Abstract: Recently, global climate anomalous changes have caused frequent storm surge disasters which cause great disaster loss to our country. It's of great significance for measuring risk, estimating storm surge disaster loss and guiding to spread disaster risk effectively to measure value at risk of storm surge disaster scientifically. In view of the characteristics of low frequency and fat tail for storm surge disaster loss data in China, this paper introduces the POT model of extreme value theory to fit the direct economic loss distribution of storm surge disaster, measure value at risk under different confidence level and pass the Kupiec failure frequency test. The study shows that the value at risk under different confidence level estimated by the POT model could become the basis of scientific measurement for storm surge disaster risk and the reasonable estimates of the maximum possible loss for single disaster in China.