Assessment and Prediction of Ecological Environment Vulnerability in China Under Rainstorm

This paper deals with the prediction and assessment methods of flood disasters due to heavy rainfall, especially focusing on analyzing the extreme weather performance of ecologically fragile zones in China. Initial discussions are devoted to the discussion of the development of flood disaster prediction methods in and outside of the country, traditional methods taking support from depth loss curves together with modern applications in the realm of machine learning, remote sensing and GIS technologies. Based on that a vulnerability assessment system focusing on natural and anthropogenic factors was developed and subject to the analysis using data normalization and a random forest model. The analysis indicates that the rainfall is the most important factor affecting flooding into the disaster area, while natural factors such as forests reduce the risk of the disaster. In forecasting, the author applied a basic linear regression model along with an LSTM neural network for predicting the type of land use change and trends in rainfall in China for the next decade. Combining these predictions and a random forest model will allow for the establishment of the potential disaster-prone areas from 2025 to 2035. The predictions reveal that the southern and south-eastern coastal China are to be concerned as high-risk areas. With the integration of heat maps and model results, a multi-faceted analysis gives out the important insights into the spatial distribution characteristics of disaster-prone areas in China and its implications with respect to disaster risk assessment, management, and responsive strategies.