Lettuce Field UAV Time Series Summer 2019

Flower opening and closure are traits of reproductive importance in all angiosperms because they determine the success of self- and cross-pollination. Different accessions of cultivated and wild lettuce were observed to flower at different times of day. An F6 recombinant inbred line population (RIL) had been derived from accessions of L. serriola accession Armenian999 x L. sativa landrace PI251246 that originated from different environments and differed markedly for daily floral opening time. The 236 RILs, both parental lines, and two controls, L. sativa cv. Salinas and L. serriola accession UC96US23, were grown in summer 2019 at the Department of Plant Sciences field facility in Davis, CA. The experiment had two complete randomized blocks, each consisting of 240 plots to accommodate the 240 genotypes. Within each block, eight individuals of each RIL or parent were planted into one 10 ft x 1 row plot.

Seven ground control points were set up in the field, four near the corners and three along the field’s East-West centerline. GPS coordinates, with an accuracy within a few centimeters, were recorded using a handheld data collector (Trimble Geo 7x Series). These coordinates were used in processing drone images to ensure that images collected at different times and dates aligned perfectly with one another.

A MicaSense RedEdge multi-spectral camera was mounted on a DJI Matrice100 drone. The camera captured images at five wavelengths: blue (475 nm center, 20 nm bandwidth), green (560 nm center, 20 nm bandwidth), red (668 nm center, 10 nm bandwidth), red edge (717 nm center, 10 nm bandwidth), and near-infrared (840 nm center, 40 nm bandwidth). In this study, only the blue, green, and red wavelengths were used for flower identification. The drone was flown over the experimental field at 9 am, 11 am, 1 pm and 3 pm on July 1st, 2019, and 10 am, 12 pm, 2 pm and 4 pm on July 9th, 2019. A DJI GS Pro app was used to plan and execute the flight. The drone flew at 15 meters above ground, and images were taken at a frequency that ensured 85% front- and side-overlaps between each pair of adjacent images. A MicaSense calibration panel was used for automated adjustment of the reflectance spectra. Raw images from the camera were stitched and processed with the Pix4DMapper Pro photogrammetry software to generate orthomosaic maps of surface reflectance at 1 cm spatial resolution. On average, 2,309 raw images were generated per time point, and 2,181 raw images were used to assemble each five-spectrum field map. With the reconstructed maps, the borders of individual plots were manually determined using the software ArcMap.