Evaluating assumptions behind design-based estimators for unreported catches

Abstract

Understanding a fishery’s impact on the marine ecosystem requires a quantification of total catches, which include unreported catches. For recent years in Norwegian waters, unreported catches have been estimated using data collected by the Norwegian Reference Fleet, a fisher self-sampling programme that regularly gathers data on catches of all species (including unwanted bycatches and discards). In this study, we focused on the use of design-based estimators for total catches in offshore fisheries, which have previously been used to estimate discards in the Norwegian coastal gillnet fisheries. After adapting the current methodology to the data available in offshore fisheries, we explored the assumptions behind both unit- and ratio-based estimators, and the effect of ignoring the cluster sampling design. Using a jack-knife resampling method to estimate the true bias in estimates of total catches and associated variability, we found that ignoring the cluster sampling design tended to underestimate the variability, which lead to occurrences where unreported catches were statistically detected when in fact there was too much uncertainty to make such a conclusion. Further validations suggested the cluster unit estimator is not unbiased as theoretically expected due to the sampling design favouring the selection of more active vessels. We therefore concluded that the unit and ratio cluster estimators are applied and compared, as per best practices.