山洪灾害是全球范围内区域社会经济发展面临的重大挑战之一。中国东南沿海山地丘陵区极易受到山洪灾害的影响，尤其随着气候变化和社会经济不断增长，将面临着山洪灾害带来的持续性威胁。科学分析历史灾害事件是了解当前灾害风险水平及其未来变化的关键。开展气候变化背景下山洪灾害风险评估及其影响因素研究，有助于山洪灾害易发区防灾减灾措施及可持续性发展规划的管理与制定。

本文基于自然灾害风险评估框架，借助数值模拟、实地调研、情景设定、地理空间统计等方法，以台风－极端降雨－山洪灾害链为例对中国东南沿海山洪灾害多发区——浙江永嘉县岩坦溪流域开展了气候变化背景下山洪灾害风险评估。首先，借助水文水动力模型构建区域化危险性评估框架，获取山洪灾害危险性变化特征；然后，以资产量作为区域承灾体价值，结合土地利用数据对其进行空间化处理，根据山洪灾害经济损失评估模型形成风险评估；进而，通过RCP4.5—SSP2和RCP8.5—SSP5组合，解析未来情景下（2052 - 2100）相同事件发生概率的山洪灾害风险变化。最后，在考虑山洪灾害特征的基础上，构建山洪灾害风险影响因子指标体系，并利用地理探测器对不同情景下山洪风险变化的主要影响因素及其影响力大小进行探测。本文主要结论如下：

（1）小流域范围内的山洪灾害风险变化可通过经济损失来定量化表达。根据自然灾害风险评估理论，将过程降雨重现期定义为山洪灾害事件的发生概率，相同事件发生概率下经济损失即为研究区所要面临的山洪灾害风险。本文构建了过程降雨量、流水冲击力、水流速度、人口密度、GDP密度、资产密度、人均资产拥有量、人均GDP和地区城镇化9个影响山洪灾害风险的可能因素。

（2）气候变化情景下（2052 - 2100）岩坦溪流域内极端降雨显著增强。通过时空要素分析，1971 - 2019年内共有153场台风对岩坦溪流域产生影响，对研究区有降雨贡献的6个气象站点的台风降雨量均值在7.94~15.05 mm之间，降雨强度在3.06~4.53 mm/h之间，据此得到研究区极端降雨特征。基于单个气象站点降雨特征对气候模式降雨进行校正，结果表明在未来情景下（2052 - 2100），6个气象站点相同重现期降雨量在RCP4.5情景中增加24.41%，RCP8.5中增加35.50%。

（3）气候变化情景下（2052 - 2100）岩坦溪流域山洪灾害风险变化呈中（2.00－3.00 m）、高（3.00－4.00 m）危险性趋势发展。选择岩坦溪流域内的2005年山洪灾害事件作为极端灾害事件代表，此灾害事件的过程降雨量为244.89 mm，超过流域内1500 a一遇降雨量的34.16 mm。HEC—HMS与FLO—2D耦合模型输出的淹没结果与实际淹没的相关系数（R²）为0.75，均方根误差（RMSE）为0.66。RCP4.5和RCP8.5情景下，相同山洪灾害事件发生概率的降雨量增加22.93%和34.57%。2005年山洪灾害事件影响面积为36.44 km²，未来情景下（2052 - 2100），岩坦溪流域内山洪灾害影响范围分别较历史情景（2005）增加3.41%和7.10%。在历史情景、RCP4.5和RCP8.5情景下淹没深度> 2.00 m的区域占比分别为8.59%，25.32%和38.36%，而淹没深度< 2.00 m的区域占比分别91.14%，74.68%和61.64%。

（4）岩坦溪流域内面临暴露风险的人口与资产量平稳化增加。将产业资本量与土地利用类型相结合，通过自上而下的方法获取承灾体空间化分布特征，依据永嘉县资本存量空间化结果，岩坦溪流域内共有24.18亿元资本暴露于2005年的山洪灾害事件中，占永嘉县暴露资本总量的3.14%。通过RCP4.5—SSP2和RCP8.5—SSP5组合得到的气候变化情景、经济变化情景和联合变化情景中，岩坦溪流域内暴露于山洪灾害风险的资产量分别较历史情景（2005）增加36.13%~51.20%、84.56%~90.34%和90.14%~94.76%。根据未来的人口预测数据，在21世纪末永嘉地区户籍人口城镇化率将达到10.97%~14.41%，较2010年地区增加0.62%~4.06%，意味着将有更多的社会经济要素暴露于山洪灾害风险中。

（5）过程降雨量与其他影响因素的交互作用推动了山洪灾害风险变化，未来情景下人类活动因素的影响力显著增加。2005年山洪灾害事件造成了岩坦溪流域8611.0万元经济损失，在未来情景下（2052 - 2100）同等发生概率的山洪灾害风险较历史情景（2005）预计增加65.00%~98.17%。过程降雨量与其他各因子的交互影响力均值在历史情景和未来情景下分别为0.140、0.353，GDP密度、人口密度单个因子对山洪灾害风险的影响力较历史情景增加了0.145，0.150。

Flash floods are one of the major challenges to regional socioeconomic development. The mountainous and hilly areas of southeastern China are vulnerable to flash floods, and the region will face a constant threat from flash floods, especially in context of climate change and rising socioeconomic growth. The analysis of historical disaster event is a critical step in understanding current rank of the risk and its changes over time. Conducting research on changes in flash flood risks and the driving factors in the context of climate change can help with disaster management and the formulation of disaster prevention and mitigation measures, as well as sustainable development planning in flash flood-prone areas.

