Neglecting spatiotemporal rainfall variability misrepresents flood hazard and risk

This study provides the first direct comparison of flood hazard maps and flood risk (damage) estimates derived from idealized design storms against those generated using stochastic storm transposition (SST), a probabilistic method that preserves the spatial and temporal structure of real rainfall. Flood hazard maps and risk products underpin land-use planning, infrastructure design and insurance decisions worldwide, yet most rely on design storms that assume rainfall is uniform in space and idealized in time, and that storm, flood and damage frequencies scale one-to-one.

Relative to SST, design storms were found to underestimate flood inundation frequency by 31 to 35 per cent across the watershed, including in densely developed areas such as downtown New Bern. Design storms also underestimated expected annual damage by 93 per cent ($6.8 million) and cumulative damage over 50 years by 21 per cent ($79 million), equivalent to roughly 5 per cent of total building value in the model domain. The results revealed a systematic decoupling between the frequency of rainfall, flooding and damage, with several of the most damaging simulated events driven by storms of comparatively short return periods whose intense rainfall was concentrated over urban areas or occurred later in the storm duration, after soils had begun to saturate.