Detecting Crested Wheatgrass Using Remote Sensing

Crested wheatgrass (Agropyron cristatum) is a perennial grass widely introduced across the North American prairies, where it has become a persistent invasive species. Initially planted to revegetate degraded rangelands after the 1930s drought, its establishment success stems from early spring growth, extensive roots, prolific seed production, and tolerance to grazing and drought. These traits enable rapid establishment, seedbank dominance, and displacement of native grasslands, reducing biodiversity and degrading plant and animal habitats. Monitoring this invasive grass remains a major challenge, as field surveys are costly and limitations of remote sensing to differentiate between native and non-native grass species. This research integrates field-based hyperspectral measurements with biophysical data to develop a spectral and trait-based approach for separating crested wheatgrass. Field surveys and hyperspectral data revealed that crested wheatgrass stands had lower bare ground cover and higher grass cover, height, leaf area index, and biomass compared to native grasses. Spectrally, crested wheatgrass exhibited significantly lower reflectance in visible and shortwave regions. Several spectral indices, Red Green Ratio (RGR), Water Stress Index (WSI), Shortwave-Infrared Vegetation Index (SWVI), and Cellulose Absorption Index (CAI), effectively separated crested wheatgrass. These indices were applied to Sentinel-2A and PlanetScope time series for classification. The 5-day RGR PlanetScope based model accurately mapped crested wheatgrass. Results emphasized the importance of satellite data with frequent revisit intervals (i.e. high temporal resolution: 5-day) and phenological windows such as early spring growth and fall regrowth. This study identifies key spectral regions and biophysical traits sensitive to species differences, offering practical insights into scalable satellite-based detection strategies to support grassland conservation and invasive species management.