Detecting Phenology Markers for Discernment of Grass Species at High Spatial and Temporal Scales

Abstract

Inexpensive near real-time environmental monitoring capabilities are needed by land managers to identify vulnerable plant communities and assess restoration efforts. Image acquisition by commercial airborne or satellite remote sensing is often cost-prohibitive. Development of relatively inexpensive, commercial multispectral unoccupied aerial systems (UAS) provides an alternative means of data collection at high spatial and temporal scales, which can be used for the detection of subtle phenological differences between plant species. Sub-meter multispectral imagery was collected using a UAS in a Southern California grassland/shrubland ecotone in June 2017, December 2017, and March 2018 to record phenological differences between study site herbaceous plants. Structure-from-motion software was used to build reflectance maps from the UAS data. An object-based image analysis framework was used to develop three vegetation cover classification experiments. Unsupervised clustering techniques were unable to correctly discern between herbaceous species. Future work will consider study sites characterized by more equal herbaceous perennial cover, increasing target spectral separability between herbaceous species while tuning the species detection framework.