Ecological Modelling of Potential Habitats for Indian Mackerel (Rastrelliger spp.) in the Western of Banda Sea using an Artificial Neural Network Approach (Early View)

Abstract

Indian mackerel (Rastrelliger spp.) is a species with high catch volumes, amounting to approximately 451.750 tonnes over five years. This substantial yield holds significant potential for local communities, making sustainable utilisation crucial. This study focuses on the fishing season of Indian mackerel (Rastrelliger spp.) and the development of a habitat suitability model in the waters of the western of Banda Sea, Indonesia. The Fishing Season Index (FSI) method identified November as the peak fishing season, with the highest CPUE recorded at 220 kg trip^-1. During this period, stable salinity levels were observed, which supported the reproductive processes of Indian mackerel. Additionally, high rainfall and strong winds facilitated local upwelling, influencing currents and bringing nutrients to the surface, which were consumed by mackerel larvae. The ANN (Artificial Neural Network) models used to estimate potential fishing zones for Indian Mackerel demonstrated high accuracy, with an error rate of just 1.12%. The analysis revealed that salinity and currents were the most influential environmental parameters, contributing 16% and 14% to catch success during the peak fishing season with salinity levels at 34.2 psu and current velocity at 3.29 cm s^-1. The implementation of this model in analyzing Indian mackerel habitats and their relationship with environmental factors supports data and technology-driven fisheries management. This study also introduces a novel integration of the Fishing Season Index (FSI) method and Artificial Neural Network (ANN) modeling to simultaneously identify peak fishing seasons and predict potential fishing zones based on dynamic oceanographic parameters. The application of machine learning in this model enables the identification of non-linear relationships between environmental variables and fish distribution with high accuracy, representing a significant advancement in predictive habitat modeling for Indian mackerel in Indonesian waters. This approach contributes to sustainable fisheries resource management and aligns with the achievement of SDG 14 in Indonesia.