In the Hawaii-based longline fishery, changes in fishing operations to target different species produce changes in the effectiveness of fishing effort units. Catch-per-unit-effort (CPUE) indices used in resource monitoring were improved by segregating dissimilar types of fishing effort. Cluster analysis was used to classify longline sets in relation to species composition of the catches. Based on proportions of eight species and three broader species groups in 46 961 longline sets from 4 years (1991-1994) of commercial fishery data, five effort clusters were identified. Spatial distribution of sets and differences in fishing operations among clusters were then compared to reveal apparent differences in fishing strategies. Three clusters comprised similar to 80% of the total sets, and the catch compositions suggested targeting for either broadbill swordfish (two clusters) or bigeye tuna. The other two clusters were most similar to the tuna cluster, but their catch compositions indicated a mixed-species fishing strategy. Fishing operations were most different between sets in the tuna and swordfish clusters. Swordfish sets were characterized by (1) the largest vessels, (2) the least number of hooks per set, (3) the greatest number of lightsticks, (4) the longest set duration, (5) the highest percentage of night sets, (6) a larger percentage of sets within the full moon phase, and (7) the lowest percentage of sets within the main Hawaiian Islands Exclusive Economic Zone. Time series of CPUE for three species (bigeye tuna, yellowfin tuna, and swordfish) based on different clusters were compared, and the most appropriate CPUE time series for resource monitoring are recommended. (C) 1997 Elsevier Science B.V.
