Predicting Superior Yielding Spring Wheat and Oat Cultivars using Past Yield Data

Agronomists use data from yield trials when recommending the greater yielding (top‐yielding) cultivars of spring wheat ( Triticum aestivum L.) and oat ( Avena saliva L.) to growers. At most land grant universities, cultivar recommendations are based on 3‐yr yield means. It is not clear, however, that this procedure is the best predictor of top‐yielding cultivars in the subsequent year. Using South Dakota cultivar performance testing data, eight different yield means were derived from various combinations of previous years' yield data. These eight predictors were assessed for their ability to identify the five top‐yielding cultivars in the subsequent year. Mean yield data were used from 1972 to 1987 for six and five locations for spring wheat and oat, respectively. The eight predictors were compared for differences in mean number of top‐yielding cultivars identified with Friedman rank sum two‐way analysis of variance. Also, predictors were compared to the expected mean number of top‐yielding cultivars identified if cultivar selection was random. The effectiveness of prediction from yield trials with varying coefficient of variation or least significant difference values was assessed. When averaged across all locations, selecting cultivars by mean yields using only the previous year, or using all yield data available for each cultivar, identified the largest number of top‐yielding cultivars in the subsequent year. These predictors were also superior to random selection in one of 11 locations. For oat, prediction was better from yield trials with CV values below 10.7 or LSD values below 8.7. For this data set, cultivar recommendations based on fewer than 3‐yr yield data were equally effective at predicting top‐yielding cultivars as those using 3 yr of yield data.