Efficiency of One-Replicate Yield Testing

Abstract

Conducting yield trials at multiple locations is an essential component of breeding programs that develop crop cultivars with stable performance across a broad range of growing conditions. Recognition that grower acceptance relates closely to stable crop performance has led to the implementation of testing programs that sample a large number of environments. Historically, resource allocation in commercial testing programs has changed from several replicates per location to more locations per year. One-replicate testing appears to offer potential for accurately predicting cultivar performance across a broad range of growing conditions. In order to be feasible, however, use of one-replicate testing must result in greater precision at equivalent or lower costs compared to programs that utilize replication. This study presents a methodology that utilizes cost constraints in conjunction with variance component estimates to determine the efficiency of one-replicate testing. Estimates of genotype × location and error variance components from grain sorghum [Sorghum bicolor (L.) Moench] yield trials, and four sets of multiples of these estimates were used in conjunction with cost estimates for four hypothetical situations that differed in the relative costs of locations and replications. Use of one replicate per location resulted in the lowest standard error of a cultivar mean when both the cost of an additional location was not excessively high relative to that of an additional replicate and the estimate of σ2e (error variance) was low relative to that of σ2gl (genotype × location variance). For other situations, use of multiple replications should result in the most efficient utilization of the research budget. The methodology presented allows the researcher to determine the efficiency of one-replicate testing for a program with any set of location and replication costs.