Stratification of SADC regional pearl millet testing sites based on grain yield performance of lines

Sequential retrospective (SeqRet) pattern analysis was used to stratify pearl millet [Pennisetum glaucum] testing sites according to their similarity of line-yield differentiation using grain yield data from 90 multi-environment trials (METs) conducted in the Southern African Development Community (SADC) region. The METs, conducted across 25 sites over 9 years were split into two sets: Set 1 (1989/90-1992/93) included introductory genetic materials, and Set 2 (1994/95-1998/99) included advanced genetic materials. Site stratification analysis from Set 1 and Set 2 partitioned the testing sites into six and five groups with R2-values of 76 and 79%, respectively. Analysis of the cumulative dataset (1989/90-1998/99) clustered the 25 sites into six groups with R2=76% and captured the major patterns of site similarities in Set 1 and Set 2. Based on our experience from running multi-year METs in the SADC region, the cumulative dataset was more informative in judging the relevance of site-stratification results. SeqRet pattern analysis, exploiting available highly imbalanced historical MET data, provided an objective basis for stratifying the test-sites to facilitate selection of a few representative test-sites for future testing of lines. The results from cumulative dataset suggest that future line testing could be restricted to a few sites picked from within each of the identified site-groups