Skip to main content
NCBI home page
Scientific Reports logoLink to Scientific Reports
. 2026 Feb 2;16:6890. doi: 10.1038/s41598-026-38381-2

Scenario-dependent northward shifts in wintering centroids of Anatidae in South Korea

Han-I Choi 1,2, SeonYong Lee 2, Hyung-Kyu Nam 2,✉
PMCID: PMC12916801  PMID: 41629466

Abstract

Climate and land-use change are reshaping the spatial ecology of migratory waterbirds, yet few studies in East Asia have quantitatively integrated ensemble species distribution models, land-cover dynamics, and multi-scenario climate projections to assess future shifts and fragmentation of wintering habitats. We quantified climate-driven changes in the wintering distributions of six rice-paddy–dependent Anatidae species across South Korea using an ensemble species distribution modelling framework integrating nine algorithms. Species occurrence data from GBIF and field surveys (2022–2024) were combined with land-cover, topographic, and bioclimatic variables, with future projections generated under four Shared Socioeconomic Pathways (SSP126, SSP245, SSP370, SSP585) and four time periods (2021–2040, 2041–2060, 2061–2080, 2081–2100). Across all scenarios, wintering distributions exhibited pronounced northward shifts, reduced spatial overlap with historical hotspots, and increasing habitat fragmentation. The loss of southern wintering grounds—historically critical as climatic refugia during extreme cold events—combined with the rapid expansion of greenhouse-based agriculture threatens the stability of emerging northern habitats. While northward shifts may facilitate adaptation to gradual warming, they also heighten vulnerability to short-term climatic extremes, forcing energetically costly movements to suboptimal sites. These findings underscore the urgency of climate-informed conservation planning.

Supplementary Information

The online version contains supplementary material available at 10.1038/s41598-026-38381-2.

Keywords: Hotspots, East-Australasian flyway, Fragmentation, SSP scenarios, Anatidae, Korea

Subject terms: Climate sciences, Ecology, Ecology, Environmental sciences

Introduction

Climate change is profoundly reshaping species distributions worldwide, with migratory waterbirds serving as sensitive sentinels of environmental shifts across the temperate zone1,2. Rising winter temperatures are increasingly influencing migratory phenology, energetic balance, and habitat use, often driving rapid wintering range shifts3. For instance, north–northeast shifts in the wintering distributions of several common waterbird species along the North-West European flyway over the past three decades correlate strongly with regional warming trends (mean increase of 3.8 °C in early-winter temperature)3.

Despite such insights from Europe and North America, studies from East Asia have largely focused on individual species or localized regions, with more limited integration of land-use change and future climate scenarios, even though it is characterized by rapid land-use change alongside accelerated climate warming4,5. In South Korea, inland wetlands and rice paddies serve as primary wintering habitats for many Anatidae species6–8. In East Asia, rice paddies function as surrogate wetlands during winter, providing abundant waste grain and shallow flooded foraging habitats that support large numbers of herbivorous and granivorous Anatidae species when natural wetlands are scarce9,10. However, these landscapes are increasingly threatened by agricultural intensification. Major drivers include habitat loss resulting from land-use conversion of rice paddies, such as the expansion of greenhouse cultivation11. The combined effects of long-term climatic warming and anthropogenic land-use change pose substantial yet incompletely quantified risks to wintering waterbird distributions in the region.

Climate change is increasingly reshaping species distributions in ways that challenge traditional single-species conservation frameworks12,13. In human-dominated agricultural landscapes such as rice-paddy systems, conservation actions are typically implemented at broad spatial scales and must accommodate multiple species with overlapping habitat requirements14,15. Under such conditions, identifying areas of shared climatic suitability across species provides a practical basis for multispecies conservation prioritization16. Overlap-based approaches have therefore been widely used to delineate zones of high conservation relevance, particularly where coordinated management is required to maximize biodiversity benefits under limited resources17,18. In this study, we adopt a multispecies hotspots framework to identify areas where climatically suitable wintering conditions converge for multiple rice-paddy–dependent Anatidae species, thereby highlighting shared landscapes that may be disproportionately important for sustaining wintering waterbird assemblages under ongoing climate change.

Moreover, existing studies seldom integrate ensemble species distribution models (SDMs), land-cover dynamics, and comparative landscape analyses across multiple emission scenarios to quantify future shifts in wintering (non-breeding) distributions of migratory waterfowl. Additionally, the consequences of such shifts for habitat connectivity and conservation planning under ongoing climate change are largely unquantified1,19. To address these gaps, we modeled changes in wintering distributions of six rice-paddy–dependent Anatidae species across South Korea using an ensemble SDM framework. We aimed to (i) quantify northward shifts, range contractions, and fragmentation under four Shared Socioeconomic Pathways (SSP126, SSP245, SSP370, SSP585), (ii) evaluate changes in the spatial configuration of key wintering areas, and (iii) identify conservation priorities to maintain habitat connectivity and ecological resilience under accelerating climate change.

