HyMap acquires images over the spectral range 450 - 2500 nm (Visible, Near Infrared and Shortwave Infrared) in 128 bands. For this study it was flown at an altitude to give 5m pixels over a 2.5 km wide swath. The image below left shows a synthetic true colour composite of the HyMap imagery for the Pyramid Hill test site acquired in May 1999. The salt scalds show up clearly as bright, white, areas. The association of soil salinity with palaeo-channels having low elevation is clearly visible.

The high spatial and spectral resolution of HyMap allows for spectral components within the image to be reliably "unmixed" and mapped (Taylor et al, 1998). Spectral feature fitting techniques were used to quantitatively assess the contribution of the salt scald spectra acquired in the field (described above) to the signature of each pixel (image right). For this analysis the data affected by uncombined water absorptions around 1400 and 1900 nm was excluded. This image is an accurate map of the salt scalds existing at the time of image acquisition.

 

Subsequent tests have shown that the subtle hydrate absorption features at 980 and 1170 nm are crucial to correct mapping of the salt-affected soils.

The Minimum Noise Transform (MNF) described by Green et al. (1988) was employed to both remove noise and to compress the hyperspectral data into a small number of significant bands.

 

Boardman (1993) has developed methods for isolating extreme pixels within an image and then visualising these in a multi-dimensional display (as shown below). The most extreme pixels are identified by repeatedly projecting n-dimensional scatter plots of the MNF values onto a random unit vector. The extreme pixels in each projection and the total number of times each pixel is marked as extreme are recorded (Boardman et al, 1995). These extreme pixels are assumed to represent spectrally homogenous examples of the mixing endmembers.

MNF dimensions 1, 3 and 5 are used to discriminate 26 endmember classes. Green vegetation, occurring as crops, and halophytic (salt-loving) vegetation occupy very different volumes within the n-dimensional space.

 

The numbered classes are the various soil endmembers. Classes 1 through to 7 are a distinct series showing decreasing salinity. Class 31 is a unique type of saline soil. Class 6 is a damp, saline, soil.

 

NB. This method cannot identify endmembers present in sub-pixel amounts as each endmember has the spectral properties of the whole extreme pixel. Those endmembers present in sub-pixel amounts can be identified and mapped by image-derived methods only as a constituent component of a composite endmember. Endmembers whose abundance varies temporally and whose surface area could become less than a whole pixel will not be seperately identified on multi-temporal images.