Projecting results of zoned multi-environment trials to new locations using environmental covariates with random coefficient models: accuracy and precision | JoVE Visualize

Projecting results of zoned multi-environment trials to new locations using environmental covariates with random coefficient models: accuracy and precision

Harimurti Buntaran¹, Johannes Forkman², Hans-Peter Piepho³

¹Biostatistics Unit, Institute of Crop Science, University of Hohenheim, Fruwirthstraße 23, 70599, Stuttgart, Germany.

TAG. Theoretical and Applied Genetics. Theoretische Und Angewandte Genetik|April 8, 2021

View abstract on PubMed

Summary

Environmental covariates improve genotype performance predictions in new locations. Random coefficient models enhance prediction accuracy and precision for agricultural breeding trials.

Area of Science:

Agricultural Science
Biometrics
Plant Breeding

Background:

Multi-environment trials (MET) are crucial for assessing genotype performance across diverse locations.
Accurate prediction of genotype performance in new, grower-specific locations is essential for agricultural applications.
Existing methods like linear mixed models offer predictions but may lack precision in novel environments.

Purpose of the Study:

To demonstrate the benefit of incorporating environmental covariates into random coefficient (RC) models for predicting genotype performance.
To improve the accuracy and precision of genotype performance predictions in new locations for Swedish winter wheat trials.
To assess the impact of environmental information on the reliability of predictions in agricultural breeding.

Main Methods:

Utilized random coefficient (RC) models incorporating environmental covariates.
Applied models to Swedish multi-environment trials (MET) data for winter wheat.
Accounted for correlations between zones and included intercepts/slopes for covariate regressions.

Main Results:

The RC model with environmental covariates significantly improved the precision of genotype performance predictions.
Prediction accuracy of the RC model was competitive with models lacking covariates.
Reduced standard errors for individual genotype predictions (30-38%) and genotype differences (12-40%) in new locations.

Conclusions:

Environmental covariates integrated into RC models enhance genotype performance predictions in new locations.
RC models offer a valuable tool for increasing the accuracy and precision of agricultural breeding decisions.
The study highlights the importance of auxiliary environmental data for robust genotype evaluation.