Methods

Study area and focal species

This study focused on the entire Republic of Korea, encompassing all major inland wetlands and agricultural landscapes that provide critical wintering habitats for migratory waterbirds. We targeted six rice-paddy–dependent Anatidae species such as Bean Goose Anser serrirostris, White-fronted Goose A. albifrons, Mallard Anas platyrhynchos, Eastern Spot-billed Duck A. zonorhyncha, Northern Pintail A. acuta, and Eurasian Teal A. crecca. These species represent the dominant guild of herbivorous and granivorous waterbirds wintering in the region based on long-term national winter waterbird census data in South Korea8. Here, “rice‐paddy–dependent” refers to wintering Anatidae species whose habitat availability and distribution in South Korea are strongly linked to rice‐paddy landscapes that provide key foraging resources during the non‐breeding season6,7,9,10.

Species occurrence data and spatial filtering

Species occurrence records were compiled from two complementary sources: (i) Global Biodiversity Information Facility (GBIF) records covering the major wintering seasons (November-February) during 2022–2024 and (ii) systematic field surveys conducted by the authors over the same period. The field surveys followed standardized winter census procedures and were conducted during the same wintering window (November–February) across major inland wetlands, riverine floodplains, and rice-paddy landscapes where Anatidae are known to concentrate. To ensure compatibility between sources, both GBIF and field records were filtered to the same season and subjected to identical quality-control steps. We did not explicitly weight records by survey effort; instead, we addressed uneven sampling and spatial clustering by applying a unified spatial thinning procedure (1-km) after merging all occurrence records, thereby minimizing differences in spatial bias between datasets. All occurrence records underwent rigorous quality control to remove duplicates, erroneous coordinates, and observations outside plausible habitat boundaries (e.g., open ocean, urban centers). Erroneous coordinates were defined as records located in clearly implausible areas for wintering Anatidae, including open ocean, dense urban cores, or locations far outside known inland wetland and agricultural habitats. To reduce the influence of potential “flying-by” records associated with citizen science data, we restricted occurrences to the core wintering season (November–February), excluded observations outside plausible wintering habitats, and applied spatial thinning. Such transient records were therefore considered rare at the national scale and unlikely to affect broad-scale SDM predictions, which focus on habitat associations rather than individual behavioral states. To reduce spatial sampling bias and minimize autocorrelation among records, we applied a spatial thinning procedure, retaining only one occurrence within each 1-km radius using the spThin R package. This ensured spatial independence among presence points while maintaining sufficient sample sizes for robust model calibration. The final dataset for each species was used to develop and evaluate species distribution models (SDMs).

Environmental and bioclimatic predictors

We assembled a comprehensive set of environmental and climatic variables hypothesized to influence the wintering distributions of six Anatidae species. Twenty environmental predictors—including land-cover variables (e.g., rice paddy extent, inland water and wetland areas, coastal wetlands, bare land, forest cover, and anthropogenic land use), distance to water bodies, and topographic elevation—were obtained from the Environmental Geographic Information Service (EGIS; https://egis.me.go.kr/, accessed on 21 July 2025). Land cover data were derived from the national land‐cover map (year 2024) and originally provided at a 5‐m spatial resolution; all layers were resampled to a 1‐km grid to ensure consistency with climate predictors and species occurrence data. A complete list and description of all environmental predictors are provided in Table S1.

Land-use variables used in the SDMs represent baseline agricultural and land-cover conditions derived from the national land-cover map for the most recent available period (year 2024) and were treated as temporally static predictors. Agricultural land-use change was therefore not incorporated as a time-varying predictor in the models but was assessed separately through post hoc spatial overlap analyses comparing projected wintering hotspots with cumulative rice paddy loss relative to the 2007 baseline (as of 2024).

Bioclimatic variables (bio1–bio19) describing annual trends, seasonality, and extremes in temperature and precipitation were obtained from WorldClim version 2 (https://www.worldclim.org/, accessed on 28 July 2025) for the baseline period (1970–2000) and for each future scenario–period combination (see below). To reduce multicollinearity among predictors, we calculated pairwise Spearman correlation coefficients and Variance Inflation Factors (VIFs), sequentially removing variables with |ρ| > 0.7 or VIF > 5 while retaining ecologically meaningful predictors (Table S2). The final set of predictors was standardized across all species, scenarios, and time periods to ensure methodological consistency and cross-species comparability, while species-specific responses were captured through independent model calibration.

Climate scenarios and time periods

Future distributions were projected under four Shared Socioeconomic Pathways (SSPs) representing contrasting greenhouse gas emission trajectories: SSP126, SSP245, SSP370, and SSP585. For each scenario, we generated projections for four future time periods: 2021–2040, 2041–2060, 2061–2080, and 2081–2100. The baseline period (1970–2000) was used as a standardized historical climatological reference. All climatic layers were obtained from WorldClim version 2 at 1-km resolution and harmonized with environmental predictors prior to modelling.

