Upskilling Low‐Fidelity Hydrodynamic Models of Flood Inundation Through Spatial Analysis and Gaussian Process Learning. Issue 8 (8th August 2022)

• Record Type: Journal Article
• Authors: Fraehr, Niels; Wang, Quan J.; Wu, Wenyan; Nathan, Rory
• Is Part Of: Water resources research. Volume 58:Issue 8(2022)
• Page Start: n/a
• Page End: n/a
• Publication Date: 2022-08-08
• Subjects: low‐fidelity -- Gaussian Process -- hydrodynamic -- inundation -- flood -- Empirical Orthogonal Function; Hydrology -- Periodicals; 333.91
• DOI: 10.1029/2022WR032248
• Languages: English
• ISSNs: 0043-1397

Abstract

Abstract: Accurate flood inundation modeling using a complex high‐resolution hydrodynamic (high‐fidelity) model can be very computationally demanding. To address this issue, efficient approximation methods (surrogate models) have been developed. Despite recent developments, there remain significant challenges in using surrogate methods for modeling the dynamical behavior of flood inundation in an efficient manner. Most methods focus on estimating the maximum flood extent due to the high spatial‐temporal dimensionality of the data. This study presents a hybrid surrogate model, consisting of a low‐resolution hydrodynamic (low‐fidelity) and a Sparse Gaussian Process (Sparse GP) model, to capture the dynamic evolution of the flood extent. The low‐fidelity model is computationally efficient but has reduced accuracy compared to a high‐fidelity model. To account for the reduced accuracy, a Sparse GP model is used to correct the low‐fidelity modeling results. To address the challenges posed by the high dimensionality of the data from the low‐ and high‐fidelity models, Empirical Orthogonal Functions analysis is applied to reduce the spatial‐temporal data into a few key features. This enables training of the Sparse GP model to predict high‐fidelity flood data from low‐fidelity flood data, so that the hybrid surrogate model can accurately simulate the dynamic flood extent without using a high‐fidelity model. The hybrid surrogate model is validated on the flat and complex Chowilla floodplain in Australia. The hybrid model was found to improve the results significantly compared to just using the low‐fidelity model and incurred only 39% of the computational cost of a high‐fidelity model. Plain Language Summary: Floods are the most common type of natural disaster and therefore it is important to predict when and where flooding occurs. This is normally done using a complex computer model that divides the area of interest into small subareas and then calculates how the water moves between each subarea. However, to predict flooding accurately over large areas, it is necessary to use millions of small subareas and it takes a long time to calculate the movement of flood water between subareas. To mitigate this issue, this study proposes an alternative approach based on a simpler computer model. This simpler model uses larger subareas to predict flooding, which makes the model less accurate but much faster. To compensate for the reduced accuracy, the results are corrected using an advanced computer method that is calibrated to predict the relationship between the predictions made using the complex and simpler models. The new approach is used to predict flooding on a large, flat floodplain in Australia. The predictions show a significant improvement compared to just using the simpler computer model. Furthermore, the calculations only take about 39% of the time taken by a complex model with the small subareas, but the accuracy is similar. Key Points: A new hybrid surrogate model for predicting the dynamic evolution of flood inundation extent is proposed The hybrid model significantly improves the accuracy of flood inundation extent predictions compared to a low‐fidelity model The computational cost is substantially reduced compared to a high‐fidelity model