Detecting crested wheatgrass using sentinel and planetscope time series

Crested wheatgrass ( Agropyron cristatum ) is a non-native grass species in North America. Crested wheatgrass was considered valuable forage but its proclivity as an invasive weed has prompted land managers to control its spread. Remote sensing is an effective tool for monitoring the distribution of exotic grasses; however, it is challenging due to their sparse distribution and spectral similarity to other vegetation. Unique phenological characteristics can be leveraged to detect invasive plants using mid to high temporal and spatial satellite data. This study aimed to identify which spectral indices, temporal resolutions, and satellite datasets can be used to accurately detect crested wheatgrass, as well as determine the most important dates for its classification. Several Sentinel-2A and PlanetScope SuperDove spectral index time series were analyzed at 5-day, 10-day, and 30-day temporal resolutions, and crested wheatgrass presence was classified using a Random forest model. The 5-day Red–Green Ratio (RGR) PlanetScope model and the 10-day RGR Sentinel model performed best, highlighting the importance of high temporal resolution and spectral indices based on visible bands. Early-season and late-season features were the most informative, reflecting the unique phenology of crested wheatgrass (early spring growth and fall regrowth). Overall, the findings demonstrated the effectiveness of high-resolution multispectral time series for detecting invasive plants and provided valuable insights for developing grassland management strategies through improved monitoring approaches. • Sentinel-2A and PlanetScope were compared in crested wheatgrass classification • Several temporal resolutions and spectral index time series were analyzed • PlanetScope SuperDove 5-day RGR and Sentinel-2A 10-day RGR models achieved best accuracy • Early and late-season phenology were critical for crested wheatgrass classification