Estimating of Soil surface moisture distribution in Arid and semi-arid region using Remote Sensing

The purpose of this paper was to make the preparation of soil surface moisture map applied in arid and semi-arid regions using satellite imagery of OLI and TIRS along with the calculated values for spectral reflectance, principal component analysis and Tasseled Cap transformation.Twenty-four predictor variables were used and the most correlated ones were identified at three moisture levels of 4 to 5 percent, more than 5 percent and less than 4 percent by exploratory regression and bivariate correlation method through calculating the inflation factor of variance, Pearson coefficient and coefficient of explanation. The MoransI index was used for geo-spatial autocorrelation. Forty-seven soil samples were collected randomly by creating 1800-meter networks in a systematic algorithm and the soil moisture was calculated by W-thermal method. The estimation functions of soil surface moisture were derived in the form of partial least square regression (PLSR), enter regression and stepwise regression. All models had acceptable calibration. Our results clearly showed that the Landsat 8 data could be useful in estimating soil surface moisture and the accuracy of the functions extracted with stepwise regression method was more than that of other methods (RMSE=0.585 to 1.425 %). The models introduced for a moisture level of ≤5%, tend to over estimate (MBE= 0.788) and for other moisture levels tend to underestimate.