Mapping Soil Salinity Using Hyperspectral Imagery
Secondary saline areas across Victoria have been produced by changes in landuse since European settlement. Traditionally these areas are mapped using ground based assessments that are difficult to implement at landscape scale due to cost. Extreme saline environments are frequently dominated by exposed saline scalds and halophytic, perennial vegetation providing opportune conditions for mapping salinity using remote sensing techniques. This study aims to develop an approach to map soil salinity using hyperspectral imagery in moderately saline environments that are typically characterised by less surficial expression of salinity (e.g. salt scalds) than extreme saline sites.
The study was conducted around Hamilton in Victoria’s Western District where salinity is characterised by low to moderately severity soil salinity with a good cover of salt tolerant grasses and herbs and an absence of salt sensitive species. A suite of whole pixel (Spectral Angle Mapper, Spectral Feature Fitting) and sub-pixel (Mixture Tuned Matched Filtering) analysis techniques were evaluated using 125 band, 3 metre hyperspectral imagery. All techniques were applied to reflectance and transformed reflectance data to map soil salinity based on the presence and absence of salt indicators. Laboratory measured reference spectra for vegetation and soils, ground truth data, soil analyses and soil XRD spectra were used to calibrate and evaluate techniques.
Results of the image analysis are compared to on-ground, vegetation based assessments of soil salinity.
This method shows some promise but there are a number of operational problems to be addressed before it could be implemented at landscape scale.