Predicting Pests: Combining Meteorology, Biology, and Information Technology
Penn State University
Krista Weldner, Penn State Staff
Checking the daily weather forecast is a matter of habit for most of us—it’s a handy tool that helps us decide which jacket to wear or whether to grab an umbrella on the way out. But for farmers, forecasting is more than a convenience: It can be an integral tool that allows for economically and environmentally sound pest management. And while forecasting the arrival of pests has always played an important role in agriculture, today’s growers can get more specific, up-to-date information than ever before through computer-based, pestprediction models.
In one ongoing, cooperative project, entomologist Dennis Calvin and his research team look at how climate and weather influence the timing of insect emergence in field crops. Focusing on European corn borer, corn rootworm, common stalk borer, and alfalfa weevil, the researchers integrate data on temperature, climate, insect development, and plant development to create computer models that predict insect phenology: the timing of biological events. Scientists have long known that insect development is related to temperature. During their life cycle, insects move through instars, the discrete stages of larval development. Each one of these stages is temperature dependent: The cooler it is, the longer insects stay in a particular stage; the warmer it is, the faster they move through that stage.
“To create models of insect development, we express this concept mathematically,” Calvin explains. “If you know an insect completes a tenth of its development every day at a certain temperature, you can say it takes ten days to go through that stage. To develop models, you have to take, say, a hundred insects and rear them in a laboratory growth chamber at a constant temperature, keeping track of how many hatch or transition to the next stage. From that information you get an average time they spend in each stage. Then you take another hundred insects and keep them at a different temperature, and so on. Gradually you build a range of points from which to build equations, and from those equations you build models.”
The next piece of the puzzle involves getting the best possible sitespecific temperature measures. In the past, temperature records from National Weather Service stations provided the best estimates for weather predictions. But because stations could be as many as 50 miles apart, and local weather varies greatly between stations, predictions often were not accurate. In the mid-1980s, former assistant professor Joe Russo developed techniques to acquire sitespecific temperature measurements between weather stations, taking into account factors such as topography. (For example, weather stations are usually near airports, and temperatures there would differ from those in farm fields.)
These site-specific temperature measurements, when merged with insect-development models, allow researchers to predict when insect pests at a given site will emerge or enter a key stage of development. ZedX Inc., an information technology company based in Bellefonte, Pennsylvania, provides daily weather data that Calvin feeds into the insect-development models. “We take the weather data from ZedX, merge it with our insect models, and estimate when, for example, European corn borer is going to hatch or transition into a stage,” he says. “Essentially, our models move the insect from stage one to stage two and on through the life cycle based on whether it’s a warmer or cooler year.”
These forecasting models span across every geographical area in North America, making the college an international leader in pest forecasting-model research. Calvin’s team uses weather data that’s been spatially interpolated, which means it has been made relevant for a specific point on the landscape by taking into account elevation, longitude, latitude, and other factors. Based on these variables, the researchers can modify temperature predictions for greater accuracy. The insect prediction models that run every day have a resolution of 10 square kilometers, or about 6.2 square miles, meaning researchers can make pest predictions for every 6.2 square miles in North America.
The information generated by forecasting models is reflected on maps created by ZedX. Users of these maps, which have been available on the Internet for the past four years, include extension educators, individual growers trying to manage crops more effectively, and various businesses such as seed corn companies and pesticide companies.
The prediction models provide a seven- to tenday forecast, making them an effective early warning system. The information from models can save time and money by allowing growers to see when an insect pest is at a stage that causes damage in the field or when the pest is gone for the season and no longer a source of worry. Maps can also tell growers specifically when they should be scouting their fields for insect eggs, other life stages, or early damage symptoms. Growers who use biological controls can determine when to release biological control agents, according to the development stage of the pest. Organic and conventional growers can avoid damage from a particular pest by knowing the stage it will be in relative to when the crop is in the field.
“We also have models that ‘grow’ the corn plant so you can see the various stages of plant development,” Calvin says. “There are changes in the plant that make it attractive to pests. For example, corn rootworm adults prefer to lay their eggs in fields that are shedding pollen. Another model depicts plant-insect interaction and can predict corn plant yield impacts according to when it is attacked by European corn borer. We’ve quantified through field research and mathematical models how much loss there will be if they attack the plant at different heights, when the ears start to form, and so forth. We can tell the stage the corn will be in on any date, and we can tell the stage the insect will be in on any date. We can then merge the data and measure impact on yield. Using that information, growers can find the ideal window in which to plant.”
