Remote-sensing-based Monitoring of Spatiotemporal Variations and Driving Forces of Terrestrial Vegetation NPP in a Subtropical Urbanizing Region.

Net primary productivity (NPP) underpins terrestrial carbon cycling, yet existing regional analyses often rely on linear regression or correlation methods that cannot resolve multicollinearity or nonlinear interactions among drivers. Here, we introduce a novel coupled Carnegie-Ames-Stanford Approach (CASA)-Geographical Detector framework that quantifies the relative spatiotemporal weights of human activities versus climatic factors without presupposed functional forms. Applying this model to MODIS and meteorological data for Guangdong Province (2001-2020), we applied Theil-Sen slope estimation, Mann-Kendall tests, and coefficient of variation to map NPP trends and then employed partial correlation and the Geographical Detector's Q-statistic to partition driver effects. Mean NPP rose to 986.7 g C m year with a clear north-high/south-low gradient. Land-use/cover change (LUCC) dominated (q = 0.86), exhibiting its strongest synergy with solar radiation (q = 0.88), whereas among climate variables, temperature (q = 0.22) exerted more spatial control than precipitation or radiation (each q ≈ 0.016). By overcoming the linearity assumption, our framework reveals how LUCC not only directly alters NPP but also amplifies or dampens climate influences and supports spatially stratified interventions-such as targeted afforestation in the north and urban greening in the Pearl River Delta-projected to boost regional carbon sequestration.