The Antarctic krill (Euphausia superba) fishery is primarily concentrated in specific areas of the Southern Ocean, where recurrent spatial and temporal patterns of fishing activity have been documented. In this study, the spatiotemporal variability of catch per unit effort (CPUE) in recurrent fishing opportunities was assessed using data from a single commercial vessel operating between 2017 and 2024 in statistical subareas 48.1 and 48.2. Fishing opportunities were identified through spatial clustering analysis of individual hauls, and CPUE was modeled using hierarchical Bayesian spatiotemporal models fitted with the Integrated Nested Laplace Approximation (INLA). Fixed effects of environmental (sea temperature, bathymetry, wind speed) and operational (trawl depth, distance, heading, trawl speed) variables were included, along with seasonal and annual random effects and a dynamic spatial field with an AR(1) temporal structure.
Results revealed marked differences between subareas. In subarea 48.1, CPUE was consistently higher and exhibited lower interannual variability than in subarea 48.2, with fishing opportunities E and F showing the highest catch rates. Conversely, in subarea 48.2, opportunity G exhibited the lowest CPUE values, consistent with the estimated spatial patterns. Trawl distance had a significant negative effect in both subareas, with a steeper slope in 48.2, suggesting greater sensitivity of fishing efficiency to this operational variable. While environmental and operational variables accounted for part of the variability, the inclusion of the spatial effect was essential to capture regional heterogeneity, indicating that catch patterns are strongly spatially structured.
Spatial field analysis revealed differences in spatial autocorrelation between subareas. In 48.1, the spatial range was 49.6 km and the standard deviation of the spatial field was 0.28. In contrast, subarea 48.2 showed higher values (56.5 km and 0.36, respectively), indicating greater spatial heterogeneity and mobility of fishing effort. In both subareas, the GroupRho parameter provided no strong evidence of spatiotemporal dependence, although it was slightly positive in 48.2.