A grasshopper adult admires the sunset on a New Mexico rangeland. Myriad variables influence when and where outbreaks of pest grasshoppers will occur in the western United States, but researchers are making strides in developing models to try to predict the swarms. (Photo by Lonnie R. Black, USDA)

By Erica Kistner-Thomas, Ph.D.; Derek A. Woller, Ph.D.; Sunil Kumar, Ph.D.; and Larry Jech, Ph.D.

Setting the Stage

The battle between humans and grasshoppers has been going on in the western United States for over 100 years. These insects are major pests of rangeland habitats as well as adjacent croplands, and cyclical population outbreaks have the potential to cost the economy millions of dollars in annual damages.

Despite this risk, these native pests (mainly 12 to 15 species out of hundreds) are often overlooked in terms of their economic and ecological impacts compared to novel invasive insect pests. Ranchers annually seek technical assistance and treatment options for managing grasshopper outbreaks from the United States Department of Agriculture (USDA) Animal and Plant Health Inspection Service (APHIS) Rangeland Grasshopper and Mormon Cricket Suppression Program.

In addition to these duties, the program also conducts bi-annual population surveys, in spring and early summer for nymphs, and in late summer and fall for adults. The primary goal of these surveys is to identify areas of outbreak potential and provide very limited population predictions for the next season using a statistical methodology known as “empirical Bayesian kriging.” We say “limited” because, from a modeling standpoint, grasshopper outbreaks are notoriously difficult to predict due to the myriad and complex variables that are involved, such as species complexes, weather patterns, soil type and moisture level, and many more.

In a new open-access article published in June in the Journal of Economic Entomology, we developed a grasshopper population density forecasting model using the most complete APHIS population survey data set (10 years) we could find that focused on four counties in north central Wyoming. To the best of our knowledge, this is the first model of this type for grasshoppers to incorporate both geographic information system-based climate variables as well as landscape variables.

Incredible Diversity and Complex Variables

The diversity of grasshopper species at the 56 survey sites was staggering: A total of 99 species was recorded over 10 years, from 2007 to 2017. That’s a lot of hoppers! The two most abundant species sampled consistently were Melanoplus sanguinipes and Ageneotettix deorum, both of which are major rangeland pests and are part of the top 14 grasshopper pests in Wyoming. In fact, across the time period, all of these species shown in the graphic below were represented, with some far more abundant than others.

In a data set used to model rangeland grasshopper outbreaks, a total of 99 species was recorded over 10 years in four Wyoming counties, from 2007 to 2017. Shown here is the relative abundance of grasshopper species recorded, with the 15 most abundant identified by name. Asterisks denote species that are one of the 14 major pest species in Wyoming. (Figure originally published in Kistner et al. 2021, Journal of Economic Entomology)

Since no one had ever developed a predictive geospatial model for U.S. grasshoppers, we were not even sure which environmental variables would be good predictors for future population densities. Past research suggests that climate, topography, soil properties, land cover and land use types, historical grasshopper densities, and remotely sensed enhanced vegetation index are correlated with grasshopper population densities. Plus, as noted earlier, the presence of species complexes also presented a unique challenge because it is far more common to focus predictive models on a single species.

We ended up examining 72 biologically relevant environmental variables as potential predictors of grasshopper density in north central Wyoming. Using these predictor variables, we created several regression models and tested their robustness using the survey data from the years 2012 to 2016 as our response variable. The best-fit model included a handful of the predictor variables (some monthly weather variables and corresponding past mean grasshopper density) and was able to explain 35 percent of the variation, which we were pleased with, all things considered. In fact, when we compared this model’s population predictions to the actual (observed) survey data, we had a pretty good match overall (see map below).

A team of researchers used historical grasshopper outbreak data, combined with geographic information system-based climate variables and landscape variables, to develop a model for forecasting grasshopper outbreaks. The map here shows observed versus predicted (outbreak risk) mean grasshopper density levels in north central Wyoming for July from multiple regression modeling for 2012–2016. (Image originally published in Kistner et al. 2021, Journal of Economic Entomology)

What’s Next?

While our forecasting models provided moderate predictive power, there was still a lot of unexplained variation that traditional statistical models could not account for. Therefore, we have decided to start incorporating machine learning techniques, which can better-handle the complex ways in which biotic factors (like past population densities) and abiotic factors (like monthly precipitation) appear to be driving grasshopper population densities. To enhance our new modeling abilities even further, we are also now focusing on specific grasshopper species (12 of the most economically important pests in the west) and across the known geographic range for each.

Erica Kistner-Thomas, Ph.D. is a national program leader at the USDA National Institute of Food and Agriculture’s Institute of Food Production and Sustainability. Email: erica.kistnerthomas@usda.gov. Derek A. Woller, Ph.D. is a supervisory entomologist and team leader of the Science & Technology Rangeland Grasshopper and Mormon Cricket Management Team at the USDA Animal and Plant Health Inspection Service (APHIS). Email: derek.a.woller@usda.gov. Sunil Kumar, Ph.D. is an ecologist and quantitative risk analyst at USDA-APHIS. Larry Jech, Ph.D. is retired from the USDA APHIS. Email: larryjech@gmail.com.