Improving Prediction of Genotypic Values in Historical Crop Trial Data via Stepwise Adjustment Method

Improving prediction of genotypic values from long-term historical crop trial data will enhance the utilization for both genetic study and crop improvement. However, because many long-term historical crop trial data are highly unbalanced due to the frequent changes in test entries and locations, it is statistically challenging to analyze the long-term historical data simultaneously without proper adjustment. In this study, we proposed a stepwise method that can be used to adjust the differences caused by environmental conditions among years. First, this method was evaluated by Monte Carlo simulation, which showed that this stepwise adjustment method can consistently improve the prediction impacted by environmental conditions among years. Second, the stepwise adjustment method was applied to a 16-year soybean trial data set in South Dakota and showed that model fitness for genetic gain over these 16 years was improved compared to the model fitness using the non-adjusted data (0.85 vs 0.48). The annual genetic gain estimated from non-adjusted data was 1.35 bushel/ac while the adjusted annual genetic gain was 0.72 bushel/ac, which was more in line with annual state soybean production from 1987-2011.