A comparative evaluation of two methods of selecting locations used for testing spring wheat cultivars

One of the considerations of regional cultivar evaluation programs is to optimize the number of locations used for testing. Although optimization of numbers of locations using cluster analysis has been previously attempted, no objective comparison of methods has yet been made. A new clustering method that uses the pairwise contribution of locations to the cultivar x location mean square as the distance measure (LB) was compared to another method that employs diallel correlations as the distance measure (CL). Data from six spring wheat (Triticum aestivum L. em Thell) cultivars grown at 13 locations for five years were used in the initial cluster analysis. Another set of data, from a separate year, consisting of yields of the original 6 cultivars and a set of 12 independent cultivars was then used to check the validity of the original groupings and to compare the two methods. When the 6 original cultivars were considered, the LB technique was found to be superior to the CL. When the 12 independent cultivars were used, neither method was considered to be superior. Because of the lack of flexibility on the part of the LB method, neither technique could be deemed as fully adequate.