Predicting salt abundance in slightly saline soils from Landsat ETM+ imagery using Spectral Mixture Analysis and soil spectrometry

The monitoring of soil salinity levels is urgent for effective ecological restoration plans in Burg Al-Arab, Egypt. In the present work, salinity levels and spectrometry measurements for 21 soil samples were carried out and used to evaluate sophisticated Spectral Mixture Analysis (SMA) techniques and the simple Wetness Index (WI) in predicting the salt abundance in the soils of the area. Mixture Tuned Matched Filtering — MTMF, Linear Spectral Unmixing — LSU, and Spectral Angle Mapper — SAM, were applied to 2003 and 2010 Landsat Enhanced Thematic Mapper plus (ETM+) images. The spectra of two samples; slightly saline (2.62dSm−1) and non-saline (0.17dSm−1) were used as end-members. Low and high salinity classes were derived applying k-means clustering to the maps. The land surface temperatures (LST) were estimated from the thermal bands of the images. Correlations among the studied maps and parameters highlighted the major factors governing the salt abundance.The prediction of salinity levels was linearly and significantly high at R2=0.88, 0.84, and 0.87 for MTMF, LSU, and WI, respectively. The extents and the spatial distribution of the predicted classes were comparable and congruent in many places. Salinity levels strongly negatively correlated with soil reflectances (av. r=−0.90) where coefficient of correlation becomes higher at longer wavelengths. This was also true for the at-sensor reflectances but at lower coefficient in the 2010 (av. r=−0.47) and 2003 (av. r=−0.32) images. The higher the salinity levels, the deeper (R2=0.55), broader (R2=0.63), and the more asymmetrical the water absorption band centered at 1925nm indicating an increase in the soil moisture content. The proximity of cultivated clay loams to the urbanized flat depressions with local high heat islands irrigated by salt-laden groundwater contributed largely to raising the levels. MTMF model was the most appropriate among the SMA techniques to predict salt abundances and to determine the distribution of the slightly saline soils in the studied arid to semi-arid area.