Separating wild versus stocking components in fish recruitment without identification data: a hierarchical modelling approach

Abstract

Salmonid juvenile stocking programs are often poorly monitored due to the lack of identification between stocked and wild fish. In this study, a hierarchical Bayesian model is developed to take advantage of spatiotemporal variations of stocking and wild recruitment for estimating these two components despite the absence of identification data. It is first tested by means of simulated data and then applied to the 37 year abundance data set of the Atlantic salmon (Salmo salar) population of the Allier catchment (France). Despite the absence of identification data, juvenile densities could be estimated and split into wild and stocked components. We found that the stocked juveniles contributed significantly to the total juvenile production, while the wild reproduction continued to provide an important contribution. This approach is encouraging and promising from a management advice perspective. It is flexible enough to accommodate for case study specificities and shows that long-term monitoring abundances can be useful to assess the impact of stocking programs even in the absence of direct means of identifying stocked versus wild fish.