Multiscale mapping of plant functional groups and plant traits in the High Arctic using field spectroscopy, UAV imagery and Sentinel-2A data

• Published in: Environmental Research Letters
• Main Authors: Thomson, Eleanor R, Spiegel, Marcus P, Althuizen, Inge H J, Bass, Polly, Chen, Shuli, Chmurzynski, Adam, Halbritter, Aud H, Henn, Jonathan J, Jónsdóttir, Ingibjörg S, Klanderud, Kari, Li, Yaoqi, Maitner, Brian S, Michaletz, Sean T, Niittynen, Pekka, Roos, Ruben E, Telford, Richard J, Enquist, Brian J, Vandvik, Vigdis, Macias-Fauria, Marc, Malhi, Yadvinder
• Other Authors: Norges Forskningsråd, SIU-foundation project, Jackson Foundation, Natural Environment Research Council
• Format: Article in Journal/Newspaper
• Language: unknown
• Published: IOP Publishing 2021
• Subjects: Arctic; Svalbard; Longyearbyen; Tundra
• Online Access: http://dx.doi.org/10.1088/1748-9326/abf464; https://iopscience.iop.org/article/10.1088/1748-9326/abf464; https://iopscience.iop.org/article/10.1088/1748-9326/abf464/pdf

Abstract The Arctic is warming twice as fast as the rest of the planet, leading to rapid changes in species composition and plant functional trait variation. Landscape-level maps of vegetation composition and trait distributions are required to expand spatially-limited plot studies, overcome sampling biases associated with the most accessible research areas, and create baselines from which to monitor environmental change. Unmanned aerial vehicles (UAVs) have emerged as a low-cost method to generate high-resolution imagery and bridge the gap between fine-scale field studies and lower resolution satellite analyses. Here we used field spectroscopy data (400–2500 nm) and UAV multispectral imagery to test spectral methods of species identification and plant water and chemistry retrieval near Longyearbyen, Svalbard. Using the field spectroscopy data and Random Forest analysis, we were able to distinguish eight common High Arctic plant tundra species with 74% accuracy. Using partial least squares regression (PLSR), we were able to predict corresponding water, nitrogen, phosphorus and C:N values ( r 2 = 0.61–0.88, RMSEmean = 12%–64%). We developed analogous models using UAV imagery (five bands: Blue, Green, Red, Red Edge and Near-Infrared) and scaled up the results across a 450 m long nutrient gradient located underneath a seabird colony. At the UAV level, we were able to map three plant functional groups (mosses, graminoids and dwarf shrubs) at 72% accuracy and generate maps of plant chemistry. Our maps show a clear marine-derived fertility gradient, mediated by geomorphology. We used the UAV results to explore two methods of upscaling plant water content to the wider landscape using Sentinel-2A imagery. Our results are pertinent for high resolution, low-cost mapping of the Arctic.