Ultra-fine grain landscape-scale quantification of dryland vegetation structure with drone-acquired structure-from-motion photogrammetry

This is the final version of the article. Available from Elsevier via the DOI in this record.Covering 40% of the terrestrial surface, dryland ecosystems characteristically have distinct vegetation structures that are strongly linked to their function. Existing survey approaches cannot provide sufficiently fine-resolution data at landscape-level extents to quantify this structure appropriately. Using a small, unpiloted aerial system (UAS) to acquire aerial photographs and processing theses using structure-from-motion (SfM) photogrammetry, three-dimensional models were produced describing the vegetation structure of semi-arid ecosystems at seven sites across a grass–to shrub transition zone. This approach yielded ultra-fine (< 1 cm2) spatial resolution canopy height models over landscape-levels (10 ha), which resolved individual grass tussocks just a few cm3 in volume. Canopy height cumulative distributions for each site illustrated ecologically-significant differences in ecosystem structure. Strong coefficients of determination (r2 from 0.64 to 0.95) supported prediction of above-ground biomass from canopy volume. Canopy volumes, above-ground biomass and carbon stocks were shown to be sensitive to spatial changes in the structure of vegetation communities. The grain of data produced and sensitivity of this approach is invaluable to capture even subtle differences in the structure (and therefore function) of these heterogeneous ecosystems subject to rapid environmental change. The results demonstrate how products from inexpensive UAS coupled with SfM photogrammetry can produce ultra-fine grain biophysical data products, which have the potential to revolutionise scientific understanding of ecology in ecosystems with either spatially or temporally discontinuous canopy cover.This research was supported by a NERC PhD studentship (NE/K500902/1) and Sevilleta LTER program research fellowship (NSF grant DEB-1232294) both awarded to AMC; neither funder had any further involvement in this experiment and the authors declare no conflict of interest. We thank Scott Collins, the Sevilleta LETR director and US Fish and Wildlife for their support during this research and for granting access to the field site. The 3D Robotics Y6 was supplied by the University of Exeter Environment and Sustainability Institute's (ESI) Environmental Monitoring DroneLab (EMDL). The authors wish to express their thanks to Leon DeBell and Agisoft's Alexey Pasumansky for the excellent technical support, to Susan Beck and Phil Cunliffe for facilitating access to archival material, and to Isla Myers-Smith and three anonymous reviewers whose comments allowed us to improve an earlier draft of this article. For access to the data presented herein please contact the first author