Improving Selection Accuracy with Post-Hoc Blocking in Turfgrass Breeding.

The Randomized complete block design (RCBD) is the most widely used experimental design in turfgrass breeding programs due to its simplicity and decent isolation of environmental variation. In early stages of breeding programs, plant breeders screen many genotypes to identify better performing breeding lines. However, as the number of breeding lines increases, the block size increases, which can cause the blocks within an RCBD to loss their original designated function, controlling the environmental variation. Thus, the estimation of parameters and prediction of breeding values is negatively impacted from the uncontrolled environmental variation. To improve the analysis accuracy, our study incorporated spatial factors in the analysis by post-hoc blocking. Post-hoc blocking involves superimposing a different blocking structure onto the original experimental design. Post-hoc blocking was tested using a series of five experiments with 80 genotypes of zoysiagrass (Zoysia spp.) established across Florida. Two post-hoc blocking designs were evaluated (incomplete block design and row-column design), and their model fitting and residual variances were compared with the original RCBD. Results indicated that the row-column design was consistently superior to the original RCBD design in terms of residual variances and data fitting. The incomplete block design performed either slightly better or similar to the original RCBD design. Most importantly, the increase in prediction accuracy was accompanied by a change in rank of breeding values of evaluated genotypes. Turfgrass breeders make selections based on the ranking of genotypes; therefore, these results can have a significant impact in the selection of genotypes for cultivar development.