Ensemble species distribution modelling

To account for algorithmic uncertainty and improve predictive accuracy, we adopted an ensemble SDM framework incorporating nine widely used modelling algorithms: Classification Tree Analysis (CTA), Multivariate Adaptive Regression Splines (MARS), Generalized Linear Models (GLM), Generalized Additive Models (GAM), Gradient Boosting Machines (GBM), Random Forest (RF), Maximum Entropy (MAXENT), Artificial Neural Networks (ANN), and Support Vector Machines (SVM). Models were calibrated using fivefold spatial block cross-validation to minimize spatial autocorrelation between training and testing data. In each cross-validation run, approximately 80% of the occurrence data were used for model training and the remaining 20% were withheld for model testing, with spatial blocks ensuring geographic independence between subsets. Model evaluation was therefore based on repeated internal validation using withheld test folds, and no separate external validation dataset was applied. Predictive performance was quantified using the Area Under the Receiver Operating Characteristic Curve (AUC). AUC values were used to assess within-model discriminatory performance for each species and scenario, and were not intended for direct comparison or ranking across models. Only models with AUC ≥ 0.7 were retained for ensemble construction. Ensemble predictions were generated as weighted averages across all retained algorithms, with weights proportional to individual model AUC scores, producing a single consensus prediction for each species, scenario, and time period. All spatial analyses were conducted in the WGS84 coordinate system to ensure consistency across occurrence data, environmental predictors, species and model outputs.

Hotspots spatial analysis

Continuous ensemble predictions were thresholded to binary presence–absence using the maximum sensitivity plus specificity criterion to balance omission and commission errors. For each scenario–period, binary maps for the six species were summed to produce species-richness layers (0–6 per 1-km cell). Hotspots were defined as cells with richness = 6. This definition represents grid cells where all six focal Anatidae species were predicted to co-occur, and was deliberately applied as a highly conservative criterion to delineate areas of maximum and unambiguous multispecies overlap. Our objective was to identify shared hotspots of high conservation relevance across species, rather than to capture all areas important to individual or range-restricted species. We acknowledge that this approach may underrepresent ecologically important locations supporting fewer species (e.g. richness = 4–5), and therefore interpret hotspots pattern as complementary to species-level distribution results.

Comparative metrics and spatial analysis

All spatial metrics described below were calculated exclusively for hotspots defined as grid cells with species richness = 6. To compare spatiotemporal dynamics across scenarios and periods, we quantified two complementary sets of spatial metrics. First, centroid-based metrics were used to characterize large-scale geographic shifts in hotspots distributions, including centroid longitude and latitude and the magnitude of northward displacement relative to the baseline. Second, patch-based landscape metrics were calculated to assess changes in spatial configuration and fragmentation, including hotspots, number of patches, largest patch area, largest patch index, total edge length, edge density, and Jaccard similarity with baseline hotspots. We computed the following metrics on harmonized 1-km grids using an eight-neighbor (Queen) connectivity rule: centroid longitude and latitude and the northward shift (km) relative to the baseline centroid (km converted as 1° lat = 111.32 km; 1° lon = 111.32 × cos φ at mean latitude φ), hotspots (km2), number of patches, largest patch area (km2), largest patch index (LPI, % of total hotspots), total edge length (km), edge density (km/km2), and Jaccard similarity with the baseline hotspots.

Software.

All analyses were performed in R (version 4.3.3)21 using the following packages: spThin20, biomod222, dismo23, raster24, sf25, terra26, dplyr27, and ggplot228.

Results

Model performance and baseline spatial distribution

Baseline distributions of the six Anatidae species were primarily concentrated in low-altitude regions along major river systems and western and southern coastal plains (Fig. 1a,b). After quality control and spatial thinning, the final sample sizes used for model calibration were n = 185 for Bean Goose, n = 331 for White-fronted Goose, n = 832 for Mallard, n = 813 for Eastern Spot-billed Duck, n = 237 for Northern Pintail, and n = 533 for Eurasian Teal. Model evaluation indicated high predictive accuracy, with mean AUC values exceeding 0.83 across all species and scenarios (Table S3). For example, the White-fronted Goose and Bean Goose models achieved AUCs of 0.896 and 0.890, respectively, demonstrating strong discrimination capacity between suitable and unsuitable habitats. Sensitivity and specificity values consistently exceeded 0.75, indicating robust predictive reliability across model runs.

Fig. 1.

Fig. 1

Open in a new tab

(a) Topographic map of South Korea with major river (Han River, Geum River, Seomjin River, and Nakdong River) systems indicated in blue, illustrating key low-altitude regions. (b) Baseline species richness map (1970–2000) for six Anatidae species, showing high-richness areas concentrated along western and southern coastal plains. All maps were generated by the authors using R software (version 4.3.3; https://www.r-project.org/).

