Investigating the Ability of Landsat 8 and Sentinel 2A Satellite Images for Estimating Soil Organic Matter and Available Phosphorus in Semnan Plain

Identification of soil quality changes, including soil organic matter (SOM), is one of the most important usages of remote sensing and geographical information system. Available phosphorus is also an important nutrient for optimal growth of plants. The purposes of this study were to investigate the capability of satellite images data and to compare the accuracy of SOM and available phosphorus maps by using Landsat 8 and Sentinel 2A satellite images. The location of sampling points was determined by using conditional Latin hypercube sampling for 84 soil samples in agricultural lands of Semnan plain. The SOM content, the particle size fractions including sand, clay, and silt were measured using wet oxidation and hydrometer methods, respectively, and available phosphorus was measured by the Olsen method. The auxiliary variables included the bands and combination of bands. The results showed that soil available phosphorus had the highest correlation with SOM content. Results of Random Forest algorithm indicated that axillary variables derived from multi-spectral instruments (Sentinel 2A satellite) evaluated the amount of SOM and available phosphorus more accurately than the axillary variables extracted by Landsat 8 satellite images. The random forest nonlinear method estimated the amounts ​​of SOM and available phosphorus with low error values ​​and a relatively high coefficient of determination. The root means square error (RMSE) and coefficient of determination (R2) for prediction of SOM were 0.413 and 0.758 for the multi-spectral instrument, and 0.432 and 0.736 for the operational land imager, respectively. Also, the RMSE and R2 for prediction of available phosphorus were 5.96 and 0.74, for the multi-spectral instrument, and 7.24 and 0.56 for the operational land imager, respectively.