INTER-ROW YIELD CORRELATIONS IN SELECTION TRIALS—ASSESSMENT AND IMPLICATIONS

THE objective in clonal evaluation trials is to use a plot format and an experimental design that maximise the broad sense heritability of the assessed trait or traits by minimising interference from non-genetic variation. This study assessed within-plot variability by examining the relationship between the yields of bordered, paired middle rows in a large number of advanced clonal trials. Yields of paired rows were correlated and the significance of the difference between the yields assessed in two crops in two programs. The length of row occupied by cane and gaps was measured in three sequential trials at each of three sub-stations in the crop improvement program based at BSES Meringa. Yields were measured in two uniformity trials conducted in commercial fields in the Mulgrave mill area. Correlations between yields of paired rows within plant and first-ratoon crops in the two programs, on a trial basis, were variable. Median values ranged from 0.421 to 0.575. Lighter and heavier row yields differed highly significantly, within crops and programs. There was large variation for actual and occupied (actual + half gap lengths) plot lengths. Correlations between row lengths within plots were weak to moderate (0.146–0.464, and 0.115–0.475, respectively). Correlations between yields were weak to strong (0.327–0.873 and 0.332–0.867, respectively). Error due to variation in yields of paired rows consistently swamped the plot-to-plot error less the sampling error. In the uniformity trials, the correlations between the yields of paired rows were 0.343 and 0.399 and the error from variation between paired, contiguous rows dominated. Variation in row length and the generally weak relationship between yields of bordered, paired rows is concerning. These features reduce trial efficacy. The precision of planted row lengths needs substantial improvement. Plot shape recommendations can be made only after further study to determine the inherent error arising from variation between or among yields of bordered, paired, or multiple rows. An optimised partition of resources between numbers of bordered rows and replicates would then be possible.