Changes in hotspots and northward shift

Across scenarios and time periods, multispecies hotspots (cells with species richness = 6; i.e., overlap of binary suitable habitat for all six species) showed consistent baseline cores with pronounced scenario- and time-dependent reconfiguration (Fig. 2; Table 1; Table S3). Baseline hotspots were concentrated along the western and southern coastal plains and major river systems—most prominently the Geum River Basin and the Nakdong Estuary. Although localized hotspot patches occurred along low-altitude reaches of the lower Han River, their spatial extent and continuity were more limited, and this region was therefore classified as a secondary concentration under baseline conditions. Under SSP585 (2021–2040), hotspots expanded northward into the west-central plains (including lower Han River areas), and total hotspots reached 9,246.69 km2 (+ 38.7% vs. baseline). In contrast, late-century SSP126 (2081–2100) yielded 6,007.92 km2 (− 9.8%) with a smaller but more consolidated configuration (highest LPI = 71.36%). Northward shifts of hotspots centroids were consistent across scenarios, peaking at 26.31 km under SSP370 (2021–2040), but diminishing in the late century to 2.22 km (SSP126, 2081–2100) and 3.48 km (SSP370, 2081–2100).

Fig. 2.

Fig. 2

Open in a new tab

Projected spatial distributions of species richness for six Anatidae species under four climate change scenarios (SSP126, SSP245, SSP370, SSP585) across four future time periods (2021–2040, 2041–2060, 2061–2080, 2081–2100). Warmer colors indicate higher species richness, with early-century expansions followed by late-century contractions and northward shifts in habitat centroids. All maps were generated by the authors using R software (version 4.3.3; https://www.r-project.org/).

Table 1.

Changes in hotspots, centroid location, patch metrics, edge characteristics, and habitat similarity of regions where all six Anatidae species co-occur under baseline (1970–2000) and four future climate scenarios (SSP126, SSP245, SSP370, SSP585) across three time periods (2021–2040, 2041–2060, 2081–2100). Values represent centroid coordination (longitude and latitude), hotspots (km2), northward shift form baseline (km), number of patches, largest patch size (km2), largest patch index (%), total edge length (km), edge density (km/km2), and Jaccard similarity relative to the baseline.

Scenario Period Centroid longitude Centroid latitude Hotspots (km2) Northward shift (km) No. of patches Largest patch (km2) Largest patch index (%) Edge total (km) Edge density (km/km2) Jaccard similarity
Baseline Baseline 126.94 36.89 6665.42 0.00 312 4255.60 63.85 7649.66 1.15 1.00
ssp126 21–40 127.04 36.77 8683.02 16.03 313 5332.97 61.42 9246.85 1.06 0.65
41–60 127.01 36.71 8357.86 20.93 323 4995.26 59.77 8649.80 1.03 0.63
61–80 126.94 36.91 6007.92 2.22 307 4287.23 71.36 6637.56 1.10 0.73
ssp245 21–40 126.97 36.82 6350.45 8.22 211 4064.02 64.00 6701.21 1.06 0.69
41–60 126.95 36.78 9190.92 12.24 316 5262.75 57.26 9477.13 1.03 0.66
61–80 126.99 36.82 9407.31 8.96 391 5886.44 62.57 9753.83 1.04 0.62
ssp370 21–40 127.05 36.67 7201.35 26.31 368 4264.41 59.22 8477.75 1.18 0.66
41–60 127.02 36.75 8310.04 17.10 292 5455.67 65.65 8513.46 1.02 0.64
61–80 126.97 36.91 7855.58 3.48 260 5107.27 65.01 8002.04 1.02 0.71
ssp585 21–40 127.04 36.70 9246.69 22.90 311 5522.90 59.73 8293.37 0.90 0.63
41–60 126.99 36.90 8104.96 4.59 305 5546.50 68.43 8525.77 1.05 0.68
61–80 126.97 36.85 8348.54 5.18 314 5295.71 63.43 8098.91 0.97 0.68
Open in a new tab

Patch structure and fragmentation

The number of habitat patches varied substantially among scenarios. The maximum number (391 patches) occurred under SSP245 (2081–2100), whereas the minimum (211 patches) was observed under SSP245 (2021–2040), highlighting substantial habitat fragmentation under certain scenarios and periods. The largest patch area reached 5,522.90 km2 under SSP585 (2021–2040) but declined to 4,064.02 km2 under SSP245 (2021–2040). The largest patch index, representing the proportion of total area contributed by the largest patch, peaked at 71.36% under SSP126 (2081–2100). Mid- to late-century SSP245 projections showed pronounced fragmentation across the mid-western coastal plain and the Geum Basin, with patch number up to 391, total edge length 9,753.83 km, and Jaccard similarity 0.62.

Edge metrics and habitat continuity

Complementary edge-based metrics indicated that the total edge length of hotspots increased relative to the baseline across most scenarios, with the maximum value (9,753.83 km) recorded under SSP245 (2081–2100). However, edge density remained relatively constant around 1.0 km/km2, suggesting that area expansion did not necessarily translate into more complex habitat boundaries. Habitat similarity, quantified by the Jaccard index, consistently declined across all scenarios. The lowest similarity (0.62) occurred under SSP245 (2081–2100).

