The grassland ecosystem, which dominates arid and semi-arid regions, is vital for maintaining ecological equilibrium, biodiversity, and critical ecosystem functions. Quantifying the spatial patterns of temporal stability in the grassland ecosystem and its responses to environmental change is therefore essential for understanding and enhancing the resilience of these landscapes. In this study, we evaluated the spatial patterns of stability across the grassland ecosystem in China. By integrating the Optimal Parameters-based Geographical Detector (OPGD), Spatial Random Forest (SRF) models, and Piecewise Structural Equation Modeling (piecewiseSEM) incorporating a Generalized Least Squares (GLS) framework, we characterized the response patterns of grassland ecosystem stability to a suite of key environmental drivers. The results indicated a significant positive spatial autocorrelation of stability within the grassland ecosystem. Abiotic factors emerged as the dominant drivers, accounting for 80.24% of the variance, with soil physicochemical properties exerting the most substantial influence, while biotic factors contributed 19.76%. Notably, after rigorously accounting for spatial autocorrelation using the spatial GLS framework, species diversity exhibited a slight negative influence on stability, a relationship mediated by the complex interplay of other environmental and structural drivers. These findings underscore the dominant role of abiotic drivers in sustaining the stability of the grassland ecosystem in arid and semi-arid climates, provide a novel framework for assessing ecosystem stability across broad spatial scales, and advance our understanding of relationships between biodiversity and ecosystem functioning in dryland ecosystems.