Monitoring Dynamics of Invasive Plants Using Remote Sensing and Machine Learning Tools

Geography and the Environment Colloquium Series.

Guest speaker Susan Meerdinck, University of Iowa.  

Invasive plant species remain one of the primary threats to ecosystems because they alter critical ecosystem functions and suppress native species' ability to respond to changes in environmental conditions. Accumulation of invaders is projected to increase over the next two decades. Detection and long-term monitoring of invasions provide valuable ecological information and ultimately guide management decisions.

One of the challenges of managing invasives across landscapes is determining where they are located, particularly at the early stages of an invasion, where they occur at low densities mixed with native vegetation. Traditional surveys through ground efforts are often infeasible due to cost and logistics, especially when considering repeat surveys are needed to determine spread. Remote sensing could improve response to invasions by providing accurate and affordable repeat imagery as a more cost-effective way to monitor spread.

This research explores our ability to detect invasives using multi-spectral remote sensing and a machine-learning approach adapted to detect invasives at low densities with imprecision in the dataset. Here we show data from the Everglades National Park, which presents an ideal case study for using remote sensing to capture invasive species spread because the majority of the park is accessible only by watercraft or helicopter, making monitoring and managing invasive plant populations exceptionally difficult and costly.