Another advantage offered by forecasting models is geographic specificity. For example, explains Calvin, “In a state like Pennsylvania, because of the climatic variation and the weather being affected by topography, events such as egg hatch could be three or four weeks earlier in the warmest part of the state compared to the coolest part. With our forecasting models, we can tell growers in, say, Potter County, which is cooler, that they’re still three weeks away from egg hatch. But if you’re down in Chester County, it’s happening right now.”
While Calvin’s research focuses on resident pests, another entomology research project deals with migratory pests, which come with their own challenges. Entomologist Shelby Fleischer has developed Pestwatch, a coordinated system of monitoring migratory pests of vegetable crops, particularly sweet corn. “Pennsylvania ranks in the top five states for producing fresh market sweet corn,” he says, “so we’re interested in optimizing production through integrated pest management. To do that, you need to monitor insect population density, insect life stage, and crop stage.
“So if we knew more about when and where these immigrants arrive each season, we could dramatically reduce insecticide use,” he continues. “Unfortunately, because of farm and crop diversity, spatial segregation of farms, the long distances that these migratory pests travel, and the smaller size of many farms in the Northeast, traditional IPM monitoring programs, such as field scouting, can be a logistic and economic challenge.”
Fleischer monitors two migratory species of lepidopterans—moths and their caterpillar larvae— that eat sweet corn: corn earworm and fall armyworm. “Fall armyworm is a true migrant that has no physical mechanism that allows it to get through harsh times,” he explains. “It doesn’t hibernate or suspend development. The corn earworm is more of a mix—we might have some overwintering here, but just how much is hard to measure.”
The female corn earworm moth, Fleischer explains, deposits a single egg on the silk of an ear of corn, then another egg on another ear of corn, and so on. “So one female doing a good job can spread out 500 eggs in a few nights. When we do the math, we can conclude that 30 to 50 females can blanket a crop-acre and cause 80 to 100 percent infestation.” Within three days after the eggs are laid, they hatch into larvae, and within another four days they manage to crawl into the ear, hidden from any pesticide application. While one management option is to blanket-spray insecticide so that it’s already on the silk when the eggs are laid, more accurate monitoring could help growers avoid possibly unnecessary blanket spraying.
One measuring method that’s had some success is pheromone trapping, which traps male insects. Presuming that females are laying eggs at the same time the males are flying, researchers can monitor egg laying indirectly by counting male insects caught the night before. Data from the pheromone traps, combined with field scouting, can provide some sense of the migratory process and how to manage crops accordingly.
“We monitor insect activity, and we want to use information technology to map this activity in real time,” Fleischer says. “For that we need people—what we call the human infrastructure of Pestwatch.” Extension educators throughout Pennsylvania play an important role by establishing and monitoring field sites, then entering data from those sites online. The data is processed in the form of maps that allow users to see a time series graphic for each site, as well as a point map of all sites.
“We worked with Doug Miller and his group at Penn State’s Center for Environmental Informatics to create this interactive cartography,” says Fleischer. “Doug saw the potential for using Macromedia Flash as an application for pulling data in and rapidly mapping it back out. He and his group created this system, which links time series graphics with a map interface so users can get up-to-date pheromone trapping data from any site. In extension, we’re seeing a shift toward dynamic information, and that’s what Pestwatch gives us.”
Today, Pestwatch extends from Virginia to Maine, and plans are under way to extend the program into the Midwest and Ontario, Canada, in response to concerns that Midwestern sweet corn pests are becoming resistant to certain insecticides. Fleischer also plans to collaborate with aerobiologist Scott Isard and plant pathologist Erick De Wolf to incorporate an aerobiology modeling component into the Pestwatch system.
Isard is using aerobiology, the study of organisms that move in the air, to monitor the spores of Asian soybean rust, an invasive plant disease that poses a threat to the nation’s soybean crop. In the 1990s, when the disease spread from Asia into Africa and South America, the USDA became concerned about potential invasion into North America. At that time, Isard had been developing theory for forecasting movement of organisms and had published a book entitled The Flow of Life in the Atmosphere. As a result of that work, he received a biosecurity grant from USDA to create a forecasting system based on field research in South America. “So in 2004, when soybean rust came into the southern United States late in the growing season—the spores came in during Hurricane Ivan— we had developed our model and it proved very helpful,” Isard says.
