New Technique Accurately Predicts THC Content Weeks Early, Study Shows

 Dr. Aaron Phillips and members of Professor Rachel Burton‘s lab had one of their recent papers picked up by Forbes Magazine. The paper will be published in Industrial Crops and Products Vol 236, 2 November 2025.

Forbes article by Benjamin Adams, 10 October 2025

A new study suggests that combining hyperspectral imaging (HSI) with machine learning is the key to predicting cannabinoid content—weeks in advance—with high accuracy. This technique would enable growers to accurately predict the final cannabinoid content in plants still in the early and later flowering stages, long before harvest, allowing them to appraise cannabis early.

With so much emphasis on cannabinoid content in cannabis products, the new technology could spell a shift in standard analysis techniques, which primarily rely on the “gold standard,” high-performance liquid chromatography (HPLC) or gas chromatography techniques. Standard techniques involve running a sample through a column to separate compounds. They are also the methods used to detect cannabinoids in oral samples.

Researchers associated with the School of Agriculture, Food and Wine at the University of Adelaide in South Australia analyzed two cannabis cultivars under seven different lighting conditions, using a handheld hyperspectral device to measure fan leaf reflectance at both early and late flowering stages. They used predictive machine learning models to predict THC content, using fan leaf hyperspectral reflectance spectra.

Flowers were later harvested, dried, and the cannabinoid content was calculated. They then compared the data with the cannabinoid content of harvested flowers. Machine learning models trained on the data achieved strong predictive accuracy, with a low variance (an R-squared variance of up to 0.77 for THC, 0.89 for CBD, and 0.8 for total cannabinoids). Predictions were also determined to be reliable for minor cannabinoids, such as CBGA and CBCA. They gathered data using fan leaf hyperspectral reflectance (FLHR) in order to make predictions.

The study is published in the November 2 issue of Industrial Crops and Products. If growers are able to determine the THC and other cannabinoid content from plants during the early and late flowering stages, they could appraise its value weeks earlier than using older techniques.