Emily Bick is an entomologist with three degrees in the field. She’s also something of a spy. Using a novel detector, Bick can listen in on insects as they nibble crops. It’s an impressive feat, a kind of superhuman hearing for researchers. A sensor aboveground can overhear an insect gnawing on the root of a plant. That ability, as well as a background in computer coding, helps Bick take data from the detectors and figure out how best to manage pest interventions.

Bick’s interest in insect detection began during a postdoctoral fellowship at the University of Copenhagen, where she worked with entomological lidar, an expensive remote sensing method that takes measurements using a laser. She quickly realized that more cost-effective pest sensors were needed.

Given her interest in detectors and experience with computer modeling, Bick found an open position for an assistant professor of entomology and pest ecology extension specialist at CALS to be the perfect fit. She joined the college in July 2022.

What aspects of insect ecology is your lab studying?
We are trying to better understand fundamental insect pest population dynamics — economically damaging insects and how they move and change over space and time. It’s distinct from life cycles, where scientists follow a single individual to see how it changes. We’re taking an almost epidemiological approach to insects and agriculture. To do that, we are developing novel tools and strategies. And the end goal is to help decision-makers make better, more sustainable, more precise decisions. Essentially, we want to help folks better time and space their pest interventions.

What novel insect sensors are you working to develop?
I was challenged by a group of sugarcane growers in Indonesia to directly measure insects as they attack plants, rather than measure insects as they immigrate into a system and then try to predict when they’ll be a problem. There’s a disconnect between the stages that we tend to observe with insect sensors — oftentimes adult moth or beetles — and the life stage that tends to impact the plant. So, I was challenged to think about how we directly measure plants as they’re being injured by pests. Better data equals better strategies. I started exploring ways that we could look at insects in plants, and I ended up on something called a Piezo electric microphone, which is a contact microphone. I found them because I was looking into how people can measure the vibrations of walls to listen in on conversations. We can use similar strategies to eavesdrop on insects as they’re feeding directly on plants. And that’s why we called the sensor the Insect Eavesdropper.

How have you tested the Insect Eavesdropper?
The hardware is simplistic, but the software and data processing side are much harder problems to solve. We first asked the question, could we detect insects? We put large-bodied insects, such as tobacco hornworm, on the leaves of tobacco plants. We also tried Colorado potato beetles. You could hear them crunch. We could, indeed, detect those insects. Then we moved to insects that would be much more a target for a sensor like this. For instance, we tested Northern corn rootworm eggs, and, almost immediately after hatching, we could hear insects feeding on the root zone of plants with a microphone clipped a couple centimeters above the ground.

Beyond hearing the insects, what do you want the Eavesdropper to do?
The next big push was to determine whether we can identify the insects present. Applying machine learning algorithms, we have been able to show that we can, at least in the lab, with upwards of 96% accuracy. The lowest accuracy we saw when trying to tell these species apart from controls was about 80%. But the variety of insects we’d actually be seeing in the field is smaller. At any given seasonal time, entomologists know what’s going on. We understand what could be there, maybe 20 species of insects at a given time. And each insect species has a distinct circadian rhythm, a time of day that they feed. So, our problem space is much smaller, and it will become easier to separate and identify species. My lab is also focused on density questions. Can we estimate how many insects are on the plant? In the lab and in the field, we’ve placed different numbers of insects from zero to as many as we think is likely double to be there. We saw some clear correlations between numbers and the rate of chomping. At this point, we have been handing these sensors off, trying to figure out the limitations and the biases of these sensors with different research groups at Penn State, Cornell, USDA, Kansas State, and more, as well as with companies like Bayer.

Does the Eavesdropper use technology that’s new to the field?
We are definitely standing on the shoulders of giants. In the ’80s, folks put gramophone needles and seismograph needles on plants to sense substrate communication. In the ’90s, they started attaching accelerometers to plants to see how fast they were moving. Later, they put $5,000 laser vibrometer systems in labs to see the same signal vibrations that we’re picking up. The thing that makes the Eavesdropper different is the fact that it’s such a cost-effective system, and that we are putting a ton of effort into the signal processing to differentiate insect species. And we know it works in the field.

How many sensors do you put out, and how long do they have to be there?
The number of sensors will be pest- and question-specific. Companies may want to put sensors on every single plant because every single individual needs to be monitored to look for resistance. If we’re talking about regional monitoring, we can have plots like weather stations with fewer sensors that help growers understand which pests are entering the region and when. As for how long the sensor has to be on the plant, we assumed that we wanted 85% or 90% confidence that we’re catching the insect during its circadian feeding time, if it’s there. Turns out having a sensor on for just 40 seconds gets you 95% accuracy for corn rootworm.

What types of projects are you using the Eavesdropper for now?
We have one project with a farmer in Watertown where we’re looking at five fields with the Eavesdropper. We’re trying to figure out the spatial dynamics of larval corn rootworm. We’re trying to tie that back into soil water content and how soil water can change insect position or egg-laying behavior. The other one that we’re working on with actually a whole range of fields — both at UW research stations and with farmers— is how time of day and sweep netting intersect. The time of day when a person uses sweep nets to collect insects can affect the number they collect. During the day, insects often come up to the upper canopy of plants. Other times, the insects may not be as available. So, the timing of those measurements could greatly impact decisions on pest management. We’re trying to untangle that for two key pests: Japanese beetles in soybean and potato leafhopper in alfalfa.

How does your lab share information about pest issues?
The lab, in collaboration with the Wisconsin Department of Agriculture, Trade and Consumer Protection, has developed a text alert system that keeps Wisconsin farmers and crop consultants informed about incoming insect pressure. Anyone can sign up for that. All of these tools help us better understand how and when pest populations affect crops. Then we can better identify management decisions to help growers.

 

This article was posted in Fall 2024, Food Systems, Healthy Ecosystems, Living Science and tagged Arlington Agricultural Research Station, colorado potato beetle, corn rootworm, Emily Bick, Entomology, Extension, Insect Eavesdropper, Japanese beetle, machine learning, pest ecology, potato leafhopper, tobacco hornworm.