全球气候变化引发极端暴雨频发加之山区河道水流运动的复杂性，使得山洪预警始终是一个科学难题。目前常用于洪水预测的各类水文、水动力模型虽各有侧重，但均难以指导地形数据要求较高的小尺度山洪预警。本研究以隐潭溪为例，基于高精度地形数据构建了水文水动力耦合模型，以提高山洪预警与风险评估能力，制定隐潭溪山洪预警指标体系并探究山洪危险区、避险路径及安置点分布。研究结果表明：新安江模型模拟2015-2021年4次台风洪水事件中，径流深和洪峰流量相对误差均低于20%，确定系数超过0.7，适用性较高，且在场次洪水模拟中效果亦优；通过HEC-RAS模型开展20210721场次洪水一维与二维非恒定流耦合模拟，依托网格单元划分，探明洪水主要淹没河道低地与泛洪区；采用水位（流量）反推法构建隐潭溪山洪预警指标体系，发现临界雨量指标随预警时段延长而增加，但随前期土壤湿润度增大而减小；河道由内向外山洪风险递减，由于各村地形、水文条件等不同，使得面临洪水时的山洪危险区、避险路径及安置点存在区域性差异。针对小流域山洪预警技术短板，本研究从模型耦合、动态预警、防灾闭环三个维度优化技术方法，构建了南方山丘区小流域山洪全流程技术体系，可为同类流域防灾减灾提供技术支撑。

Frequent extreme rainstorms driven by global climate change, together with the complex flow dynamics in mountainous river channels, have rendered flash flood early warning a persistent scientific challenge. While various hydrological and hydrodynamic models commonly adopted for flood prediction each have distinct focuses, none can effectively underpin small-scale flash flood early warning that demands high-resolution topographic data. Taking the Yintanxi Watershed as a case study, this study developed a coupled hydrological-hydrodynamic model based on high-precision topographic data to enhance the capacity of flash flood early warning and risk assessment, established a flash flood early warning indicator system for the watershed, and clarified the distribution of flash flood hazard zones, evacuation routes and emergency shelters. The results demonstrate that: the Xin'anjiang Model was applied to simulate four typhoon-induced flood events from 2015 to 2021, yielding relative errors of less than 20% for both runoff depth and peak discharge with coefficients of determination exceeding 0.7, which verifies its high applicability and superior performance in event-based flood simulation; coupled 1D and 2D unsteady flow simulations were conducted for the flood event on 21 July 2021 via the HEC-RAS Model, and grid cell discretization revealed that the primary flood inundation areas were low-lying river reaches and floodplains; a flash flood early warning indicator system for the Yintanxi Watershed was constructed using the water level (discharge) back-calculation method, and the results show that critical rainfall indicators increase with the extension of the warning period but decrease with higher initial soil moisture content; flash flood risk exhibits a gradual decline from the river channel outward, and regional differences exist in the distribution of flash flood hazard zones, evacuation routes and emergency shelters among villages during flood events due to variations in their topographic and hydrological conditions. Addressing the technical deficiencies of flash flood early warning in small watersheds, this study optimized the technical methodologies from three dimensions—model coupling, dynamic early warning and closed-loop disaster prevention—and developed a full-process technical system for flash floods in small watersheds of southern hilly and mountainous areas, which can provide technical support for disaster prevention and mitigation in similar watersheds across China.