ABSTRACT
A major challenge facing ecologists studying the earth as a system is the mapping of vegetation quantities over large areas. In the absence of remotely sensed data, only point-based interpolation methods, such as kriging other distance-related methods, are possible. The availability of remotely sensed data allows the use of whole-area approaches, using classification, statistical or physical models. Geostatistical methods such as co-kriging and conditional simulation may allow the combination of point-based and whole-area approaches to exploit more fully the available information. These alternatives may have very different consequences depending on the purpose of the vegetation map, as shown by an example constructed from a 36 km segment of imaging spectrometer data. The consequences of these methods must be considered before such maps are provided to geographic databases.
