A Deep Learning Model for Forecasting Sea Surface Height Anomalies and Temperatures in the South China Sea

Abstract

The field of forecasting oceanic variables has traditionally relied on numerical models, which effectively consider the ocean's dynamic evolution and are of physical importance. However, to make the models more realistic, complicated processes need to be considered, which is difficult because their calculations are complex. In fact, information on the internal dynamic mechanisms and external driving forces of the ocean are already embedded in the time series of observations. Therefore, we can determine the patterns of ocean variations through data mining of these series to achieve forecasting. Furthermore, to predict variations in ocean processes more realistically, interactions between variables and spatial correlations should be effectively considered. Thus, inspired by available remote sensing data and advancements in deep learning technologies, we develop a hybrid model based on a statistical method and a deep learning model to predict multiple sea surface variables. A case study in the South China Sea shows that this model is highly promising for short‐term daily forecasts of the sea surface height anomaly (SSHA) and sea surface temperature (SST). When the forecast time is 10 days, the root mean square errors of this model forecasts for SSHA and SST are approximately 0.0276 m and 0.46°C, respectively, which are much smaller than those of persistence, climatology and linear regression predictions. The anomaly correlation coefficients for SSHA and SST are approximately 0.864 and 0.633, respectively. The model performs satisfactorily under both normal and typhoon weather conditions.

Composite Analysis-Based Machine Learning for Prediction of Tropical Cyclone-Induced Sea Surface Height Anomaly

“…The RF methods' performance is timedependent because the pre-storm oceanic averaged variables are used as input features. Previous studies have also demonstrated that ML-based methods exhibit superior performance in shortterm predictions when compared to long-term predictions [40,41]. It can be observed that both types of approaches have their own advantages in predicting the temporal evolution of TCinduced SSHA in terms of the time scales.…”

Seven-day sea surface temperature prediction using a 3DConv-LSTM model

“…Compared with the M-LCNN, our model's accuracy in the 7-day forecasting period is improved in the Yellow Sea (Xu et al, 2020). The Conv1D-LSTM model takes the SST and sea surface height anomaly after multivariate empirical orthogonal function decomposition as input (Shao et al, 2021). Since the model only predicts SSTs in areas with water depths greater than 200 m, we also excluded nearshore areas in our statistics, but included the Sulu Sea.…”

“…This approach is efficient in training and prediction but also requires deep mining of the spatio-temporal relationships in the regional SST. Shao et al (Shao et al, 2021) first put sea surface height anomaly and SST through multivariate empirical orthogonal function analysis to establish the spatial relationship between the discrete points. Then the principal components were used as model inputs.…”

A hybrid deep learning model for predicting the Kuroshio path south of Japan

“…The long short-term memory (LSTM) neural network, as one of the essential variants of the RNN, can detect even minor changes from the time series and avoids the problem of vanishing gradient and exploding gradient (Hochreiter and Schmidhuber, 1997). In recent years, the LSTM neural network had a good performance in the time series prediction of ocean variables (Liu et al, 2018;Xiao et al, 2019;Shao et al, 2021b). However, few deep learning models are used to predict the Kuroshio path south of Japan.…”