Thematic Mapper (TM) data were tested in a classification of general land use, but especially semi-natural habitats, in an intensively-farmed area of lowland Britain. The range of classes which could be identified, and the accuracy of the classifications which were much enhanced by the use of multitemporal (summer and autumn) data. Results were assessed by comparisons with census statistics, by scoring classifications field-by-field, and by comparing the c1assmap, pixel-by-pixel, with check-maps. In the cereal farming areas which make up most of the arable lowlands of Britain, TM classifications can provide accurate cover and distribution data, at the scale of the individual field, for most major land uses. No other method could realistically do this. With some caution, such methods might be adopted for routine operational use.