洪水是全球最严重的自然灾害之一，每年都会造成严重的人员伤亡和经济损失，威胁着人类的生命安全和社会的发展。随着全球气候变暖，极端降水的频率和强度在全球大部分地区逐渐增加，由极端降水引起的洪水频率和强度也随之发生变化，加之世界人口的增长与社会经济的发展，全球变化导致的洪水灾害风险也进一步增加，探究未来洪水灾害死亡人口风险对区域洪水风险管理和防灾减灾工作具有重要的意义。因此，本文利用未来不同代表性浓度路径（RCP4.5和RCP8.5）下洪水淹没数据、未来不同共享社会经济路径（SSP2和SSP5）下的人口预估数据、历史时期的洪水损失数据等对未来全球洪水的危险性、洪水人口暴露度和脆弱性进行了分析，并进一步评估了未来不同组合情景（RCP4.5-SSP2和RCP8.5-SSP5）下的河道型洪水灾害人口死亡风险并分析了不同要素对风险的贡献率水平。

论文基于CaMa-Flood模型得到的全球洪水淹没数据，对不同淹没水深下的栅格数量及其相对于历史时期的变化进行了分析；综合淹没深度和淹没范围信息，用洪水总量指标，在大洲、国家（地区）和流域三种单元上定量分析了2016–2035年（记为2030s）和2046–2065年（记为2050s）RCP4.5和RCP8.5情景下的全球河道型洪水危险性及其相对于历史时期的变化。根据GPW人口网格数据，对ISI-MIP人口数据进行降尺度，并结合洪水淹没数据，在三种地理单元上定量分析了RCP4.5-SSP2和RCP8.5-SSP5组合情景下2030s和2050s的年均洪水人口暴露度。基于已有的脆弱性曲线和历史洪水淹没数据，估算历史时期（1986-2005年）的死亡人口并统计到国家（地区）单元，对比估计的死亡人数与历史统计死亡人数之间的无偏性和误差，选择无偏性和误差较小脆弱性曲线的为基准；然后在国家（地区）单元上构建调整系数K_c，对基准脆弱性曲线进行调整，从而构建适用于各个国家（地区）的脆弱性曲线。基于未来的洪水淹没数据、未来人口预估数据和脆弱性模型，估算了未来两种组合情景下的死亡人口，用情景分析法分解了气候要素变化、人口变化对风险变化的贡献水平，并在大洲和国家（地区）单元上进行了分析。结果表明：

（1）在同一时段内，RCP8.5和RCP4.5情景下洪水危险的空间分布差距不大，但RCP8.5情景下的危险性高于RCP4.5情景，高危险区域主要分布在亚洲和非洲；2030s和2050s的年均洪水危险性相对历史时期分别增加了约0.14~0.19倍和0.22~0.34倍，年均危险性降低的区域主要在欧洲和非洲，年均危险性增加幅度较大的区域主要在亚洲；淹没水深在1~2米之间的栅格数量最多，约占淹没栅格总数的26%。

（2）2030s在RCP4.5-SSP2和RCP8.5-SSP5情景下，全球年均洪水暴露人数分别约为4460万人和4360万人，相对历史时期分别增加了0.90倍和0.86倍；到2050s两种情景下的年均暴露人数分别约为6470万人和6200万人，相对于历史时期分别增加了1.76倍和1.64倍；淹没水深在3~4米之间的暴露人数最多，约占暴露总人数的21%；高暴露度及相对于历史时期暴露度增加幅度较大的地区位于亚洲和非洲，相对于历史时期暴露度降低的区域主要在欧洲。

（3）2030s在RCP4.5-SSP2和RCP8.5-SSP5情景下，全球洪水年平均死亡人数分别约为1.05万人和0.99万人，相对历史时期的年平均死亡人数分别增加了1.05倍和0.93倍；到2050s两种情景下的年均死亡人数分别约为1.48万人和1.64万人，相对于历史时期分别增加了1.89倍和2.20倍；高风险区域主要位于东亚、东南亚和南亚；相对历史时期风险增加幅度较大的区域位于南亚、东南亚、北美洲和非洲东西海岸小部分地区，风险降低的区域主要位于欧洲和非洲；全球洪水死亡人口风险变化主要受气候变化影响较大，贡献率在38%~55%之间，其次是气候-人口共同作用，人口变化影响最小，贡献率在11%~25%之间。

   本文的创新之处在于构建了不同国家（地区）单元上的脆弱性曲线并利用脆弱性曲线在不同地理单元上系统地评估了河道型洪水死亡人口风险，改善了目前洪水风险研究中死亡人口风险较少且脆弱性曲线缺乏的问题。论文开展的全球河道型洪水死亡人口风险评估工作，可为未来气候变化背景下全球洪水灾害风险评估提供典型案例，研究结果可为全球洪水灾害风险防范提供科学依据。

