Advancing Soil and Environmental Analysis with Dual-Wavelength Raman Spectroscopy and Machine Learning

Advanced environmental measurements require versatile, high-throughput methodologies that can analyze complex and heterogeneous systems. Raman spectroscopy presents a promising solution as an optical measurement technique, owing to its minimal sample preparation requirements, real-time and non-destructive measurements, and its potential for field deployment. However, its adoption in environmental applications has been limited by challenges such as fluorescence interference and sample heterogeneity. Here, we describe a dual-wavelength Raman spectroscopy approach that overcomes these challenges, enabling precise and reliable measurements of soil. Central to our approach is a custom Shifted-Excitation Raman Difference Spectroscopy (SERDS) instrument, which integrates advanced optical design, signal processing, and machine-learning multivariate analysis. 

We utilize our SERDS methodology to measure soil organic carbon (SOC) in agricultural soils and tire wear particles. By leveraging custom spectral collection strategies and signal processing tools, such as common-mode rejection (CMR) along with hyperspectral data fusion techniques, we effectively mitigate fluorescence interference, particle size variations, and nonlinear optical behavior in soils for accurate SOC and tire wear quantification. Nonlinear machine-learning regression techniques, including tree-based models and a custom Partial Least Squares Regression algorithm, enhance predictive accuracy and validate the methodology. 

While the measurement of SOC and tire wear particles in soil highlight the potential of our SERDS methodology in advancing real-time and high-throughput soil measurements, its versatility extends to a broad range of environmental sensing applications, including water quality monitoring, pollutant detection, and the analysis of complex environmental systems. This research presents an in-depth examination of the design and implementation of the SERDS instrument and methodology, showcasing its potential for advancing environmental measurement and its adaptability for addressing a wide range of analytical challenges in environmental science.