A novel deep learning approach for typhoon-induced storm surge modeling through efficient emulation of wind and pressure fields

Abstract

Modeling typhoon-induced storm surges requires 10-m wind and sea level pressure fields as forcings, commonly obtained using parametric models or a fully dynamical simulation by numerical weather prediction (NWP) models. The parametric models are generally less accurate than the full-physics models of the NWP, but they are often preferred owing to their computational efficiency facilitating rapid uncertainty quantification. Here, we propose using a deep learning method based on generative adversarial networks (GAN) to translate the parametric model outputs into a more realistic atmospheric forcings structure resembling the NWP model results. Additionally, we introduce lead-lag parameters to incorporate a forecasting feature in our model. Thirty-four historical typhoon events from 1981 to 2012 are selected to train the GAN, followed by storm surge simulations for the four most recent events. The proposed method efficiently transforms the parametric model into realistic forcing fields by a standard desktop computer within a few seconds. The results show that the storm surge model accuracy with forcings generated by GAN is comparable to that of the NWP model and outperforms the parametric model. Our novel GAN model offers an alternative for rapid storm forecasting and can potentially combine varied data, such as those from satellite images, to improve the forecasts further.

A Review of Application of Machine Learning in Storm Surge Problems

“…While numerical models require more computation resources and time to update the forecasts in real-time forecasting. Mulia et al [139] showed that the computation time for a 73-h simulation for GAN and parametric model is 13 s and 8 s, respectively, with a single processor, while that for a 45-h simulation of NWP is 4200 s on a supercomputer with 256 processors. Though during the training phase, ML models can also be computationally expensive, especially when the model architecture is more sophisticated and dealing with large datasets, ML models are more efficient, especially for site-specific forecasting as a whole.…”

“…In terms of timeliness, an operational forecast usually can give a longer forecast horizon of over 72 h. However, the forecast horizons of ML models are usually up to 12 h ahead [10,49,102,139,149]. Only a small part of studies extended the lead time to 24 h [44,83], and fewer studies attempted to output the forecast water level time series with a length of over 24 h [38,48,62,78,79].…”

“…For the parameterization in preprocessing status, it was demonstrated that the improved cyclone wind fields by deep CNN performed well in real-time storm surge forecasts and were compared to the conventional blended wind fields [138]. Furthermore, to maintain the high fidelity of full-physics models of Numerical Weather Prediction (NWP) and bring into the computational efficiency of parametric models, Mulia et al [139] utilized GAN to translate the parametric model outputs into a more realistic atmospheric forcings structure similar to the NWP results. The proposed method was favorable for operational forecasting with an acceptable accuracy for lead times up to 12 h. Additionally, the ML methods can be used to provide a boundary condition for a high-resolution numerical model suitable for areas with sophisticated geometrics.…”

Prediction of storm surge in the Pearl River Estuary based on data-driven model

“…Moreover Machine learning offer a clear advantage in capturing the nonlinear interactions between typhoon influencing factors and storm surge magnitude. These models necessitate a one-time investment in training time, enabling swift forecasting using the pre-trained model, thus promoting widespread adoption (Tadesse et al, 2020;Ayinde et al, 2023;Ku and Liu, 2023;Li et al, 2023;Mulia et al, 2023). Ayyad et al (2022a) employed a coupled ADCIRC and SWAN model to generate a dataset comprising 10,300 typhoon-induced storm surge events.…”