Logo image
AI-driven ocean-colour prediction of potential fishing zones for blue-economy resource sustainability in the northern Bay of Bengal
 

AI-driven ocean-colour prediction of potential fishing zones for blue-economy resource sustainability in the northern Bay of Bengal

Ismail Mondal, SK Ariful Hossain, Felix Jose, Javed Akhter, Mofareh D. Qoradi Mukhiddin Juliev
Physics and Chemistry of the Earth Parts A/B/C, Vol.144, 104655
10-01-2026
Potential fishing zones Ocean colour Artificial intelligence Sentinel-3 Bay of bengal Blue economy SDG 14
Predicting Potential Fishing Zones (PFZs) is increasingly important for data-limited coastal fisheries, particularly where fishery advisories are constrained by sparse field observations, optically complex estuarine waters, and limited uncertainty validation. This study develops an integrated Sentinel-3- Artificial Intelligence-based framework for seasonal PFZ suitability prediction in the northern Bay of Bengal, with emphasis on the dynamic Sundarbans estuarine shelf. Sentinel-3 OLCI/SLSTR observations were combined with 150 in-situ measurements of chlorophyll-a (Chl-a), phytoplankton biomass, sea surface temperature (SST), nitrite, total dissolved solids (TDS), and biochemical oxygen demand (BOD) to generate a multi-source environmental predictor space. Four machine learning (ML) models, Artificial Neural Network (ANN), Random Forest (RF), Support Vector Machine (SVM), and K-Nearest Neighbors (KNN), were trained using 4700 georeferenced samples and evaluated through class-wise accuracy, ROC/AUC, confusion matrices, balanced accuracy, Cohen's kappa, Matthews correlation coefficient, spatial block validation, and external consistency with operational PFZ advisory patterns. The validation showed strong but realistic model skill, with RF and ANN providing the most stable seasonal performance, whereas SVM and KNN were more sensitive to mixed estuarine fronts and transitional PFZ boundaries. Chl-a, phytoplankton biomass, and SST emerged as dominant ecological signals, while nitrite, TDS, and BOD improved the interpretation of nutrient enrichment and water-quality stress. Post-monsoon PFZs showed greater spatial coherence, whereas monsoon suitability was fragmented by runoff, turbidity, and hydrodynamic mixing. The framework advances ocean-colour-based PFZ prediction from simple fish-location mapping toward transparent habitat-suitability intelligence for blue-economy resource sustainability.

(1)

url
Link to journal article.

1
Logo image