Techniques for computer-assisted mapping of rangeland change.

Abstract

Management of grasslands subject to replacement by woody species requires an understanding of the scales and patterns of change and how to detect and express them. We used the Jornada del Muerto Basin of southern New Mexico as a case study, testing the suitability of Landsat Multi-Spectral Scanner (MSS) data for detecting vegetation changes. Cycles of drought and heavy grazing have apparently changed the once extensive grasslands of the upland Jornada and surrounding areas gradually into a patchwork of shrublands and relict grasslands. Integrated remote sensing and geographic information system (GIS) techniques can facilitate automated detection of these rangeland changes. A GIS was used to store and process two 4-band, co-registered multi-temporal Landsat MSS scenes collected in July 1983 and August 1992. Scene-to-scene radiometric calibration was performed using a regression technique. The data were then evaluated for changes 3 different ways using algorithms based on differences between the 'red' (chlorophyll absorption) bands for the 2 scenes; the Euclidean distances between the 'red' and 'near-infrared' bands for the 2 scenes; and a standardized principal components analysis using all 8 MSS bands. A threshold of 3 standard deviations above the mean was applied to each of the 3 resulting 'change' images to represent areas of extreme change. Correlations among these images ranged between 0.83 and 0.95. We conclude these techniques can identify successfully the patterns and extent of extreme change, and thus have potential value for management of our rangelands resources.