Document Type : Research Paper

Authors

• Hadi Poorbagher ¹
• Soheil Eagderi ²
• Fateh Moezzi ²

¹ Department of fisheries, Faculty of Natural Resources, University of Tehran

² Department of Fisheries, Faculty of Natural Resources, University of Tehran

Abstract

The statistical models with the best performance in habitat suitability studies of fish species are of high importance. The present study compared the performance of linear (linear regression model (LM), generalized linear model (GLM)) and non-linear models (artificial neural networks (ANN), support vector machine (SVM)) in estimating habitat suitability index (HSI) for Capoeta razii in a southern Caspian Sea basin. The environmental parameters were altitude, depth, width, velocity, temperature, pH, electrical conductivity (EC), total dissolved solids (TDS), bottom stone diameter and total count of stones with diameter > 15 cm per m2. The linear models had weak predictive performance (higher RMSE values) compared to ANN and SVM models. The SVM was the best model with the predictors of altitude, pH, temperature and stone diameter and ANN was the best model using the rest of the parameters. The arithmetic mean model (AMM) showed better performance in estimating HSI compared to the geometric mean model (GMM). The distribution of HSI values along the sampling stations in the Caspian Sea basin (the Taleghan River) showed high diversity in the habitat condition of the fish species.

Keywords

• Species distribution
• Taleghan River
• Habitat variables
• Modelling