Remote Sensing Data from Satellites in Assessing Soil Productivity

Relations are established between various indices for the condition of the plant cover and humus content of the soil. Spectral reflective characteristics of vegetation are determined by analysis of images in different channels from the Sentinel-2 satellite using linear regression and random forest algorithms. It is found that the vegetation indices NDI45 and PSSRa are more sensitive to the humus content in the soil than the indices RVI and NDVI. The determination coefficient (R² = 0.59) and mean square error (MSE = 0.00 014) indicate that the predictive power of the proposed model is good.