In our recent paper in Communications Biology, our research team made a perhaps startling prediction: if nothing is done to control the spread of the invasive insect spotted lanternfly in the United States, the species will likely invade California in nine years or less, placing the state’s multi-billion-dollar grape industry at risk. There’s even a chance that the pest could arrive in California in less than five years from now.
How did we arrive at this prediction? And how can such a forecast help prepare for, or even prevent, pest invasion?
Our paper’s lead author, Chris Jones of the Center for Geospatial Analytics, is the lead developer of a modeling framework called PoPS, or Pest or Pathogen Spread. He and his team have been working closely with personnel from the US Department of Agriculture’s Animal and Plant Health Inspection Service (USDA APHIS) to make sure that PoPS can do two things very well: predict the spread of invasive species across both space and time, and serve as a virtual testing ground for control strategies.
Working with resource managers improves predictive models
Using information about the pest’s native range, researchers had already identified where in the US spotted lanternfly (SLF, Lycorma delicatula) would likely thrive. When SLF could be expected to invade “highly suitable” areas like California, though, remained unknown. Hence the need for PoPS, which, for every pixel in a landscape, can simulate the processes of pest reproduction, dispersal, and establishment.
To predict the spread of a pest species, PoPS uses the locations and timing of where the pest has been found, the locations of host plants it feeds on, and information about its biology, such as the impact of precipitation and temperature on survival and reproduction and how it disperses to reach new places.
In the case of SLF, this last point is crucial. Spotted lanternflies are weak fliers, so they rely on accidental transport by people to carry them far. The species lays its eggs on anything from pallets of cut stone to cars. Natural resource managers on the ground, however, began to notice a pattern in their SLF surveys––the insect was particularly abundant in the rights-of-way along train tracks.
After conversations with APHIS personnel, Jones adjusted the design of PoPS so that it could capture the main way that SLF travels long distances––hitching a ride on railcars. Not coincidentally, rail lines often pass through prime habitat for tree of heaven (Ailanthus altissima), another invasive species and SLF’s primary host. Adjusting the PoPS model to account for dispersal along rail networks improved its accuracy by 8%.
PoPS continues to be modified to suit the observations of managers in the field. For example, US states at the current fringe of SLF invasion have observed the pest at truck stops, and Jones reports that trucking networks will soon also be integrated into PoPS to capture local patterns of spread and fine-tune predictions at various spatial scales.
Virtually simulating pest control for optimal impact
Our paper described a scenario in which no efforts were made to stop the spread of SLF in the US. But of course, government agencies at both the national and state levels are currently trying to control SLF, and new strategies for doing so are being brainstormed all the time. Our predictions therefore provide a baseline, or benchmark, against which control efforts can be compared, to see how effective they would be (or, in retrospect, were) versus doing nothing.
Every six months or so, Jones and his team meets virtually with APHIS and multiple state Departments of Agriculture to run new simulations with PoPS at the state level, using an interactive, web-based dashboard (see video below) to simulate and visualize how SLF might be managed using a realistic budget. These meetings have had a big impact on the way that PoPS works and the look and feel of its interactive dashboard.
PoPS allows users to compare a forecast without management to “what-if” scenarios that involve, for example, applying insecticide or removing host plants. Users can even simulate adaptive management in multiple years to examine the potential cumulative impacts of interventions on spread through time.
As SLF continues to be detected in new places, Jones and his team feed that info into PoPS, regularly updating the spread simulation. Because the species is hard to find when it’s at low numbers, detection usually lags behind true spread, and so PoPS’ predictions also guide survey strategies, indicating where more search effort would be helpful.
Using PoPS to forecast when SLF can be expected to reach new places also means that preventative action can be taken sooner. Already, Jones reports, some states in the Northeast are trying to get ahead of SLF invasion by removing the insect’s invasive host, tree of heaven, from transportation hubs like ports and airports, and from around vineyards and orchards vulnerable to SLF damage.
Ecological forecasting platforms like PoPS provide a win-win-win for scientists, natural resource managers, and growers, through knowledge-sharing about the likely spread of pests and through virtually testing how to slow or stop them. Although our paper forecasted that spotted lanternfly will reach California within a decade if nothing is done to stop it, we are hopeful that this future will not come to pass.
We encourage others to use the open source PoPS platform for their own species of interest. The newest version (PoPS 2.0) is now available on GitHub.