Image-derived spectral endmembers as indicators of salinisation

Abstract

The objective of this research is to evaluate the utility of HyMap hyperspectral imagery for characterising and mapping irrigation-induced salinisation. Strategies for extracting and mapping spectral endmembers from HyMap imagery are assessed and a mixture-tuned matched filter approach adopted. Soil and vegetation indicators of salinisation are identified. Disturbed bare ground due to ploughing operations can be differentiated from bare ground due to salinisation. Three saline soil endmembers can be mapped and these relate well to the surface expressions of soil salinity as measured by ground geophysics. Increased salinisation is shown in the spectra of image-derived soil endmembers by the appearance of an infrared reflectance plateau between 800 nm and 1100 nm, hydrate-related absorption features at 1010 nm, 1180 nm, 1780 nm, 1950 nm and 2210 nm, the decrease in intensity (depth) of the hydroxyl absorption feature at 2200 nm and the widening of the water absorption features at 1950 nm and 1400 nm. Image-derived vegetation indicators of salinisation correspond with halophytic vegetation comprising the 'succulent' species of Samphire and Sea Blite and two species of native grasses. Classification of pixels according to the highest mixture-tuned matched filter score corresponds well with an independent assessment of the groundcover conditions prevailing at the time of image acquisition.