Scarf: Soil Carbon Sensing with Wi-Fi and Optical Signals

Soil carbon is a key soil property for soil health management and a crucial part of the global carbon cycle. Existing methods for soil carbon determination are too expensive and time-consuming. This paper introduces Scarf, a novel soil carbon sensing technique that leverages Wi-Fi and images and eliminates the need for specialized hardware. Our analysis reveals a strong correlation between soil carbon content and both the soil permittivity obtained from Wi-Fi signals and the soil lightness captured in soil surface images. We develop mathematical models to quantify the relationships between soil carbon content and the two signal modalities, and use the models to estimate soil carbon. We apply machine learning to help handle relationships that mathematical models do not capture. Our experiments demonstrate that Scarf delivers highly accurate soil carbon estimations.