Estimation of catch parameters from a fishery observer program with multiple objectives

Observer programs around the world collect information about fishing activities used to manage species of commercially important fish species as well as untargeted co-occuring species that are impacted by fishing activities. The U.S. government oversees several observer programs including the North Pacific Groundfish Observer Program (NPGOP) that monitors fishing activities in Federal waters off of Alaska. The NPGOP is the largest observer program in the world with respect to yearly numbers of deployed observers and the extent of data collected. Observers working in the NPGOP undergo extensive training and use sampling methods that are rigorous with respect to randomization, however, there are some inadequacies in both the size of samples and the way data are ultimately archived that inhibit unbiased estimation of various catch parameters and associated variances. Others have made important strides in illuminating areas where the data collection procedures can be improved, but there has been no systematic description of appropriate estimators and their corresponding assumptions nor has there been exploration of the levels of sampling effort in the various sectors of the Alaska groundfish fishery that would be optimal with respect to the many catch parameters estimated from the resulting data. This dissertation provides answers to these important gaps in knowledge and in doing so, also provides a general construct that can be applied in other observer programs with similar deficiencies. Using these derived estimators I provide methods and examples for optimizing observer coverage to different vessel size classes in the Alaskan groundfish fisheries. The optimization can account for any number of catch parameter estimates and provides a transparent and methodical procedure for managers to determine observer coverage allocation. I also recommend changes in data collecting and reporting procedures that will require fewer model assumptions for making inferences from NPGOP data.