Relative Efficiency of Spatial Analyses for Non-replicated Early-stage Sugarcane Field Experiments

In the early stages of sugarcane (Saccharum spp.) selection programs, large numbers of clones are tested in non-replicated plots. Field trends are likely to affect the performance of these non-replicated experimental genotypes and mask their true genetic potential. The purpose of this study was to evaluate different spatial analyses for their relative efficiency in accounting for field trends in early-stage sugarcane selection trials. A Moving Means (MM) and Autoregressive Spatial (SP) analysis methods were used to adjust genotype values in the 2000 and 2002 nonreplicated Stage-II field experiments within the Canal Point (FL, USA) sugarcane breeding program. Variogram plots showed more complex patterns for fields of larger size and suggested the existence of both local trends and gradients across the fields. Based on reduced error variance for the checks, the relative efficiency (RE) obtained with MM (quadratic method) was inconsistent and ranged from 4 to 188%. Based on the average standard error of differences, gains in RE ranged from 95 to 492% with the SP method. SP was, therefore, the preferred method for giving consistent and greater gains in RE, thereby indicating its suitability for analysis and selection of early-stage sugarcane trials. Of the 135 clones normally advanced to the next testing stage, 75% were congruent between the standard and SP. Adjusting for field trends also resulted in the improvement of phenotypic correlations among the traits, even annulling some of the negative associations between cane yield and quality traits. These results suggest that spatial methods could be suitable for the analysis and selection of sugarcane clones planted in non-replicated early-stage field trials.