Choosing trial sites to maximize selection response for grain yield in spring wheat

Abstract

Nine sites used in a wheat breeding programme in South Australia were investigated. Of the six major sites, four occupy farmers' fields and two are located on experiment stations. The data analysed comprised grain yields of 31 genotypes grown in 31 trials over a 5-year period. Relative to that grown at non-station sites, wheat grown at the experiment-station sites generally had higher mean yields, higher coefficients of variation, and lower heritabilities. Data from a set of independently conducted trials were used to estimate the true yields of the 31 genotypes. A heritability of the correlated response to selection was then calculated for each breeding trial, and was generally found to be low for the experiment-station sites. Further investigation showed that the non-station sites are suitable for the selection of genotypes intended for low-yielding environments, and the station sites are better suited for the selection of genotypes intended for high-yielding environments. The optimum selection scheme was found to consist of several unreplicated trials per year, and at least four trials per year are necessary to avoid the possibility of a negative heritability of the correlated response to selection.