Grassland systems play a critical role in regulating soil carbon dynamics and delivering essential ecosystem services, particularly under the pressures of global climate change. Effective monitoring of soil physical and chemical properties is therefore critical to support sustainable grassland management. However, conventional laboratory-based analyses are often costly, time-consuming, and difficult to implement at large scales.
Spectroscopy, combined with machine learning, offers a promising alternative by enabling rapid and cost-effective prediction of soil properties. This study evaluates the potential of a commercially available near-infrared (NIR) spectrometer, NeoSpectra, to support practical soil monitoring in grassland systems. The use of an accessible device aims to simulate real-world conditions, potentially allowing farmers to perform measurements directly in the field or in simple laboratory settings. NIR spectroscopy operates in a wavelength range that is highly sensitive to organic carbon and N–H bonds, which reflects the change of C and N content and making it a suitable tool for monitoring grassland soils. The approach integrates spectral data with multiple machine learning models to predict soil physical and chemical properties, with a focus on optimizing model performance and assessing its robustness under conditions that reflect realistic application scenarios.
This study highlights the potential of combining NIR spectroscopy with machine learning as a scalable tool for soil health monitoring. The approach provides a rapid and accessible method for evaluating grassland soils and shows potential to support decision-making in sustainable land management under changing environmental conditions.