| Abstract | In the early stages of selection, a sugarcane breeder tests a large number of genotypes. As the amount of planting material for each genotype for testing is typically limited, selections are usually made on small, unreplicated, single-row plots. Unfortunately, such designs are prone to errors arising from spatial variability and interplot competition, which, unless accounted for, can seriously bias variety estimates and reduce genetic progress.In this project, an approach to the simultaneous modelling of spatial variability and interplot competition is developed. This approach combines nearest-neighbour techniques to model spatial variability, together with the genotypic and phenotypic interference models to estimate interplot competition. The joint modelling and standard approaches are compared using 23 sugarcane data sets for cane yield. Agreement between the two approaches varied from approximately 38% to 90%. Hence, for some trials there would be large differences in the selections to be advanced to final assessment trials. Additionally, for two trials, the joint modelling approach was applied to cane yield and CCS data. The number of selections in common for sugar yield for the two approaches was compared. Approximately 43% and 75% of the clones were in common, indicating that appropriate modelling of interplot competition and spatial variability can have a very large effect on the varieties to be advanced to final assessment trials.This project has resulted in an improved selection system, and this is likely to result in increased genetic gain through the advancement of superior varieties to later stages.The project has formed the basis of a PhD thesis submitted to the University of Queensland. |
