Efficient sampling design for larvae of American lobster

For 13 years, lobster larvae were collected from the same 23 stations at ap proximately weekly intervals for the duration of the larval season (Scarrat t, 1964). Using the most complete 8 years of collections, we produced sever al data subsets for each year by reducing the number of sampling dates and the number of stations per date. Our analyses were confined to the fourth a nd final larval stage, the most likely stage to be correlated with fishery yields. The means and confidence intervals for annual larval abundance were calculated on the data subsets and compared with the full dataset. We conc lude that the usual lobster larval sampling frequency of weekly, or more of ten, was unnecessary. Tows taken at 10 or 14 stations per week for 4 weeks of the 9 week larval season represented well the mean and confidence interv als on the full dataset. These choices remained the best after the costs of sampling were included. Further cost reduction could be achieved without l oss of precision by removing collections from the net after every second, r ather than after every tow. Mean annual density was found to be a good inde x of annual larval production. Thus, the collections during 4 weeks represe nted annual abundance and annual production.