In‐Field Whole‐Plant Maize Architecture Characterized by Subcanopy Rovers and Latent Space Phenotyping

Abstract

Core Ideas Subcanopy rovers enabled 3D characterization of thousands of hybrid maize plots. Machine learning produces heritable latent traits that describe plant architecture. Rover‐based phenotyping is far more efficient than manual phenotyping. Latent phenotypes from rovers are ready for application to plant biology and breeding. 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 subcanopy rovers are an efficient and scalable way to generate sensor‐based datasets of in‐field crop plants. Rovers equipped with lidar can produce three‐dimensional reconstructions of entire hybrid maize ( Zea mays L.) fields. In this study, we collected 2103 lidar scans of hybrid maize field plots and extracted phenotypic data from them by latent space phenotyping. We performed latent space phenotyping by two methods, principal component analysis 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 that 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 that 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.

Plant phenotyping relevance

LiDAR搭載ローバーと潜在空間解析により、トウモロコシの全草型・バイオマス分布を定量化する手法が研究の中心である。

Code and data availability

The paper's raw lidar point clouds are deposited at a public DOI, and the authors' analysis code plus phenotypic data (including the trained autoencoder HDF5 model) are available in a public Bitbucket repository. Both are paper-specific, public, and directly actionable.