Multi-Scale Dynamics of Permanent Surface Water in the Lower Yangtze River Basin: Basin-Scale Expansion and Pixel-Scale Fragmentation

Permanent surface water is an important component of regional hydrological systems, yet basin-scale area statistics alone may overlook structural changes in highly modified basins. This study investigated long-term permanent surface-water dynamics in the Lower Yangtze River Basin (LYRB), China, during 2000–2020 using a basin-to-pixel remote sensing framework that combined annual water mapping with multi-scale structural diagnostics. Annual permanent surface-water maps were derived from Landsat imagery using adaptive thresholding. The reliability was evaluated through product-level comparison, reference-based accuracy assessment, and representative complex-scene comparisons with machine learning and deep learning baselines in urban-shadow, turbid-water, and algae-affected environments. Basin-scale landscape metrics, centroid migration, and pixel-level Spatial Water Variation (SPAT) indices were constructed to characterize structural change, while generalized additive models were applied to examine changing associations with hydro-climatic, land-surface, and anthropogenic factors. The extracted dataset achieved high overall accuracy (OA = 98.43%) and improved delineation of narrow channels and fragmented littoral boundaries. Permanent surface water in the LYRB followed a decline–recovery–stabilization trajectory, but basin-scale expansion often co-occurred with increased pixel-scale fragmentation and reduced aggregation. Expansion was mainly expressed through scattered small patches, whereas shrinkage was concentrated along the margins of existing water bodies. Driver analysis further showed that NDVI remained the strongest explanatory variable throughout the study period, while land-surface and anthropogenic variables became more prominent in later periods relative to hydro-climatic variables. These results indicate that basin-scale area statistics alone are insufficient to describe the full evolution of permanent surface water and provide a useful multi-scale perspective for understanding permanent surface-water change in highly modified river–lake basins.