An Adaptive Sampling Design for Estimation of Thresher Shark Catch/Effort in a California Recreational Fishery

A stratified adaptive cluster sampling design is described to estimate the fishing effort and catch of thresher sharks in a recreational fishery along the coast of California. The use of a cluster sampling design is based on data collected in the NOAA SWFSC juvenile thresher shark survey, and research by the Pfleger Institute of Environmental Research and Scripps Institution of Oceanography. These studies and anecdotal reports from fishermen demonstrate aggregation of threshers into clusters. Anglers exploit these aggregations by departing in abundance from nearby marinas and launch sites and returning to the same locations after fishing. Since aggregation of threshers occurs over relatively large spatial scales and relatively small periods of time, capture of a thresher can be considered a rare event, the sampling of which has been identified as a challenge in the larger sampling context.

Data will be collected by interviewing anglers from returning recreational boats. The objective of the adaptive sampling procedure is to maximize the number of observations that contain a thresher shark relative to the number of vessels sampled. The design takes advantage of “prior” information, based on 5 years of sampling effort from the California Recreational Fishery Survey (CRFS) at selected landing sites for recreational boats. These landing sites include private marinas, public launch ramps, piers, breakwaters, and beaches.

A metric based on the number of sample sites per mile is used to define stratum boundaries. Boat trip effort, angler effort, and catch estimators are developed for stratified adaptive cluster sampling and used to estimate catch-per-unit-effort (CPUE), along with their variance estimates. The exact variance of angler effort is estimated using the product of two dependent random variables, the estimators of mean number of anglers per boat and of boat trip effort per day. A new sample data collection form designed specifically for this survey is proposed that quickly captures the needed data for the sample design, minimizing the number of “missed boats.” A theorem-based argument is provided to describe sufficient conditions for the adaptive cluster sampling design to outperform a simple random sampling design. In comparison to simple random sampling, this adaptive cluster sampling design is expected to provide more precision in catch and effort estimation where target species and fishing activities are patchily distributed.