Estimating soil salinity using hyperspectral data in the Western Australian wheat belt

With the aim of early detection of soil salinity, this study developed a method to quantitatively estimate soil salinity in arid and semi-arid environments using SWIR reflectance spectroscopy. Focusing on the soil spectral characteristics occurring around 2000 nm, we determined, using subset analysis, that the spectral wavelength ranges which provide the greatest diagnostic information relating to soil salinity occur at 1996 and 2025 nm. An index was created using the normalized difference between reflectance values at these two wavelengths. Using this index a robust estimation equation (R2=0.91) was established with laboratory measurements of soil salinity. The derived estimation equation can provide quantitative estimates of soil salinity at an early stages across a variety of soil types. Our proposed method would significantly contribute to soil restoration by providing a means to take countermeasures before the situation of soil salinity gets worse and irreversible.