SELECTION INDEX BASED ON GENETIC CORRELATIONS AMONG ENVIRONMENTS

Plant breeders often conduct preliminary tests of genotypes at one or two primary locations due to limited seed supply and/or limited resour ces. The location of these tests may be somewhat arbitrary, and in gen eral, it is not known how performance at these locations relates to pe rformance in a much broader target environment. The objective of this study was to develop a method for weighting performance at primary loc ations, based on the genetic correlations between these locations and the target environment, that would lead to greater predictive value of the primary locations. Using data from wheat (Triticum aestivum L. em . Thell) and soybean (Glycine mar L.) cultivar trials in Kentucky from 1987 to 1991, we developed a selection index based on the genetic cor relations of cultivar performance in the primary locations, Lexington (LEX) and Princeton (PR), and the target environment, represented by a s many as seven locations from the cultivar trials (($) over bar Y). A s expected from theory, predicted response to index selection exceeded predicted response to selection at either LEX or PR, and it was super ior to predicted response of selection based on the unweighted mean of LEX and PR. In all years, a lower number of statistically significant wheat genotype rank changes occurred when ranks across the primary lo cations were based on the index than when ranks were based on the unwe ighted means. For soybean, this was true in 4 of 5 years. Although the use of cultivar trial data for parameter estimation violates some of the usual assumptions of traditional quantitative genetic investigatio ns, we suggest the potential benefits to breeding programs outweigh th ese concerns.