Climate & Sustainability
Pest prediction and management improved with better modeling
With new research, applied mathematicians at UC Santa Cruz introduce methods to improve the forecasting of pest populations
The researchers studied the Lygus bug, a common cotton pest in California. (Scott Loarie/ iNaturalist)
Press Contact
Media Access
Key takeaways
- A new strategy helps choose the best timing and strength for interventions to control insect pest populations
- The method is generalizable to any pest population and can be used to understand the impact of interventions beyond traditional pesticides
Insect infestations pose danger to humans by destroying crops, spreading disease, and damaging ecosystems, but efforts to manage them are often reactive, coming into effect only after an outbreak has begun.
With new research, applied mathematicians at UC Santa Cruz introduce methods to improve the forecasting of pest populations, and use this to make predictions about an ideal management and intervention strategy. The researchers show that their method can help choose the best timing and strength for taking action to control pests, reducing the overall costs of these interventions. The approach can help to understand the impact of interventions beyond traditional pesticides.
The study, published in the journal Ecological Monitoring, was led by Adjunct Professor of Applied Mathematics Stephan Munch, Associate Professor of Applied Mathematics Marcella Gomez, and Cal Poly Humboldt Assistant Professor of Mathematics Bethany Johnson, a former UCSC Ph.D. student. The research was supported by the National Oceanic and Atmospheric Administration, with Johnson supported by a Fisheries-Sea Grant Joint Graduate Fellowship.
“Using a method like this tends to reduce costs both in terms of the damage a pest can do and their overall population levels — and in the amount of control that you have to apply. It reduces those costs compared to a reactive approach,” Johnson said.
Understanding and preventing pests
Worldwide, pest outbreaks spread diseases that lead to hundreds of thousands of human deaths per year, and are estimated to cost at least $76 billion annually in healthcare, goods, and services costs. Damage from pests is expected to increase as the climate crisis continues to progress.
Pest management strategies are often reactive, meaning that outbreaks are addressed with pesticides or other interventions after an issue has already started, which are often not ideal for containment. While some take proactive approaches, these tend to be very specific to an individual pest species and can’t be easily applied to other systems.
In contrast, the researchers have devised a method that is generalizable to any pest population, and is especially useful for the many insect species that scientists have not already studied in a rigorous manner to understand their biology and dynamics.
“This method and the results in this paper show that even when we pretty much know nothing about the pest that we’re working with, we can make reasonable predictions which tend to have better performance than strictly reacting to whatever the insects are doing,” Johnson said. “This is a method that would be very useful for those understudied pests where people are still trying to figure out that biological information.”
The method is also compatible for those studying ways to control pests beyond traditional pesticides.
“There’s a lot of research effort being done in this idea of integrated pest management, looking into alternatives to chemical insecticides that we can use to suppress pests,” Johnson said. “This is a method that can be pretty flexible when it comes to using combinations of intervention methods, or favoring one over another. It’s easily transferable to different types of interventions that we might want to make — it’s not something that will become obsolete if chemical insecticides become obsolete.”
The project prompted the researchers to consider which strategies might be realistic for the agricultural industry to pursue.
“It was interesting to try to figure out where there’s mutual interest and alignment here,” Gomez said. “You need to try to align what’s in the interest of the AgTech farmworkers and what’s low barrier for them to implement with the sustainability goals this method is trying to achieve.”
Better models for ecological forecasting
Munch is a leading researcher of an approach called empirical dynamic modelling, a method of using historical data to make forecasts about the future of a species’ population dynamics. Munch typically applies this method to problems of fisheries management, but this project opened up another highly relevant area of ecological forecasting.
This method enables researchers to make predictions about a population without needing data on all of the ecosystem factors that affect it, such as the populations of their predators and prey. Instead, it relies on applying mathematical theory to historical data — known as time series data — about a population to make predictions of the future.
“It turns out that by incorporating the history, you can account for some of those unobserved components of the ecosystem that we aren’t able to take measurements of, or that we don’t have access to data for,” Johnson said.
To make predictions for the best way to control pest populations, researchers paired the predictions of population size with a second modeling method called stochastic dynamic programming, which predicts the possible effects of possible intervention such as applying pesticides. This allows the researchers to forecast the effects of any potential intervention for each population size and decide what an optimal action might be.
“Using those two methods together, we’re able to come up with a policy for what we should do in any given state we might find ourselves in,” Johnson said.
This approach is immediately transferable to many species, but works best with short-lived species, as Munch’s research with fish and other populations has shown. This is because the models need several population-cycles-worth of data to make predictions, which is easier to get for species with short life cycles.
“The methods require recurrences to make a statistically reasonable prediction,” Munch said. “The recurrence time is short for the short-lived species, so the same number of years worth of observations gets us to a much more informative time series for things like pests or plankton or diseases, than it does for long-lived species like a desert tortoise.”
Real-world pests
To test the effectiveness of these models, the research team applied their methods to pests that impact Californians, particularly the cotton crop pest lygus. This was challenging for the researchers because of the lack of publicly available agricultural data, and if there is data, it has not been collected with mathematical modelling in mind.
Lygus bugs have done a great deal of damage to cotton plants in central California, which led a group of researchers at UC Davis to collect extensive long-term monitoring data — data that is essential for powering the researchers’ models. This rigorous record of data allowed the researchers to verify that their models make accurate predictions of population size.
This example also demonstrated how data from multiple sites — in this case, multiple cotton fields — can be used to fill in gaps in the time series data, providing more robust information from which to make predictions.
Going forward, the researchers are interested in finding opportunities to further study this method in collaboration with scientists working in the field to study specific species.
“I’m definitely keen to apply this to some real system where we can actually do it and get results and do it some more and get results and have some real world applications,” Munch said. “I think that really the next step is to find a partner, and actually put this into practice.”