Скачать или смотреть Mapping of Invasive Species with Unoccupied Aerial Systems

Dr. Megan Cattau, Assistant Professor in Human-Environment Systems at Boise State University, will present how the use of modern technology (UAS) can aid in mapping invasive species.
Invasive plant species are one of the most prominent threats to native vegetation in the Great Basin. Detecting and mapping invasive species is a requisite first step toward understanding changes in their distribution and abundance and supports management interventions. The use of remote sensing technology to detect and map invasive species over large spatial extents is limited, in part, by the spectral similarity of different species, meaning by how similar they look in remotely sensed imagery. However, differences in phenology between species can result in unique spectral and structural signatures over time; for example, we can leverage species-specific timing of green-up to differentiate species from one another in remotely sensed imagery. The use of Unoccupied Aerial Systems (UAS), or drones, is a particularly suitable tool to detect species-specific phenology, as data can be collected on-demand and at a sufficiently fine spatial grain, even in communities of diverse vegetation. We conducted repeat UAS flights in an arid system near Kuna, Idaho to capture the spectral and structural signatures over time of invasive species (Cheatgrass, Russian Thistle, and Bur buttercup), native grass species (Squirreltail and Poa secunda), planted vegetation species (Russian wildrye and Forage Kochia), and other background land cover types (bare soil, rock, and moss/biocrust). A 10-band multispectral sensor was flown approximately every 2 weeks March - November 2021, and a red, green, blue sensor was flown during select moments in the growing cycle to derive spectral and structural information, respectively. We found that nearly every species is uniquely distinguishable based on the date when a metric of live vegetation, Normalized Difference Vegetation Index (NDVI), reached its peak and on the magnitude of peak NDVI. Further, classification of the UAS data using NDVI from all dates resulted in 95.7% calibration and 89.8% validation accuracies. These efforts to resolve species-specific phenologies will improve mapping of these species. They will ultimately benefit mitigation efforts that leverage phenological information (e.g., post-emergent herbicide) and/or that require spatially-explicit knowledge of species occurrence.