Agricultural land-use change between 2007 and 2024

Analysis of agricultural land cover between 2007 and 2024 revealed substantial losses of rice paddies across southern and western Korea (Fig. 3 and Table S4). District-level comparisons showed that several key wintering regions, particularly in the Geum River Basin, Nakdong Estuary, and western coastal wetlands, experienced losses exceeding 240 ha over the 17‐year period. Moderate losses (121–180 ha) occurred across southern inland areas and parts of the west coast, whereas northern regions experienced comparatively smaller reductions (< 60 ha). Spatially, the largest paddy losses overlapped with areas of historically high waterbird richness.

Fig. 3.

Fig. 3

Open in a new tab

Spatial distribution of paddy field loss (ha) across South Korea from 2007 to 2024 at administrative level (si–gun–gu). Loss areas are classified into five categories (0–60, 61–120, 121–180, 181–240, and ≥ 240 ha). Darker shades indicate greater loss, with the most substantial declines concentrated in southwestern and southern coastal regions. All maps were generated by the authors using R software (version 4.3.3; https://www.r-project.org/).

Discussion

Our study demonstrates that climate change is driving a progressive northward shift in the wintering distributions of six rice-paddy–dependent Anatidae species in South Korea, fundamentally reshaping long-established migratory patterns. Across all future scenarios, wintering centroids shifted northward by up to 26.3 km, while hotspots contracted substantially under several scenarios, declining to as low as 5,633.3 km2 and spatial overlap with current hotspots wintering sites (Jaccard index) dropped to 0.62. These findings indicate that range shifts are not merely positional adjustments but are accompanied by considerable habitat contraction and fragmentation, carrying significant ecological implications3.

Multiple ecological mechanisms likely contribute to these changes. Rising temperatures alter energetic balances during migration and wintering, while phenological mismatches between peak food availability and migratory timing may further accelerate redistribution2,29. Beyond these general processes, our results provide structural evidence of habitat degradation: patch numbers increased, the Largest Patch Index (LPI) declined, and edge-related metrics indicated increased fragmentation across scenarios—patterns consistent with fragmentation-driven loss of stable refugia and reduced adaptive capacity.

Importantly, such redistributions are not unique to Korea. In Europe, waterbird wintering centroids have shifted northeastward in response to milder winters3,30, while in North America, climate-driven changes in waterfowl wintering distributions have also been documented, including northward redistribution31,32, altered migration timing, and associated morphological responses such as reductions in body size. The convergence of these findings across continents underscores the role of migratory waterbirds as sentinel species of global climate change1.

Beyond climate forcing, land-use change acts as a compounding stressor within the EAAF context8,33. The rapid northward expansion of greenhouse agriculture in southern Korea is reducing the availability of traditional rice-paddy wintering habitats, constraining the capacity of birds to exploit newly suitable northern areas9,34. Spatially, districts with the largest 2007–2024 paddy losses in southern and western Korea overlapped baseline hotspots, coinciding with the weakening of southern hotspots and the emergence—but reduced connectivity—of more northerly configurations. Our findings suggest that this agricultural transformation amplifies the negative impacts of climate change, reflecting not a purely climate-driven redistribution but a synergistic effect of climate and anthropogenic land-use change with profound conservation consequences.

Although extreme climatic events were not explicitly examined in this study, climate variability—particularly episodic cold spells—may nevertheless interact with the observed redistribution of wintering habitats. Historically, southern wintering grounds have served as climatic refugia during severe winters30,35,36. Their ongoing loss creates a paradox: while northward shifts may be adaptive under gradual warming, they increase vulnerability to acute cold spells, forcing energetically costly long-distance movements to suboptimal habitats31,32. These considerations are presented here as a conceptual implication rather than a modeled result, underscoring the need for future studies to explicitly incorporate climate variability and extreme events, in addition to mean warming trends, into conservation planning37.

Taken together, our results demonstrate that climate change is not only shifting species distributions but also contracting, fragmenting, and degrading critical wintering habitats. When compounded by agricultural transformation and climate variability, these changes pose a substantial threat to the resilience of migratory waterbird populations. Future conservation planning must therefore move beyond site-based protection alone, integrating climate adaptation, land-use policy, and transboundary cooperation across the East Asian–Australasian Flyway (EAAF) to ensure long-term population persistence1,33.

It is important to note that SDM-based hotspot projections implicitly assume that species can occupy newly suitable areas without substantial ecological or energetic constraints. In reality, the formation of new wintering hotspots may be limited by dispersal barriers, energetic costs, habitat accessibility, and strong site fidelity, particularly in highly modified agricultural landscapes. Consequently, reduced spatial overlap with historical hotspots should be interpreted as potential shifts in climatic suitability rather than guaranteed redistribution or successful establishment of new wintering areas. At the same time, the hotspots analyzed in this study should be interpreted as analytical constructs rather than discrete or static habitats, reflecting prioritization outcomes rather than ecologically bounded units14,15. Hotspots, defined model-derived areas of overlapping climatic suitability across all six focal species, represent zones of shared multispecies importance derived from stacked species distribution models, and should be interpreted as spatial prioritization outcomes rather than realized ecological co‐occurrence driven by species interactions38. Because species distribution models were developed independently for each species, these overlap areas do not explicitly account for interspecific interactions, competition, or facilitation, which are well-recognized limitations of correlative SDM approaches39,40. The term “co-occurrence” is therefore used strictly in a spatial and predictive sense, denoting geographic overlap in predicted suitability without implying biological interaction, demographic coupling, or simultaneous site occupancy41. Nevertheless, focusing on areas of highest species overlap provides a conservation-relevant perspective by highlighting shared wintering landscapes where coordinated, landscape-scale management—rather than site-based protection alone— may yield disproportionate benefits for multiple species under constrained resources14,42. Changes in the extent, coherence, and fragmentation of these overlap areas thus offer insights into how climate change may erode the capacity of agricultural landscapes to support multispecies assemblages, beyond what species-specific or land-use analyses alone can reveal43,44.

