Estimation of soil organic carbon using data synergy of Landsat 8 satellite imagery and MODIS sensor through machine learning models

The main objective of this study was to develop machine learning models for predicting soil organic carbon (SOC) content in agricultural soils enriched with calcium-rich materials in the central region of Iran. Soil samples were collected from 336 points in agricultural lands with two land uses, namely cropland and orchard, from Alborz Province. The soil organic carbon was measured using the Walkley-Black oxidation method. Remote sensing data, obtained from various sources such as Landsat 8 imagery and MODIS sensor, were utilized for soil organic carbon analysis. Three machine learning models, including Artificial Neural Networks (ANN), Support Vector Regression (SVR), and Gene Expression Programming (GEP), were employed in this study to estimate the soil organic carbon content. The results of this research demonstrated positive outcomes, indicating that the simultaneous use of Landsat 8 imagery and MODIS sensor data through the SVR model yielded better performance (R2 = 0.62, RMSE = 0.63, R2/RMSE = 0.98) compared to the use of each image separately. Therefore, this study concluded that the simultaneous synergy of remote sensing data from different sources (referred to as data synergy) can significantly improve the accuracy of predicting soil organic carbon content in calcium-enriched agricultural soils in the central region of Iran. It is recommended to use advanced fusion techniques and deep learning methods for combining features at a higher level.