01-05 December 2025
INCOIS, Hyderabad, India.
| Abstract Submission No. | ABS-01-0372 |
| Title of Abstract | Using Long Short Term Memory networks to predict daily-averaged sea level anomaly and surface currents |
| Authors | Athul C R*, Balaji B, Vinod Daiya, Arya Paul |
| Organisation | INCOIS |
| Address | B5 Quarters, INCOIS Residential Quarters Hyderabad, Telangana, India Pincode: 500090 E-mail: athulcr604@gmail.com |
| Country | India |
| Presentation | Oral |
| Abstract | Short-term forecasts of sea level anomalies (SLA) and surface currents are traditionally generated using ocean general circulation models. In this work, we introduce a univariate machine learning approach based on Long Short-Term Memory (LSTM) networks for predicting daily averaged SLA in the north Indian Ocean, with a lead time of three days, utilizing historical satellite altimetry observations at a spatial resolution of approximately 13 km. By considering SLA reanalysis products from advanced dynamical systems as benchmarks, our results reveal that the proposed model delivers superior predictive performance. The forecast errors remain below 0.04 m, and correlation values are consistently near unity, across most of the study region. Furthermore, surface currents predicted from the SLA forecasts, using geostrophic and Ekman balance relations, show comparable skill to state-of-the-art reanalyses when validated against both coastal and open ocean in-situ observations. When these predicted currents are treated as synthetic observations for data assimilation, the accuracy of subsurface current forecasts improves notably⿿correlations become statistically significant at the 99% confidence level across depths, and errors decrease by about 0.1 m·s⁻¹. Our study demonstrates that the short-term forecast of daily-averaged sea level and surface currents can be approached as a collection of localized low-dimensional independent univariate systems, leading to substantial reductions in computational demand. This framework highlights the potential of machine learning to reshape the landscape of operational ocean forecasting. |
| Are you part of IIOE-2 endorsed project | no |
| Keywords | Sea Level Prediction, Machine Learning, Neural Networks, LSTM |
| For Awards | no |