Spatial Variability Models to Improve Dryland Field Trials

Abstract

Spatial variability in field trials is a reality. A proportion of this is accounted for as inter-block variability by using block (complete or incomplete) designs. A large amount of spatial variability still remains unaccounted for, however, and this may lead to erroneous conclusions. To capture this inexplicable variation (which is mainly due to intra-block variation), yield data from a series of variety yield trials, using cereals and legumes, were analysed using various spatial models. The most suitable of these, selected on the basis of the Akaike Information Criterion, were used to assess the relative performance of genotypes. Although incomplete-block designs have been found to be effective in variety trials, spatial models have added considerable value to trials with legumes and cereals. The ‘best’ spatial models gave efficiency values of over 330% in winter-sown chickpea (Cicer arietinum), 140% in lentil (Lens esculenta), and 150% in barley (Hordeum spp.) trials. Furthermore, the use of these best models resulted in a change in the ranking of genotypes (on the basis of mean yield), which resulted, therefore, in a different set of genotypes being selected for high yield. It is recommended that: (i) incomplete block designs be used in variety trials; (ii) the Akaike Information Criterion be used to select the best spatial model; and (iii) genotypes be selected after the use of this model. The selected model would account most effectively for spatial variability in the field trials, improve selection of the most desirable genotypes and, therefore, improve the efficiency of breeding programmes.