New Spectroscopy Method Offers Rapid, Reliable THC Classification for Cannabis Samples

Recently, a research collaboration between the National Institute of Standards and Technology (NIST) and the University at Albany, State University of New York, demonstrated the potential of near-infrared (NIR) spectroscopy as a fast, nondestructive method for identifying and quantifying THC levels in cannabis plant materials (1). The work could provide laboratories and regulators with a rapid screening tool to help distinguish legal hemp from regulated cannabis products, which is an increasingly important distinction since the legalization of hemp in the United States.

Following the 2018 Farm Bill, hemp was removed from Schedule I of the Controlled Substances Act, defined as Cannabis sativa containing no more than 0.3% total delta-9-tetrahydrocannabinol (Δ9-THC) on a dry weight basis (2). However, distinguishing between hemp and higher-THC cannabis varieties remains a complex analytical challenge. Traditional laboratory methods, such as liquid chromatography (LC) or gas chromatography (GC), though highly accurate, are time-consuming, require solvents and sample preparation, and depend on trained analysts (1). The NIR spectroscopy method developed by Urbas and colleagues aims to simplify this process by providing rapid, real-time THC assessment with minimal sample handling.

What did the researchers do in their study?

In their study, published in Forensic Chemistry (1), the researchers used NIR diffuse reflectance spectroscopy to analyze 75 cannabis flower samples. Then, to classify and predict THC concentrations, they developed two chemometric models, specifically partial least squares discriminant analysis (PLS-DA) and partial least squares (PLS) regression (1). The classification models were designed to separate samples into “low-THC” (less than 2% total Δ9-THC) and “high-THC” (2% or more total Δ9-THC) categories (1). With the PLS-DA model, the researchers achieved a 98.9% accuracy during cross-validation and 96.7% accuracy in the test set, misclassifying only one high-cannabigerol (CBG) sample (1).

What were the two quantitative prediction models the team constructed?

The researchers created a full-range quantitative prediction model and a low-range quantitative prediction model. The full-range model covered the entire spectrum of THC concentrations, yielded a root mean square error of prediction (RMSEP) of 0.741%, while the low-range model (focused on samples under 1% THC) achieved an RMSEP of 0.073% (1). However, two samples high in cannabidiol (CBD) and moderate in THC (1–2%) were not well captured by the low-range model, suggesting that complex THC–CBD interactions influence spectral interpretation (1).

There were a couple key observations the researchers highlighted in their study. For one, the low-THC model relied heavily on CBD-related spectral features because of the strong chemical correlation between THC and CBD. This interdependence may actually enhance the ability of NIR-based models to differentiate cannabis types, as high-CBD, low-THC profiles are typical of hemp products (1).

Second, NIR spectroscopy proved to be an efficient and simple technique to use for this purpose. Unlike chromatographic approaches, which require chemical solvents and calibration standards, NIR spectroscopy requires little to no sample preparation, can analyze ground cannabis directly, and produces immediate spectral data (1). The subsequent statistical analysis can be automated, meaning results can be obtained without expert interpretation, which is a major benefit for field applications, law enforcement, and agricultural testing laboratories (1).

As a result, the researchers emphasize that NIR spectroscopy could serve as a rapid, first-line screening tool, flagging samples that require confirmatory analysis through more complex laboratory methods (1). Its nondestructive nature also preserves evidence integrity, a critical factor in forensic contexts.

What are the key takeaways from this study?

While further refinement of the models and larger data sets will be needed before full regulatory adoption, the findings suggest a promising step toward standardized, high-throughput THC classification. As cannabis regulation continues to evolve globally, this study positions NIR spectroscopy as a powerful, accessible technology for distinguishing hemp from marijuana and ensuring compliance with legal THC thresholds (1).

With continued development, this collaboration between NIST and the University at Albany could help reshape how cannabis materials are tested, offering a faster, greener, and more efficient approach to forensic and agricultural THC analysis.

References

1. Mistek-Morabito, E.; Wilson, W. B.; Lednev, I. K.; Urbas, A. A. Analysis of Cannabis Plant Materials by Near-infrared (NIR) Spectroscopy and Multivariate Data Analysis for Differentiating Low-THC and High-THC Cannabis. For. Chem. 2025, 46, 100698. DOI: 10.1016/j.forc.2025.100698
2. Kafka, D. C. The 2018 Farm Bill’s Hemp Definition and Legal Challenges to State Laws Restricting Certain THC Products. Congress.gov. Available at: https://www.congress.gov/crs-product/R48637 (accessed 2025-10-31).