Increasingly frequent natural hazards put food systems under constant stress, and there is an urgent need to improve the resilience of food systems. Quantitative assessments of food system resilience (FSR) and research on the factors that influence it are important ways to achieve this goal. In this research, we use probabilistic modeling to quantitatively assess the FSR of forty counties in the southeastern Tibetan Plateau from 2000 to 2020, and explore the spatiotemporal dynamics characteristic of FSR. The Geographically and Temporally Weighted Regression (GTWR) model is then used to examine the drivers of FSR from four dimensions: shocks, natural ecology, production conditions, and socioeconomics. The focus is on the spatial and temporal characteristics of the effects of drought, flood, wind hail, snow disaster, earthquake, and geological hazards on FSR. The results of this study highlight the spatiotemporal heterogeneity of FSR and its differential response to different hazards: 1) the assessment of FSR accurately pinpoints areas of low resilience at both spatial and temporal scales, and reveals a significant temporal phase characteristic, with most regions experiencing a notable increase in FSR since 2010; 2) High-intensity droughts and snow disasters, and frequent floods and wind hails significantly affect FSR, so it is necessary to plan targeted measures to mitigate the negative impacts of different natural hazards on food production systems. Our findings enhance the understanding of the relationship between food systems and natural hazards, and contribute to the planning of measures to safeguard food system stability and improve food system resilience to natural hazards in natural hazard-prone regions.