Separation of Cotton Cultivar Testing Sites based on Representativeness and Discriminating Ability Using GGE Biplots

Abstract

An important task of multienvironment trials (MET) analysis is evaluation of testing sites for megaenvironment differentiation and selection of “ideal” candidate location to improve the efficiency of cultivar selection and recommendation. The objectives of this research work were (i) to divide the Spanish cotton (Gossypium hirsutum L.) testing locations into megaenvironments and (ii) to separate the testing locations based on their distance to the “ideal” location, discriminating ability, representativeness, and uniqueness. GGE biplot was employed to analyze eight 1‐yr and two multiyear (3‐yr, 4‐yr) balanced datasets from 1999 to 2006 cotton trials of Delta & Pine Land Co. in Spain for yield, fiber quality traits, a selection index (SI) based on yield and quality, and Verticillium wilt (Verticillium dahliae Kleb.) disease infestation level. Yearly GGE biplots revealed crossover genotype × location interactions, but not large enough to divide the area into different megaenvironments. Therefore, the Spanish cotton region may be considered as a complex megaenvironment and cultivar recommendation may be based on both mean performance and stability. Las Cabezas location was the closest to an ideal based on both yield and the SI regardless of the change from plastic to nonplastic mulching cultural practice. Aznalcazar did not provide unique information and could be dropped as a test site. The separation of test locations for their discriminating ability and representativeness provided useful information on the effectiveness of each testing location for developing and/or recommending cultivars with specific or broad adaptation. In this sense, Lebrija could be considered as trait‐specific selection environment for early screening of verticillium tolerant genotypes.