A Non-intrusive Quantification Method for Biochar Water Retention Capacity Using A Portable Microwave Sensor and Machine Learning

Abstract

Biomass from agriculture and forestry can be converted to biochar by pyrolysis gasification process and when applied to soil, this becomes a carbon negative technology. How to quantify the effect of biochar applied into soil in terms of water retention capacity accurately, quickly, cost-effectively and in large scale mapping are still a challenging problem. This work for the first time, reported extensive lab tests to demonstrate that a non-intrusive quantification method for biochar water retention capacity using a portable microwave sensor and machine learning is feasible. Scaling up can be achieved by putting the sensor on a mobility platform, either ground or aerial. We reported in this paper details of our materials and methods, and machine learning workflow and results. The significance of this work lays the foundation for future carbon negative technology applications assessment such as soil carbon accounting. We evaluated our nonintrusive quantification method with SVM and LDA classifier algorithm on the data collected from real-world experiments. Results show that our method achieved a 74.0% recognition accuracy over six different kinds of biochar footprints.