Abstract
Use of Unmanned Aircraft Systems (UAS) gives the opportunity to carry out research with a reduced environmental footprint. Unmanned aircraft, including both fixed wing and multi rotor types (helicopters) allow us to collect very high resolution image data for vegetation mapping without the need for any personnel walking into the site and thereby potentially disturbing the sensitive Arctic ecosystems. The main aim of this project was to explore the feasibility of UAS-based vegetation mapping and extraction of vegetation indices (NDVI) for a range of different Arctic plant communities including dense marshes, moss tundra communities and different tundra heaths. The study area of Adventdalen valley on Svalbard, Arctic Norway, is located at 71.2°N 16°E and experiences a dry Arctic climate with a mean July temperature of about 6°C. The UAS was a fixed wing aircraft instrumented with a Red, Green, Blue (RGB) compact camera and a Normalized Difference Vegetation Index (NDVI) camera taking pictures from 100 metres altitude with highest ground resolution of 2.5 cm capable of mapping 2-3 km² per flight. The study area's two main plant communities; the Arctic bell heath and a graminoid rich Polar Willow heath were easily detected both in the NDVI and RGB images. In addition, wet moss tundra and mires were separated from the heath communities. In the NDVI image the moss-dominated mires were difficult to separate from the graminoid dominated mires in most cases, but they were well separable in RGB colour space. Also in situ NDVI measurements by a handheld passive proximal sensor were simultaneously done during the flight campaign. These measurements were analysed in order to correlate the species level NDVI and community level NDVI measurements with the NDVI images acquired at a variety of spatial resolutions by the UAS. The analysis shows that NDVIs of four main plant species at in situ leaf and community levels were significantly correlated (R² = 0.60, p<0.01). The correlation between the surface (in situ) NDVI community level and the UAS NDVI community level acquired from 100 metres above the surface of four main plant communities was R² = 0.75 (p<0.01), and these two scales are considered to be best for extraction of NDVI observations in Arctic areas like Svalbard.