A spatially adaptive synthetic vegetation index for monitoring ecosystem changes in climatically heterogeneous basins.

The vegetation in climatically heterogeneous regions exhibits significant spatial variability and temporal succession characteristics. It is crucial to obtain consistent vegetation characteristics in this region over time. Traditional single indices such as NDVI (Normalized Difference Vegetation Index), LAI (Leaf Area Index), and NPP (Net Primary Productivity) each have their own advantages, but they often show inconsistent trends when applied to complex vegetation. To effectively capture spatial heterogeneity and enhance the ecological interpretability, we propose a Dynamic Spatially Variable Weighted Synthesis Vegetation Index (DWS-SVI). Based on four Global Land Surface Satellite Dataset (GLASS) vegetation parameters (FVC (Fractional Vegetation Cover), LAI, NDVI, and NPP) and land cover types, this method employs the CRITIC method to perform dynamic weighting and generate continuous weight surfaces, ultimately synthesizing a comprehensive vegetation index at the pixel level. It combines "global trend and local adaptation" by integrating spatial heterogeneity modeling and multi-variable dynamic weighting, thereby overcoming the limitations of traditional methods in terms of spatial heterogeneity and ecological interpretability. Results show that over the past two decades, more than 69.4% of the area in the YRB has witnessed a significant improvement in vegetation conditions. The improvement was most notable in the summer, and it was mainly attributed to the improvement in the temperature and humidity conditions in this region. Compared with a single indicator, the DWS-SVI index can reflect the coordinated evolution of ecosystem structure and function, and can effectively suppress the observation errors caused by the bias of a single vegetation index, especially the "false greening" signals in transition zones and arid areas. Furthermore, the dominant factor map constructed based on DWS-SVI further reveals the differentiated driving mechanisms of ecosystems such as farmland, grassland, and forest, demonstrating that it has superior interpretability. This study provides a transferable framework for constructing spatially adaptive vegetation indices, enabling more reliable monitoring of ecosystem changes in large river basins and other climatically heterogeneous regions.