A chip-based radio frequency sensor for soil moisture measurements: A machine learning and deep learning calibration approach

Accurate soil moisture monitoring is important for precision agriculture, irrigation scheduling, and sustainable water resource management. To address the limitations of high cost and the complex, specific calibration processes inherent in soil moisture sensors, this study investigates a novel chip-based radio frequency (RF) sensor designed to operate as a fully passive system. Transforming the complex, multi-parameter RF signal into reliable volumetric moisture content (VMC) data requires a robust calibration framework. This study introduces and evaluates six machine learning (ML) and deep learning (DL) calibration models using soil samples representing diverse soil textures and physicochemical properties. Among the ML models, Polynomial Regression Model achieved the highest accuracy (R 2 = 0.75, RMSE = 2.04 % VMC). While the DL models, specifically the Convolutional Neural Network (CNN), demonstrated the best overall performance with an R 2 of 0.78 and an RMSE of 1.92 % VMC, followed by the Dense Neural Network (DNN) (R 2 = 0.76, RMSE = 2.02 % VMC). These models confirm their effectiveness in capturing the nonlinear relationships between RF signals and diverse soil properties. Feature importance analysis identified bulk density and the raw RF signal as the most influential predictors, and 5-fold cross-validation confirmed the high stability of the CNN and DNN models. These findings establish a strong proof of concept for integrating low-cost, chip-based RF sensing with data-driven modeling to deliver a highly accurate, scalable, and generalized solution for real-time soil moisture estimation in precision agriculture. • Machine learning calibrates a battery-free chip for soil moisture sensing. • Deep learning improves accuracy across diverse soil properties. • Models remain robust when inputs are pruned to core features. • Bulk density and radio frequency signal strength dominate predictions. • Passive sensing architecture supports scalable deployment in the field.