Modelling habitat preference of Caspian Kutum, Rutilus kutum, using non-linear habitat suitability indices and generalized additive models

• Published in: Regional studies in marine science 2022-11, Vol.56, p.102715, Article 102715
• Main Authors: Moëzzi, Fateh, Poorbagher, Hadi, Eagderi, Soheil, Feghhi, Jahangir, Dormann, Carsten F., Nergi, Sabah Khorshidi, Amiri, Kaveh
• Format: Article
• Language: English
• Subjects: aerosols; Caspian Sea; chlorophyll; coasts; fish; Fish distribution; fisheries; GAM; Habitat modelling; habitat preferences; habitats; HSI; inorganic carbon; marine science; Rutilus kutum; surface water temperature
• ISSN: 2352-4855; 2352-4855
• DOI: 10.1016/j.rsma.2022.102715

The distribution of Caspian Kutum, Rutilus kutum, an economically important fish species with a limited understanding of its ecology, was investigated along the southern Caspian Sea coast to identify the environmental drivers of its occurrence. The environmental predictors including sea surface temperature, chlorophyll-a concentration, particulate organic and inorganic carbon, aerosol optical thickness, depth, bottom slope, coastline aspect and distance to rivers, and long-term monthly commercial beach seine catch data, procured from 2002 to 2012, were analysed. Using two alternative approaches to describe catch per unit effort (CPUE), a multiplicative effect of predictors was found that is often being used in fishery studies (the so-called continued product model, HSICPM) to perform weaker than a Generalized Additive Model (GAM). The highly variable CPUE was strongly related to sea surface temperature, bottom slope, aerosol optical thickness and distance to rivers using HSICPM, but coastline aspect, particulate inorganic carbon and bottom slope in the GAM. The steps involved in computing the HSICPM led to a biased fit. This study provides a robust quantification of habitat characteristics of Caspian Kutum that can be used to inform management plans with both commercial and conservation goals. •Fish abundance was strongly related to SST and bottom slope using HSI.•Coastline aspect and PIC were the main habitat drivers with the GAM model.•Kutum distribution was predicted more accurately with lower bias by the GAM model.•The GAM model detected narrower optimum ranges of habitat variables for the fish compared to HSI.