Analysis Of Spatiotemporal Characteristics Of Tropical Cyclones Affecting China And Construction Of An Assessment Mode

In the context of global warming climate,tropical cyclones as a frequent extreme weather event have brought serious disaster impacts to China,and it is of great research significance to statistically analyze tropical cyclone disasters,and constructing tropical cyclone post-disaster models can improve the ability to respond to tropical cyclone disasters,reduce post-disaster losses,and accelerate recovery and reconstruction efforts in the affected areas.This paper analyzes the spatial and temporal characteristics of tropical cyclone population,asset exposure and disaster loss indicators(number of deaths,area of flooded farmland,direct economic loss)affecting China based on different data sets provided by various official agencies,and constructs a classification model and a quantitative model of tropical cyclone population exposure and disaster loss indicators direct economic loss affecting the Chinese region.The main findings of the full paper are as follows:(1)Regional exposure to tropical cyclone hazards in China varied in the context of tropical cyclones of different intensities during 1950-2015.The number of tropical cyclone exposure events,population exposure and asset exposure affecting China all show a decreasing trend,but the population and asset exposure slowly increase when only China is affected by tropical cyclones;the average frequency of tropical cyclone movement paths in the northern South China Sea and southeastern Taiwan waters shows an increasing trend with the chronological change;the regression analysis of the population and asset exposure of tropical cyclones in China shows that the relationship between the two There is a strong linear relationship between the two.(2)Typhoon disasters during 1993-2016 mainly occurred in April-November,and the active peak season July-September resulted in three types of disasters with a high percentage of fatalities,inundated farmland area,and direct economic losses of 80% and above,but all three types of disasters showed a decreasing trend;the degree of local disaster was closely related to regional economic development and landfall frequency.In addition,the changes in the average percentage values of typhoon disaster losses in the coastal region in each era are inextricably linked to the number and intensity of landfalling typhoons.(3)A CART classification model and a BP neural network regression model were constructed for the population exposure of tropical cyclones affecting the Chinese region.The prediction accuracy of the CART model for the training and test sets was 76.63% and 75.31%,respectively;the goodness of fit of the BP neural network with the structure of 14-5-7-1 for the training and test sets was about 82.64% and 68.46%.(4)A combined classification model and a BP neural network regression model were constructed for the direct economic losses caused by typhoons.The average prediction accuracies of the combined classification model for the training and test sets were about 86%and 67% after cross-validation;the fitting accuracies of the BP neural network model with structure 13-5-7-1 for the training and test sets were about 95% and 80%.