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

Area of Science:

• Agricultural Science
• Genetics
• Food Security

Background:

• Smallholder agricultural systems are crucial for global food security, especially in challenging environments.
• Conventional breeding methods often overlook the diversity within these systems.
• Genomics, farmer knowledge, and environmental data integration is needed for effective crop improvement.

Purpose of the Study:

• To test a data-driven, decentralized approach (3D-breeding) for crop improvement.
• To compare 3D-breeding with conventional genomic prediction in durum wheat.
• To enhance local adaptation and productive performance of crops in diverse farming systems.

Main Methods:

• Implemented a decentralized trial of durum wheat across 1,165 farmer-managed fields in the Ethiopian highlands.
• Utilized an incomplete block design for trial distribution.
• Compared 3D-breeding performance against a benchmark of genomic prediction in conventional breeding.

Main Results:

• 3D-breeding demonstrated a twofold increase in prediction accuracy compared to the benchmark.
• Identified durum wheat genotypes with superior local adaptation and performance across seasons.
• Successfully leveraged farmer field diversity for crop improvement.

Conclusions:

• 3D-breeding offers a powerful decentralized strategy to complement conventional plant breeding.
• This approach enhances local adaptation in crops for challenging production environments.
• Integrating genomics, farmer knowledge, and environmental data is key to future food security.