ABSTRACT
Tracking of changes on the landscape is central to assessment of ecosystem health and sustainability, although implications of change phenomena are not always straightforward (Costanza, 1991; Jones et al., 1997; O’Neill et al., 1997; Picket and White, 1985). Determining and recording changes on a parcel basis at local scale is a fairly mechanical matter when modern geographic information systems (GISs) are utilized, but becomes much more problematic for large areas when funds are limited. In the latter case, it becomes a practical necessity to exploit remotely sensed data obtained electronically from instrumentation carried by satellites and transmitted to strategically located receivers. Remotely sensed data are typically recorded in a grid arrangement with a set of spectral reflectance readings for each cell of the grid. The readings correspond to energy of selected wavelengths that is integrated (averaged) over the sector of Earth’s surface encompassed by the cell. Remote sensing specialists use the technical term pixel (for picture element) instead of the more generic cell terminology (Wilkie and Finn, 1996). The area covered by a pixel (cell) is usually too large to provide high resolution regarding geometric detail of objects on the ground (Lillesand and Kiefer, 1994). Differences in landscape features must be inferred from computed differences among spectral readings, as opposed to detailed visual examination.
