|
Arctic vegetation mapping using UAV RGB imagery scaled up to AVIRIS imagery using ground spectroradiometry and ocular vegetation estimates
David
Paradis, University of Maine - Fort Kent, david.p.paradis@maine.edu
(Presenter)
We are using AVIRIS ng imaging spectrometer data as a means of scaling up cover estimates of lichens and other plant functional types to a fine taxonomic resolution using UAV-based RGB mosaics and ground measurements. We flew our commercial UAV within 2-3 weeks and under AVIRIS ng flight lines over seven 100x100 m areas (plots) in Alaska wherein we systematically sampled vegetation on 100m transects using target plant functional groups. We scanned each ground sample unit (1m2 quadrat where vegetation cover was ocularly estimated) with a spectroradiometer (PSR-3500 spectral evolution) in four 30cm diam FOVs, which were averaged to the quadrat. We used a contact probe with an active light source to make high quality scans of individual species from which we built a spectral library of arctic plants and lichens. We made lichen fractional cover maps produced from ML classification of UAV orthomosaics, located across Alaska, as training data for scaling up to AVIRIS flight lines. We also use the spectral library as the spectral endmember parameters which are passed to filtering algorithms which ultimately produce binary lichen/no lichen classification using AVIRIS data. Results from AVIRIS vegetation mapping are tested against fractional lichen cover maps derived from the independent UAV imagery. Our end product will be a large number of AVIRIS flight lines with high precision vegetation cover estimates that will be scaled to Landsat and larger extents in Alaska and NW Canada. Associated Project(s): Poster Location ID: 95 Session Assigned: Vegetation Dynamics and Distribution
|