Phenotyping studies of spring wheat by multispectral image analysis

Abstract

To meet an increasing demand for food production there is a need for faster genetic gains in Norwegian cereal breeding through more precise phenotyping. High-Throughput Phe- notyping (HTP) and genomic selection through multispectral imaging and statistical anal- ysis offer possibilities of yield gains. Several indices have been tested to indicate grain yield, such as the Normalized Differential Vegetation Index (NDVI), MERIS Terrestrial Chlorophyll Index (MTCI) and the Enhanced Vegetative Index (EVI). These indices uti- lize the difference in reflectivity in different spectral bands. The indices can indicate differences between healthy plants, stressed plants or non-plants.

The research revolves around 96 plots of 24 historical wheat cultivars and 602 plots of 301 younger breed lines. Both sites planted at Vollebekk research farm at Ås in Norway, laid out in an alpha-lattice split plot design. The design for the 24 historical cultivars had two levels of nitrogen (N) fertilization, 75 and 150 kg N/ha, applied at sowing. There were two replicates of wheat cultivars of each fertilization level. The set of 24 spring wheat represents the yield progress over the last 40 years in Norway.

Multispectral images were taken in the wavebands green (GRE) (550nm), red (RED) (660nm), red edge (REG) (735nm) and near-infrared (NIR) (790nm) with a Parrot Se- quoia multispectral camera combined with a sunshine sensor. The spectral band images were stitched together using Pix4D software by utilizing GPS coordinates and image fea- tures. To aid stitching of the multispectral and RGB images, tie point objects were laid out in the field. Maps of vegetation indices were computed in Python, by forming linear combinations and ratios of sums and differences in the multispectral reflection. In addi- tion,3D models and Digital Surface Models (DSM) of the area were calculated from RGB images using Pix4D, which were used to indicate plant height. All cameras and sensors were mounted on a light Unmanned Aerial Vehicle (UAV). Images were taken throughout the season of growth at regular intervals.

The time series of the vegetation indices showed peak values during the period of grain filling before declining when plants approached maturity. Values where slightly higher for wheat plots that received a higher dose of fertilization throughout the season of sampling. By combining the digital measurements with manual measurements of grain yield, kernel weight, and plant height, the statistical significance of separating cultivars is explored.