The adjustment of frequency distribution of cover types within sections of the land cover map of Great Britain

The Land Cover Map of Great Britain consists of a mosaic of 30 classified map sections derived from 46 Landsat Thematic Mapper images captured between 1987 to 1990 (Fuller et al. 1994a). Some 88 per cent of Britain was mapped from combined summer (mid-May to late July) and winter (mid-October to mid-March) images. Persistent local cloud cover meant that 11 per cent of Britain was mapped using only single season imagery, and 1 per cent was not mapped. The target summer and winter periods had to be extended up to six weeks outside the ideal range for 32 per cent of Britain (Fuller et al. 1994b). Use of suboptimal imagery causes inconsistencies and inaccuracies between classifications. When data is summarized onto a coarser grid differences between sections can sometimes become intrusive where they join. For many users it is important to estimate any known bias and, as far as possible, compensate far it in summary statistics. For such users removing the ostensibly discontinuous image boundaries may produce summary maps with improved appearance and allow better integration with other data sets when trying to produce correlative statistics. We present a method which can be used to removed local bias in summary data due to known misclassifications in the original data.