Image acquisition for detection of vegetation change based on long-term rainfall in an arid rangeland in Western NSW, Australia

Research on vegetation change, rangeland assessment or desertification modelling in drylands using remotely sensed image acquisition normally ignores long-term rainfall as a key criterion in image acquisition. This article will present a novel procedure for image acquisition to investigate vegetation change in a degraded rangeland located in Western New South Wales (Western NSW) Australia. Western NSW experienced an unusually prolonged period of rainfall deficit during the 2000s compared to the 1970, 1980 and 1990s. For this purpose, vegetation changes were assessed using Landsat images supported by field survey. The long-term rainfall variability (42-year) was regarded as a key element in image acquisition. Within the timeframe of the 2000s, 2 years with 25 % lower than the 42-year mean annual rainfall were selected. These images were then compared to an image captured in a year (1988) with rainfall closer to the 42-year mean annual rainfall. Two change detection techniques were used, namely univariate image differencing and GIS approaches. Classification of the produced images was pursued based on the digital numbers (supervised) of ground-checked points within the reference image whilst considering the histogram (unsupervised) of each digital number of the produced image. This research emphasized rainfall as a key variable in image acquisition for vegetation change analysis in rangelands. Image acquisition based on long-term rainfall data allowed for the assessment of changes in perennial plant cover by eliminating the effects of extreme rainfall variation on annual grass dynamics and removing extreme reflections caused by their temporary high photosynthetic activity.