Nonlinear relationships between vegetation dynamics and ecosystem services in the Yangtze River Basin: insights from random forest and GTWR analyzes

Understanding the nonlinear relationships between vegetation dynamics and ecosystem services (ES) is essential for sustainable ecosystem management in the Yangtze River Basin (YRB). Using the Google Earth Engine platform, we analyzed NDVI dynamics in the YRB from 2000 to 2020. The InVEST model was used to assess four key ES, including water yield (WY), carbon storage (CS), habitat quality (HQ), and soil retention (SR). Random forest and spatiotemporal geographically weighted regression models were further applied to explore the nonlinear relationships and spatial heterogeneity between NDVI and ES. The results showed that: (1) mean NDVI in the YRB was 0.692, 0.697, and 0.724 in 2000, 2010, and 2020, respectively, exhibiting clear spatiotemporal heterogeneity. (2) WY showed a non-monotonic trajectory of initial increase followed by decline, SR increased markedly, HQ remained relatively stable, and CS showed a slight decrease. (3) CS and SR showed more pronounced positive associations with NDVI, whereas WY and HQ exhibited stronger interval variability or weaker localized effects. (4) NDVI generally had positive local effects on CS, HQ, and SR, while the NDVI–WY relationship exhibited both positive associations and localized negative effects. These findings provide valuable insights for ecological restoration and ecosystem service management in the YRB.