Predicting abundance indices in areas without coverage with a latent spatio-temporal Gaussian model

Abstract

A general spatio-temporal abundance index model is introduced and applied on a case study for North East Arctic cod in the Barents Sea. We demonstrate that the model can predict abundance indices by length and identify a significant population density shift in northeast direction for North East Arctic cod. Varying survey coverage is a general concern when constructing standardized time series of abundance indices, which is challenging in ecosystems impacted by climate change and spatial variable population distributions. The applied model provides an objective framework that accommodates for missing data by predicting abundance indices in areas with poor or no survey coverage using latent spatio-temporal Gaussian random fields. The model is validated, and no violations are observed.