Advancing Sea Surface Temperature Forecasting with Deep Learning Techniques on Copernicus Data: An Application in the Gulf of Trieste

Accurate forecasting of the sea surface temperature (SST) is essential for understanding and mitigating the impacts of climate change and other anthropogenic/natural risks on our natural heritage. However, the nonlinear temporal dynamics with changing, complex factors and the inherent difficulties in long-scale predictions lead to some issues in accomplishing the task. This paper proposes a Deep Learning-based approach for predicting SST over a seven-day temporal horizon, exploiting the comprehensive datasets provided by the Copernicus Marine Environment Monitoring Service (CMEMS). The proposed model leverages a Continuous Wavelet Transform (CWT) to obtain a hybrid time-frequency transform of the 1D time-series signal, which is further displayed as a 2D image. In this preliminary setting, the model is trained on the Copernicus Hub subset which includes the historical series of SST. The promising results obtained by the experiments confirm the validity of the proposed approach, which could be further improved by integrating different variables data and by developing ad-hoc models for the task.