Using Machine Learning Algorithms Based on GF-6 and Google Earth Engine to Predict and Map the Spatial Distribution of Soil Organic Matter Content

Abstract

The prediction of soil organic matter is important for measuring the soil’s environmental quality and the degree of degradation. In this study, we combined China’s GF-6 remote sensing data with the organic matter content data obtained from soil sampling points in the study area to predict soil organic matter content. To these data, we applied the random forest (RF), light gradient boosting machine (LightGBM), gradient boosting tree (GBDT), and extreme boosting machine (XGBoost) learning models. We used the coefficient of determination (R2), root mean square error (RMSE), and mean absolute error (MAE) to evaluate the prediction model. The results showed that XGBoost (R2 = 0.634), LightGBM (R2 = 0.627), and GBDT (R2 = 0.591) had better accuracy and faster computing time than that of RF (R2 = 0.551) during training. The regression model established by the XGBoost algorithm on the feature-optimized anthrosols dataset had the best accuracy, with an R2 of 0.771. The inversion of soil organic matter content based on GF-6 data combined with the XGBoost model has good application potential.