The development of image processing techniques to assess changes in green vegetation cover along a climatic gradient through Northern Territory, Australia

Abstract

The green vegetation cover gradient in three regions referred to as subtropical, semi-arid and arid in Northern Territory was assessed in terms of spectral reflectance characteristics. Incremental vegetation cover over different background components (combined spectra) resulted in three vegetation types being defined, Type 1, darkening Type 2 partial darkening and Type 3, highly near-infrared (NIR) reflective. Spectral plots on field sites showed the subtropical region as being typified by high NIR reflectance, the semi-arid curves showed extensive darkening, while the arid region curves show less darkening due to a lower vegetation cover and concomitant high soil reflectance. Results of analysis of variance of regression using different combinations of single bands, band transforms and principal component data resulted in the single band, MSS-5 providing the most useful predictive data in the subtropical and semi-arid areas. Normalized albedo on MSS-5 proved most useful in arid areas. Transforms involving relationships between near-infrared and visible wavebands were not ranked as suitable predictive tools for the semi-arid region. Prior to classification, single bands and band transforms were ranked. Only those showing high significance levels in relation to green vegetation cover were used as input data. Results showed that classification maps using maximum likelihood classifiers can be uniformly applied throughout all three regions resulting in high levels of accuracy in terms of map areas and community types. The classification technique separates out darkening (Type 1) and NIR reflective (Type 3) classes with the threshold of separation for the two types standing at 45-50 per cent green vegetation cover.