01-05 December 2025
INCOIS, Hyderabad, India.
| Abstract Submission No. | ABS-08-0092 |
| Title of Abstract | Uncovering the Spatial and Temporal Distribution of Vertical Freshwater Intrusion Using an Unsupervised Machine Learning Approach |
| Authors | ASISH K SASIDHARAN*, B PRAVEEN KUMAR, ROHITH BALAKRISHNAN |
| Organisation | INDIAN NATIONAL CENTRE FOR OCEAN INFORMATION SERVICES, HYDERABAD |
| Address | INDIAN NATIONAL CENTRE FOR OCEAN INFORMATION SERVICES HYDERABAD, TELENGANA, India Pincode: 500090 E-mail: asishksasidharan@gmail.com |
| Country | India |
| Presentation | Oral |
| Abstract | Decoding freshwater fingerprints of the Bay of Bengal (BoB) is paramount for understanding regional climate patterns and ocean dynamics. Five major rivers flowing into the BoB north of 15°N create Sea Surface Salinity (SSS) values as low as ~28 psu, establishing the BoB as one of the world's most freshwater-influenced basins. Present-day ocean forecast models show considerable bias in reproducing currents, especially in the northern BoB, due to inaccurate representation of freshwater inputs and associated mixing processes. Past studies have primarily focused on surface freshwater transport, utilizing observation and model simulations. However, understanding vertical thermohaline structure during freshening events using observations remains limited. A Gaussian Mixture Model (GMM), an unsupervised clustering algorithm, was applied to over 5,000 in-situ low salinity profiles (SSS < 32 psu) and their corresponding temperature profiles to characterize the spatio-temporal variability of vertical thermohaline structures in the BoB. Our analysis identifies five distinct hyposaline classes (HSCs) and five associated temperature classes (ATCs) during freshening events. Seasonal freshwater advection begins with riverine inputs during the second half of the summer monsoon season, when mesoscale eddies and anticyclonic gyres in the north-central BoB control freshwater evolution and drive salinity to its lowest annual values. When the East India Coastal Current intensifies, most of the freshwater is advected along the western periphery of the BoB towards the equator. The rest stays in the north-central bay, creating favourable conditions for inversion layer formation. During winter, strong convective mixing causes freshwater to mix and evolve in the north-central bay. Our findings reveal three dynamic freshwater zones in the North BoB, which evolve seasonally, highlighting an interplay between freshwater input, stratification, and mixing. These results offer a new direction to validate and improve the freshwater forcing in the ocean model. |
| Are you part of IIOE-2 endorsed project | no |
| Keywords | Machine learning, Ocean Observation, Bay of Bengal |
| For Awards | yes |
| Date Of Birth | 05-08-1994 |
| ECSN Registration Number | IIOE2-ECSN-0114 |