|TANG, ZHOU - Washington State University|
|PARAJULI, ATIT - Washington State University|
|CHEN, CHUNPENG - Washington State University|
|HU, YANG - Washington State University|
|REVOLINSKI, SAMUEL - Washington State University|
|MEDINA, CESAR CULMA - Washington State University|
|LIN, SEN - Washington State University|
|ZHANG, ZHIWU - Washington State University|
Submitted to: Scientific Reports
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 1/25/2021
Publication Date: 2/8/2021
Citation: Tang, Z., Parajuli, A., Chen, C., Hu, Y., Revolinski, S., Medina, C., Lin, S., Zhang, Z., Yu, L. 2021. Validation of UAV-based alfalfa biomass predictability using photogrammetry with fully automatic plot segmentation. Scientific Reports. 11(1):3336. https://doi.org/10.1038/s41598-021-82797-x.
Interpretive Summary: Alfalfa (Medicago sativa L.) is an important legume forage grown worldwide. Alfalfa is called “Queen of Forage” due to its high production and nutrition values. One of the major bottlenecks for phenotypic selection is the phenotyping burden for complex traits such as biomass. In this work we used a drone to take images in field plots one day before harvesting, followed by analyzing images using a series of software to quantify biomass for each plot. We then measured the havested yield of fresh weight and analyzed the correlation with the estimated biomass by drone. We obtained 60% (R sequare or coefficient of determination) can be explained by drone. These findings suggest that high throughput phenotyping of drone image could be used to estimate alfalfa biomass yield in large field effectively.
Technical Abstract: Alfalfa is the most widely cultivated forage legume, with approximately 30 million hectares planted worldwide. Genetic improvements in alfalfa have been highly successful in developing cultivars specialized in traits related to winter hardiness and disease resistance. However, genetic improvements have been limited for other economically important traits such as biomass, which are barely under either natural or artificial selections. One of the major bottlenecks for artificial selection is the phenotyping burden for biomass. In this study, we planted 200 alfalfa lines in the field with three replicates. Each replicate was accompanied by 60 plots as covariates using a common variety across the three replicates. Three cuttings were harvested and measured biomass in May, July and September in 2019, the entire field was imaged one day prior to harvesting with DJI Phantom 4 drone carrying an additional Sentera multispectral camera. Alfalfa plots images were extracted by GRID (GReenfield Image Decoder) software package to quantify the vegetarian area. The results found that nearly 60% (R square) of biomass variation was explained by four features from drone images, including the number of pixels, height, and normalized Green-Red difference index (NGRDI). These findings suggest that high throughput phenotyping could be used to phenotype alfalfa biomass toward genetic improvement effectively.