Gridded Grazing Intensity Based on Geographically Weighted Random Forest and Its Drivers: A Case Study of Western Qinghai–Tibetan Plateau

ABSTRACT Overgrazing affects the grass‐livestock balance and endangers grassland ecological security. Despite extensive studies conducted on identifying and quantifying grazing intensity, there is still room for improvement in the research on gridding grazing intensity, particularly in areas with limited data on the Qinghai–Tibet Plateau. Therefore, we proposed a grazing intensity spatialization method using geographically weighted random forest (GWRF) to gain further insights into the spatial heterogeneity of alpine grassland grazing intensity. This method incorporates multiple remote sensing data related to human activities and natural factors, as well as annual livestock statistics at the township level over several years, while adequately considering the spatial autocorrelation of grazing intensity. Additionally, we employed Lindeman Merenda Gold (LMG), the geographical detector model, and the structural equation model (SEM) to assess the contribution and influence path of driving factors to grazing intensity. We also utilize partial correlation analysis and dual‐phase mapping to examine the impact of natural and human activities on the spatial distribution of grazing intensity. The results demonstrate that the GWRF‐based grazing intensity spatial model accurately predicts grazing intensity by demonstrating its consistency with township‐scale livestock data ( R 2 = 0.92 ( p < 0.01), RMSE = 1.07). This provides valuable technical support for quantifying grazing intensity in alpine pastoral areas with limited data availability. We evaluate trends in grazing intensity and observe an increase in Gar and Purang counties. Furthermore, population density, normalized difference vegetation index (NDVI), and temperature are identified as three influential factors affecting grazing intensity in alpine pastoral areas. Additionally, other factors indirectly impact grazing intensity by influencing population density and NDVI levels, while their interactions amplify their overall influence. The dual‐phase mapping technique has demonstrated a significant impact of population density on 45.92% ( p < 0.01) of the study area, emphasizing the substantial influence of human activities on grazing intensity. Our study provides a novel framework for spatially analyzing grazing intensity and unraveling the intricated driving mechanisms behind spatiotemporal changes, particularly in areas with limited data availability.