Evaluating testing strategies for plant breeding field trials: redesigning a CIMMYT international wheat nursery
- Additional Document Info
- View All
Genotypes produced from plant breeding programs for annual crops are evaluated in multiyear, multilocation field trials where a large number of genotypes grown in a few locations in the first year are progressively reduced by selection and tested in a greater number of locations in successive years. Provided fixed resources are specified, the efficiencies of testing strategies differing in the number of locations and replications used and in the number of genotypes tested in successive years can be evaluated by three statistics (genetic repeatability, acceptance probability, and potential gain) calculated from five variance components applicable to the field trials. The five variance components required are deployed genetic variance, genotype-by-year, genotype-by-location, and genotype-by-year-by-location interaction variances and the residual (error) variance. Strictly, known variance components are required, but the statistics and their standard errors can be calculated using robust estimates of the variance components obtained from a combined analysis over years of a large set of historical data from field trials that are connected by shared genotypes across years, as genotype-by-year and genotype-by-year-by-location variance components are estimable. Single-year analyses are unsuitable, as no estimate of the interactions with year are available, and such analyses produce biased estimates of genetic variance and genotype-by-location interaction variance. These procedures are illustrated using an example from the CIMMYT international wheat breeding program to evaluate 2-yr testing strategies for CIMMYT's Elite Spring Wheat Yield Trials that has been distributed and evaluated internationally since 1979.
has subject area