In-field whole plant maize architecture characterized by Subcanopy Rovers and Latent Space Phenotyping

Authors: Gage, Joseph; RICHARDS, ELLIOTT - Cornell University; Lepak, Nicholas; KACZMAR, NICHOLAS - Cornell University; SOMAN, CHINMAY - Earthsense, Inc; CHOWDHARY, GIRISH - University Of Illinois; GORE, MICHAEL - Cornell University; Buckler, Edward - Ed

Submitted to: The Plant Phenome Journal
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 11/5/2019
Publication Date: 11/1/2019
Citation: Gage, J.L., Richards, E., Lepak, N.K., Kaczmar, N., Soman, C., Chowdhary, G., Gore, M.A., Buckler IV, E.S. 2019. In-field whole plant maize architecture characterized by Subcanopy Rovers and Latent Space Phenotyping. The Plant Phenome Journal. 2(1):1-11. https://doi.org/10.1101/763342.
DOI: https://doi.org/10.1101/763342

Interpretive Summary: Plant breeding and biological studies rely on accurate measurements of various plant traits. These traits are chosen for their relevance to the study at hand, but also for the feasibility of measuring them. Measuring these traits on large populations of plants can be time consuming and labor intensive, which limits the number of plants that can be measured and constrains measurements to a limited number of time points during development. Using data collected by semi-autonomous rovers, we applied machine learning algorithms to detect traits that can be recorded and processed with less effort and resources than traits that are traditionally measured by hand. These results will be useful by allowing automated measurement of crop traits. The type of traits that are described in this study are not possible for humans to measure and are a more complete representation of plant size and shape than the traits normally measured by hand (e.g., plant height or leaf count). These traits can be recorded approximately twice as fast as a single, manually measured trait, and represent more than ten-fold greater useful information than a single manually measured trait. Because of the efficiency with which plants can be measured in the field, this technology will allow assessment of plant growth on a daily or weekly basis throughout their life, rather than at a single time point as is traditionally measured. The increase in efficiency, and potential for more frequent measurement, will enable scientific studies of the interplay between plant variety, weather, and plant development. These methods can also be used to make plant breeding more efficient or to facilitate assessments of crop health by farmers and crop scouts.

Technical Abstract: Collecting useful, interpretable, and biologically relevant phenotypes in a resource-efficient manner is a bottleneck to plant breeding, genetic mapping, and genomic prediction. Autonomous and affordable sub-canopy rovers are an efficient and scalable way to generate sensor-based datasets of in-field crop plants. Rovers equipped with light detection and ranging (LiDar) can produce three-dimensional reconstructions of entire hybrid maize fields. In this study, we collected 2,103 LiDar scans of hybrid maize field plots and extracted phenotypic data from them by Latent Space Phenotyping (LSP). We performed LSP by two methods, principal component analysis (PCA) and a convolutional autoencoder, to extract meaningful, quantitative Latent Space Phenotypes (LSPs) describing whole-plant architecture and biomass distribution. The LSPs had heritabilities of up to 0.44, similar to some manually measured traits, indicating they can be selected on or genetically mapped. Manually measured traits can be successfully predicted by using LSPs as explanatory variables in partial least squares regression, indicating the LSPs contain biologically relevant information about plant architecture. These techniques can be used to assess crop architecture at a reduced cost and in an automated fashion for breeding, research, or extension purposes, as well as to create or inform crop growth models.