We present a statistical method for intercalibration of fishery surveys methods, i.e. determining the difference in catchability and size selectivity of two methods, such as trawl gears or vessels, based on data from paired fishing operations. The model estimates the selectivity ratios in each length class by modelling the size distribution of the underlying population at each station and the size-structured clustering of fish at small temporal and spatial scales. The model allows for overdispersion and correlation between catch counts in neighbouring size classes. This is obtained by assuming Poisson-distributed catch numbers conditional on unobserved log-Gaussian variables, i.e. the catch is modelled using log-Gaussian Cox processes. We apply the method to catches of hake (Merluccius paradoxus and M. capensis) in 341 paired trawl hauls performed by two different vessels, viz. the RV Dr Fridtjof Nansen and the FV Blue Sea, operating off the coast of Namibia. The results demonstrate that it is feasible to estimate the selectivity ratio in each size class, and to test statistically the hypothesis that the selectivity is independent of size or species. For the specific case, we find that differences between size classes and species are statistically significant.
- Log-Gaussian Cox processes
- Mixed-effects models