Abstract for presentation at The 13th Australasian Remote Sensing and Photogrammetry Conference

There is a growing demand for up-to-date mapping of perennial pastures, in particular dryland lucerme (Meticago sativa). Dryland lucerne is a deep rooted perennial pasture that is recognised as an important pasture for controlling rising ground waters associated with dryland salinity. Satellite imagery has the potential to provide an objective repeatable method for mapping perennial pastures at moderate catchment scales. This study evaluated multispectral imagery for mapping dryland lucerne southern Australia. It aimed to develop a mapping methodology that might be more widely applicable, and to identify the limitations to image-based discrimination.
Summer was considered the optimum time of year to spectrally distinguish paddocks of green dryland lucerne from annual pastures, and SPOT imagery was acquired in February 2006 over two study areas in South Australia, Jamestown and Upper South East (USE). Normalised Difference Vegetation Index (NDVI) and multispectral classifications were compared for their discrimination of summer-green lucerne from other pastures, weeds and native vegetation. NDVI values were calibrated against field measurements of lucerne cover collected near the time of image acquisition. Lucerne maps were validated on a per-paddock basis using roadside observations through the study areas.
The mapping of dryland lucerne was more accurate in the Jamestown area than in the USE, where a wider range of perennial vegetation, samphire and summer weeds created spectral confusion with lucerne. Climate, soil moisture and land management practices all contribute to regional vegetation cover characteristics that influence the ability to discriminate dryland lucerne on the basis of summer spectral contrast.