Digital Mapping of Topsoil Salinity Using Remote Sensing Indices in Agh-Ghala Plain, Iran

Soil salinization is a world-wide land degradation process in arid and semi-arid regions that leads to sever economic and social consequences.
We analyzed soil salinity by two statistical linear (multiple linear regression) and non-linear (artificial neural network) models using Landsat OLI data in Agh-Ghala plain located in north east of Iran. In situ soil electrical conductivity (EC) of 156 topsoil samples (depth of 0-15cm) was also determined. A Pearson correlation between 26 spectral indices derived from Landsat OLI data and in situ measured ECs was used to apply efficient indices in assessing soil salinity. The best correlated indices such as blue, green and red bands, intensity indices (Int1, Int2), soil salinity indices (Si1, Si2, Si3, Si11, Aster-Si), vegetation Indices (NDVI, DVI, RVI, SAVI), greenness and wetness indices were used to develop two models.
Comparison between two estimation models showed that the performance of ANN model (R2=0.964 and RMSE=2.237) was more reliable than that of MLR model (R2=0.506 and RMSE=9.674) in monitoring and predicting soil salinity. Out of the total area, 66% and 55.8% was identified as non-saline, slightly and very slightly saline for ANN and MLR models, respectively.
This shows that remote sensing data can be effectively used to model and map spatial variations of soil salinity.