SCIENCE
Drones with special cameras can estimate hemp's chlorophyll and photosynthesis from the air
North Carolina A&T researchers flew a drone over hemp plots and trained models to estimate leaf traits linked to nitrogen from the images, without picking a leaf.
Hemp'in
8 November 2026 · 2 min read

The research
AI-driven prediction of plant physiological traits in hemp using UAV-based multispectral imagery, by Harjot Sidhu, Arnab Bhowmik and Harmandeep Sharma, North Carolina Agricultural and Technical State University. Published in Frontiers in Remote Sensing, August 2026. Peer-reviewed and open access.
What they did
Nitrogen strongly shapes how hemp grows, how much it yields and the quality of its fibre, the authors write. Too little and the plants grow poorly; too much wastes money and can harm the environment. Checking a crop's nitrogen today usually means taking plant samples and sending them to a lab. It is slow, and it only tells you about the spots you sampled.
The team tested a faster way. Over two seasons, 2024 and 2025, they grew fibre and dual-purpose hemp on their research farm with different amounts of nitrogen. They flew a drone (a WingtraOne GEN II) carrying a multispectral camera, which sees light colours the eye can't, such as near-infrared and "red edge". From the images they computed 33 vegetation indices, numbers that describe how green and active plants look to the camera, and kept the seven that matched best.
On the ground they measured the real thing: chlorophyll with a handheld SPAD meter, and photosynthesis, leaf pore opening (stomatal conductance) and water loss (transpiration) with a LI-COR 6800 instrument. Then they trained several statistical and machine-learning models to predict the ground measurements from the drone data.
What they found
- Chlorophyll was easy to read from the air. Red-edge and green-based indices tracked SPAD chlorophyll very closely, with correlations of 0.91 to 0.94. A simple linear model predicted it with an R² of 0.88.
- So was photosynthesis. A model called a generalised additive model predicted photosynthetic rate with an R² of 0.87.
- Water-related traits were harder but workable. Stomatal conductance and transpiration were predicted with R² of 0.82 and 0.75.
R² runs from 0 to 1 and says how much of the real variation a model explains; 0.88 means the drone-based estimate captured most of the variation in the ground readings.
What it means, and what it doesn't
The authors present this as a scalable framework to support precision nitrogen management in hemp. The idea behind precision management is to fly the field, see where plants are short of nitrogen, and fertilise only there. The authors note that hemp has been underrepresented in drone studies of nitrogen, so this work fills a gap.
It is one research farm in one state over two seasons, with the cultivars the team chose. The models predict leaf traits linked to nitrogen, not the fertiliser dose a farmer should apply; turning a map into a recommendation is a further step. The equipment is research-grade, and the abstract gives no costs for a working farm.
Words to know
UAV: uncrewed aerial vehicle, a drone. Multispectral: a camera that records several bands of light, including some we can't see. Vegetation index: a number computed from those bands that tracks how green and active plants are. R²: how much of the real variation a model explains, from 0 to 1.
How did this land?
Ask the desk
The conversation
Open in LIFENo one has replied yet. Start the conversation.
WRITE FOR HEMP'IN
Have a story from the hemp industry?
Join Hemp'in and publish your own articles to everyone who reads News.
MORE FROM HEMP'IN NEWS

A wall's direction mattered more than its insulation, a year in an old French building shows
Hemp'in · 16 November 2026

Hemp oil and ground hurd did not improve wet alfalfa silage, but orange powder did, a study finds
Hemp'in · 15 November 2026

Hemp hurd processor Tiger Fiber files for Chapter 11 and keeps operating
Hemp'in · 14 November 2026