Tracking vegetation dynamics in the Upper Yellow River Basin: An analysis using kNDVI and multi-model trajectory detection

The Upper Yellow River Basin (UYRB) contains vast areas of alpine and arid grassland, making it a typically fragile and sensitive ecosystem. Effective methods for monitoring vegetation change trajectories in this region, especially nonlinear abrupt changes, are still lacking. To address this, we analyzed the kernel Normalized Difference Vegetation Index (kNDVI) from 2000 to 2024 and employed a multi-model trajectory identification technique using three statistical models (linear, curvilinear, and step) to detect these changes. Furthermore, we combined Random Forest (RF), Shapley Additive Explanations (SHAP), and Generalized Additive Model (GAM) with the Geographical Convergent Cross Mapping (GCCM) model to quantitatively identify the drivers of vegetation change and their causal relationships. Our results showed that approximately 60.7% of the UYRB exhibited a positive vegetation trend, with 58.92% of this improvement being abrupt. Conversely, 78.93% of the areas with negative trends experienced abrupt decline. Elevation was the dominant factor controlling these abrupt vegetation shifts, with a critical threshold similar to 1500 m. Above this threshold, a warming trend of less than 0.02 degrees C/yr triggered positive abrupt changes. Below it, vegetation was more susceptible to negative abrupt changes. The GCCM analysis further revealed complex causal relationships within the "topography-climate-vegetation-human activity" system. We found that the drivers of vegetation abrupt changes form a coupled system with a clear hierarchy and key feedback loops. Elevation acts as the key factor, governing the distribution of water and heat and influencing climate trends, which in turn affect vegetation growth. Ecological project and the human footprint then modulate how vegetation responds to these climatic changes. This study not only establishes a robust method for monitoring vegetation trajectories in fragile alpine and arid regions but also clarifies the internal mechanisms of ecological change in the UYRB, offering new perspectives for consolidating vegetation restoration efforts.