More Than Meets the Eye
Using light reflectance to monitor crops indoors
Staying out in front of plant health matters a lot to farmers, especially when they raise crops in controlled environment agriculture (CEA). When it comes to plant health, we may think that “seeing is believing,” but some things we can’t see can also tell us a lot.
In this article, we want to look at the field of photometry as a robust crop monitoring technique that’s attracted some promising research.
What Is Photometry?
Photometry measures light waves, which include visible light we can see ourselves as well as invisible light, such as ultraviolet light and infrared light, that can’t be seen by the naked eye. The most common application of photometry to monitor plant health is called NDVI. NDVI (normalized difference vegetation index) is an equation that uses the light reflected off of plants to determine how healthy they are on a scale from -1 to 1.
Not surprisingly, plants reflect more green light than red and blue light. But did you know they also reflect significantly more near-infrared (NIR) light than visible light? This is important because visible light is absorbed and converted into energy by the plant, while NIR light is not absorbed but reflected back. This produces the unique circumstances in which healthy plants reflect a lot of NIR and not much visible light, and unhealthy plants reflect less NIR and more visible light. (See Figure 1.)
Figure 1: This graph shows the percent of light reflected by a healthy plant vs. a stressed plant for each wavelength of light. Notice how healthy plants reflect more near-infrared light and less blue, green, and red light.
Now, look at this series of images below. The top one shows leaves becoming progressively more withered from left to right. The bottom image shows the same leaves with their NDVI visualized. NDVI often appears like a heat map, except heat is replaced by plant health.
Both conventional and CEA growers stand to benefit from the information provided using NDVI. Conventional farmers rely on NDVI to estimate crop yield in large tracts of land using satellite imaging. They also use drone imaging to collect data on demand in a higher resolution, allowing them to see precisely where in their fields they have stressed crops.
However, CEA environments are where early detection of crop stress by NDVI can really shine because growth there takes place faster, leaving less time to adjust. In a natural environment, for example, the soil serves as a “regulator,” retaining nutrients and water that plants can feed on to regulate their condition. CEA environments are soilless — that’s one of their strengths — making them less forgiving.
How Useful is Photometry in Practice?
Given the importance of early detection, researchers want to learn how photometry can be used effectively in CEA to detect stress early enough to minimize crop loss and identify the cause(s) of loss.
One study showed that NDVI could quickly detect salt stress even if there were no clear, visual signs, a feat even an experienced plant biologist can’t manage!
Two other studies look at how photometry might be able to predict the cause(s) of stress. For example, plants may show different light reflectivity patterns based on their composition, which means the light a plant reflects could be used to predict whether it has too little of a given nutrient. Although NDVI looks at a few specific “colors” of light, this type of measurement needs access to the whole spectra, or at least a wide section of it. Different plants will likely have different reflectivity patterns, a point that requires further study.
Finally, NDVI could team up with artificial intelligence to deliver a solid one-two punch, with NDVI targeting areas where crops are stressed and AI distinguishing among shapes and patterns to predict the potential causes of stress, which has been demonstrated for disease and nutrient deficiency.
Lot’s of Promise . . . More Study Needed
There’s so much promise in photometry’s potential to be a game changer that more experimentation is needed. Which crop stresses can be detected accurately, treated (and to what degree), and how much testing and AI training is required to get us to a point of acceptable accuracy?
Sounds pretty high-tech, huh? Well, it is, and that’s why we’re working on an article going into more depth on machine learning in agriculture. Stay tuned!
