Machine Learning-based Soil Property Prediction for Remediation of Radioactive Contamination in Agriculture

The Joint FAO/IAEA Division of Nuclear Techniques in Food and Agriculture launched a new Coordinated Research Project (D1.50.19) called "Monitoring and Predicting Radionuclide Uptake and Dynamics for Optimizing Remediation of Radioactive Contamination in Agriculture'', in October 2019. Within the CRP, the high-throughput characterization of soil properties and the estimation of soil-to-plant transfer factors of radionuclides are of critical importance. As already highlighted in Soils Newsletter Vol. 43, No. 1, July 2020, for several decades, soil researchers have been successfully using near and mid-infrared spectroscopy (MIRS) techniques to estimate a wide range of soil properties (Carbon, Nitrogen, CEC, Clay, Sand, pH, ...). In recent years, soil science researchers are increasingly shifting their focus from traditional modeling techniques such as PLSR (Partial Least Squares Regression) to new classes of algorithms, such as Ensemble Learning (Random Forest, Boosting, …) or Deep Learning (Convolutional Neural Networks), that have proven to outperform PLSR on most (if not all) soil properties prediction in a large data regime.