[Objective] Scientifically assessing the spatiotemporal evolution and spatial association between ecosystem service value and ecological risk in national parks provides an important foundation for supporting differentiated ecological management and zoned restoration.[Methods] Taking Sanjiangyuan National Park as the study area, this study utilized land cover data from 2000 to 2020 and integrated spatial autocorrelation analysis, geographically weighted regression with optimal bandwidth, random forest, and an optimal-parameter geographic detector to analyze the dynamics of different ecosystem types and to reveal the spatiotemporal evolution, spatial association, and driving mechanisms of ecosystem service value and ecological risk. [Results] (1) From 2000 to 2020, significant differences were observed in the area changes among ecosystem types in the study area, with shrubs and cropland showing the largest changes (-3.19% and -2.69%, respectively), while grassland exhibited the smallest change (-0.13%). (2) Ecosystem service value showed an overall increasing trend. The area of high-level ecosystem service value decreased the most (-2.60%), whereas the area of relatively low-level ecosystem service value increased the most (1.57%). (3) Ecological risk exhibited an overall decreasing trend. The area of high-level ecological risk decreased the most (-3.05%), while the area of moderate-level ecological risk increased the most (1.42%). (4) Ecosystem service value and ecological risk were closely spatially associated and could be classified into four spatial clustering types: high value - high risk, high value - low risk, low value - high risk, and low value - low risk. The high value - high risk area decreased by 0.50%, while the low value - high risk, low value - low risk, and high value - low risk areas increased by 1.41%, 1.11%, and 0.09%, respectively. (5) Driving factor analysis indicated that the human footprint index and precipitation seasonality were the most important individual driving factors. The interactions between NDVI and annual mean temperature, NDVI and the human footprint index, and railway - road distance and the human footprint index showed strong explanatory power, revealing a coupled driving mechanism of natural processes and human activities. [Conclusion] Based on spatial association characteristics, Sanjiangyuan National Park can be divided into ecological restoration-dominated zones, ecological protection-dominated zones, and ecological protection - restoration balanced zones. Considering the differences and transitions among these zones, differentiated ecological protection strategies—including zoned management, key area regulation, and targeted governance—should be implemented to support the scientific conservation and restoration of the national park ecosystem, enhance ecological functions, and promote high-quality and sustainable development of national parks.