Remote mapping of leafy spurge (Euphorbia esula, L.) in Northwestern Colorado

Abstract

Leafy spurge (Euphorbia esula L.) has been introduced to the Yampa River in Northwestern Colorado for over 40 years and flood and runoff events transport leafy spurge propagules onto adjacent landscapes. The spread of leafy spurge beyond the river channels has yet to be mapped and recorded, and this research was conducted to map leafy spurge occurrence in the Yampa River Valley. Significant stakeholder mapping efforts took place in the summer of 2019–2021, leading to excellent spatial data on leafy spurge presence and absence along the main channel. In summer 2019, multispectral SPOT seven satellite imagery, stakeholder ground mapping efforts, and bright yellow-green leafy spurge bracts were used to interpret imagery, identify dense, unobscured patches of leafy spurge, and digitize them. Spectral signatures from training samples for leafy spurge and other land cover classes (generalized as “not leafy spurge”) were then used to train a Random Forest machine learning classification. In the summer of 2021, generated classification maps were compared to multispectral satellite imagery and stakeholder ground mapped leafy spurge presence. Mismatches were identified, and 271 validation locations were identified, navigated to, and evaluated for leafy spurge presence. Leafy spurge training samples were classified with 96% accuracy. Correctly classified leafy spurge locations had higher leafy spurge coverage and lower overstory canopy than missed leafy spurge locations. Leafy spurge growing beneath shrub canopy or growing as individual plants along the riverbanks were more likely to be missed. A frequency analysis for other plant species found at validation locations determined that smooth brome (Bromus inermis Leyss.), dandelion (Taraxacum officinale L.), and willow (Salix sp.) were most frequently misclassified as leafy spurge. In conclusion, multispectral satellite imagery was useful at remote detection of leafy spurge in open areas with dense leafy spurge coverage, but more work must be done for identification of sparse and diffuse leafy spurge infestations.