Aboveground biomass corresponds strongly with drone-derived canopy height but weakly with greenness (NDVI) in a shrub tundra landscape

Abstract

Arctic landscapes are changing rapidly in response to warming, but future predictions are hindered by difficulties in scaling ecological relationships from plots to biomes. Unmanned aerial systems (UAS, hereafter 'drones') are increasingly used to observe Arctic ecosystems over broader extents than can be measured using ground-based approaches and facilitate the interpretation of coarse-grained remotely-sensed datasets. However, more information is needed about how drone-acquired remote sensing observations correspond with ecosystem attributes such as aboveground biomass. Working across a willow shrub-dominated alluvial fan at a focal study site in the Canadian Arctic, we conducted peak season drone surveys with a RGB camera and multispectral multi camera array to derive photogrammetric reconstructions of canopy and normalised difference vegetation index (NDVI) maps along with in situ point intercept measurements and biomass harvests from 36, 0.25 m2 plots. We found high correspondence between canopy height measured using in situ point intercept compared to drone-photogrammetry (concordance correlation coefficient = 0.808), although the photogrammetry heights were positively biased by 0.14 m relative to point intercept heights. Canopy height was strongly and linearly related to aboveground biomass, with similar coefficients of determination for point framing (R2 = 0.92) and drone-based methods (R2 = 0.90). NDVI was positively related to aboveground biomass, phytomass and leaf biomass. However, NDVI only explained a small proportion of the variance in biomass (R2 between 0.14 and 0.23 for logged total biomass) and we found moss cover influenced the NDVI-phytomass relationship. Biomass is challenging to infer from drone-derived NDVI, particularly in ecosystems where bryophytes cover a large proportion of the land surface. Our findings suggest caution with broadly attributing change in fine-grained NDVI to biomass differences across biologically and topographically complex tundra landscapes. By comparing structural, spectral and on-the-ground ecological measurements, we can improve understanding of tundra vegetation change as inferred from remote sensing.