Salinity modelling by inverted Gaussian parameters of soil reflectance spectra

Abstract

This paper presents an approach to estimate soil salinity through modelling of soil spectra using an inverted Gaussian (IG) function. The approach is tested on experimental datasets including measurements of soil physicochemical properties and their spectral reflectance which are obtained under controlled laboratory conditions. The near-infrared (NIR) and shortwave infrared (SWIR) region of the salt-affected soil spectra were fitted to an inverted Gaussian curve. Parameters of the fitted curve, such as functional depth, distance to the inflection point and area under curve, were then used as predictors in regression analysis to estimate soil salinity levels. The results suggest a successful estimation of salinity levels, especially, for soil samples treated with epsomite and bischofite solutions. Amongst the calculated IG parameters, the area under fitted curve resulted in the most accurate estimations. The results demonstrate the relative utility of high spectral resolution data for estimating soil salinity under laboratory controlled conditions.