Abstract
International centers, such as CIAT, routinely conduct crop performance tests across years and locations internationally, generating valuable multienvironment trial (MET) data. These data must be properly evaluated and interpreted to enhance efficiency of breeding/ testing programs. The purposes of this study were to identify superior bean (Phaseolus vulgaris) cultivars from MET conducted from 1995–2002 in eight southern African countries and to evaluate test locations to improve testing efficiency. Analyses of variance by year across locations revealed significant variation among cultivars in seven years as well as significant variation associated with cultivar-by-location interaction in all eight years. Kang's YS; statistic and GGE biplot methodology based on principal-component analysis helped identify cultivars that had general and/or specific adaptation across/to locations. These methods were also useful in identifying cultivars whose stability/instability was influenced by the linear effect of environmental index. The GGE biplot methodology helped identify redundant and/or non-informative locations. We conclude that the number of test sites used for screening of germplasm in CIAT's southern African bean program can easily be reduced. Elimination of some of the redundant locations could save the CIAT's common bean testing program valuable resources, such as time, labor, and money. Greater use should be made of those test locations that showed greater differentiation among cultivars, e.g., Bembeke, Chitedze, and Delmas. The use of environmental index as a covariate provided only limited amount of information. In half of the trials, the linear effect of environmental index affected the stability of certain cultivars.