Prediction of Yield for Specific Locations in German Winter-wheat Trials

The interaction between varieties and environment can make the prediction of crop performance difficult. The farmer is obviously interested in achieving optimum production within the particular growing environment of his farm, thus, this paper investigates if crop performance predictions can be improved by making information from other locations into account. The investigation is based on yield data from winter wheat resulting from official German variety-performance tests. The predictors are based on the single location (control), unweighted means, principal components, and weighted means, produced using regression coefficients as weights. These predictors were tested on an independent data set from another year. For the given yield data in winter wheat, the overall mean proved to be the best, which suggests that prediction for all locations should be the same, and predictors for specific locations cannot, therefore, be recommended. The main reason for this is the relatively small interaction between genotype and location, in comparison to the second-order interaction between genotypes, locations and years.