H193-0002
A probabilistic methodology for pluvial flood hazard mapping with the Safer_RAIN rapid flood model
Abstract:
Synthetic design storms associated with given return periods are usually used for assessing pluvial flood hazards, assuming that the T-year design storm generates the T-year pluvial flood. However, such an assumption is incorrect, as the T-year water depth in a given location should be estimated from the frequency curve of water depths generated by a long series of possible rainfall events. Therefore, stochastic approaches should be required for pluvial flood hazard mapping.
Pluvial floods are simulated accurately in urban areas with 2D hydrodynamic models (2DHM), though they need high computation times not compatible with stochastic approaches. Rapid flood models (RFMs) can reduce computation times to few minutes or seconds, identifying depressions and their links from a digital terrain model (DTM) and using the continuity equation to simulate how such depressions fill and spill. In this study, the RFM Safer_RAIN developed under the SAFERPLACES project funded by the EIT Climate-KIC is used. The Safer_RAIN model has been benchmarked with the 2DHM IBER in Pamplona (Spain) by using 10-min precipitation fields available for three real pluvial flood events. A long set of real rainfall events has been extracted from a 15-min rainfall gauging station located close to the case study by using a POT analysis. Rainfall events have been characterised by their total rainfall and storm duration. A stochastic rainfall generator has been developed based on a bivariate copula approach, to generate 10 000 rainfall-duration pairs, which are the input data of Safer_RAIN. Safer_RAIN preprocessing was done in 90 seconds and each simulation lasted 20 seconds. Pluvial flood hazards maps were obtained from a frequency analysis of water depths in each cell of the grid.