Abstract
Conventional classification of remotely-sensed data seeks to allocate observations into distinct classes. For many regions, such as lowland heaths, this approach may be inappropriate since the vegetation types lie along continua and do not exist in a well-defined mosaic. In conditions like these the adoption of ordinatory techniques may be preferable to those of classification and the results may be displayed as probability surfaces. This paper presents the results of a preliminary investigation which demonstrates the use of such an approach for heathland in Surrey.