Article ID: 26053
Phenotyping is the rate-limiting step in crop breeding under climate change; the bottleneck lies upstream in data acquisition, reference definition, and preservation rather than in algorithmic accuracy. Using rice heat-tolerance breeding as a case study, this review briefly surveys the maturity of the technical foundation, from grain-quality imaging to unmanned aerial vehicle (UAV)-borne multimodal sensing, and analyzes why mature technologies fail to reach breeding pipelines through four questions: what to measure, how to measure, how to evaluate, and how to preserve. After critically examining spatial aggregation loss, deficiencies in calibration standardization, the asymmetric pace of algorithmic versus data-acquisition improvement, and the Gold Standard Paradox, this review presents three recommendations: a redefinition of ground truth that places well-calibrated sensor measurements at the center of the reference value; a multiscale phenotyping framework that interprets grain quality across spatial scales; and a shared phenomic data archive accompanied by provenance and calibration metadata. Further, this review considers how these recommendations may contribute to genetic gain by integrating them with genomic selection and discusses the practical potential of large language models for data infrastructure development. Whether breeding can keep pace with a warming climate depends not on new sensors but on the community’s willingness to invest in data infrastructure.