Stress testing approaches for High-Impact-Low-Probability floods: A review

In Central Europe, climate change contributes to an increasing frequency of devastating flood events, such as those observed in Western Europe in July 2021. While floods with return periods of up to 200 years have been relatively well studied, understanding and preparedness for more extreme, less frequent High-Impact-Low-Probability (HILP) floods remain limited. A key tool to assess the potential consequences of such events is the stress-test scenario. More specifically, these are hypothetical yet plausible simulations of a very low-likelihood flood events.

This review systematically analyses scientific studies that apply stress-testing approaches to HILP floods. The focus lies on research examining the impacts of very extreme pluvial and fluvial floods on humans, the built environment, and critical infrastructure. A systematic SCOPUS keyword search initially identified ~12,000 studies, which were reduced to 137 relevant publications using a filtering process assisted by a large language model. The selected studies are differentiated by how physical boundary conditions are derived, floods are modeled, and impacts are quantified.

The analysis shows that most studies use univariate statistical methods to derive hypothetical rainfall events, while more complex approaches such as climate model reforecasts or multivariate weather generators are employed far less frequently. A wide range of techniques is used to modify historical events to simulate unprecedented flooding. Counterfactual scenarios in flood modeling mainly investigate the effects of reservoirs and similar structures, whereas other simulations explore the potential of early-warning systems to reduce exposure. In terms of impact modeling, the reviewed literature examines a broad range of system components. About 60% of studies employ simple GIS overlays to assess the number of structures affected by floodwaters, while more advanced modeling tools include agent-based models, cascading impact models, network theory, and multi-criteria decision models. Only a few studies assess multi-sectoral impacts, and their analyses are often shallow or overly simplified. Future research should address this gap to achieve a more comprehensive understanding of the potential damages caused by HILP floods.