Despite these advances, several limitations should be acknowledged. First, our analyses were constrained by the geographic coverage of occurrence data. The absence of species records from North Korea prevented us from evaluating the potential expansion of wintering grounds beyond the current northern boundary. Consequently, the projected northward shifts of up to 26.8 km reported here should be interpreted as minimum estimates, as wintering distributions may extend further north under future climate scenarios. Second, although we incorporated multiple climate scenarios and time periods, land-use projections were based on current agricultural patterns, limiting our ability to fully capture future interactions between climate change, agricultural intensification, and habitat loss. Addressing these data gaps will require cross‐border monitoring programs and integrated land‐use–climate modeling to better quantify the spatial dynamics of migratory waterbird populations under accelerating environmental change. Accordingly, model performance metrics such as AUC are interpreted as indicators of relative discrimination within each species–scenario combination, rather than as strictly comparable measures across different models. Third, the baseline bioclimatic variables used in this study represent long-term climatological normals (1970–2000), whereas bird occurrence data were collected during 2022–2024. This temporal mismatch may limit the ability of the models to capture the influence of recent climate variability and extreme events on wintering distributions. Future studies could reduce this uncertainty by reconstructing bioclimatic variables from temporally matched, high-resolution climate datasets.

Conservation implications

This study provides quantitative evidence that climate change is simultaneously driving northward shifts, habitat contraction, and fragmentation in the wintering grounds of rice-paddy–dependent Anatidae species in South Korea. These findings highlight that the impacts of climate change extend beyond simple spatial redistribution to include severe habitat degradation, offering critical insights for future conservation strategies:

(1) Maintain and protect southern wintering grounds as functional climatic refugia.

Southern wintering hotspots are rapidly shrinking, with Jaccard similarity indices declining to as low as 0.62, indicating a severe loss of habitat continuity. Rather than restoring historical habitat conditions lost to climate change, conservation efforts should focus on maintaining and enhancing the functional capacity of remaining southern wintering grounds, which continue to serve as critical climatic refugia buffering populations against extreme events.

(2) Integrate agricultural policy with habitat management.

The northward expansion of greenhouse agriculture restricts the use of newly suitable northern habitats, compounding habitat loss of over 1,000 km2 in some scenarios. Agricultural policies must explicitly incorporate the ecosystem services of rice paddies as key foraging habitats.

(3) Establish a north–south habitat network.

Increased patch numbers and declining LPI values indicate that single-site conservation is insufficient. A coordinated network linking southern refugia with emerging northern habitats is critical for enhancing spatial resilience.

(4) Incorporate climate variability and extremes into planning.

Focusing solely on mean warming overlooks vulnerabilities during acute cold events, when populations may be forced into energetically costly relocations. Conservation planning must explicitly integrate climate extremes as key drivers of population risk.

(5) Promote transboundary cooperation along the EAAF.

All focal species migrate along the East Asian–Australasian Flyway. Effective conservation therefore requires cross-border collaboration involving China, Japan, Russia, and other range states to secure a coordinated network of wintering and stopover sites33.

Supplementary Information

Below is the link to the electronic supplementary material.

Supplementary Material 1 (40KB, docx)
Supplementary Material 2 (22.2KB, docx)
Supplementary Material 3 (49.9KB, docx)

Author contributions

Conceptualization — H‑I Choi, S Lee, H‑K Nam; Data curation — H‑I Choi; Methodology — H‑I Choi, H‑K Nam; Formal analysis — H‑I Choi; Visualization — H‑I Choi; Writing—original draft — H‑I Choi, H‑K Nam; Writing—review & editing — S Lee, H‑K Nam; Supervision — H‑K Nam.

Funding

This research was supported by grants from the National Institute of Biological Resource (NIBR), funded by the Ministry of Environment, Republic of Korea (NIBR202617102).

Data availability

The datasets generated and/or analysed during the current study are available in the zenodo repository, https://doi.org/10.5281/zenodo.17337531.

Declarations

Competing interests

The authors declare no competing interests.

