在全球气候变暖和经济全球化背景下，全球暴雨洪涝灾害损失呈现增长的趋势。中国是受气候变化影响较为严重的国家之一，特殊的地理环境以及人口的快速增长、经济与城市化的快速发展，导致中国具有较高的暴雨灾害风险。开展中国暴雨影响人口风险评估，可为中国应对气候变化减轻人口风险提供依据。 根据区域灾害系统理论，本文从暴雨致灾因子危险性、承灾体脆弱性以及暴露性三个方面，结合未来情景数据，对中国暴雨影响人口风险进行了系统评估和分析。本文基于历史气象数据对暴雨变化趋势值的空间格局进行了分析，并且利用典型浓度路径（RCPs）下的气候模式模拟数据对中国未来的暴雨进行了时空分析；同时，基于辅助变量的人口预测模型，结合中国二孩政策和共享社会经济的参考路径情景（SSPs），研究了中国暴雨人口暴露性；选择暴雨影响人口作为因变量，年暴雨量、区域人口、高程和河网长度等因子作为自变量，基于广义相加模型，构建影响人口的脆弱性模型，在此基础上结合未来气候变化和人口发展情景开展了中国暴雨影响人口风险定量评估，结果表明： 在暴雨区，除河北、河南、四川盆地以及内蒙古西部等地区的年暴雨量有减少趋势外，其余相当多的地区年暴雨量有增多趋势；在RCP4.5和RCP8.5情景下，2030s和2050s的中国年暴雨量和年暴雨频次均由东南沿海地区向西北内陆地区逐步减少。从2000s到2030s和2050s，年暴雨量大于50mm县域的总面积呈现缓慢增加的趋势，但年暴雨量大于250mm县域总面积在RCP4.5情景下呈现先减再增的趋势，在RCP8.5情景下呈下降趋势，表明未来年暴雨量极端高值区域有所减少；两种情景下年暴雨频次大于3的区域面积同年暴雨量大于250mm的区域面积有相同的增减趋势。 在RCP4.5-SSTC和RCP4.5-SRES B2两种组合情景下，2000s到2050s，中国暴雨区总人口增加；2000s到2030s和2050s，年暴雨量大于250mm的区域，其总人口数呈现先降后升的趋势。 影响人口脆弱性模型显示了解释因子与影响人口之间的非线性关系。整体上影响人口数随着年暴雨量的增大而增加，随高程增加呈现先增后减的趋势，随人口的增多呈增加趋势但起伏较大，随河网长度的增加而增加但有较小波动。 未来中国暴雨影响人口较多的地区集中分布在浙江、福建、广东等东南沿海地区；在RCP4.5-SSTC和RCP4.5-SRES B2两种组合情景下，从2000s到2030s和2050s，中国暴雨总影响人口数呈现先降后增的趋势，2050s暴雨总影响人口数仍低于2000s。这与年暴雨量大于250mm区域面积的变化趋势一致，结合暴雨危险性、暴露性和脆弱性，推测中国总影响人口水平的高低很大程度上取决于年暴雨量大于250mm区域的面积及其人口总数。 基于未来预估情景的暴雨灾害影响人口风险评估，同时考虑了致灾因子和承灾体的变化和未来情景，为灾害风险评估提供了一种新的思路。以影响人口作为灾害损失，考虑致灾因子危险性、环境胁迫、社会经济水平几个方面的要素，应用广义相加模型构建的暴雨影响人口脆弱性曲线，能反映解释变量与响应变量之间的非线性关系，较以往直接基于历史致灾因子强度和对应损失关系而言，考虑因素更为全面。论文结果可为区域人口风险防范提供决策依据。

In light of global warming and economic globalization, the loss of rainstorm floods has become increasingly. China is one of the countries significantly influenced by climate change, the special geographical environment and the rapid growth of population, economy and urbanization make the country to be seriously affected by rainstorm. The affected population risk assessment to rainstorm can provide the basis for China to mitigate the risk of population in response to climate change. In this study, the affected population risk to rainstorm in China is quantitatively assessed and analyzed by considering the hazard, exposure and vulnerability based on histirical data combining with future scenario. Firstly, this thesis analyzes the spatial pattern of rainstorm tendency based on historical meteorological data, and then the spatial-temporal analysis of future rainstorm is conducted by using climate model simulation data under the typical concentration path scenarios (RCPs). At the same time, combining with China's two-child policy and shared socio-economic reference path scenario (SSPs), the population exposure within the rainstorm hazard area is studied based on an auxiliary variable-based population prediction model. Then, the affected population is selected as dependent variables, and annual rainstorm, regional population, elevation and river network length as independent variables. The relationship between rainstorm and affected population is carefully studied, and a vulnerability model for the affected population is constructed based on the generalized additive model (GAM). Finally, the affected population risk to rainstorm is quantatively assessed and mapped. The results are as following: Most of regions in the rainstorm hazard areas expect Hebei, Henan, Sichuan, and western Inner Mongolia have an increasing trend of annual rainstorm precipitation. Under the RCP4.5 and RCP8.5, both the annual precipitation and frequency of rainstorm gradually decrease from the southeastern coastal areas to the northwestern inland region in the 2030s and 2050s. The total area of counties with annual rainstorm precipitation over 50 mm shows a slow increase trend from 2000s to 2050s. However, the total area of counties with annual rainstorm precipitation over 250 mm shows a decrease trend from 2000s to 2030s and an increase trend from 2030s to 2050s underRCP4.5, and a decrease from 2000s to 2050s trend under RCP8.5. The area with rainstorm frequency greater than 3 has similar change tendency with annual rainstorm precipitation greater than 250 mm in the two scenarios. Under two combined scenarios of RCP4.5-SSTC and RCP4.5-SRES B2, the total population in rainstorm area in China will increase from 2000s to 2050s. From 2000s to 2030s and 2050s, the number of the total population in the area with annual rainstorm greater than 250mm decreases from 2000s to 2030s and then increase from 2030s to 2050s. The vulnerability model of affected population shows a non-linear relationship between the interpretive factor and the affected population. As a whole, the affected population shows an increasing trend with the increase of annual rainstorm, and first increase and then decrease trend with the increase of elevation, and an increasing trend and large fluctuation with the increase of the population, and an increasing trend and slight fluctuation with the increase of river length. In the future, counties with larger number of affected population risk will be distributed in the southeast coastal areas, such as Zhejiang, Fujian, Guangdong. In the two combined scenarios, the total affected population to rainstorm of China has a decreasing trend from 2000s to 2030s and an increasing trend from 2030s to 2050s, while the total affected population of 2050s is still lower than 2000s. This is consistent with the trend of the change of the area with annual rainstorm precipitation greater than 250mm. Considering with the hazards, exposure and vulnerability, the levels of affected population to rainstorm in China may depend largely on the area and the total population of counties with annual rainstorm precipitation greater than 250mm. The affected population risk assessment to rainstorm considers both the change of hazards and exposure in the future, which provides a new idea for disaster risk assessment. The vulnerability model which comprehensively considers the factors of hazards, environment and social economic level based on generalied additive model, is more resonable to relect the relationships between disaster loss and their affected factors compared with the vulnerability based on the relationship between loss and hazard. The results of this thesis can provide a good reference for populationr risk management and governance.