ST-LSTM-SA：A new ocean sound velocity fields prediction model based on deep learning

Abstract: The scarcity of in-situ ocean observations poses a challenge for real-time information acquisition in the ocean. Among the crucial hydroacoustic environment parameters, ocean sound velocity exhibits significant spatial and temporal variability and it is highly relevant to ocean research. In this study, we propose a new data-driven approach, leveraging deep learning techniques, for the prediction of sound velocity fields (SVFs). Our novel spatiotemporal prediction model, ST-LSTM-SA, combines Spatiotemporal Long Short-Term Memory (ST-LSTM) with a self-attention mechanism to enable accurate and real-time prediction of SVFs. To circumvent the limited amount of observation data, we employ transfer learning by firstly training the model using reanalysis datasets, followed by fine-tuning with the in-situ analysis data to obtain the final prediction model. By utilizing the historical 12-months SVFs as input, our model predicts the SVFs for the subsequent 3-months. We compare the performance of five models: Artificial Neural Networks (ANN), Long Short-Term Memory (LSTM), Convolutional LSTM (ConvLSTM), ST-LSTM, and our proposed ST-LSTM-SA model in the test experiment spanning from 2019 to 2022. Our results demonstrate that the ST-LSTM-SA model significantly improves the prediction accuracy and stability of sound velocity in both temporal and spatial dimensions. The ST-LSTM-SA model not only predicts the ocean sound velocity field (SVF) accurately, but also provides valuable insights for spatiotemporal prediction of other oceanic environmental variables.

Keywords:
• sound velocity filed,
• spatiotemporal prediction,
• deep learning,
• self-attention