Method for estimating variability due to sampling of catches on a trawl survey

Abstract

Many catches arising on trawl surveys of fish stocks are sampled for the measurement of fish lengths and the collection of otoliths for age determination, thereby increasing the variability of the estimated abundance indices, expressed as catch per hour by species and year-class. This paper demonstrates how catch-sampling variability can be assessed, using the English groundfish survey of the North Sea (EGFS) as an example for analysis. Unbiased estimators are derived for catch-sampling covariance matrices for numbers-at-length of a species by size category, by haul, and by sampling stratum and for numbers-at-age by stratum. The same statistical principles could be applied to derive appropriate estimators for other surveys having different catch-sampling procedures. A method is also put forward for estimating the relative importance of variability due to catch sampling and that due to the design of the survey, i.e., due to the number, location, and timing of fishing stations and the method of fishing. Analysis of a small set of EGFS data for haddock (Melanogrammus aeglefinus) is included for illustrative purposes. Implications for survey design and catch-sampling efficiency are briefly discussed.