Monitoring Vegetation Biomass in Continental Antarctica: A Comparison of Hyper- and Multispectral Imagery

• Main Author: White, Megan
• Other Authors: Colesie, Claudia
• Format: Master Thesis
• Language: English
• Published: The University of Edinburgh 2020
• Subjects: Remote Sensing; Sentinel-2; Biological Soil Crusts; Antarctica; Supervised Classification; Random Forest; Spectral Unmixing; Antarctic; The Antarctic; The Sentinel; Antarc*
• Online Access: https://hdl.handle.net/1842/37639; https://doi.org/10.7488/era/920

In the last few decades Antarctica has come under intense scrutiny as an area that could potentially provide insight into climate change as an early warning system for the rest of the world. This is due in part to the vegetation that inhabits the area which includes populations of lichen, algae and moss. Also known as biological soil crusts, lichen, algae and moss have all been proven to be indicators of climate change and pollution. The Antarctic environment has the advantage of being mostly untouched by the influence of humanity and other environmental factors. This allows for a pure environment for the study of how climate change affects the distribution of vegetation. The recent availability of the Sentinel-2 satellite constellation provides researchers with an opportunity to increase the scale of vegetation surveys beyond what manual surveys can conduct. In this study a random forest algorithm is applied to UAV hyperspectral imagery and spectral unmixing is applied to Sentinel-2 imagery. A spectral library extracted from the UAV imagery is used to conduct the spectral unmixing of 10m resolution satellite imagery. The outcome was a comparison of the effectiveness of the two different resolutions and the creation of a classification map for the diversity of Biological Soils Crusts in the Antarctic environment.