Based on a natural disaster risk assessment framework, this paper used numerical simulation, field research, scenario setting and geospatial statistics, and took tropical cyclones–precipitation–flash floods as an example to carry out a risk assessment of flash floods under climate change in the Yantanxi River Basin of Yongjia County, Zhejiang Province, southeastern China, which is prone to flash floods. Firstly, a regionalised hazard assessment framework was constructed with the help of a hydrological-hydraulic model, in order to obtain the characteristics of inundation changes of flash flood. Secondly, carried out the spatial calculation and analysis of asset value and population in the region, and assessed the flash flood risk through the regional damage model. Thirdly, combining the RCP4.5—SSP2 and RCP8.5—SSP5 scenarios to project changes in flash flood risk with the same probability occuring in future scenarios (2052 - 2100). Finally, constructed an indicator system of flash flood risk impact factors is based on the characteristics of flash floods, and the main driving factors and their impact on the change of flash flood risk under different scenarios were detected using the GeoDetector. The main conclusions are listed as follows:

(1) Economic damages can be used to quantify the changes in flash flood risk at a river basin scale. According to the natural disaster risk assessment framework, the rainfall return period was defined as the probability of a flash flood disaster event occurring, and the flash flood risk indicates the variation of economic damage under a certain probability of occurrence. In this paper, nine driving factors of the flash floods risk were constructed: process rainfall, flow impact, flow velocity, population density, GDP density, asset density, per capita asset value, per capita GDP and urbanization rate.

(2) Extreme rainfall in the Yantanxi River Basin is significantly increased under the climate change scenario (2052 - 2100). Through spatial and temporal element analysis, a total of 153 typhoons impacted the Yantanxi River Basin between 1971 and 2019, and the mean values of typhoon rainfall at the six meteorological stations which contributing to rainfall in the study area ranged from 7.94 to 15.05 mm, with rainfall intensities ranging from 3.06~4.53 mm/h, stronger than Non-typhoon rainfall by 55.81%~72.42% and 26.66%~36.20%. The correction of the climate model rainfall based on the rainfall characteristics of individual meteorological stations, and the result indicated that under the future scenario (2052 - 2100) the rainfall with same return period at six meteorological stations increased by 24.41 % in the RCP4.5 scenario and 35.50 % in the RCP8.5 scenario.

(3) Under the climate change scenario (2052 - 2100), the change of flash floods risk in the Yantanxi River Basin with a high-hazard trend. The 2005 flash flood event within the Yantanxi River Basin was selected as a representative extreme disaster event, and the rainfall is 244.89 mm during this period, and exceeding the 1500a rainfall within the area by 34.16 mm. The correlation coefficient (R²) between the inundation results output from the coupled HEC-HMS and FLO-2D models and the actual inundation was 0.75, and the root mean square error (RMSE) was 0.66. Under the RCP4.5 and RCP8.5 scenarios, the rainfall increases by 22.93% and 34.57% with the same probability of flash flood events, and the impact area of flash flood in the Yantanxi River Basin increases by 3.41% (RCP4.5) and 7.10% (RCP8.5) respectively compared to the historical scenario (2005). Under the historical, RCP4.5, and RCP8.5 scenarios, the percentages of areas with inundation depth > 2.00 m are 8.59%, 25.32%, and 38.36% respectively, while the percentages of areas with inundation depth < 2.00 m are 91.14%, 74.68%, and 61.64%, respectively.

(4) The amount of exposure within the Yantanxi River Basin shows a smoothed increasing trend. By combining the industrial asset value with land use types and obtaining the spatialised distribution characteristics of exposure through a top-down approach. According the results of the spatialisation asset value in Yongjia County, the total of 2.42 billion of asset was exposed to the 2005 flash flood event within the Yantanxi River Basin, accounting for 3.14% of the total exposed asset in Yongjia County. In the climate change scenario, economic change scenario and combine change scenario that obtained from the combination of RCP4.5—SSP2 and RCP8.5—SSP5, the amount of assets exposed to flash flood risk within the Yantanxi River Basin increased by 36.13%~51.20%、84.56%~90.34% and 90.14%~94.76% respectively compared to the historical scenario (2005). Based on future population projection data, the urbanisation rate will reach 10.97%~14.41% by the end of the 21st century in the Yongjia region, an increase of 0.62%~4.06% compared to 2010, indicates that more socioeconomic factors will be exposed to the risk of flash flood disaster.

(5) Changes in flash flood risk are driven by the interaction of process rainfall with other impact factors, and the influence of human activity factors increases significantly under the future scenarios (2052 - 2100). The 2005 flash flood event resulted in 86.11 million in economic damage in the Yantanxi River Basin, and flash flood risk with the same probability occurring under the future scenario is expected to increase by 65.00%~98.17% compared to the historical scenario (2005). For the historical and future scenarios, the mean impact of the interaction effects of process rainfall with other factors are 0.14 and 0.35, respectively, and the impact of individual factors of GDP density and population density on the flash flood risk are 0.140 and 0.353, respectively.