Fisheries management is a domain where decisions under uncertainty carry enormous ecological and economic consequences. Overfishing can collapse stocks and devastate coastal communities; underfishing leaves economic value on the table. Bayesian methods have become standard in fisheries science because they propagate uncertainty from noisy data through complex population models to the management quantities that matter — sustainable yield, stock status, and the probability of overfishing under proposed harvest rules.
Bayesian Stock Assessment
Stock assessment models estimate the current biomass, fishing mortality, and recruitment of a fish population from a combination of catch data, fishery-independent surveys, age composition, and biological information. Bayesian stock assessments produce posterior distributions of key management quantities rather than point estimates, enabling risk-based decision-making.
Key Parameters r = intrinsic growth rate
K = carrying capacity (unfished biomass)
Cₜ = catch in year t
MSY = r · K / 4 [maximum sustainable yield]
B_MSY = K / 2 [biomass at MSY]
The posterior distribution of B_current / B_MSY tells managers whether the stock is overfished, while the posterior of F_current / F_MSY tells them whether overfishing is occurring. These probabilistic statements directly inform harvest control rules.
Catch-at-Age and Statistical Catch-at-Age Models
Modern statistical catch-at-age (SCAA) models track the abundance of each age class through time, fitting to age-composition data from commercial catches and surveys. Bayesian implementations like Stock Synthesis (SS3) and CASAL estimate dozens of parameters — selectivity curves, natural mortality, stock-recruitment relationships — with full uncertainty quantification. Hierarchical priors on natural mortality, informed by meta-analyses across species, help stabilize estimates for data-poor species.
Many of the world's fisheries lack the long time series and age-composition data that conventional assessments require. Bayesian data-poor methods use informative priors derived from life-history theory — relating natural mortality to growth rate, maximum age, or temperature — to constrain stock assessments when data are scarce. The prior makes the borrowed information explicit and allows its influence to diminish as local data accumulate.
Management Strategy Evaluation
Management strategy evaluation (MSE) is a simulation framework that tests the robustness of harvest control rules against a range of uncertainties — in population dynamics, observation error, and implementation. Bayesian operating models generate plausible alternative realities by sampling from the posterior distribution of stock assessment parameters. The performance of each harvest rule is then evaluated as a distribution of outcomes (yield, biomass, probability of collapse) across these scenarios, providing managers with a risk profile rather than a single projection.
"The question in fisheries is not what the stock size is, but what is the probability that the stock is below the level that can sustain fishing. The answer demands Bayesian thinking." — Ray Hilborn, The Nature of Natural Resources
Ecosystem and Multi-Species Models
Bayesian ecosystem models — such as Ecopath with Ecosim fitted via MCMC — estimate trophic interactions, predator-prey dynamics, and the ecosystem effects of fishing. Multi-species virtual population analysis uses Bayesian methods to estimate predation mortality, recognizing that the fate of a herring stock depends not just on the fishing fleet but on the abundance of cod. These models inform ecosystem-based fisheries management, which considers the broader ecological context of harvest decisions.
International Applications
The International Whaling Commission uses Bayesian stock assessments to set catch limits for aboriginal subsistence whaling. The Commission for the Conservation of Antarctic Marine Living Resources uses Bayesian decision rules for krill management. Tuna regional fisheries management organizations worldwide use Bayesian SCAA models. In each case, the Bayesian framework enables transparent treatment of uncertainty and precautionary management in the face of incomplete data.