Remote and Vis-NIR spectra sensing potential for soil salinization estimation in the eastern coast of Urmia hyper saline lake, Iran

Abstract

Soil salinization is an important threat for agriculture and environment in the eastern coast of Urmia hyper saline Lake, a lake in the western part of Iran. Predicting soil salinization requires rapid and low-cost measurement tools of soil salinity. It is hypothesized that remote sensing and visible near-infrared spectroscopy may offer a feasible method for that purpose. To test this hypothesis, 96 soil samples were collected using Latin hypercube sampling method. Soil salinity (ECe) and spectral reflectance at various bands were measured in all samples in the laboratory. The spectral indices derived from Landsat OLI, Sentinel-2 MSI and visible near-infrared (Vis-NIR) spectroscopy data, as input variables, along with partial least squares regression (PLSR) method were used for predicting contents soil salt. The results revealed that the PLSR derived regression models of the spectral indices accounted for 42.7 and 45.3% of the soil salinity variation by Landsat OLI and Sentinel-2 MSI data, respectively. And PLSR derived model using Vis-NIR spectroscopy data explained 59.4% of the variation in soil salt contents. Compared to derived models using Vis-NIR spectroscopy, Landsat OLI and Sentinel-2 MSI data, the derived PLSR models (which were developed as a result of integrating Vis-NIR spectroscopy and remote sensing data) increased the power of soil salinity prediction Finally, the combined of Vis-NIR spectroscopy + Sentinel-2 MSI data was found to provide the best input data for the derived regression model to estimate soil salt contents (R2 = 0.713 and RMSE = 21.76 dS m−1).