Utilization of Multiyear Plant Breeding Data to Better Predict Genotype Performance

Despite the availability of multiyear, multicycle, and multiphase data in plant breeding programs for annual crops, selection is often based on single-year, single-cycle, and single-phase data. As genotypes in the same fields are usually grown under the same management practice, data from these fields can and should be analyzed together. In Monsanto's North American maize (Zea mays L.) breeding program, this approach enables a spatial model to be fitted in each field, providing an estimate of spatial trend and a better estimate of residual variance in each field. Multiyear, multicycle analysis showed that the estimates of genotype x year variance (V-GY) and genotype x year x location variance (V-GYL) were still the largest components of the estimated phenotypic variance. Analysis of any single-year subset of the data inflated the estimate of genotypic variance (V-G) by the size of the estimate of V-GY, resulting in potential bias in the estimates of genotype performance. These results demonstrate the advantage of a combined analysis of data across years and cycles to make selection decisions for genotype advancement.