Abstract
Accurate, low-cost, and field-deployable soil moisture content (SMC) sensing remains a major challenge in precision agriculture and environmental monitoring. This study presents a compact, energy-efficient approach for in-situ SMC estimation based on discrete near-infrared (NIR) spectroscopy combined with machine-learning (ML) algorithms. A portable transfer-standard sensor was developed using seven carefully selected NIR light-emitting diodes (LEDs) targeting strong water absorption bands (970, 1150, 1450, and 1900 nm) and weak-absorption reference wavelengths (1100, 1300, and 1650 nm). The sensor design complies with recent international soil reflectance spectroscopy standards and enables direct comparability between laboratory reference measurements and field observations. SI-traceable calibration was performed using thermogravimetric methods over an SMC range from dry conditions to approximately 25% below the critical saturation level. <div> Linear physical models based on relative absorption depth showed good accuracy for soil-specific calibration but degraded significantly across heterogeneous soil types, reflecting sensitivity to texture, particle size, and composition. To overcome these limitations, six ML regression models were evaluated using stratified K-fold cross-validation, feature selection, and normalization. All ML approaches outperformed physical models, with Gaussian Process Regression performing best, achieving errors below 0.5% for soil-specific calibration and below 2% in generalized field conditions, with agreement with gravimetric references for SMC values below ~25%. Wavelength subset analysis indicated diminishing returns beyond five LEDs, with the combination of strong and weak wavelengths emerging as the most influential feature. Overall, the results demonstrate that LED-based NIR spectroscopy combined with ML provides a robust, energy-efficient, and scalable alternative to conventional spectroscopic systems for soil moisture monitoring. </div>
Keywords
Subject Areas
Related YTU Activities
Sustainability activities and projects connected to this article.
