New research shows growers can predict final cannabinoid yields weeks before harvest using non-destructive hyperspectral imaging

Australian researchers have developed a breakthrough method that could transform how cannabis cultivators manage crops, predict yields, and ensure regulatory compliance, all without damaging a single plant.

In a study published in Industrial Crops & Products, scientists from the University of Adelaide demonstrate that hyperspectral measurements of cannabis fan leaves taken during early flowering can accurately predict final cannabinoid concentrations in mature flowers.

The Game-Changing Technology

Using a hand-held hyperspectral device, researchers achieved prediction accuracies with R² values up to 0.89 for CBD, 0.77 for THC, and 0.8 for total cannabinoids. These results significantly outperform previous approaches.

The method works by measuring how light reflects off intact fan leaves at wavelengths invisible to the human eye. These spectral "fingerprints" correlate with the plant's biochemical composition. This allows machine learning models to predict what cannabinoid profile the flowers will ultimately produce.

Lead researcher Aaron Phillips and his team at the University of Adelaide's School of Agriculture, Food and Wine tested the technology across two cannabis cultivars under seven different lighting regimes. They took measurements both early and late in the flowering period.

Why This Matters for Australian Growers

The implications for Australia's emerging cannabis industry are substantial.

For Industrial Hemp Producers: Exceeding legislated THC levels necessitates destruction of entire crops, posing substantial economic risks. Early prediction enables identification and removal of non-compliant plants before they jeopardise an entire harvest.

For Medicinal Cannabis Operations: High-performing plants can be identified and prioritised early in the growth cycle. This minimises resources wasted on inferior genetics. The technology could also help optimise harvest timing to maximise cannabinoid yields.

For Breeders: Promising plants can be selected for crossing programmes before flowering even begins. This dramatically accelerates breeding timelines.

Completely Non-Destructive

Unlike previous methods that required removing inflorescences and leaves prior to measurement, this approach enables rapid, in-situ assessment of intact fan leaves. No sacrificial sampling or laboratory analysis is required.

This is crucial. There is evidence that wounding or removing plant material can alter cannabinoid profiles, potentially confounding predictions.

The hand-held device used in the study takes measurements directly in the growing environment, whether glasshouse or field. It provides instant analysis without the need for HPLC or GC-MS laboratory testing.

How It Works

The research team took hyperspectral measurements at two critical timepoints:

• Early: Two weeks after initiating flowering, before flowers emerged
• Late: Four weeks after flowering initiation

Measurements were taken from five fan leaves per plant across the vertical length of the canopy. The device captured wavelengths from approximately 350 to 2500 nanometres.

Machine learning models trained on these spectral profiles, combined with actual cannabinoid measurements from harvested flowers, learned to predict final concentrations with remarkable accuracy.

Beyond Cannabinoid Prediction

The technology offers additional benefits.

Hyperspectral measurements successfully differentiated between cannabis cultivars and lighting treatments, offering a tool for germplasm classification. This could help growers verify cultivar authenticity. Breeders could use it to select diverse parent genetics for crossing programmes.

Looking Forward

While the study focussed on controlled indoor environments, the technology shows promise for field applications. The researchers suggest that measurements taken from fan leaves proximal to sampled flowers may provide even better predictions. This opens avenues for further refinement.

For Australia's cannabis industry, navigating strict regulatory frameworks while pursuing economic opportunities in both industrial hemp and medicinal cannabis, this non-destructive, early-prediction technology represents a significant step forward. It promises more efficient, compliant, and profitable cultivation practices.

The research was conducted at the University of Adelaide's Waite Campus. The team collaborated with German analytics firm Compolytics GmbH and South Australian lighting company VAILO.

The study "Hyperspectral measurements of Cannabis sativa fan leaves during early floral development predict final cannabinoid yield" is available as open access in Industrial Crops & Products.