Prediction of Sugarcane Cross Potential at Early Selection Stages.

The primary goal of this study was to develop a faster sugarcane cross appraisal method. The effect of intrarow plant spacing on trait variability was also investigated. The study evaluated 1800 progeny from 15 crosses among 23 parents at two intrarow plant spacings in 1989 and 1990. Data were collected on plant cane (PC) and first ratoon (FR) single stool seedlings, and first clonal (FC) plots. The family mean, the normal probability of an elite proportion (PROB), the observed elite proportion, and the best linear unbiased predictors (BLUPs) were estimated for each trait. The calculated statistics were strongly correlated within the PC, FR and FC tests (0.69 $\u3c$ r $\u3c$ 1.00). The study suggested that the potential to produce an elite clone with a specific superior trait could be accurately predicted by the cross mean with accuracy similar to BLUP. The mean was the most easily obtained statistic and hence would be most practical to use in the breeding program. Correlations of the PC and FR tests with the FC test suggested that the PC estimates could for most traits be used to make cross appraisal estimates. This would enable selection among families before the normal selection within families occurs in the first ratoon crop. The research also suggested use of wider intrarow spacing may improve selector ability to discern among seedlings due to its enhancement of stool weight variability. Families were additionally evaluated with three statistics for their potential to produce elite progeny with two superior traits. BLUPs and RANK showed good repeatability among tests while the PROB generally demonstrated poorer repeatability among tests. RANK was the sum of rankings for individual traits. The predictions were compared to progeny selection rate within the crosses. When the predictions were compared to seven of 15 families, where over 1000 seedlings had been evaluated for each cross, the results illustrated that joint prediction of Brix and stool weight by means of BLUP and RANK identified the better crosses. Initial PROS was not consistent in this regard. The comparative ease to calculate the RANK estimate versus the BLUP suggested that the RANK method would be the best statistic to make bivariate predictions