Can the data from at-sea observer surveys be used to make general inferences about catch composition and discards?

Some fishery characteristics such as total discards are often inferred from data collected by at-sea observers on a subset of fishing trips. Such inference is predicated on the assumption that observed and unobserved trips are statistically exchangeable. There are two principal reasons why this may not be so. A deployment effect results from nonrandom distribution of observers among sampling units. An observer effect results from changes in fishing practice or location when observers are present. Both effects can impact the precision and accuracy of fishery-level inferences drawn from observer data, though this is rarely addressed quantitatively. We found evidence for deployment and observer effects in Gulf of St. Lawrence fisheries. The impact of deployment bias was further quantified by resampling from fisheries data collected with 100% observer coverage. We conclude that the nature of the effects observed in our study preclude merely correcting observer-collected catch data for possible biases and imprecision. Furthermore, regulatory compliance monitoring by observers in the existing program may not be completely effective. Modifications to program structure would therefore be beneficial and some suggestions are evaluated in this paper.