High-resolution UAS multispectral imaging for cultivar selection in grain sorghum breeding trials

Abstract

As the global human population continues to increase, there is an increased responsibility on plant breeders to develop varieties with improved productivity. Current yearly yield improvement rates are not completely estimated to meet future demands, so new technologies that allow for rapid cultivar screening and selection in large-scale breeding trials are warranted. High-resolution imagery data collected with unmanned aerial systems (UAS) shows potential to greatly assist breeders. A crop that could greatly benefit from this technology is grain sorghum (Sorghum bicolor (L). Moench). Grain sorghum is an important food, fuel, forage, and livestock feed source for many people across the world. In addition, it is well-suited to be grown in climates with limited precipitation, providing a means of food security for nations in such agro climatic regions. As global climate becomes increasingly warmer, more grain sorghum is predicted to be grown in areas that have traditionally grown with more water-dependent crops. To maximize productivity, sorghum not only needs to be selected for higher yielding cultivars, but also cultivars that can withstand abiotic stresses such as herbicide application and drought. Focusing on these two stresses, the objectives of this project were to: i) evaluate the effectiveness of UAS imagery in quantifying, detecting, and differentiating sorghum spectral response to herbicide, mesotrione, and ii) develop and evaluate a methodology to collect, process, extract, and compare UAS data to select for traits related to drought tolerance in grain sorghum. For the first objective, a field experiment was sown in the 2019 growing season (Ashland Bottoms, Manhattan, KS) consisting of a mesotrione tolerant and susceptible genotypes, and a commercial grain sorghum hybrid for comparison. Plots were sprayed with 0, 105, 420, and 840 g ae (acid equivalent) ha-1 of mesotrione, and weekly flights were flown over the experiment up to 35 days after treatment (DAT). Ground-measured herbicide damage ratings were taken, and were compared to vegetative indices (VIs) derived from the imagery. Results showed highly-significant relationships between VIs and ground ratings. For the second objective, an experiment with 20 commercial hybrids was planted in 2019 (Manhattan, KS). Flights were flown at the flowering (F), soft dough (SD), hard dough (HD), and physiological maturity (PM) growth stages. Ground-samples that were collected included whole plant biomass, leaf biomass, stem biomass, senescence scores, leaf area index, and final grain yield. Results showed that the near infrared (NIR) spectral band was the most significant to plant traits related to biomass, the green normalized difference vegetation index (GNDVI) was highly significant to yield, and the visible atmospheric resistant index (VARI) was the most related to both senescence at PM and senescence rates. A hierarchical clustering analysis showed that through all stages, significant differences among groups could be detected. These results suggest that multi-spectral imagery data collected via UAS could be very useful for sorghum breeders to differentiate between grain sorghum hybrids in large-scale breeding trials, particularly for breeders looking to increase herbicide tolerance in grain sorghum and to develop more drought-resistant cultivars.