Abstract: Taking Sichuan province as the research region, this paper selects the initial evaluation indicators for social vulnerability of natural disaster from three aspects of sensitivity, coping capacity and resilience. The correlation coefficient method is adopted to filter the indicators and the variation coefficient method is applied to determine indexes' weights. The TOPSIS model is used to evaluate the social vulnerability of natural disasters in cities (prefectures) of Sichuan province. The results are manifested as follows. The social vulnerabilities of Chengdu and Panzhihua are the lowest, while those of Neijiang, Meishan and other two cities are the highest. The spatial distribution pattern is generally the fact that the social vulnerability of northwest Sichuan is lower than that of southeast Sichuan. Combining the specific situation of cities in Sichuan province, it proposes the countermeasures and suggestions of reducing social vulnerability for cities with different vulnerability levels, in order to provide scientific foundations and references for disaster prevention, mitigation capabilities and sustainable development of Sichuan province.