Testing various geostatistical models to combine bottom trawl catches and acoustic data

Abstract

The aim of the CATEFA project is to combine information on demersal fish stock abundance from acoustic and bottom trawl surveys. While acoustic data are collected continuously while the research ship is underway, it is likely that combining two sources of information on the same variable should improve abundance estimation. A variety of geostatistical models are compared, contrasted and the output described. In this study, twenty scientific surveys from three areas (the North Sea, the Irish Sea and the Barents Sea) are analysed. These 3 zones have diverse species assemblages and hydrographic environments. Nevertheless we manage to find models relevant to most of these different situations. Unfortunately, however, we show that this enhancement can increase the variance of the estimation because of the inherently high variability of acoustic recordings. The purpose of this paper is to compare the results of three geostatistical models (i.e. co-kriging, model with orthogonal residuals, kriging with external drift) in terms of precision, details of maps, variance local and global of the estimation’s errors and cross validation. The role of the acoustic in each model and the precision it brings, is then discussed.