Trawl survey estimates of the distribution of fish population characteristics are not likely to give an absolute true reflection of the population. Errors and biases can be introduced in sampling and calculation techniques of the population characteristics. Analysis done on trawl survey sampled length data of Namibian Merluccius capensis indicates a better precision of length frequency distributions if spatial covariances were included in the calculations. Spatial covariance was detected using experimental semi-variograms, from which variogram models were derived and used in kriging. Arithmetic and kriged means of the density (#/nm2) of M. capensis' length groups were calculated, and 97% of the length groups gave lower coefficient of variation using the geostatistical method. Hence, the geostatistical approach provided improvements of the population characteristic estimates for a low cost.