Analysis of Series of Cultivar Trials with Perennial Grasses for Subdivided Target Regions

Field trials with perennial grasses may often be conducted at several locations with different starting years. A key issue in the analysis of such trials is the distinction between effects of calendar years, which are associated with external environmental variation, and harvest years, which represent internal yield formation processes of the perennial crop. Furthermore, analysis of field trials with perennial grasses needs to account for serial correlation of observations on the same plot from consecutive harvest years. Such analyses are conveniently implemented using mixed models. Here, we consider series of trials when the target region is subdivided into several zones. We show how cultivar yield means per zone can be estimated borrowing strength across zones. The proposed mixed models are illustrated using simulated data generated by employing variance component estimates from real experiments. It is shown in simulations that best linear unbiased prediction (BLUP) can provide more precise zone-specific mean estimates than alternative methods.