Footnotes

Publisher’s note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

6. References

  • 1.Robinson, R. A. et al. Travelling through a warming world: climate change and migratory species. Endang Species Res.7, 87–99. 10.3354/esr00095 (2009). [Google Scholar]
  • 2.Both, C., Bouwhuis, S., Lessells, C. M. & Visser, M. E. Climate change and population declines in a long-distance migratory bird. Nature441, 81–83. 10.1038/nature04539 (2006). [DOI] [PubMed] [Google Scholar]
  • 3.Lehikoinen, A. et al. Rapid climate-driven shifts in wintering distributions of three common waterbird species. Glob Change Biol.19, 2071–2081. 10.1111/gcb.12200 (2013). [DOI] [PubMed] [Google Scholar]
  • 4.Duan, H., Yu, X., Xia, S. & Zhang, G. Waterbird habitat loss: fringes of the yellow and Bohai seas along the East Asian–Australasian flyway. Land. Degrad. Dev.32, 4174–4182. 10.1002/ldr.4023 (2021). [Google Scholar]
  • 5.Wang, C., Xia, S., Yu, X., Duan, H. & Qi, G. Stage-specific impacts of climate change on greater White-fronted geese along the East. Asian Flyway Biology. 14, 1050. 10.3390/biology14081050 (2025). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 6.Kim, M. R. et al. Status of birds using a rice paddy in South Korea. Korean J. Environ. Agric.32, 155–165. 10.5338/KJEA.2013.32.2.155 (2013). [Google Scholar]
  • 7.Son, S. J. et al. Spatial and Temporal patterns of waterbird communities and impact of landscape features at multi-scale in rice fields of the Korean Peninsula. Landsc. Ecol.40, 151. 10.1007/s10980-025-02165-2 (2025). [Google Scholar]
  • 8.Nam, H. K. et al. Long-term changes in population trends of wintering waterbirds in the Republic of Korea. Bird. Conserv. Int.35, e30. 10.1017/S0959270925100178 (2025). [Google Scholar]
  • 9.Fujioka, M., Lee, S. D., Kurechi, M. & Yoshida, H. Bird use of rice fields in Korea and Japan. Waterbirds33, 8–29. 10.1675/063.033.s102 (2010)
  • 10.Tajiri, H. & Ohkawara, K. The effects of flooding and plowing on foraging site selection by wintering dabbling ducks in rice fields. Ornithol. Sci.12, 25–34. 10.2326/osj.12.127 (2013). [Google Scholar]
  • 11.Choi, S. H., Choi, G. & Nam, H. K. Impact of rice paddy agriculture on habitat usage of migratory shorebirds at the rice paddy scale in Korea. Sci. Rep.12, 5762. 10.1038/s41598-022-09708-6 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 12.Parmesan, C. Ecological and evolutionary responses to recent climate change. Annu. Rev. Ecol. Evol. Syst.37, 637–669. 10.1146/annurev.ecolsys.37.091305.110100 (2006). [Google Scholar]
  • 13.Chen, I. C., Hill, J. K., Ohlemüller, R., Roy, D. B. & Thomas, C. D. Rapid range shifts of species associated with high levels of climate warming. Science333, 1024–1026. 10.1126/science.1206432 (2011). [DOI] [PubMed] [Google Scholar]
  • 14.Margules, C. R. & Pressey, R. L. Systematic conservation planning. Nature405, 243–253. 10.1038/35012251 (2000). [DOI] [PubMed] [Google Scholar]
  • 15.Kukkala, A. S. & Moilanen, A. Core concepts of Spatial prioritisation in systematic conservation planning. Biol. Rev.88, 443–464. 10.1111/brv.12008 (2013). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 16.Lehtomäki, J. & Moilanen, A. Methods and workflow for Spatial conservation prioritization using zonation. Environ. Model. Softw.47, 128–137. 10.1016/j.envsoft.2013.05.001 (2013). [Google Scholar]
  • 17.Meller, L. et al. Ensemble distribution models in conservation prioritization: from consensus predictions to consensus reserve networks. Divers. Distrib.20, 309–321. 10.1111/ddi.12162 (2014). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 18.Velazco, S. J. E. et al. Overprediction of species distribution models in conservation planning: a still neglected issue with strong effects. Biol. Conserv.252, 108822. 10.1016/j.biocon.2020.108822 (2020). [Google Scholar]
  • 19.Fan, R. et al. Species distribution modeling of the breeding site distribution and conservation gaps of lesser white-fronted goose in Siberia under climate change. Land11, 1946. 10.3390/land11111946(2022) .
  • 20.Aiello-Lammens, M. E. et al. SpThin: an R package for Spatial thinning of species occurrence records for use in ecological niche models. Ecography38, 541–545. 10.1111/ecog.01132 (2015). [Google Scholar]
  • 21.R Core Team. R: A language and environment for statistical computing (R Foundation for Statistical Computing, 2023). https://www.R-project.org/
  • 22.Thuiller, W., Georges, D., Engler, R. & Breiner, F. biomod2: Ensemble platform for species distribution modeling. R package version 3.5-1 (2021). https://CRAN.R-project.org/package=biomod2
