Causal modelling of climate-fish-fisheries: confronting models with data

Understanding the interplay of environmental and anthropogenic drivers is essential to interpret past changes and anticipate future dynamics of marine fish stocks. Using Norwegian spring-spawning herring (NSSH) as a case study, we applied an iterative structured causal modelling framework combining causal knowledge analysis and inference-using structural equation models-to quantify the respective effects of climate and fisheries on stock biomass and catches. Initial simple models, grounded in prior ecological understanding, yielded implausible estimates when confronted with observational data, underscoring the necessity to treat data structure and content as integral to model design. We gradually incorporated multiple causal pathways, feedback loops, and confounders. Final models explained over 90% of observed variation in both biomass and catch and showed strong support for the roles of recruitment, prior biomass, and total allowable catch (TAC) advice. We found no causal effect of ocean climate variables (Relative Heat Content and Arctic Water Content) on year-to-year changes in NSSH biomass, nor of Spawning Stock Biomass on catches. Instead, catch levels were primarily governed by TAC advice and the degree of agreement between fishing nations. We conclude that reliable causal inference in ecological systems requires iterative model refinement, causal knowledge analysis, and explicit accounting for data structure. Adopting and transparently reporting this approach would benefit ecological studies used to support fisheries management.