Multi-Algorithm Species Distribution Approach in Modeling Suitable Habitat Distribution for Skipjack Tuna (Katsuwonus pelamis) in FMA 713 and 714, Indonesia

Muhyun, Ainun Apriliyani; Pratama, Gilar Budi; Hidayat, Rian; Mujahid, Muhammad; Chuzaimah, Syakirah; Zamira, Nasywa Aurellia; Sianturi, Oksto Ridho

doi:10.21608/ejabf.2025.416360.6452

Multi-Algorithm Species Distribution Approach in Modeling Suitable Habitat Distribution for Skipjack Tuna (Katsuwonus pelamis) in FMA 713 and 714, Indonesia

Document Type : Original Article

Authors

¹ Institut Teknologi Sains dan Bisnis Muhammadiyah Selayar

² Departement of Fisheries, Faculty of Fisheries and Marine Sciences, Universitas Padjadjaran

³ Data Desa Presisi, IPB University, Bogor, West Java, Indonesia

⁴ Department of Fisheries and Marine, State Polytechnic of Lampung

⁵ Capture fisheries practitioners, West Java, Indonesia

⁶ Marine Fisheries Technology Study Program, Graduate School of IPB University, Bogor, Indonesia

⁷ Research Center for Oceanograpy, National Research and Innovation Agency, Jakarta, Indonesia.

10.21608/ejabf.2025.416360.6452

Abstract

This study aims to model the potential habitat distribution of skipjack tuna (Katsuwonus pelamis) in Fisheries Management Areas (FMA) 713 and 714 using a multi-algorithm Species Distribution Modeling (SDM) approach. Catch logbook data from 2020–2024 were combined with key oceanographic parameters, namely sea surface temperature (SST), chlorophyll-a (CHL), salinity (SAL), sea surface height (SSH), and ocean currents (CUR). Four modeling algorithms (Generalized Additive Model (GAM), Multivariate Adaptive Regression Splines (MARS), Maximum Entropy (MAXENT), and Support Vector Machine (SVM)) were applied and evaluated using two commonly used metrics, the Area Under Curve (AUC) and True Skill Statistics (TSS). The evaluation results showed that most algorithms performed well, with average AUC values > 0.7 and TSS > 0.5, making them suitable for habitat distribution analysis. Variable importance analysis revealed that SAL and SSH were the dominant factors influencing habitat distribution. The contribution of SAL was most prominent in the MAXENT algorithm (47.47%), while SSH was dominant in MARS (34.23%). Prediction overlays generated spatial habitat suitability index (HSI) maps, with a suitability threshold of ≥ 0.6 indicating optimal habitats. The results indicated that the skipjack tuna habitats are seasonal, being most extensive and stable in the first half of the year (April–July) and at the end of the year (November–December), aligning with the skipjack fishing seasons.

Keywords