Fitting data that vary continuously over an area of land into a discrete data model can introduce a high level of error. The work done in this thesis deals with this problem by exploring the use of alternative data structures and processing methods to represent better those features of the environment being analysed. The fuzzy c-means (FCM) classification algorithm has been used to measure the variation of geographic features over a spatial domain, and to output this information in the form of a membership raster for each feature. A membership raster is similar to a raster-based digital elevation model (DEM). Therfore a triangular irregular network (TIN) can also be used to represent a membership raster. A TIN representation requires less storage space than a membership raster, and allows the algorithms that have been developed for processing terrain information to be used for processing membership values.