Mapping arid landscapes with multispectral and hyperspectralimagery

Abstract

This paper draws on several studies that have applied multispectral and hyperspectral imagery to the task of discriminating landscape composition in southern Australian rangeland environments. Imagery from Landsat Thematic Mapper, Geoscan II, Casi and Airborne Multispectral Scanner were compared and several feature extraction and mapping techniques evaluated. Hyperspectral imagery enabled more distinct vegetation and landscape components to be mapped, improved calibration of imagery enabled comparison with reference spectra and more certainty about these components, while both VNIR and SWIR regions yielded useful information about arid landscapes. Classification of imagery produced a thematic map suitable for land or habitat inventory, while mapping of individual landscape components is more likely to be useful for land condition assessment and monitoring.