The formulation of policies influencing land cover requires accurate information on the distributions of competing land cover types. This paper investigates the use of a numerical land classification (derived by the Institute of Terrestrial Ecology, Merlewood, UK) in the survey and prediction of land cover. The classification groups 1-km grid squares with similar environmental characteristics into distinct landscape types (ITE land classes). The ITE land classes were used to stratify sampling effort in a field survey of land cover within the catchment of the River Tyne in northeast England in 1991. Variation in land cover within and between land classes is analysed and the principal patterns of land cover within the catchment are described. The use of the land classes in the prediction of land cover within unsurveyed 1-km grid squares is discussed in light of the observed variation in land cover. It is concluded that stratification of field surveys using land classification can aid the identification and interpretation of broad patterns of variation in land cover, but that intra- and inter-land class variation in land cover is too great to allow the use of ITE land classes in the prediction of land cover within individual unsurveyed grid squares.