Bugs Smell Success
Human technology may be taking a backseat to nature in the detection of crop disease.
The score so far is Wasps 1, Electronic Noses 0. In a test of technologies for diagnosing plant disease by odor, a team of trained wasps detected compounds up to a million times fainter than those that alerted a commercial sniffer.
Scientists at the University of Georgia, Tifton, and at the adjacent unit of the U.S. Department of Agriculture, who together staged the contest, believe odor could allow farmers to identify crop ailments before the plants show visible signs. This ability would let them use costly and potent chemicals such as fungicides only as needed.
Project leader Glen Rains, an engineer at the University of Georgia, briefed members of the American Society of Agricultural Engineers on the showdown between insect and human technology at the society’s annual meeting in August.
Something in the Air
It was the wasps that inspired the idea of using smell, says USDA entomologist Joe Lewis. Certain species are drawn to chemical signals that plants issue through the air-for example, in response to the nibble of a caterpillar. An arriving wasp then injects an egg into the nibbler and the hatchling eats it alive from the inside out.
Unseemly? Perhaps, but the Tifton researchers recognized talent when they spotted it. Rearing and teaching such wasps in the lab, Lewis and his coworkers found they had an aptitude for odors of all kinds.
“The capability is there for them to do all the things that dogs do,” says Lewis.
The wasps smell through their antennae and remember scents that accompany a reward of food. Three rewards in ten minutes, and a wasp has a scent down cold, says Rains.
On the Field of Honor
Rains conceived the idea of benchmarking the wasps against a commercial “e-nose”-a flexible, handheld device that can learn odors as wasps can, but accepts the push of a button as a reward.
For the test, wasps were trained to recognize one of two chemical odors and competed in teams of six. The e-nose worked alone and learned both odors: a volatile compound that plants release when injured and another produced by a crop-afflicting mold.
Testing the wasps meant tallying how many danced around the perforated opening of a bottle-which the insects only did when it contained odors they’d been taught. In contrast, the e-nose was allowed to poke its tubular “snout” into the bottle itself, causing the machine’s verdict and confidence rating to show on the unit’s liquid crystal display.
The match was not even close: the e-nose was unable to recognize either compound in traces less than a few parts per million. The wasps detected traces in parts per billion or even trillion, depending on the molecule.
The View from the Sidelines
California Institute of Technology chemist Nathan Lewis says the trouncing is not a crushing blow for e-noses. The fundamental polymer-based technology was not itself on trial, he maintains, since the device’s internal sensors were buried down a deep snout. That said, he might still have put his money on the wasps.
“Biological systems are far better, when they’re tuned, than any analytical device humans have today,” Lewis says.
Yet Ohio State University entomologist Brian Smith says Lewis’s proviso “when they’re tuned” is the crux. Insects are famously sensitive to their own chemical mating cues, yet the talent may be too specific. Bugs might not be nearly so sensitive to odors they have to learn. In theory, either competitor could have won by a nose, says Smith.
Encouraged by the wasps’ performance, Rains says he’s continuing efforts to integrate the insects into a portable device that inspectors might take into fields or that might ride on a tractor.
The goal stands some years distant, but he has worked several potential elements of the technology into a tabletop system, he says. The apparatus signals electronically when a wasp pursues an odor down a hole, and a rotating cylinder like the one in a revolver brings fresh wasps up to bat.
But the insects haven’t been tested yet in field conditions, where the task is to discern the target amid a bouquet of other odors.
The ultimate aim of diagnosing plant diseases will depend on identifying telltale odors, which researchers are only beginning to catalog. Recent studies suggest that plants have a repertoire of explicit chemical distress signals, Rains says, besides the ones that summon wasps. He plans to carry out his own search for a plant cry indicating aflatoxin, a potent carcinogen that often makes peanuts unsaleable.
Greg Carter, a plant physiologist at NASA, notes that crop stresses are difficult to tell apart with imaging and light-based sensing, so using wasps and odor might well have advantages. “It sounds great to me,” says Carter.
AI is here.
Own what happens next at EmTech Digital 2019.