Evaluating High-Frequency, Moderate-Resolution Satellite Imagery for Assessment and Monitoring of Invasive Annual Grasses

Invasive annual grasses (IAGs) represent an unprecedented threat to native plant communities in rangeland ecosystems, displacing native vegetation, altering fire regimes, affecting wildlife species that depend on native perennials for food and shelter, and causing immense economic costs. Remote sensing allows the monitoring of IAGs through multispectral imagery, using phenological differences to separate invasive species from native vegetation. Our research goal was to evaluate the use of high-frequency, moderate-resolution multispectral Planet imagery and machine learning in detecting IAG species – specifically ventenata (Ventenata dubia [Leers] Coss.), medusahead (Taeniatherum caput-medusae [L.] Nevski), cheatgrass (Bromus tectorum L.), and Japanese brome (Bromus japonicus Thunb.). Our research questions were: 1) do IAG species groupings influence remote detection rates? and 2) can high spatio-temporal resolution imagery accurately identify IAG response to large-scale herbicide applications? Ventenata was best represented in our invasive grass training data and by classification models and accuracy indices. Models detected differences in IAG presence likelihood in pastures treated aerially with indaziflam at 73 g · ai · ha⁻¹ and untreated pastures, even in treated locations where no ground mapping took place. This suggests the possibility of extending these invasive grass prediction models beyond ground-mapped training areas in similar mixed-grass prairie environments to increase the efficiency of IAG monitoring.