promote higher levels of benthic macroinvertebrates and are preferred by tactile-foraging shorebird species like godwits [52]. On the other hand, the human disturbance associated with the maintenance and harvesting of algae negatively influences godwit foraging efficiency and leads to decreases in body condition and survival in individuals using highly disturbed sites [53–55].

For godwits, the energetic expenditures and time associated with avoiding heavily disturbed areas could be especially costly in the days leading up to departure, which is immediately followed by a 7 day non-stop trans-Pacific flight [24,44]. To sustain a flight of this magnitude, individuals must acquire substantial nutritional reserves in the weeks prior to departure [56]. Godwits have thus been documented foraging throughout the day and night during this period [57], and disturbances may limit any surplus preparation time that godwits might possess [53]. As a consequence, we might expect to see significant shifts, or in this case, delays, in migratory timing at the population level.

Irrespective of the cause, was this shift in migration timing adaptive? It may depend on your perspective. The consequences for godwit reproductive success, at least, have not been positive. From 1974 to 2010, while godwit first arrival dates shifted 9 days earlier on average, the date of peak insect abundance in Beluga shifted 11 days earlier, meaning that godwit arrival dates closely tracked this rate of change [22]. Since 2012, however, godwits have become increasingly mismatched with local insect abundance—a key prey source for growing chicks [23]. The period from 2021 to 2023, in particular, witnessed some of the lowest chick survival rates since our study began (Huang, Stager, Puleo, and Senner, 2025, unpublished data). At the same time, other similarly sized shorebird species that also breed in the Upper Cook Inlet region but do not spend the non-breeding season in southern Chile, such as greater yellowlegs (Tringa melanoleuca), have advanced their first arrival dates by approximately 6 days since 2010 and now arrive two weeks ahead of godwits [58]. From the adult perspective, if conditions on the non-breeding grounds have deteriorated, a delay might indicate that adults are prioritizing their own survival. Godwits therefore may now be facing a tradeoff between arriving earlier to the breeding grounds and maintaining high adult survival, which is key to the population stability of a species with such a slow life-history strategy [59].

(d). Into the future

We have documented an unexpected and rapid shift in migratory timing for a breeding population of an extreme long-distance migratory bird. While our evidence suggests that this shift may be driven by phenotypic flexibility, we cannot make firm conclusions with our current sample size of repeated individual tracks. Nonetheless, differentiating between developmental plasticity, phenotypic flexibility and evolutionary processes remains essential to assessing the rapidity with which migrants can respond to changes in their environments. Therefore, studies that employ a bottom-up approach—focusing on repeated individual-level measures, ideally from across replicated populations—should be a high priority and coupled with larger-scale, multi-species studies (e.g. [13]).

Our data also suggest that conditions on the non-breeding grounds are changing and influencing godwits breeding in Beluga in unpredictable ways. We propose two potential factors that could be responsible for this pattern—increased anthropogenic disturbances and shifting climatic conditions. Regardless, we emphasize that both factors have the potential to constrain entire populations from responding to climatic changes occurring elsewhere. Thus, both factors warrant future investigation. Further, delays in departure timing make stopover site quality another potential limiting factor: degraded sites can prolong refuelling duration [10], which could further delay breeding-ground arrival and potentially affect subsequent reproductive success. Future studies should therefore focus on assessing and monitoring sites not just at the beginning and end of migratory journeys but also in the middle.

Currently, much of our knowledge regarding the vulnerability of long-distance migrants to climate change is derived from studies that only explore the influence of climatic factors on migratory timing and not other anthropogenic and environmental factors (but see [60]). Looking to the future, we see the need for a more holistic approach—one that quantifies a diversity of variables from across the entire annual cycle—when assessing shifts in migratory timing and investigating their potential adaptiveness. For long-distance migrants, this holistic approach requires multi-hemispheric efforts and collaborations that, while ambitious, are imperative for addressing conservation concerns in a rapidly changing world.

Ethics

Animal care and use approval was provided by the University of South Carolina (no. 2449-101417-042219) and University of Massachusetts Amherst (no. 4136). Scientific permits were provided by the Alaska Department of Fish & Game (no. 19-133, 20-024, 21-025, 22-033 and 23-113) and the US Geological Survey (no. 24191).

Data accessibility

All data are publicly available and can be accessed at Dryad [61].

Declaration of AI use

We have not used AI-assisted technologies in creating this article.

Authors’ contributions

L.P.: conceptualization, data curation, formal analysis, investigation, methodology, writing—original draft, writing—review and editing; F.H.: investigation, writing—review and editing; M.S.: investigation, writing—review and editing; N.R.S.: conceptualization, formal analysis, funding acquisition, investigation, methodology, project administration, supervision, writing—review and editing.

All authors gave final approval for publication and agreed to be held accountable for the work performed therein.

Conflict of interest declaration

We declare we have no competing interests.

Funding

Funding was provided by NSF-PCE 2318983 and 1110444, NSF-DGE 1144153, USFWS-NMBCA 7320, the Association of Field Ornithologists, Wilson Ornithological Society, American Ornithological Society, Arctic Audubon Society, Cornell Lab of Ornithology, Faucett Family Foundation, David and Lucille Packard Foundation, US Fish and Wildlife Service Migratory Bird Management Office (via J. A. Johnson) and University of South Carolina.