Author(s): Yonghong Zeng; Qingqing Ma

Linked Author(s):

Keywords: Economic loss estimate; Typhoon disaster; Principal component analysis; Support vector regression

Abstract: A recent statistics shows that typhoon is the worst hazard sources among the reasons on flood disaster in Zhejiang province, PRC. From 1987 to 2009, above half a loss of flood disaster is derived from typhoon. The direct annual economic loss induced by typhoon is up to 0.83 billion dollars and presents clear ascendant trend. IATD, an index to describe the degree of disaster caused by typhoon, is used to clarify the causes of typhoon disaster in Zhejiang province in this paper. Using principal component analysis method, the assessment factors including disaster inducing factor, disaster formative environment and disaster affected body are analyzed, and the principal component is extracted as the input of the support vector regression model, thus a statistical prediction model for disaster, including IATD, economic loss, damaged and collapsed houses and flooded farm land caused by typhoon is developed. Based on the prediction result, disaster situation grade is determined. In order to decrease the complex degree of training in support vector machine, a smoothing technique is employed to change the quadratic programming with constraints to a convex counterpart without constraints. The fitting results for historical and a single typhoon in 2019 show that the prediction model performs so well that it can be used in early warning of typhoon disaster in practice.

DOI: https://doi.org/10.3850/IAHR-39WC2521711920221312

Year: 2022