The utility of catch-per-unit-effort when assessing and managing long-lived fish stocks

Abstract

The research in this Ph.D. dissertation focuses on the relationship between fishery-dependent catch-per-unit-effort (CPUE) and abundance and how its use when assessing and managing fisheries may be affected by some of its shortcomings. CPUE data are often used to provide information on the historical and current stock size, but changes in CPUE may not be proportional to abundance. Hyperstability is when CPUE declines slower than abundance and can result in optimistic estimates of abundance and risk of overfishing. Hyperdepletion is when CPUE declines faster than abundance and using these data in an assessment can produce pessimistic estimates of abundance. Assuming a proportional relationship between CPUE and abundance when CPUE actually has a nonlinear relationship with abundance can misinform management and result in overfishing or lost yield. Instead, the nonlinear relationship can be modeled and estimated. Three stand-alone chapters are presented in this dissertation with methods and results from analyses estimating the amount of nonlinearity in the relationship between CPUE and abundance and its implications on fisheries management. The first chapter investigated the nonlinearity between empirical CPUE data and abundance from orange roughy fisheries in Australia and New Zealand. Data from four orange roughy stocks were integrated using a Bayesian hierarchical state-space model to estimate the hyperparameters of a distribution for a nonlinearity parameter in the relationship between CPUE and abundance. Hyperdepletion was more probable than hyperstability and a prior distribution for an unknown stock showed a 83\% probability of hyperdepletion. This study is unique because it used data from the beginning of a fishery and created a prior distribution that may be used in future assessments. The second chapter used simulation to look at the ability to estimate nonlinearity in CPUE data using different models, assumptions, and types of data. A deterministic delay-difference model, a deterministic age-structured model, and a state-space delay-difference model were used to estimate the nonlinearity parameter of simulated CPUE data with a hyperstable, proportional, or hyperdepleted relationship to abundance. Estimates of the nonlinearity parameter were mostly unbiased, but highly variable. Using an informative prior distribution resulted in lower variance, but higher bias when the prior was not congruent with the true level of nonlinearity. Overall, the state-space model showed the best performance, and an informative prior distribution was useful as long as it is appropriate, justifiable, and wide enough to support all possible values of the nonlinearity parameter. The final chapter performed a management strategy evaluation to determine the consequences of managing a fish stock with and without estimating a nonlinearity parameter between CPUE and abundance. An age-structured operating model was used to simulate a true population from which CPUE and survey data were generated, where CPUE data were either hyperstable, proportional, or hyperdepleted. Estimation models with the nonlinearity parameter estimated or fixed at proportionality were used to estimate population trajectories from the simulated data. Two-area age-structured operating models with movement dependent on fishing effort and density-dependence were also investigated. The benefits to estimates of stock status and yield objectives when estimating a nonlinearity parameter were dependent on the true underlying relationship between CPUE and abundance. Hyperstable scenarios were especially risky, and should always be accounted for, or at least acknowledged, in a management strategy. Hyperdepletion scenarios, on the other hand, may seem to be less of a concern due to reduced conservation risk, but the negative consequences of reduced yield, at least in the short term, and a pessimistic view of the stock, and thus the management system, make it worthwhile to at least acknowledge the potential for hyperdepletion and that not accounting for it is a chosen management strategy. In the long-term, yield was similar whether or not nonlinearity was estimated, but if when estimating nonlinearity, risk to the stock was reduced when CPUE were hyperstable and although risk increased when CPUE showed hyperdepletion, stock status was typically always higher than the hyperstable scenarios. Nonlinearity in CPUE is common in many fisheries and it was seen that CPUE declined faster than abundance during the initial development of orange roughy fisheries. The presence of hyperdepletion can result in pessimistic views of stock status which may result in lost yield, the closure of fisheries, and conflict amongst user groups. Estimating a nonlinearity parameter in assessments and acknowledging that nonlinearity exists can improve management by bringing catches and stock depletion closer to target levels.