While research on quantitative measures of vulnerability is increasing, current studies may be criticized for failing to consider the spatiotemporal dynamics of vulnerability, ignoring uncertainty in the assessment process, and lacking validation. Using the index-based method, this study quantitatively assesses the spatiotemporal dynamics of flash flood vulnerability in the Hengduan Mountain Areas (HMA) of the southeastern Tibetan Plateau. First, a vulnerability index system is constructed based on existing vulnerability assessment studies, and the uncertainty is controlled by introducing statistical and machine learning methods. Then, we explore the spatiotemporal change trend of vulnerability and the characteristics of spatiotemporal hotspots, and validate the vulnerability results with a spatial matching method based on the actual flash flood frequency and damage records. The results show that our method is effective in capturing the spatial and temporal patterns of flash flood vulnerability, and the validation shows that the assessment results can reflect the relationship between flash flood vulnerability and observed flash flood impacts. Our study highlights the importance of spatiotemporal dynamics, uncertainty control, and validation for quantitative vulnerability assessment, and proposes methods to achieve these goals. The spatiotemporal dynamics assessment of flash floods vulnerability in the HMA is an empirical application of our methods, showing the process and results of spatiotemporal vulnerability assessment. The study of spatiotemporal dynamics of flash flood vulnerability is of great significance for predicting the impacts of future flash floods, which is important for policy makers to plan flash flood prevention and response measures, reduce flash flood impacts, and improve resilience.