Abstract: Historical sequences of Mango yield reduction caused by cold damage were calculated based on the data of yield during 1990—2010 and the data of weather of 18 cities/countries in Hainan Island from 1961 to 2010, and the distribution models of yield risk were constructed by applying extreme value theory to determine the pure premium rates of the study areas. Then, according to Mango cold damage risk and yield risk division, regional risk factors were calculated by clustering analysis to revise the pure premium rates, and the revised pure premium rates were the pure premium rates of cold damage of Mango under different deductibles in the study areas. The result showed that under different deductibles, the areas with the highest pure premium rates were Baisha, Qiongzhong, Wuzhishan and Chengmai that located in the middle and north of Hainan Island, indicating the highest cold damage risk in these areas. The pure premium rates in Danzhou, Lingao and Tunchang were slightly lower. This result agreed with the risk of cold damage in weather analysis, showing that Mango planting in the middle and northwest of Hainan Island suffered more from cold damage and was suitable for the standard rates of high deductibles. In this study, the pure premium rates were calculated based on long sequences of meteorological data, which could avoid the problem of insufficient and unstable output data, and effectively prevent the adverse selection and moral hazard. The study can provide certain references for the design of meteorological index insurance product for other tropical fruits, and encourage more farmers to be involved in the insurance.