Abstract
Over the past 10 years, fisheries scientists gradually adopted geostatistical tools when analysing fish stock survey data for estimating population abundance. First, the relation between model‐based variance estimates and covariance structure enabled estimation of survey precision for non‐random survey designs. The possibility of using spatial covariance for optimising sampling strategy has been a second motive for using geostatistics. Kriging also offers the advantage of weighting data values, which is useful when sample points are clustered. This paper discusses, with fisheries applications, the different geostatistical models that characterise spatial variation, and their variance formulae for many different survey designs. Some anticipated developments of geostatistics related to multivariate structures, temporal variability and adaptive sampling are discussed.