Monica L Vermillion, Josh Enterkine, Lucas Spaete, Dr. Nancy Glenn Boise State University
13P
The semi-arid ecosystem of the Great Basin range is vulnerable to an increasing fire cycle due to invasive species such as cheatgrass. The rapidly changing fire regime places additional demands on land management planning. Vegetation maps generated from remote sensing can play a critical role in identifying at-risk areas, areas of high conservation value, and invasive species encroachment. However, accurate high resolution vegetation maps of large areas can be difficult to create and update with traditional data processing techniques. Emerging remote sensing platforms (Sentinel-2) when combined with cloud-based computing (Google Earth Engine, GEE) can be used to create high resolution maps of dominant species over large areas of land. Since the process is cloud-based it allows for large amounts of data to be computed in minimal time.
For this project we collected field data on the Mountain Home Air Force Base and Range Complex consisting of 20 m plots of native and non-native  vegetation. The field data, consisting of coordinates taken with real-time kinematic (RTK) GPS, are used as training data for Random Forest, an ensemble machine-learning model, to classify Sentinel-2 imagery, in GEE. Sentinel-2 provides high temporal and spatial imagery with multiple RedEdge bands at 20 m resolution, making it ideal for this application. Our model is readily-updated with additional satellite data as it becomes available, enabling rapid updating of vegetation maps. Since vegetation is only one factor in ecological site response, climate and topography can be added to create a more cohesive picture of the study area.
Once created, a large area vegetation map can be used to identify potential areas that are at-risk of disturbance or invasive species, areas that are likely resilient, and areas that are in transition or are vulnerable. These maps can be of use to inform land management decisions of ideal invasive treatment methods. Then future maps can be used to assess the treatment effectiveness and for change detection given updated field data each year.
13:20 Applying Cloud-Based Computing and Emerging Remote Sensing Technologies in Vegetation Management Decisions, Monica L Vermillion