  • 23.Hijmans, R. J., Phillips, S., Leathwick, J. & Elith, J. dismo: Species distribution modeling. R package version 1.3–14 (2023). https://CRAN.R-project.org/package=dismo
  • 24.Hijmans, R. J. & raster Geographic data analysis and modeling. R package version 3.6–20 (2023). https://CRAN.R-project.org/package=raster
  • 25.Pebesma, E. Simple features for R: standardized support for Spatial vector data. R J.10 (1), 439–446. 10.32614/RJ-2018-009 (2018). [Google Scholar]
  • 26.Hijmans, R. J. terra: Spatial data analysis. R package version 1.7–39 (2023). https://CRAN.R-project.org/package=terra
  • 27.Wickham, H. et al. dplyr: a grammar of data manipulation. R package version 1.1.2 (2023). https://CRAN.R-project.org/package=dplyr
  • 28.Wickham, H. ggplot2: Elegant graphics for data analysis. (Springer, 2016). https://ggplot2.tidyverse.org
  • 29.Gill, J. A. et al. Why is timing of bird migration advancing when individuals are not? Proc. R. Soc. B281, 20132161. 10.1098/rspb.2013.2161 (2014). [DOI] [PMC free article] [PubMed]
  • 30.Pavón-Jordán, D. et al. Habitat- and species-mediated short- and long-term distributional changes in waterbird abundance linked to variation in European winter weather. Divers. Distrib.25, 225–239. 10.1111/ddi.12855 (2019). [Google Scholar]
  • 31.Schummer, M. L. et al. Weather-related indices of autumn–winter dabbling Duck abundance in middle North America. J. Wildl. Manag. 74, 94–101. 10.2193/2008-524 (2010). [Google Scholar]
  • 32.Notaro, M. et al. Projected influences of changes in weather severity on autumn–winter distributions of dabbling ducks in the Mississippi and Atlantic flyways during the twenty-first century. PLOS ONE. 11, e0167506. 10.1371/journal.pone.0167506 (2016). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 33.Studds, C. E. et al. Rapid population decline in migratory shorebirds relying on yellow sea tidal mudflats as stopover sites. Nat. Commun.8, 14895. 10.1038/ncomms14895 (2017). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 34.Seoul, L. A. N. Greenhouse Horticulture in South Korea. (Netherlands Enterprise Agency (RVO), 2021). https://www.rvo.nl/sites/default/files/2021/08/Greenhouse-Horticulture-in-South-Korea.pdf (Accessed: 13 Oct 2025).
  • 35.Adam, M. et al. Long-term changes in habitat selection of wintering waterbirds: high importance of cold weather refuge sites. Acta Ornithol.50, 127–138. 10.3161/00016454AO2015.50.2.001 (2015). [Google Scholar]
  • 36.Ridgill, S. C. & Fox, A. D. Cold weather movements of waterfowl in western Europe. IWRB Special Publication 13, 89 (Slimbridge, 1990).
  • 37.Rahmstorf, S. & Coumou, D. Increase of extreme events in a warming world. Proc. Natl. Acad. Sci. U S A. 108, 17905–17909. 10.1073/pnas.1101766108 (2011). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 38.Schmitt, S. et al. Ssdm: an r package to predict distribution of species richness and composition based on stacked species distribution models. Methods Ecol. Evol.8, 1795–1803. 10.1111/2041-210X.12841 (2017). [Google Scholar]
  • 39.Araújo, M. B. & Guisan, A. Five (or so) challenges for species distribution modelling. J. Biogeogr.33, 1677–1688. 10.1111/j.1365-2699.2006.01584.x (2006). [Google Scholar]
  • 40.Wisz, M. S. et al. The role of biotic interactions in shaping distributions and realised assemblages of species: implications for species distribution modelling. Biol. Rev.88, 15–30. 10.1111/j.1469-185X.2012.00235.x (2013). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 41.Dormann, C. F. et al. Biotic interactions in species distribution modelling: 10 questions to guide interpretation and avoid false conclusions. Glob Ecol. Biogeogr.27, 1004–1016. 10.1111/geb.12759 (2018). [Google Scholar]
  • 42.Schwenk, W. S. & Donovan, T. M. A multispecies framework for landscape conservation planning. Conserv. Biol.25, 1010–1021. 10.1111/j.1523-1739.2011.01723.x (2011). [DOI] [PubMed] [Google Scholar]
  • 43.Sinclair, S. J., White, M. D. & Newell, G. R. How useful are species distribution models for managing biodiversity under future climates? ECOL SOC.15, art8. http://www.ecologyandsociety.org/vol15/iss1/art8/ (2010)
  • 44.Pecl, G. T. et al. Biodiversity redistribution under climate change: impacts on ecosystems and human well-being. Science355, eaai9214. 10.1126/science.aai9214 (2017). [DOI] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

Supplementary Material 1 (40KB, docx)
Supplementary Material 2 (22.2KB, docx)
Supplementary Material 3 (49.9KB, docx)

Data Availability Statement

The datasets generated and/or analysed during the current study are available in the zenodo repository, https://doi.org/10.5281/zenodo.17337531.

Articles from Scientific Reports are provided here courtesy of Nature Publishing Group

RESOURCES