Estimating relative abundance from catch and effort data, using neural networks

摘要

We develop and test a method to estimate relative abundance from catch and effort data usingudneural networks. Most stock assessment models use time series of relative abundance as theirudmajor source of information on abundance levels. These time series of relative abundance areudfrequently derived from catch-per-unit-of-effort (CPUE) data, using general linearized modelsud(GLMs). GLMs are used to attempt to remove variation in CPUE that is not related to the abundanceudof the population. However, GLMs are restricted in the types of relationships between theudCPUE and the explanatory variables. An alternative approach is to use structural models basedudon scientific understanding to develop complex non-linear relationships between CPUE and theudexplanatory variables. Unfortunately, the scientific understanding required to develop theseudmodels may not be available. In contrast to structural models, neural networks uses the data toudestimate the structure of the non-linear relationship between CPUE and the explanatory variables.udTherefore neural networks may provide a better alternative when the structure of the relationshipudis uncertain. We use simulated data based on a habitat based-method to test the neuraludnetwork approach and to compare it to the GLM approach. Cross validation and simulation testsudshow that the neural network performed better than nominal effort and the GLM approach. However,udthe improvement over GLMs is not substantial. We applied the neural network model toudCPUE data for bigeye tuna (Thunnus obesus) in the Pacific Ocean.