Abstract Three methods of processing LANDSAT 5 Thematic Mapper data were used to map the surface sediment distribution on the intertidal areas of The Wash, east England. One method used a maximum likelihood classification to define each picture element (pixel), that equated to a 30 m square on the ground, as having a muddy or sandy sediment. The other methods used multiple regression (MR) and spectral mixture modelling to quantify the amount of each sediment type within a pixel. All methods produced an acceptable description of sediment when compared with the known sediment composition of 192 sample sites. Muddy sites were more accurately defined than sandy ones by all methods. Image-derived estimates of sediment composition obtained by the MR method proved to be a robust substitute for field-derived values in MR models relating macrobenthos density to sediments and other site characteristics.