Abstract: This study investigated the core grassland distribution areas within the Nanshan National Park Candidate to explore the spatiotemporal dynamics of grasslands and their driving mechanisms. Based on Landsat remote sensing data from 1990 to 2020, and utilizing the Google Earth Engine (GEE) cloud computing platform, this research systematically characterized grassland change patterns and influencing factors through an integrated application of random forest classification, the dimidiate pixel model, and partial correlation analysis. This multi-method framework enabled robust land cover mapping, quantitative assessment of vegetation conditions, and effective resolution of the relative roles of environmental and anthropogenic factors. The results indicate a significant reduction in grassland area within the study region over the past three decades. From 1990 to 2020, the total grassland area decreased by 29.7 km², representing a reduction of 33.0%. Among this reduction, 28.33 km² (31.5% of the original grassland area) transitioned to forestland, spatially concentrated in areas with elevations of 1500—1700 m and slopes of 16—25°. These areas represent transitional terrains where land use and vegetation succession were most evident. The remaining grasslands that did not undergo type conversion covered 49.84 km² (55.6% of the original area), primarily distributed across the Nanshan mountaintop and surrounding high-elevation tablelands. Although their land cover type remained unchanged, the fractional vegetation cover (FVC) in these areas exhibited an overall declining trend, indicating ecological degradation. Topographic factors indirectly drove grassland dynamics by regulating the intensity and spatial patterns of human activities, particularly grazing. The areas with lower elevations and steeper slopes experienced constrained grazing activities due to limited topographic accessibility, thereby facilitating natural succession from grassland to forestland. Conversely, gentle slopes at higher elevations remained suitable for sustained grazing, allowing grasslands to maintain their type stability while gradually undergoing vegetation degradation. These findings reveal the critical regulatory role of topography in mediating the relationship between human land use and ecosystem dynamics. Environmental factors exhibited significant spatial heterogeneity in their effects on grassland FVC. However, partial correlation analysis revealed that correlations between vegetation cover and major environmental factors lacked statistical significance in most areas (P>0.1), suggesting that environmental factors had limited direct effects on grassland vegetation cover compared to anthropogenic pressures. Overall, grassland dynamics in the Nanshan region result from the interplay between natural succession and human activities, with topographic differentiation and overgrazing serving as the dominant driving forces. The research findings can provide a scientific basis for understanding grassland change patterns in protected mountain areas of southern China, offer theoretical support for the precise restoration and sustainable management of southeastern tropical shrub-grassland ecosystems, and provide data support for scientifically informed conservation planning of grassland ecosystems within the Nanshan National Park Candidate.