U.S. Department of Commerce


Date of this Version



Fisheries Research 174 (2016) 219–233


U.S. Government Work


Estimation of at-sea discards is an issue that has received considerable attention worldwide. With this increasing focus, there is a need for greater precision of weight estimates for less common and fishery-limiting species. While one solution is to mandate full (100%) observer coverage to reduce uncertainty in estimation at the trip level, variance from on-deck sampling methods (e.g. within-haul) should also be properly addressed. Commercial fishing vessels are not perfect sampling platforms and all sampling methods suffer from implementation issues that potentially impact the quality of the data collected and the resulting estimates. We conducted a cooperative study with industry to evaluate two observer sampling methods on trawl vessels delivering their catch to shore side processors. The alternative observer sampling method that targets the portion of the haul that would be discarded directly at-sea, relying on shore side reports of retained catch to generate total catch estimates, was tested against the standard methods currently used by the NMFS North Pacific Groundfish and Halibut Observer Program that sample the entire catch, both retained and discarded portions (combined). Methods were tested simultaneously by deploying two observers to sample each haul on study trips within three Gulf of Alaska trawl fisheries that varied widely in amount and species composition of discards: Rockfish Program, arrowtooth flounder, and shallow water flatfish. Although the alternative method was successfully implemented in two of the three fisheries, logistical constraints decreased sampling effectiveness in the third. In some situations, observers were unable to collect multiple samples under both methods, preventing variance estimation. This occurred more often for the observer using standard methods. Detection of less common and rare species was higher using the alternative sampling method. Discard estimates from the two methods were found to be significantly different in two of the fisheries examined (Rockfish Program and arrowtooth flounder). Discard estimates under the alternative method tended to have smaller variances than for the standard method, although this was not universally the case. These results provide an important comparison of the relative performance of different on-deck sampling methods under varying catch conditions and fisheries.