Spatial awareness: Good practices and pragmatic recommendations for developing spatially structured stock assessments

Abstract

Spatial population structure is a fundamental aspect of marine populations, yet it is rarely incorporated in stock assessment models. The unit stock assumption, a common feature of many assessments, is an axiom of the single-species modeling convention, which developed in the mid-20th century when the spatial resolution of fishery data was coarse and uncertain, computing power limited, and the impact of biocomplexity on sustainable harvest levels not well understood. Despite rapid advances along all of these research fronts in the 21st century, spatial assessments remain rarely utilized as the basis of management decision-making. A potential hindrance to broader utilization of spatial stock assessments is the lack of guidance on how to choose an appropriate spatial modeling framework for a given application. Thus, we review the types of spatial assessment models available, summarize options to parameterize population structure, offer guidance to promote the development of candidate spatial assessment models for application in management procedures, and provide a pragmatic guide for choosing a spatial assessment model given observed spatial structure, data limitations, and management concerns. A spatial assessment should match the assessment unit(s) to interdisciplinary stock identification of unit populations, adequately represent the spatial structure within an assessment unit, simultaneously model multiple interacting population units, when appropriate, and provide outputs for managers that can be used to prevent local depletion of spawning populations. As data permits, higher resolution intra-population spatial structure can be addressed to increase homogeneity within the modeled unit, while connectivity among population units may also be explicitly incorporated. Small sample sizes can limit spatial assessment applications, but incorporating novel data sources and parameterizing models efficiently (e.g., sharing parameters among spatial units, implementing habitat preference functions to improve movement dynamics, and including spatial autocorrelation) can resolve some practical constraints. Management strategy evaluation should be more widely utilized to identify minimally complex management procedures that provide robust advice, while imprecision of spatial models should be weighed against the inherent bias of spatially aggregated assessments.