A review of the journey of field crop phenotyping: From trait stamp collections and fancy robots to phenomics-informed crop performance predictions

Abstract

Crop phenotyping encompasses methodologies for measuring plant growth, architecture, and composition with high precision across scales, from organs to canopies. Field-based phenotyping is pivotal in bridging genomic data with crop performance, offering a promising pathway for predictive modeling in diverse environments. This review traces the evolution of phenotyping from high-throughput sensor data for trait extraction to advanced modeling approaches that integrate multi-temporal data, latent space representations, and learned crop models. This evolution is exemplified mostly by morphology- and growth-related examples from the core expertise of the authors. High-throughput trait extraction, facilitated by advanced imaging and sensor technologies, has enabled rapid and accurate characterization of complex traits essential for crop improvement. Carrier platforms, such as drones, rovers, and gantries, have played a critical role in capturing high-resolution data across large fields, enhancing the spatial and temporal resolution of phenotypic data. Publicly available datasets have further accelerated research by providing standardized, high-quality data for benchmarking and model development beyond the realm of crop growth as for example in crop photosynthesis. These advancements are transforming phenotyping into a predictive science capable of informing breeding and management decisions. As phenotyping methodologies continue to evolve, the integration of machine learning and data-driven approaches offers new opportunities for enhancing prediction accuracy and understanding genotype-environment interactions. While challenges such as data heterogeneity, scalability, and cost remain, we highlight key gaps and propose solutions, underscoring phenotyping's critical role in future agricultural innovation.