ABSTRACT Flooding is one of the most devastating natural disasters worldwide. It is also expected to become more severe as climate change impacts are realised. Two‐dimensional (2D) hydrodynamic models are used to obtain reliable inundation estimations. However, these models are computationally expensive and time‐consuming, making rapid and ensemble flood scenario assessment challenging. This study investigates the application of hybrid hydrodynamic‐machine learning techniques to develop two rapid flood inundation scenario assessment models in a study catchment in Aotearoa New Zealand. This approach aims to reduce the numerical modelling load to rapidly make robust predictions of potential flooding events from an ensemble of previously assessed events. The proposed framework is based on an innovative combination of a flexible climate‐informed synthetic storm approach and a hybrid hydrodynamic (BG‐Flood 2D hydrodynamic model)—machine learning (Random Forest algorithm) approach. A catalogue of synthetic storms was created based on the characteristics of the main inundation driver (heavy rainfall) and the synoptic conditions in the New Zealand region using statistical techniques. Two hybrid models based on the Random Forest algorithm were developed to emulate the BG‐Flood outputs and efficiently predict flood/non‐flood and maximum inundation depth in the floodplain catchment. The storms' features and the geographic characteristics of the catchment were used as predictor features, and the inundation maps produced by BG‐Flood were used as the target data to develop the hybrid models. Validation results show that the hybrid models can rapidly (under 1 s) and accurately predict flood/non‐flood (mean Recall = 0.878, Precision = 0.895 and F1 score = 0.886) and maximum inundation depth (mean RMSE = 0.0610 m and NSE = 0.866) in the catchment, producing reliable and valuable flood hazard information for flood risk management.
Pozo et al. (Mon,) studied this question.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: