Spatially integrated modelling of data-limited orange roughy (Hoplostethus atlanticus) using environmental covariates
Spatial stock assessment models are recognised as increasingly important for estimation of stock status and a sustainable exploitation rate. The inclusion of movement between spatial units within a model is difficult, because the data requirements are high. However for populations with low levels of spatial exchange it is possible to reduce the data requirements by distributing information on biological parameters between neighbouring units, or units with shared environmental conditions. This can allow spatial modelling to be applied even in data-limited situations. We develop this approach here through application to orange roughy (Hoplostethus atlanticus) sub-populations inhabiting neighbouring seamounts in the South Pacific. Despite limited data for each seamount, we were able to simultaneously fit multiple, localised, process-based models of the depletion dynamics. This was achieved by sharing information on the unexploited population size via known environmental covariates, with the relationship estimated in a hierarchical and integrated manner during the model fit. Cross-validation demonstrated that this approach can compensate for a lack of seamount-specific abundance data and improve ability of the model to estimate localised depletions.