Roth, L.; Herrera, J. M.; Levy Häner, L.; Pellet, D.; Fossati, D.; Boss, M.; Chen, X.; Nousi, P.; Volpi, M.

To address challenges in food security, a better understanding of crop performance under varying environmental conditions is required. Plant Time Bender (PTB) is a deep learning model that integrates high-throughput field phenotyping data with genomic and environmental information to predict wheat yield. PTB leverages image time series, genetic markers, and environmental covariates to learn genotype-specific responses to temperature and vapor pressure deficit. Compared to mere genomic prediction models, PTB demonstrates superior performance when predicting yield in unseen environments across 48 year-locations in Europe. The model captures non-linear growth responses varying with phenological stages and identifies distinct patterns associated with yield performance and stability. Specifically, varieties with higher yield stability exhibit reduced sensitivity to vapor pressure deficit around 1.5 kPa and distinctive temperature responses during emergence and senescence. PTB enables retrospective yield predictions across 20 years, providing a foundation for location-specific variety recommendations and targeted breeding strategies.