Flood, which is one of the most serious disasters in the world and has caused huge mortality and economic losses in the past decades greatly threatens the development of human society. With the global warming, the frequency and intensity of extreme precipitation show an increasing trend in most parts of the world, and the frequency and intensity of floods caused by extreme precipitation also change accordingly. The total population of the world also shows an increasing trend. It is of great significance for risk management and disaster prevention and reduction to explore the impact of future flood disasters on population death. Therefore, based on future flooding inundation dataset for different representative concentration pathway (RCP4.5 and RCP8.5) produced by CaMa-Flood model, projected population dataset for different shared socioeconomic pathways (SSP2 and SSP5), historical deaths data and other data, the paper analyzed the flood hazard, population exposure and vulnerability, and assessed future mortality risk for different combination scenarios (RCP4.5-SSP2 and RCP8.5-SSP5) as well as analyzing the contribution rates. The main research work and results of this paper are as follows:

The outliers of the flood inundation data produced by the CaMa-Flood model were corrected by masking method and threshold method. Based on the corrected inundation data, the flooding hazard for RCP4.5 and RCP8.5 from 2016 to 2035 (shorted as 2030s) and from 2046 to 2065 (shorted as 2050s) was quantitatively analyzed at the units of continent, country and river basin by composite index considering both depth and fraction factors. The ISI-MIP population data was downscaled according to GPW data. Then based on the corrected flood inundation data and the downscaled future population estimation data, the population exposure for RCP4.5-SSP2 and RCP8.5-SSP5 scenarios in 2030s and 2050s were analyzed quantitatively in different geographical units using the total population number index. Based on the existing vulnerability function, historical flood inundation data and population data, the annual death tolls were estimated in historical period (1986-2005) and national deaths were computed. Then the error between estimated deaths and statistical deaths was calculated so as to select the optimal vulnerability curve. Finally, adjustment coefficient K_c was constructed by estimated deaths and statistical deaths in country units and adjusted the selected vulnerability function. Adjusted vulnerability function was used for risk assessment. Based on the future flood inundation data, the future projected population data and the adjusted vulnerability function, future death tolls were estimated for RCP4.5-SSP2 and RCP8.5-SSP5 in 2030s and 2050s.

(1) The spatial contribution of flood hazard is similar between RCP4.5 and RCP8.5 scenario during a period. However, the hazard of RCP8.5 scenario is higher than that of RCP4.5. The high hazard areas are mainly distributed in Asia and Africa. The flood hazard increases by 0.14~0.19 times in 2030s and 0.22~0.34 times in 2050s relative to historical period. Hazard reduces mainly in Europe and Africa and increases mainly in Asia. There are about 26% of grids inundated with water depth between 1~2 meters.

(2) The results show that: there are 4.45×10⁷ persons exposed for RCP4.5-SSP2 and 4.36×10⁷ persons for RCP8.5-SSP5 in 2030s. They increase by about 0.90 times and 0.86 times respectively compared with the historical period. By 2050s, about 6.47×10⁷ persons are exposed for RCP4.5-SSP2 scenario and about 6.19×10⁷ persons are exposed for RCP8.5-SSP5 scenario, which increase by 1.76 times and 1.64 times respectively compared with the historical period. There are about 21% people are exposed with water depth between 3~4 meters. The areas with high exposures and large growth are located in Asia and Africa, and the areas with reduced exposure are mainly in Europe.

(3) The results show that global death tolls are about 10.5k persons for RCP4.5-SSP2 and about 9.9k persons for RCP8.5-SSP5 in 2030s, which increase by 1.05 times and 0.93 times respectively compared with the historical period. By 2050, the death tolls are about 14.8k persons for RCP4.5-SSP2 scenario and about 16.4k persons for RCP8.5-SSP5 scenario, which increase by 1.89 times and 2.20 times respectively compared with the historical period. The high-risk regions are mainly located in East Asia, Southeast Asia and South Asia. The regions with large risk growth are located in the East Asia, South Asia, Southeast Asia, North America and a small part of the east and west coasts of Africa. The regions with reduced risk are mainly located in Europe and Africa. The change of global mortality risk is mainly affected by climate change, with the contribution rate from 38% to 55%, followed by the climate-population joint change. Population change causes least impact, with the contribution rate from 11% to 25%.

       The innovation of this paper is to construct the vulnerability curves at country units and evaluate systematically the mortality risk from river flood at different geographical units by using the vulnerability curves. It enriches the flood risk assessment research works, especially the study of vulnerability of mortality risk from river flood. The global mortality risk assessment of river flood can provide a typical case for future global flood risk assessment under climate change, and the results in this paper can provide a scientific basis for global flood risk prevention.