Abstract

 https://doi.org/10.1080/1755876X.2021.1946240

Demersal species play a fundamental role in fisheries, thus understanding their distribution and abundance through bottom trawl surveys is crucial for stock and fisheries management. Oceanographic (e.g. biogeochemical, physical) and fishing covariates might be considered, in addition to spatio-temporal variables (latitute, longitude, depth, year and month), to better explain trawl survey data. Here, we analyse biomass indices (kg/km2) for European hake, common sole, mantis shrimp, red mullet and common cuttlefish from scientific trawl surveys carried out in the Adriatic Sea and the Western Ionian Sea. We used three different Generalised Additive Model (GAM) approaches (Gaussian, Tweedie and Delta) to fit and predict species biomass distribution. In order to evaluate trade-offs in using different covariates, we compared the results obtained from GAM approaches based only on spatiotemporal variables and GAMs including also oceanographic and fishing effort covariates.
The Delta-GAM approach performed better for European hake, mantis shrimp and common cuttlefish, while GAMs based on Gaussian and Tweedie were performing better for the red mullet and common sole, respectively. The results highlighted that adding specific oceanographic and effort covariates to spatiotemporal variables improved the performances of spatial distribution models especially for European hake, mantis shrimp and red mullet. Significant additional explanatory variables were bottom temperature, bottom dissolved oxygen, salinity, particulate organic carbon, and fishing effort for European hake; the same variables and pH for mantis shrimp; chlorophyll-a, pH, sea surface temperature, bottom dissolved oxygen, nitrate and effort for the red mullet; phosphate and salinity for common sole; bottom temperature, bottom dissolved oxygen, and phosphate for the common cuttlefish.
The findings highlight that more accurate estimates of spatial distribution of demersal species biomass from trawl survey data can generally be obtained by integrating oceanographic variables and effort in GAMs approaches with potential impacts on stock assessment and essential fish habitats identification.