| Abstract Submission No. | ABS-4.1-0115 |
| Title of Abstract | Regionalization of water masses over the Indian Ocean for optimal deployment of ocean observation platforms |
| Authors | M. Mahendar*, Pritam Kar, R. Venkat Shesu*, J.V.S. Raju, B. Venkateswara Rao, M. Ravichandran |
| Organisation | Indian National Centre for Ocean Information Services |
| Address | INCOIS, Pragati Nagar Hyderabad, Telangana, India Pincode: 500090 E-mail: m.manda-p@incois.gov.in |
| Country | India |
| Presentation | Oral |
| Abstract | The Indian Ocean exhibits strong spatial heterogeneity in sea surface temperature (SST), salinity due to freshwater stratification, monsoon forcing, and mesoscale eddies, complicating water mass classification and observing system design. A long-standing need is to partition the water masses optimally based on statistical criteria rather than subjective expert judgment. We are focusing on regionalizing SST data from 1950 to 2024, subgrouping into contiguous clusters, and understanding the spatial patterns. A graph-based clustering framework has been utilized to regionalize the water masses with iterations leading to an ensemble cluster. Clustering produces different results primarily due to randomness, especially in partitional clustering algorithms. To minimize these errors, the iterative ensemble is one of the easiest solutions, although there are limitations to this approach. Then, the Jaccard coefficient measures pairwise similarity between the water masses. The resulting Jaccard coefficient partitions reveal distinct water masses, e.g.. like those freshwater masses primarily from the riverine flows, and those water masses influenced by monsoons and other large-scale climate drivers. Iterative ensemble sampling yields stable regional boundaries and resolves temporal shifts linked to decadal climate variability. The partitioning of the water masses is particularly useful in planning the deployment of buoys, Argo floats, and other ocean observation platforms that rely on the characteristics of the water masses for optimal operations. For future studies and applications, different ocean parameters, such as salinity, chlorophyll, etc, can also be studied to understand the ocean better. |
| Keywords | Regionalizaton, Clustering, Jaccard Similarity Matrix, Hierarchical Clustering, Graph-based Clustering |