01-05 December 2025
INCOIS, Hyderabad, India.
| Abstract Submission No. | ABS-04-0389 |
| Title of Abstract | Deep Learning for Sea Surface Temperature Prediction in the Indian Ocean: A Comparative Study Using 1D-CNN, LSTM, and CNN-LSTM Architectures |
| Authors | B. Sheela Rani*, D. Sathya Narayanana, M.B. Salma Jasmine, N.R. Krishnamoorthy, M. Roja Raman, Aneesh A. Lotliker, Abhisek Chatterjee |
| Organisation | Centre for Ocean Research, Sathyabama Institute of Science and Te |
| Address | No 51/2, Valmiki street, Thiruvanmiyur Chennai, Tamil Nadu, India Pincode: 600041 E-mail: officemailsathya@gmail.com |
| Country | India |
| Presentation | Poster |
| Abstract | Anthropogenic activities had led to major climate changes, causing the oceans to warm rapidly. Subsequently, there is a need for accurate prediction of sea surface temperature (SST). The dynamic Indian Ocean climate, weather and marine ecology is significantly affected by SST changes. In this study, hourly and daily SST data, specifically from 15 and 7 RAMA stations (1m deep) respectively, at different parts of Indian Ocean is utilized to forecast SST using three deep learning techniques; 1D Convolutional Neural Network (1D-CNN), Long Short-Term Memory (LSTM) network, and Hybrid 1DCNN-LSTM network. The SST data obtained from RAMA buoy is trained using a sliding window method and model performance is evaluated based on evaluation matrices like Mean Squared Error (MSE), Root Mean Squared Error (RMSE), and Mean Absolute error (MAE). The results indicated that LSTM outperforms 1D-CNN at most sites, consistently achieving lower MSE, RMSE and MAE. The findings also revealed that in places where SST shifts significantly, the hybrid 1DCNN-LSTM model often competes with or does better than LSTM in MAE. Distinctive physical phenomena like equatorial jets, thermohaline stratification, monsoon, Indian Ocean Bipolar (IOD) and El Niño-Southern Oscillation (ENSO) bring about geographical difference in performance. Hence, the study illustrates how deep learning frameworks make better SST predictions in complex ocean basins and thereby aids in anticipating climate change. |
| Are you part of IIOE-2 endorsed project | no |
| Keywords | Ocean warming; Sea Surface Temperature; Machine learning; 1D-CNN; LSTM; 1DCNN-LSTM |
| For Awards | yes |
| Date Of Birth | 10-07-2001 |
| ECSN Registration Number | IIOE2-ECSN-0219 |