A Soil Carbon Content Quantification Method Using A Miniature Millimeter Wave Radar Sensor and Machine Learning

Abstract

Soil carbon content plays an essential role in combating climate change, water cycling, and sustaining soil biodiversity. However, the conventional way of quantifying soil carbon content is labor intensive, lack of precision, slow, and costly. On large spatial scale, assessment of the effect of carbon (biochar) applied to the soil for soil health conditioning, remains to be very difficult. This paper for the first time demonstrates the viability using a millimeter-wave sensing method for quantifying soil carbon content. It can also distinguish biochar types from different biomass species. Furthermore, soil moisture monitoring, and biochar water retention capacity can also be quantified by utilizing the same miniature millimeter wave radar sensor empowered by machine learning. Specifically, in this study, we present our research materials, methodology, machine learning workflow, results, and the explanation and interpretation based on the physical principles of the millimeter wave radar array sensor in the context of soil carbon content. We validated our quantification method with supervised machine learning algorithm using real soil data collected in the field mixed with known biochar contents. The results show that our technique achieved a 95.7 per cent recognition accuracy across seven different biochar types. The work laid the foundation for future real-time, large spatial-scale evaluation and assessment of soil carbon content using biochar amendments or other related carbon-negative technologies. Thus, soil carbon content site-specific management can be made possible.