ABSTRACT
Different representations of the natural environment form the foundation of any spatial information system for natural resource management. However, the development and selection of appropriate representations of the natural environment present an array of diffi culties to the researcher or system designer. Many of these diffi culties relate to classical problems in geographic information science (GIScience). In the context of the natural environment, these problems can be classifi ed into three categories. The fi rst of these categories is context and meaning, where the meaning of terms used to describe the natural environment may be ill-defi ned, or vary between specifi c information communities. The second category is uncertainty in information, in particular to imprecision and inaccuracy. Any measurement of the natural environment is subject to imperfection, which must be taken into account when developing different representations. The third category is dynamism, since the natural environment is constantly changing, presenting substantial problems to any attempt to represent it. Effective representation of dynamism must go beyond mere snapshot models of change over time, to explicit modeling of processes and events. This chapter explores these three key issues, using as an example the automated fusion of heterogeneous land cover data sets.
