The research combined land-use maps from 2000, 2005, 2010, 2015 and 2020 with waterbird surveys conducted during the early and late stages of lake recession. Surveys of 22 fixed sites recorded species presence or absence in mid-November 2019 and mid-November 2020. These observations were matched to habitat and landscape variables within three-kilometre buffers around the sites. The study used negative-binomial generalized linear models to examine waterbird species richness and partial permutational multivariate analysis of variance to test associations with community composition. It also evaluated how sensitive the results were to alternative buffer sizes of one and five kilometres.

Landscape metrics showed that an increase in the area of a wetland class did not necessarily mean that its dominant patches became larger or more connected. For lakes, the percentage of the landscape occupied by the class fell from 9.9142% to 7.3225%, while the largest patch index declined from 6.6455% to 4.4133%. Lake patch density, edge density and mean nearest-neighbour distance also decreased. Marsh moved in the opposite direction: its landscape share rose from 3.4371% to 4.5411%, and its largest patch index increased from 0.2734% to 1.4822%. Reservoir and pond cover increased from 3.3318% to 4.8741%, but patch density remained the highest among the wetland classes, reflecting the dispersed distribution of numerous small water bodies.

The InVEST Habitat Quality model provided a second view of these changes. Mean habitat quality across the study area declined from 0.47827 in 2000 to 0.45371 in 2020, while mean degradation rose from 0.17985 to 0.19445. The most pronounced interval of change occurred from 2005 to 2010, when habitat quality fell by 0.01029 and degradation increased by 0.00670. Using a change threshold of 0.01, habitat quality declined in 40.94% of mapped pixels, remained stable in 49.66% and improved in 9.40%. Degradation increased in 43.07%, remained stable in 47.37% and decreased in 9.56%.

Responses differed among wetland types. Lake and marsh retained comparatively high mean habitat-quality values in 2020, at 0.76680 and 0.82985, respectively, but lake quality declined from 0.79376 in 2000. River and reservoir or pond habitat were more exposed to modeled disturbance associated with agriculture, construction and transportation. Reservoir and pond habitat quality fell from 0.60925 to 0.59294, while degradation rose from 0.23388 to 0.24637, demonstrating that expansion of that land-cover class did not translate into better modeled habitat conditions. Marsh was the notable exception: quality increased from 0.80198 to 0.82985 and degradation decreased from 0.20300 to 0.17843.

Those modeled trends survived parameter sensitivity tests. The researchers separately varied each threat weight by plus or minus 20% in 2020 and jointly varied maximum threat distances and all nonzero habitat-sensitivity coefficients across the five years. Although absolute habitat-quality and degradation values changed with model settings, every configuration retained the same long-term direction. Spearman correlations between baseline and perturbed outputs ranged from 0.9550 to 0.9729 for habitat quality and from 0.9931 to 0.9967 for degradation. The results therefore support the broad temporal trends within the tested parameter range, while also showing that the precise index values depend on assumptions in the InVEST model.

Waterbird richness remained broadly similar between recession periods despite major local redistribution. Forty-eight species were recorded in each survey period. Mean richness was 13.09 species per site during early recession and 12.77 during late recession, with site values ranging from 3 to 29 and 2 to 32, respectively. The average change across sites was only minus 0.32 species, yet individual sites changed by as much as 14 species in either direction. East Dongting Lake had the highest mean richness in both periods, at 17.56 and 17.78 species, while South and West Dongting Lake had lower averages. Similar total richness at two sites therefore did not imply similar ecological communities.

Community-composition analysis was more sensitive to landscape structure than the richness models. During early recession, the largest wetland patch index was significantly associated with Jaccard community dissimilarity after accounting for lake area and subregion. Its semi-partial R-squared was 0.1027, meaning that dominant patch structure accounted for about 10.27% of the unique compositional variation in the model. The permutation p-value was 0.0078, and the false-discovery-rate-adjusted q-value was 0.0468. The association was not significant during late recession, when the semi-partial R-squared fell to 0.0571, the permutation p-value rose to 0.2386 and the adjusted q-value was 0.7158. Habitat quality, degradation, shoreline proportion, wetland patch density and construction-land proportion were not significantly associated with community composition in either period.

These contrasts identify a period-specific ecological signal, but they do not demonstrate that large wetland patches caused particular bird communities to form. The largest-patch association was strongest only at the primary three-kilometre scale: its semi-partial R-squared was 0.0890 at one kilometre and 0.0678 at five kilometres, with adjusted q-values of 0.1060 and 0.4932. The early-recession community also showed unequal dispersion among subregions, which may partly influence the PERMANOVA result. Moreover, the models used 2020 landscape and InVEST layers as a shared background for both surveys rather than conditions matched to each survey year. Water depth, inundation duration, discharge, vegetation structure and food availability were not included, leaving important hydrological and ecological mechanisms unresolved.

Within those limits, the study offers a layered way of evaluating a dynamic wetland. InVEST can screen broad habitat condition, landscape metrics diagnose the arrangement of dominant patches and habitat mosaics, and bird monitoring tests whether biological communities respond. For Dongting Lake, the results favor conserving connected sequences of open water, shallow water, mudflats, bottomland and marsh rather than assuming that gains in one mapped class compensate for the loss of another. The authors also argue that management should account for the different roles of East, South and West Dongting Lake and for the stage of seasonal recession, when water depth and resource availability may become more important than static landscape structure. The central conclusion is not merely that the wetland changed: the way its habitat pieces were arranged was associated with how waterbird communities assembled, especially during early recession.

Subject of Research: Long-term wetland landscape change, modeled habitat quality, and waterbird community responses in Dongting Lake, China

Article Title: Period-specific waterbird associations with wetland landscape configuration and InVEST-based habitat quality in Dongting Lake, China

Article References: Xu, D., Chen, H., & Xue, Z. (2026). Period-specific waterbird associations with wetland landscape configuration and InVEST-based habitat quality in Dongting Lake, China. Frontiers in Ecology and Evolution, 14, Article 1976232. https://doi.org/10.3389/fevo.2026.1976232

Image Credits: AI Generated

DOI: 10.3389/fevo.2026.1976232

Keywords: Dongting Lake, waterbird ecology, wetland landscapes, habitat quality, InVEST model, land-use change, community composition, habitat degradation, seasonal recession, China, Period-specific, waterbird