The relationship between multi-sensor satellite data and Bayesian estimates for skipjack tuna catches in the South Brazil Bight.

In this study we tested a Bayesian model based on a conjugate gamma/Poisson pair associated with environmental variables derived from satellite data such as sea surface temperature (SST) and its derived gradient fields from Moderate Resolution Imaging Spectroradiometer (MODIS)/Terra, chlorophyll-a concentration from Sea Viewing Wide field of View Sensor (SEAWiFS)/SeaStar and surface winds and Ekman pumping from SeaWinds/Quick Scatterometer (QuikSCAT) to predict weekly catch estimates of the skipjack tuna in the South Brazil Bight. This was achieved by confronting the fishery data with model estimates and regressing the results on the satellite data. The fishery data were expressed by an index of catch per unit effort (CPUE) calculated as the weight of fish caught (in tonnes) by fishing week, and were divided into two series, called historical series (1996-1998; 2001), and validation year (2002). The output of model CPUE estimates is in good agreement with the historical weekly CPUE and generated updated weekly estimates that explained up to 62% of weekly CPUE from 2002. In general, the best proxy for the Bayesian weekly estimates is the gradient zonal SST field. The results refined previous knowledge of the influence of SST on the occurrence of skipjack tuna. [ABSTRACT FROM AUTHOR]