Modeling Soil Organic Carbon Variations Using Remote Sensing Indices in Ardabil Balikhli Chay Watershed

Modeling and providing accurate information on the spatial distribution of soil properties is a key factor in many environmental and agricultural applications. Therefore, the purpose of the present study was to model and prepare a digital map of soil organic carbon using remote sensing indices in the Balikhli Chay watershed. At first, topographic and spectral characteristics affecting soil organic carbon content were extracted from digital elevation model and Landsat 8 satellite image. Then the performance of soil organic carbon modeling for different states was evaluated and compared based on random forest models. The states including 1) terrain covariates, 2) spectral indices, and 3) combination of terrain and spectral covariates, were evaluated and compared together. To this end, the correlation coefficient (R2) between the estimated and measured soil organic carbon and root mean square error (RMSE) were calculated for the different states. The results showed that the amount of organic carbon in the study area varied from 0.32 to 6.98 and the mean value was 3.04%. Carbon changes in the study area mostly dependent on changes in spectral indices. Elevation and Emissivity were respectively the most important terrain and spectral covariates in soil organic carbon modeling. The R2 values in the three models were 0.61, 0.62 and 0.75 and the RMSE values were 0.88, 0.67 and 0.57, respectively, which indicates the better performance of the third model. The use of a combination of terrestrial and spectral variables significantly increases the accuracy of soil organic carbon modeling.