Study Of Flash Flood Disaster Warning Based On FloodArea

Natural disasters occur frequently in China.The flash flood is one of the most dangerous natural disasters.Therefore,conducting flash flood warning services has become one of the most significant and urgent tasks for governments at all levels.Besides,the analysis and calculation of rainfall threshold is the most important part of warning service because the flash flood normally happens when the area rainfall reaches or exceeds a certain critical value within a specific period of time.Hence,in order to provide the warning and forecast of mountain flood disasters with technical support,this article is devoted to effectively determining the rainfall threshold within this study area.This paper constructed two models for determining the rainfall threshold:the first one is the linear regression model which is for determining the rainfall threshold;the second one is Neural Network Model which is also to determine the rainfall threshold.The main concept of the regression model is to first use the FloodArea model to simulate the flooded area in the study area and adjust the input parameters of the FloodArea model by comparing the actual flooded data of the warning spot with the simulated flooded data.Then,using the modified FloodArea model and the statistical method where a linear regression model is built to accumulate area rainfall and inundation depth at different patterns of hourly precipitation distributions of rainfall for the warning spot,thereby calculating the rainfall threshold of the warning spot under different patterns of hourly precipitation distributions of rainfall.The main idea of the neural network model is based on the FloodArea model' s submergence data for the study area simulation.This model utilises the multi-scaled analysis of submergence depth and six factors including the rainfall,terrain slope,terrain relief,terrain roughness,digital elevation value,and plane curvature as the influencing factors of submergence depth to explore the relationship between submergence depth and these influencing factors.Firstly,during the test,BP neural network was used to fit,but it failed due to the large fitting error.Then,the genetic algorithm was used to optimize the double-layer BP neural network.After many trainings,a better prediction effect is achieved.The GA-BP neural network model not only predicts the submerged depth of each site under a given rainfall,but also obtains the rainfall threshold for each site in the study area,thereby effectively flooding the study area.Finally,the linear regression model and neural network model have been applied to determine the rainfall threshold in Anhua County,Hunan Province,and provides technical support for the early warning and forecast of the mountain flood disaster in Anhua County.