To be able to forecast movement of disease spores in the air, Isard needed to begin with source populations. “Forecasting movement in the air is a challenge, especially when you have a new species in a new environment,” he says. “You don’t understand the biology until you have enough time to do research.” To track progression of the disease, Isard and his colleagues established sentinel plots of early-maturing soybean varieties in each state in the South. Because these early-maturing varieties become infected before commercial fields do, sentinel plots can indicate where soybean rust is likely to appear.
“With the information we get from sentinel plots, we develop a model,” Isard says, “taking into account factors such as temperatures, winds, solar radiation, and precipitation. The model, which we run daily, helps us determine where we should expect infection to develop. And so far, this pathogen has not spread as rapidly as we expected. It needs green soybean or kudzu leaves to grow, and therefore it can survive only in the deep South, where winters are warm. Each year the pathogen is blown northward during the summer. So far we’ve been fortunate—it has not yet appeared in Pennsylvania, although in fall 2006 it came close, making it as far as Virginia and central Indiana. We speculate that in years when the deep South has a warm and wet spring, this pathogen will be a serious, widespread problem.”
Isard worked with the information technology company ZedX to develop the forecasting system, which operates on a national basis through USDA and is available on the Web. Each day, researchers, extension educators, observers of sentinel plots, USDA officials, and industry representatives enter data. Producers can access up-to-date maps, current surveillance reports, county-level information on disease status, and disease-management guidelines developed by county-based extension educators. Through this framework, soybean producers have easy access to the most current information on soybean rust.
An important aspect of the soybean rust forecasting system is the risk-assessment component. “One of the biggest impacts we have is the ability to forecast that the risk for the pathogen is low or nonexistent, so growers don’t have to spread fungicide,” Isard says. “If you’re a soybean producer in Pennsylvania, there’s no risk now. That’s valuable information, and it has a tremendous economic and environmental impact. In fact, the Economic Research Service published a report stating that information provided by the USDA Soybean Rust Information System Web site increased U.S. soybean producers’ profits by as much as $5 million in 2005. These savings, and the benefit to the environment of not spraying millions of acres with fungicides, demonstrates the value of a coordinated national pest management framework.”
The success of the soybean rust project stimulated the development of the 2006 Pest Information Platform for Extension and Education, which has been adopted by USDA as a new paradigm for pest management. It is now supported by the USDA Risk Management Agency as a central platform that facilitates communication among state specialists across the nation and is expanding to include soybean aphid as well as other agricultural pests.
“This has grown way beyond Penn State,” Isard says. “We built the platform— the concepts came from here. And we still run the system from an operational point of view—the technology is here. While Penn State has been the center of innovation, there are now thousands of people involved in this platform.”
In recognition of its work, the Asian soybean rust team was selected to receive one of the 2006 Department of Agriculture (USDA) Secretary’s Honor Awards in the category of Enhancing Protection and Safety of the Nation’s Agriculture and Food Supply. These awards, the highest given by USDA, acknowledge outstanding contributions to agriculture and consumers of agricultural products.
Next door to Isard, plant pathologist Erick De Wolf is part of another USDA project—the U.S. Wheat and Barley Scab Initiative. This initiative features an online disease-forecasting system that helps wheat and barley growers nationwide predict the onset of fusarium head blight, a fungal disease that affects both wheat and barley. De Wolf is part of a five-state effort, including Pennsylvania, Ohio, Indiana, North Dakota, and South Dakota, to develop forecasting models for this disease.
The Web-based forecasting system, combining the weather expertise of climatologist Paul Knight and the Web-mapping technology from the Center for Environmental Informatics, incorporates data from the National Weather Service, including temperature, humidity, and rainfall. Growers provide information on their crops, such as whether they have winter or spring wheat, expected flowering date, and production practices. The forecasting system then assesses disease risk according to the grower’s location.
“Fusarium head blight can lead to yield losses of 50 percent in many areas,” De Wolf says. “This disease clearly has a major economic impact. The goal of our forecasting efforts is to predict when and where disease outbreaks will occur and give that information to growers so they have time to factor the risk into their management practices. The other side, of course, is to tell growers when a fungicide isn’t necessary. In most cases that’s the most important use of forecasting. Use chemicals when you need to, but if you don’t need to, save money and be more efficient.”