Prediction of F3 row performance from F2 individual plant data in oats

Parameters estimated from a Gardner-Eberhart analysis of the F2 generation of a six-parent diallel in oats (Avena sativa L.) were used to compare methods for predicting the performance of F3 row plots. The prediction methods were: (1) individual F2 plant performance (F2I), (2) parent average plus F2 plot deviations (PF2), (3) parent average plus weighted F2 plot deviations (PF2P), (4) best linear unbiased prediction (BLUP) of parent average plus F2 plot deviations (BPF2), and (5) BLUP plus weighted F2 deviations (BF2). The F2 single-plant traits used for prediction were biological yield to predict F3 biological yield, whole plant and primary tiller grain yield for prediction of F3 grain yield, and whole plant and primary tiller harvest index (HI) to predict F3 HI. Prediction methods were evaluated by correlations between predicted and observed F3 performance. Prediction methods and traits for which correlations were greater than for F2I included: BF2 for biological yield, PF2, PF2P and BF2 for whole plant grain yield, PF2, BPF2, and BF2 for primary tiller grain yield. None had a correlation significantly greater than F2I for either measure of HI, where heritability was large. PF2 is the recommended method for traits with low heritability because of its simplicity and because it had the largest or nearly the largest correlation for each of the yield traits. F2I is the recommended method for traits with larger heritability.