Abstract
A computer simulation model of a fish population was developed, capable of tracking individual fish life histories, including monthly growth and natural mortality. A simulation model of the fishing activity of a fleet was developed which incorporates details of individual vessel effort and catches (including fish size and age information). Simulated data from these models were used to evaluate three algorithms to estimate catch and effort in recreational fisheries: (1) a form of the traditional, multiplicative approach which is typically used to estimate catches for recreational surveys, (2) a newer approach which uses survey data to drive a simulation model and which has been used to estimate catches for the National Marine Fisheries Service Large Pelagic Survey, and (3) a new approach developed for this study which uses bootstrapping methodology in a multiplicative framework.The accuracy and precision of the approaches were evaluated for different sampling levels and given that various proportions of the fleet were known. For survey conditions where a large fraction (about 60%) of the fleet was known and where weekly effort and catch rate sample sizes were reasonably large (40 observations for each), all three approaches produced nearly identical, unbiased point estimates. However, confidence interval widths for the simulation approach were underestimated. For conditions where a smaller fraction of the fleet was known, a positive bias in the point estimates was introduced, particularly for the multiplicative approach, and confidence interval widths for the multiplicative approach were greatly overestimated. The bootstrapping approach produced fairly robust point estimates and valid confidence intervals for nearly all scenarios, although the bootstrap point estimates became somewhat biased and confidence intervals were underestimated when sample sizes were low. Nevertheless, the bootstrap method was clearly the best performing approach for the conditions tested.Techniques were developed to assess the contribution of each estimation component to overall variance as a means to develop guidelines and procedures for optimal stratification within a limited sampling program. These survey design techniques showed that the multiplicative approach was highly sensitive to the known fleet fraction and that weekly estimates for all methods were imprecise for all conditions tested.