Comparing the reliability of maize variety data from on-farm trials and experimental stations

Abstract

While current variety recommendations are based on replicated small-plot trials in a specific region for which the trial is thought to be representative, genotype × envi­ronment interactions, genotype × management interactions, and increasing weather fluctuations make it increasingly difficult to predict which variety will be best in a given environment. An additional approach is therefore to decentralize variety trials and place them on working farms. However, although on-farm trials offer potentially more relevance for direct variety selection on site, they are also likely to be subject to more noise and trial entries can often not be fully replicated. To evaluate the relative merit of on-farm trials vs. fully replicated trials conducted at experimental stations, we tested 6 maize varieties at four farms and at two stations in a region dominated by sandy soils. The variance of variety rankings over the years within each site was used as proxy to evaluate the consistency of variety information gained at each location. For dry matter yield, on-farm trials showed both the highest and the lowest consistency of variety ranking, with the consistency being intermediate at the experimental stations. For some quality parameters, namely non-fibre carbohydrate and starch content, the majority of on-farm trials showed more consistent variety ranking over the years than the more consistent of the two replicated trials. This suggests that in terms of year-on-year reliability of maize yield and quality, on-farm trials may have the potential to complement replicated variety trials. For both types of trials, however, there is also scope for decreasing technical sources of variation.

DOI: 10.5073/JfK.2014.11.03, https://doi.org/10.5073/JfK.2014.11.03