Remote sensing analysis of spatiotemporal habitat evolution in the water-land ecotone of a typical large river-connected lake

The water-land ecotone of Dongting Lake is one of the most sensitive zones to river-lake interactions and one of the most complex areas in terms of coupled ecological processes. Using the Google Earth Engine platform and six phases of Landsat and Sentinel-2 images acquired during 1993–2023, this study constructed a multi-source feature set and applied a random forest classifier to identify habitats in the Dongting Lake water-land ecotone. The spatiotemporal evolution of habitats was analyzed for the whole lake and for the eastern, southern, and western sub-lakes over the past 30 years. The results show that the water area in the dry season decreased from 775.36 km2 in 1993 to 554.41 km2 in 2023, a reduction of 28.5%. Grass shoals exhibited a pronounced V-shaped trajectory, emergent vegetation showed an inverted V-shaped trajectory, and poplar forest underwent three stages of expansion, decline, and stabilization. Marked spatial differences were found among the three sub-lakes. These results provide a data basis for wetland conservation, ecological restoration, and river-lake regulation in Dongting Lake.