Assessing the sensitivity of global flood loss estimates to terrain data, defences, and climate change.

Global flood catastrophe models underpin decisions in insurance, infrastructure planning, and climate adaptation, yet they integrate multiple uncertain components, including terrain representation, flood defences, and climate-driven hazard changes. While each of these elements is known to influence flood risk estimates, there is limited quantitative evidence on their relative importance in controlling loss outcomes at global and regional scales. 

Here we apply global sensitivity analysis to a large-scale flood catastrophe modelling framework to assess how loss estimates respond to key modelling and data choices. We systematically vary terrain data type (ASTER/SRTM-derived DSM versus LiDAR-derived DTM), terrain resolution (30 m and 5 m), flood defence representation (defended and undefended views, legacy and updated defence datasets), and climate-driven event sets (baseline, 2°C, 4°C, and 6°C warming scenarios). The analysis is conducted across multiple geographic contexts, including Canada, South Africa, Slovakia, and Germany, capturing a range of topographic, vegetative, and urban conditions. 

In our presentation, we highlight the role of sensitivity analysis in flood catastrophe modelling, with a particular focus on terrain data representation. We examine how the choice of terrain data, specifically the transition from DSM to LiDAR-derived DTM, influences variability in modelled flood losses, and how this sensitivity compares with other key assumptions, including climate warming scenarios and flood defence representation. By considering these interacting sources of uncertainty side by side, we demonstrate the value of a multi-parameter sensitivity framework for understanding and prioritising model development choices in flood risk assessment.

These findings demonstrate the value of sensitivity analysis for prioritising data investment and model development in global flood risk modelling. In particular, they suggest that improvements in terrain data quality can yield disproportionately large benefits for loss estimation, with implications for risk pricing, adaptation planning, and climate resilience assessment.