​​​​​Accounting for trophic interactions is becoming essential in the context of ecosystem-based fisheries management and ecosystem approaches are now an integrated part of strategic science plans for international organizations, including ICES.

Members of ICES Working Group on Multispecies Assessment Methods (WGSAM) recently published a paper in the Journal of Applied Ecology “Performance of a state-space multispecies model: What are the consequences of ignoring predation and process errors in stock assessments?". This was motivated by the wish to contribute to the understanding of multispecies systems. The group brings together experts in the field of multispecies interaction and food web modelling. According to lead author Vanessa Trijoulet, this paper is not only directly relevant to WGSAM, but also to other assessment working groups that want to understand the consequences of certain assumptions taken in fisheries models.

Currently, there are only a few multispecies stock assessment models in existence. Most stock assessments use single-species models where predation is subsumed into natural mortality, which is often assumed known. Little is known about the consequences of ignoring predation in fisheries models.

In parallel, single-species assessment models are becoming more and more advanced. For instance, many single-species assessments are now state-space assessment models, which model random variations in unobserved processes (for example variations in fish survival) as process errors.

Consequences of ignoring predation

In this paper, the authors investigate how ignoring predation and process errors in stock assessment models may impact our perception of fish stocks and as a consequence how this may affect fisheries management. The authors use a state-space multispecies model to simulate a true multispecies fisheries system and fitted different assessment models to the simulated data that differed according to whether or not they accounted for predation or process errors.

The study shows that while ignoring process errors led to limited bias, ignoring predation had a larger impact on stock perception. It can lead to large bias in model outputs and subsequent reference points. Accounting for unobserved random variations (process errors) in single species models was not enough to limit the bias due to ignoring trophic interactions.

Overfit data

“Fit to observed data" is one diagnostic used by stock assessors to check the performance of an assessment model. Often, single-species models tend to fit observed data (commercial and survey catches) as well, if not better than multispecies models. An interesting result of the study shows that when predation is not explicitly accounted for, single-species models can overfit the data and lead to biased outputs. This could go unseen by stock assessors if other diagnostics are not investigated. In the paper, looking at the predictive ability of the models enabled to easily identify this type of overfitting problems.

Ignoring trophic interactions that occur in marine ecosystems can induce bias in stock assessment outputs and result in biased reference points. While it may be difficult to estimate predation mortality, especially when no data on trophic interactions exist, severe misspecification of the natural mortality on fish could have large consequences on stock perception and reference point estimates and affect resulting management advice.

WGSAM enables research on predator-prey interactions to support the development of advice on the ecosystem approach to fisheries management. Their work relates to Ecosystem Science, Impacts of human activities, Emerging techniques and technologies, Seafood production, Conservation and management science, and Sea and society, which are six of ICES seven science priorities that support our Strategic Plan. Discover our seven interrelated scientific priorities and how our network will address them in our Science Plan: “Marine ecosystem and sustainability science for the 2020s and beyond" .