Unraveling Opportunities and Challenges for Impact-Based Forecasting and Warning of Flash Floods and Surface Water Flooding

Flash floods and surface water flooding triggered by short-duration, high-intensity precipitation account for a substantial share of flood-related damage in Switzerland. Their rapid evolution in time and space makes them particularly difficult to represent in operational warning systems and poses important challenges for emergency management. In the context of the ongoing shift from hazard-based warnings toward more impact-based forecasting and warning approaches, this study explores how past events can support the development of warning products with higher operational relevance for stakeholders. Initial meetings with Swiss first responders helped to identify events ranging from purely surface-runoff-driven to more flash-flood-dominated events. These reference cases provide the basis for simulating such events using both observed and forecast precipitation data. Flood hazard is represented through two approaches: the physics-based hydrodynamic model LISFLOOD-FP and a data-driven approach that estimates precipitation return periods at the watercourse level and combines them with national and local flooding intensity maps to infer flood extent. Both are applied using the same event-specific precipitation forcing. The resulting hazard outputs are first assessed qualitatively in collaboration with Swiss first responders, including flood experts and representatives from local communities. In a subsequent step, impact estimates are derived for each approach by combining hazard outputs with building and road exposure data and a regional vulnerability curve for Switzerland. This allows us to assess differences between the hazard outputs, examine whether expressing results in terms of impact increases their usefulness for stakeholders, and evaluate the respective strengths and limitations of the different modelling approaches. Initial results indicate that model performance depends on the dominant event process, while the data-driven approach has limitations in capturing small events and small-scale effects. For one event in the Canton of Zurich, the resulting impact estimates are of the same order of magnitude as insurance records. These initial findings highlight both the potential of integrating hazard and impact information and the challenges that remain in developing more operationally relevant impact-based forecasting and warning approaches.