Enhancing trait measurement in forage sorghum using unoccupied aerial systems

Abstract Forage sorghum ( Sorghum bicolor L. Moench) can be classified into several different types, ranging from sudangrass to silage sorghum. The exact type of forage sorghum is defined by basic plant characteristics (maturity, height, and tillering), the end use, and type of biomass produced. Many of these traits must be determined in hybrid trials where the traditional collection methodology of these traits can be time consuming and prone to bias. Unoccupied aerial systems (UAS) are a technology that has been proven to collect numerous traits with high accuracy across various crops, while requiring fewer resources. The objective of this study was to determine if plant height, days to mid‐anthesis, and biomass yield could be estimated in forage sorghum breeding trials via UAS. Each case study used a subset of data derived from 12 different hybrid forage sorghum trials. Plant height was estimated via one flight, resulting in an average correlation of 0.87 and root mean square error (RMSE) of 18 cm. Days to mid‐anthesis was estimated through a random forest model using 20 vegetative indices and percent of pixels classified as panicles; producing an average correlation of 0.87, and an RMSE of 2.6 days. Lastly, biomass yield was estimated using 20 vegetative indices and the percent of pixels classified as canopy within a gradient boosting model; producing an average correlation of 0.64, and RMSE of 5.4 t/ha. Overall, the results describe methodology to estimate important forage sorghum traits via a UAS and improve the repeatability of traits in at least 50% of the trials assessed.