Grüner, Esther; Wachendorf, Michael; Astor, Thomas (2020): UAV-based multispectral data in legume-grass mixtures [dataset]. PANGAEA, https://doi.org/10.1594/PANGAEA.914667

Multispectral data from two legume-grass mixtures (clover- and lucerne-grass) were collected in the year 2018 for aboveground biomass and nitrogen fixation (NFix) estimation. In addition to the mixtures, pure stands of legumes and of grasses of the two mixtures were sown in order to represent variable conditions in practical farming (0-100% legumes). All six treatments were cultivated in four replicates and harvested three times within the year (plot size: 1.5 x 12 m). Destructive biomass samples for fresh (FM) and dry matter (DM) and NFix determination were taken three times at harvest. To cover the entire vegetation season, sub-sampling for DM and FM was done five times between the harvests. Flight missions were carried out one day before each of the eight sampling dates. A multispectral sensor (Parrot Sequoia, MicaSense Inc, Seattle, USA) with four spectral bands (green, red, red edge, near infrared) was mounted on a low-cost unmanned aerial vehicle (UAV; DJI Phantom 3, Advanced, Shenzhen, China). Eight black and white ground control points (GCPs) were distributed in the pathways. Coordinates of the plot corners and GCPs were measured by a Leica real time kinematic global navigation satellite system (Leica RTK GNSS). Orthomosaics were created by the overlapping images with a photogrammetric processing software (Agisoft PhotoScan Professional, Agisoft LLC, St. Petersburg, Russia). The orthomosaics were georeferenced using the coordinates of the GCPs. The mean reflectance value of the four bands was extracted by zonal statistics in QGIS (Quantum Geografic Infromation System) using the four plot corners of each plot as boundaries. Furthermore, eight texture features of every band were calculated, provided by the processing tool HaralickTextureExtraction of the Orfeo Toolbox library (OTB) in QGIS.