Spatiotemporal Forest-Cover Dynamics and Associated Factors in Yunnan, China, 1990–2024: Multiscale Trends, Land-Cover Turnover, and Contrasts Between Forest State and Change

Long-term forest recovery is often summarized by net changes in forest extent, but net gains may mask substantial land-cover reorganization, and the factors associated with current forest distribution may differ from those associated with long-term change. Using annual 30 m China Land Cover Dataset (CLCD) data for 1990–2024, we quantified forest-cover dynamics across Yunnan, China, and compared the spatial explanatory structures of 2024 forest state and long-term forest-cover change using trend, stage-endpoint transition, and optimized GeoDetector analyses. Forest area increased by 9938 km2, from 255,543 to 265,481 km2. A direct 1990–2024 endpoint comparison identified 32,209 km2 of forest gain and 22,271 km2 of forest loss, so gross endpoint reorganization substantially exceeded the net gain; cropland was the principal land-cover counterpart of forest transitions in both directions. Long-term change was also spatially heterogeneous: 49 of 129 counties showed significant increases and 27 significant decreases after serial-correlation-adjusted trend testing. The explanatory structures of forest state and change differed markedly. Slope had the highest explanatory power for 2024 forest cover (q = 0.397), whereas mean temperature ranked highest for long-term change (q = 0.102), and their leading interactions also differed. These findings show that forest recovery should be evaluated beyond net area change and that forest state and long-term change should be treated as distinct spatial processes.