The experiment was conducted during January-September, 2024 at the Visakhapatnam Fishing Harbour, Andhra Pradesh, India to evaluate Artificial Neural Network (ANN) models to predict sardine catches by integrating key satellite-derived environmental variables. Daily catch data were collected during this period, complemented by Aqua MODIS-derived CHL-a and SST, and meteorological observations of AT and WS. A 200 km buffer zone polygon shapefile, encompassing the study area, was created using ArcGIS 10.7.1. A total of fourteen Feed Forward Neural Network models were developed and classified into With Ban (WB) and Without Ban (WOB) categories to account for the annual fishing ban from April 15 to June 14, 2024. Among these, the WOB model incorporating SST exhibited the highest predictive performance, indicating the strong influence of environmental parameter variations on sardine availability. Fluctuations in CHL-a, primarily driven by monsoon-induced upwelling and wind-mediated nutrient enrichment, were also closely associated with catch trends. Sensitivity analysis further validated model robustness, with the comprehensive WB model demonstrating reliable predictive capability despite seasonal constraints. Overall, the findings highlighted the effectiveness of ANN models in interpreting non-linear catch dynamics and emphasized their potential as valuable tools for forecasting fishery yields. The study underscored the importance of integrating environmental indicators into fisheries management to support sustainable exploitation of the Indian Oil Sardine along the Andhra Pradesh coast.
Swapna et al. (Thu,) studied this question.