Precision in systematic trawl surveys as assessed from replicate sampling by parallel trawling off Namibia

Abstract

Following independence of Namibia in 1990, abundance of that country's hake stocks was monitored by trawl surveys conducted by the Norwegian F.R.V. Dr Fridtjof Nansen using a systematic survey design. Precision from such designs is considered to be better than with a random design, but it cannot be quantified through standard analysis of variance. In 1998 and 1999, the trawl surveys were duplicated in full using a commercial trawler that operated in parallel with the Dr Fridtjof Nansen. Both vessels had the same fishing gear and rigging. In both years, biomass and distribution patterns estimated by the vessels were similar. The paired datasets collected from the surveys were used to analyse variability in the point samples. Absolute differences between pairs of catches were, in general, proportional to the catch, but they varied randomly following a normal distribution around the catch level. Local variability was analysed as random noise, modelled and later reapplied on single vessel data series to evaluate the effect of sample size and replicates on the survey mean. In a survey with 200 stations, local sample noise accounted for about 4% of the variability in the survey mean. Alternatively, running series of simulated surveys by bootstrapping on pairs of catch data gave similar results, when a small systematic vessel effect was adjusted for. The main statistical techniques applied were less susceptible to outlier catches than straightforward correlations or regressions and could therefore, perhaps with some advantage, also be used to estimate the vessel factor when intercalibrating trawl survey vessels.

Keywords: intercalibration, point precision, replicate sampling, trawl surveys

African Journal of Marine Science 2001, 23: 385–396