Catch estimation in the federal trawl fisheries off Alaska: a simulation approach to compare the statistical properties of three trip-specific catch estimators

Abstract

Quantifying catch has been recognized worldwide as a critical component in fisheries management. Assessment of discard is challenging because of the requirement for at-sea observation, which is both logistically difficult and costly to fishery agencies. Statistical estimators using robust sampling methods may yield accurate and imprecise estimates given the variability associated with many at-sea discard species and inability for agencies to obtain high sampling fractions. However, biased estimates occur if an inappropriate estimator is used. Using Alaska trawl fisheries as an example, we investigated the statistical properties and implementation issues for three commonly used estimators: the ratio estimator; a simple mean estimator; and a deterministic imputation method currently in use in federal fisheries off Alaska. We used a simulation approach to evaluate the performance of these estimators to estimate trip-specific catch. Several statistical properties were evaluated: bias of the estimators, variability of the estimators, and accuracy of the variance estimators. The simple mean estimator had the best performance for vessels landing catch at shoreside processors. The choice of estimator was less clear for vessels processing catch, owing to sensitivity associated with species composition and implementation issues for the simple mean and ratio estimators.