Development of a Fast and Accurate Hybrid Model for Floodplain Inundation Simulations

• Published in: Water resources research Vol. 59; no. 6
• Main Authors: Fraehr, Niels, Wang, Quan J., Wu, Wenyan, Nathan, Rory
• Format: Journal Article
• Language: English
• Published: 01.06.2023
• Subjects: EOF analysis; flood inundation; hydrodynamic; Sparse Gaussian Process; surface water; surrogate
• ISSN: 0043-1397; 1944-7973
• DOI: 10.1029/2022WR033836

High computational cost is often the most limiting factor when running high‐resolution hydrodynamic models to simulate spatial‐temporal flood inundation behavior. To address this issue, a recent study introduced the hybrid Low‐fidelity, Spatial analysis, and Gaussian Process learning (LSG) model. The LSG model simulates the dynamic behavior of flood inundation extent by upskilling simulations from a low‐resolution hydrodynamic model through Empirical Orthogonal Function (EOF) analysis and Sparse Gaussian Process learning. However, information on flood extent alone is often not sufficient to provide accurate flood risk assessments. In addition, the LSG model has only been tested on hydrodynamic models with structured grids, while modern hydrodynamic models tend to use unstructured grids. This study therefore further develops the LSG model to simulate water depth as well as flood extent and demonstrates its efficacy as a surrogate for a high‐resolution hydrodynamic model with an unstructured grid. The further developed LSG model is evaluated on the flat and complex Chowilla floodplain of the Murray River in Australia and accurately predicts both depth and extent of the flood inundation, while being 12 times more computationally efficient than a high‐resolution hydrodynamic model. In addition, it has been found that weighting before the EOF analysis can compensate for the varying grid cell sizes in an unstructured grid and the inundation extent should be predicted from an extent‐based LSG model rather than deriving it from water depth predictions. Plain Language Summary Every year, lives are lost, and infrastructure is destroyed due to floods. This highlights the need for fast and accurate flood predictions to inform flood forecasting and risk assessments. However, predicting flood inundation in high resolution is often not practically feasible due to the high computational cost involved in running complex computer models. Simplified computer models can be used to provide faster flood predictions, but they lack the accuracy provided by complex models. To address this issue, this study evaluates an alternative method based on the combination of a fast simple model together with an advanced spatial feature matching method. The advanced spatial feature matching method is used to convert the predictions obtained from the simple model to accurate predictions of flood inundation depth and extent. The new approach is applied to a large floodplain in Australia and different adaptations are explored to optimize the procedure and ensure robust performance. The new approach is compared to the use of a traditional complex model and a previous approach that only predicted inundation extent. The new approach shows similar accuracy to the traditional complex model while being 12 times faster, thereby making it more practically useful for flood risk assessments. Key Points A flood inundation model is developed, based on Low‐fidelity hydrodynamic modeling, Spatial analysis, and Gaussian Process learning (LSG) The LSG model predicts water depths with a mean Root Mean Square Error (RMSE) of 4 cm and a standard deviation of 5 cm The LSG model is 12 times faster than a traditional high‐resolution hydrodynamic model