Spatial Distribution Prediction of Flash Flood Disaster in Longnan City Based on Particle Swarm Algorithm Combined with MaxEnt Model

Objectives: Flash floods are natural disasters caused by the sudden rise of water levels in mountainous rivers, which are transient and destructive in nature. In recent years, the frequent occurrence of flash floods in Longnan City has posed a serious threat to the lives and property safety of local people, and it is urgent to carry out flash flood risk assessment in Longnan City. Methods: This study takes Longnan city as the study area, and uses the MaxEnt model combining by the particle swarm algorithm to evaluate the vulnerability of the study area based on the 834 flash flood hazard points investigated and 32 disaster-caused factors, and also predicts the spatial pattern changes and mass migration trends of the future flash flood potential vulnerability areas based on three periods of climate data in the 2050s (2041-2060), 2070s (2060-2080) and 2090s (2081-2100). Results and Conclusions: The spatial distribution of flash flood prone areas in the study area obtained by using MaxEnt model to obtain the spatial probability of flooding and combining it with Particle Swarm Algorithm (PSO), the AUC value under the ROC curve of the results of the study in each period is greater than 0.85, which indicates that the precision of the results of the method in this paper is good; The combined contribution, replacement significance and Pearson correlation coefficients and causal response curve analyses identified the main causal factors in this study area as the driest month precipitation, monthly mean diurnal temperature difference, coefficient of variation of precipitation, warmest month maximum temperature, land use, distance from the river, soil texture, profile curvature, elevation, and topographic relief; The flash flood-prone areas in the study area varied in different periods, but were mainly distributed in parts of Wudu District, Wen County and Tanchang County, and the simulation results for the three future periods (2050s, 2070s and 2090s) reflected a decreasing trend compared with the current (2030s) period.