Data from: Drone‐based structure‐from‐motion photogrammetry captures grassland sward height variability

Grasslands deliver a range of ecosystem services, including the provision
of food and biodiversity, and regulation of soil carbon storage and
hydrology. Monitoring schemes are needed to quantify spatial changes in
these multiple functions alongside ecosystem degradation. Sward height is
widely recognised as a key spatial variable in the provision of these
services. Current manual monitoring approaches are labour intensive, and
often fail to capture spatial patterns of important features, including
sward height. Proximal sensing from small aerial drones carrying
lightweight cameras can be transformed into surface height models using
image‐based structure‐from‐motion and Multi‐View Stereo‐based approaches;
this presents a new opportunity for monitoring the spatial structure of
grassland sward height. We combined aerial photographs with field survey
data and an open‐source image‐based modelling‐processing workflow to
generate sward height measurements for a field comprising mainly Lolium
perenne (perennial ryegrass) and Trifolium pratense (red clover). We
compared the derived measurements with in situ data captured on the same
day using traditional agronomic sward height techniques to determine the
quality of the drone‐derived surface model product for sward
characterisation. The SfM and Multi‐View Stereo‐based surface model had a
mean absolute sward height measurement error of between 3.7 and 4.2 cm. To
produce field observations with equivalent quality would require up to 550
sward height measurements for the study site (area: 8,059 m2), which is
not feasible over larger extents required for conservation of key species
or agronomic purposes. Synthesis and applications. We demonstrate how the
collection of precise and detailed information on the spatial structure of
grasslands can be made over management‐relevant extents. Aerial digital
photographs can be transformed into surface models using an image‐based
modelling approach: structure‐from‐motion and Multi‐View Stereo
techniques. Image‐based measurements of sward heights were compared with
manual sward height data captured on the same day. This novel source of
vegetation spatial information could improve sward management for
conservation and agronomy applications. The approach supports frequent
surveys, at user‐controlled revisit times, and delivers data for spatial
monitoring of key grassland functions and services.