Flash flood dynamics in complex terrain system and its impact mechanism based on non-linear regression and...

Flash floods pose significant threats to lives and property, especially in complex terrain regions. This study aims to reveal the spatial distribution, dynamics, and driving factors of flash floods in Yunnan Province, China, using disaster point survey data (since 1955), geodetector analysis, and function fitting simulations. Results indicate that flash floods are concentrated in central Yunnan between 24°N and 27°N, particularly in areas with elevations of 1500-2000 m, slopes ≤25°, and annual rainfall between 680 and 1410 mm. Nonlinear curve-fitting analyses revealed two distinct driving mechanisms. Precipitation and anthropogenic factors are best represented by a double-exponential model (y = a ∗ e ^b ∗^x + c ∗ e ^d ∗^x), which delineates a "maximum sensitivity window" where flood probability peaks near specific thresholds. In contrast, surface environmental factors follow a power-law decay window (y = l ∗ x ^m + k), indicating a risk-decay mechanism in which susceptibility decreases as factor intensity increases. The influence of driving factors decreases in the order: disaster-causing factors, disaster-bearing body factors, and disaster-prone environment factors. Synergistic effects among factors exhibit nonlinear enhancements. In northern watersheds, flash floods are primarily driven by the combined effects of precipitation and surface conditions; in central watersheds, precipitation and human activities dominate, while precipitation alone is the primary driver in southern watersheds. This study provides novel insights into the mechanisms and spatial dynamics of flash floods in complex terrain systems. The findings offer valuable scientific support for flash flood risk assessment and the development of mitigation strategies to safeguard vulnerable communities in Yunnan and similar regions.