Can spatio-temporal characteristics of precipitation affect the critical rainfall in mountainous watersheds of the eastern Tibetan Plateau?

Accurate forecasting of flash floods is critical for regional disaster prevention and mitigation. However, the high spatio-temporal heterogeneity of precipitation in complex mountainous terrain poses a significant challenge to conventional flash flood warning methods. This study therefore investigates the influence of precipitation spatio-temporal structure on critical rainfall thresholds. Gauge-calibrated satellite quantitative precipitation estimation (QPE) products demonstrated reliable performance in hydrological simulation. We therefore used these products to determine the critical rainfall of Flash Flood Guidance (FFG) and evaluate their binary classification accuracy across various watersheds over the eastern Tibetan Plateau. Results show that for watersheds smaller than 1000 km², critical rainfall perform well, particularly at 3‑hour and 6‑hour durations, with Critical Success Index (CSI) values exceeding 0.7. For larger watersheds (>1000 km²), better performance is observed at 6‑hour and 12‑hour durations, yielding CSI values around 0.6. Furthermore, compared to other spatio-temporal precipitation metrics, the spatial variability of precipitation (Cvp) exhibits the most significant impact on peak discharge, especially in larger basins. Incorporating Cvp as a classification metric improves the accuracy of binary classification and reduces the number of missed flood events.This study enhances the precision of critical rainfall classification, thereby improving the effectiveness of flash flood forecasting triggered by short-term heavy rainfall and providing scientific support for related emergency decision-making.