COST-EFFICIENT SURVEY DESIGNS FOR ESTIMATING FOOD-CONSUMPTION BY FISH

This study examines the effect of survey design on the precision of estimates of average weight of stomach contents of fish. Since prey distribution is often patchy, fish at a station tend to have more similar stomach contents than do fish from different stations. Theory for evaluating the effects of intra-cluster correlation and variable density on the variance of estimates of average stomach contents is developed, and the implications for stomach sampling programs are examined. It is demonstrated that because of intra-cluster correlation the number of stations, more so than number of stomachs collected, determines the precision of estimated average stomach contents. As an example, the survey design for estimating the average amount of capelin (Mallotus villosus) in Barents Sea cod (Gadus morhua) stomachs in the winter is assessed, using survey data from years with high and low capelin abundance. The results, in agreement with the theoretical findings, show that precision would be maximized for fixed cost if fewer fish were collected from each station, and the freed resources were used to collect fish from as many stations as possible. In 1992, the survey design was changed accordingly; stomachs are now collected from twice as many stations, but the total number of stomachs collected has been reduced. Sampling fewer stomachs at each station enables most stomachs to be analyzed at sea.