Exploring deep learning capabilities for surge predictions in coastal areas

Abstract

To improve coastal adaptation and management, it is critical to better understand and predict the characteristics of sea levels. Here, we explore the capabilities of artificial intelligence, from four deep learning methods to predict the surge component of sea-level variability based on local atmospheric conditions. We use an Artificial Neural Networks, Convolutional Neural Network, Long Short-Term Memory layer (LSTM) and a combination of the latter two (ConvLSTM), to construct ensembles of Neural Network (NN) models at 736 tide stations globally. The NN models show similar patterns of performance, with much higher skill in the mid-latitudes. Using our global model settings, the LSTM generally outperforms the other NN models. Furthermore, for 15 stations we assess the influence of adding complexity more predictor variables. This generally improves model performance but leads to substantial increases in computation time. The improvement in performance remains insufficient to fully capture observed dynamics in some regions. For example, in the tropics only modelling surges is insufficient to capture intra-annual sea level variability. While we focus on minimising mean absolute error for the full time series, the NN models presented here could be adapted for use in forecasting extreme sea levels or emergency response.

Using Neural Networks to Predict Hurricane Storm Surge and to Assess the Sensitivity of Surge to Storm Characteristics

“…The ANN models trained in this study show significant skill in predicting peak storm surge levels and surge time series. In contrast to prior data-driven studies that use course resolution reanalysis (Bruneau et al, 2020;Tiggeloven et al, 2021) or that train ANN models with a limited number of hurricane scenarios (Lee et al, 2021;Ramos Valle et al, 2021), our ANN models do not under-predict extreme storm surge levels. This may be due to the fact that we use a much broader set of physically plausible hurricane and storm surge events to train the ANN models and that we allow the structure of the ANN models (e.g., the number of hidden layers) to vary between for each site.…”

“…Tiggeloven et al. (2021) expanded upon the NN models by using more complex deep learning models (such as convolutional NN) to construct ensembles of water levels at global tide gauge stations. Lee et al.…”

Recent Developments in Artificial Intelligence in Oceanography

“…Jiang [110] used the beetle antenna search algorithm to optimize a BP NN storm surge forecast at the next moment based on 20 prediction parameters; this algorithm had higher accuracy than a single BPNN algorithm. Tiggeloven et al [111] used ANN, LSTM, and convLSTM to construct NN ensembles at 736 global tide stations for the prediction of sea levels. The authors found that LSTM outperformed the other two approaches, but when more predictor variables were added to the forecast models, the computational time correspondingly improved.…”

Artificial Intelligence for Earth System Predictability (AI4ESP) (2021 Workshop Report)

“…This allows researchers to characterize TCs through a small set of parameters and thus form a map from these inputs to the modeled surge. Emerging areas for ML-based coastal flooding include: incorporating sea-level rise (Kyprioti et al 2021b), coupling to and including rainfall and hydrological processes for compound flooding prediction (Bass and Bedient 2018;Li, Kiaghadi, and Dawson 2021), and using more advanced ML techniques such as deep learning (Tiggeloven et al 2021) and Fourier Neural Operators ). An additional component to sea levels and coastal flooding originates from wind waves and associated wave setup, runup, and overtopping.…”