Machine Learning-Based Prediction of Sardine (Sardinella lemuru) Fishing Grounds Using Oceanographic Parameters in the Bali Strait

I Made Radiansyah (Department of Marine Science, University of Mataram, Mataram, Indonesia)
Rahmawati Astuti (Department of Oceanography, Diponegoro University, Semarang, Indonesia)
Juliatin Harahap (Department of Marine Resource Management, Hasanuddin University, Makassar, Indonesia)

Journal of Marine Fisheries
Vol. 2 No. 2 (2026): Journal of Marine Fisheries, June 2026

The distribution of small pelagic fish is strongly influenced by oceanographic variability. This study aims to synthesize and model the prediction of fishing grounds for Sardinella lemuru in the Bali Strait using machine learning-based habitat approaches. The study integrates sea surface temperature (SST), chlorophyll-a, and climate variability as primary predictors. A systematic literature-based modeling framework is developed using findings from previous studies and advanced habitat modeling approaches such as Maximum Entropy (MaxEnt) and Generalized Additive Models (GAM). Results show that SST and chlorophyll-a significantly control spatial fish distribution through upwelling dynamics and primary productivity enhancement. Climate variability, particularly the Indian Ocean Dipole (IOD) and ENSO, intensifies seasonal shifts in fish aggregation patterns. Machine learning-based models consistently improve prediction accuracy of potential fishing zones compared to conventional approaches. Studies by Akita et al. (2023), Semedi et al. (2022), and Gustantia et al. (2022) confirm that environmental variables can explain spatial habitat suitability with high reliability. The study concludes that integrating satellite remote sensing with machine learning models provides a robust framework for sustainable fisheries management in the Bali Strait.