A future potato field skynet?
“Forewarned means forearmed” is a good principle to follow when facing an adversary. There is a reason why the word “intelligence” refers both to gathering information about a foe as well as to having an ability to think. Flailing in the dark, trying to strike at random, is not the smartest approach.
This applies to pest management as much to any other hostile encounter. Proper scouting is an essential part of efficient crop protection. Missing a window of opportunity to knock down a pest population may result in a lot of damage.
Unnecessary pesticide applications, on the other hand, waste money, time, pollute surface and ground waters, and aggravate environmentally conscious neighbors. So, to bring up the wisdom of Mr. Miyagi from the timeless classic movie The Karate Kid, it is all about finding the right balance.
Similar to most things in life, maintaining a good monitoring program is easier said than done. Walking potato fields looking for diseases and insects is time- and labor-consuming. It can also be rather mind-numbing, as I can attest based on many years of personal experience.
Finding the right people to do it can be a challenge, because they need to be qualified enough to know what to look for and ambitious enough to want to find it. Plus, they must be willing to do it for a salary that can also be earned by flipping burgers at a fast-food joint. So, this kind of job is a good candidate for automation.
It would be cool to see androids canvassing potato fields looking for the Colorado potato beetle larvae. However, it is probably not going to happen, at least not soon. Presently, remote sensing using drone-mounted cameras appears to be a more promising approach. Therefore, most research that I know of is heading in that direction.
DIGGING INTO THE DETAILS
There are two main ways to detect that something may be going wrong within a field. The first one is to look at the spectra of reflected light. This is a fancy way to say that we are trying to figure out whether the plants have changed their color. Some of such change can be detected by the naked eye.
For example, plants that are stressed turn yellowish, while PVY infection produces characteristic mottling on leaves in some potato varieties.
However, modern cameras are significantly more sensitive than a human eye. They can detect a slight variation in color that is not obvious to a human. Also, they cover a significantly larger light spectrum, including infrared and ultraviolet light that we cannot see. Therefore, cameras can detect subtle changes that may be indicative of disease or feeding damage by herbivorous insects.
The challenge is to determine which color corresponds to which problem. In some cases, there could be an overlap large enough to make a separation very difficult. Simplistically speaking, we will know that a plant is yellow but will not know whether it lacks nitrogen, grew under insufficient light or is infected by a virus.
Finding spectral signatures corresponding to different stressors is currently the focus of a significant research effort. In our research, this technology worked well for detecting nitrogen and phosphorous deficiency. There is also some promise for detecting the damage caused by the Colorado potato beetle feeding.
Reflectance spectrum may also change depending on the size of the potato canopy. Smaller leaf area due to defoliation by Colorado potato beetles and other herbivores, or due to retarded plant growth, means less green foliage and more brown soil. The same applies to a reduced stand count because of poor germination or damage by soil-dwelling insects.
Another approach is to take aerial photos of the field and then use image recognition software to detect and count insects or disease symptoms, such as leaf lesions. This works well with large objects that have a contrasting coloration, such as adult Colorado potato beetles. However, it is less effective for something like potato aphids, which are small, green and hide underneath potato leaves.
Perhaps this will be addressed as technology advances. However, we are not there yet, at least not in a cost-effective way.
For both methods, data analyses are increasingly reliant on the use of artificial intelligence. AI is an overhyped buzzword right now, but it is a powerful technology that has numerous practical applications. Remote scouting for pests and diseases is already being actively commercialized. I fully expect it to become the mainstream soon. Therefore, it is a good time to start planning accordingly.
Andrei Alyokhin is a professor of applied entomology at the University of Maine.
