Data-driven decentralized breeding increases prediction accuracy in a challenging crop production environment

Crop breeding needs to adapt to diverse smallholder farming systems to ensure food security. This study introduces a data-driven decentralized approach (3D-breeding) combining genomics, farmers' knowledge, and environmental analysis. A durum wheat trial in 1,165 Ethiopian farmer fields showed that 3D-breeding doubled prediction accuracy compared to a genomic prediction benchmark from conventional breeding. 3D-breeding identified locally adapted genotypes with superior performance across seasons, suggesting its potential to enhance local adaptation in challenging environments.

Kauê de Sousa, Jacob van Etten, Jesse Poland, Carlo Fadda, Jean-Luc Jannink, Yosef Gebrehawaryat Kidane, Basazen Fantahun Lakew, Dejene Kassahun Mengistu, Mario Enrico Pè, Svein Øivind Solberg, Matteo Dell'Acqua