Vegetation EVI changes and response to natural factors and human activities based on geographically and temporally weighted regression

The research on vegetation changes plays a crucial role in the assessment of ecosystem health, monitoring environmental changes, providing early warnings for natural disasters, and supporting decision-making for sustainable development. However, understanding the nonstationary characteristics of drivers affecting vegetation change remains challenging. This study used Enhanced Vegetation Index (EVI) data obtained through Google Earth Engine (GEE), Theil-Sen, and Mann-Kendall methods to analyze the spatial-temporal patterns and trends of vegetation changes in Sichuan, western China from 2000 to 2020. The Geographical and Temporal Weighted Regression (GTWR) method was applied to deal with spatial and temporal nonstationarity simultaneously. Results showed that vegetation cover in Sichuan was good overall, with medium and high vegetation covering more than 78% of the area. About 72.75% of the area showed an increasing trend in vegetation cover, and areas with extremely significant and significant EVI growth (p < 0.01 and 0.01 ≤ p < 0.05) accounted for 23.94% of the total area. The areas with significant increases in vegetation EVI were mainly distributed in northeast, east, southeast, central, and southwest in Sichuan, while the areas with significant decreases were mainly distributed in the central Sichuan plain urban agglomeration and western Sichuan plateau. GTWR addressed the nonstationary effect of the temporal dimension on the drivers of natural and human activities, with a fitted R2 of 0.846. The study identified climate, terrain, and human activities as the primary driving factors behind vegetation EVI fluctuations. Annual average temperature and precipitation, human activities, and slope had a positive impact on vegetation EVI changes, while solar radiation and aspect had a negative inhibitory effect. The effects of climate, terrain, and human activities on EVI changes exhibited significant spatial heterogeneity and clustering, resulting in either positive promotion or negative inhibition. This study provides an additional methodology to solve the nonstationary problem of vegetation change trends and their response mechanisms. The revealed changes in vegetation EVI and the spatiotemporal heterogeneity characteristics of their driving factors are important for fragile ecosystems to adapt to and mitigate the effects of natural changes and human activities. Revealing the variations in vegetation EVI and their underlying drivers can showcase diverse characteristics across regions and time periods, the presence of spatiotemporal heterogeneity holds great significance in comprehending the adaptive strategies employed by fragile ecosystems to mitigate the effects of natural fluctuations and human-induced activities.