Enhancing Soybean Breeding Through Optimized Testing Networks and Sparse Designs

Authors

• Arthur Martins Almeida Bernardeli, University of Nebraska-LincolnFollow

First Advisor

George Graef

Degree Name

Doctor of Philosophy (Ph.D.)

Committee Members

David Hyten, Jinliang Yang, Kaio Olimpio G. Dias, Réka Howard

Department

Agronomy

Date of this Version

4-22-2026

Document Type

Dissertation

Citation

A dissertation presented to the faculty of the Graduate College at the University of Nebraska in partial fulfillment of requirements for the degree Doctor of Philosophy

Major: Agronomy

Under the supervision of Professor George Graef

Lincoln, Nebraska, May 2026

Comments

Copyright 2026, Arthur Martins Almeida Bernardeli. Used by permission

Abstract

This dissertation addresses a central challenge in soybean breeding: how to make better selection decisions across diverse environments while working within the practical limits of field testing. Multi-environment testing is essential for evaluating breeding lines, but large testing networks are costly to maintain and difficult to interpret because genotype performance often changes across locations and years. At the same time, breeders must decide how to allocate limited field resources while still capturing the environmental and genetic variation needed for effective selection. Using yield, pedigree, genomic, and environmental data from the University of Nebraska–Lincoln Soybean Breeding Program, this dissertation explores practical strategies to improve both the structure of testing networks and the efficiency of selection pipelines.

In Chapter 1, I evaluate how environmental covariates and genotype-by-environment signals can be combined to define target populations of environments and identify representative hub locations within the breeding program’s testing network. I found that integrating weather-related variables with genotype-by-location interaction patterns allowed clearer grouping of locations and more informative hub identification. These results showed that a reduced set of representative locations could preserve much of the interaction structure present across the broader network.

In Chapter 2, I build on that framework by evaluating sparse testing and genome-wide selection strategies designed to reduce phenotyping demands while maintaining useful predictive ability. I found that prediction performance depended strongly on the level of connectivity among breeding populations and environments. Predictions benefited from the inclusion of environmental kernels and optimized training populations, and in some cases, pedigree-based approaches also served as a proxy to enhance selection in sparse designs. These results demonstrated that sparse testing can support breeding decisions, but its success depends on how testing resources are allocated and how prediction models are matched to the population and environmental connectivity.

Overall, this dissertation provides practical tools for redesigning soybean multi-environment trials, improving the use of environmental information, and integrating sparse testing into breeding pipelines. These findings contribute to more efficient cultivar development and offer a path for balancing operational constraints with stronger data-driven selection decisions.

Recommended Citation

Martins Almeida Bernardeli, Arthur, "Enhancing Soybean Breeding Through Optimized Testing Networks and Sparse Designs" (2026). Dissertations and Doctoral Documents, University of Nebraska-Lincoln, 2023–. 455.
https://digitalcommons.unl.edu/dissunl/455