BIOMETRICAL CHARACTERIZATION OF TEST SITES FOR MAIZE BREEDING
2003
Domagoj Šimić | Jerko Gunjača | Zvonimir Zdunić | Ivan Brkić | Josip Kovačević
Yield stability of genotypes and analysis of genotype×environment interaction (GEI) as important objects in analyses of multienvironment trials are well documented in Croatia. However, little is known about suitability and biometrical characters of the sites where genotypes should be tested. Objectives of this study were in combined analysis of balanced maize trials i) to compare test sites in joint linear regression analysis and ii) to compare several stability models by clustering test sites in order to assess biometrical suitability of particular test sites. Partitioning of GEI sum of squares according to the symmetrical joint linear regression analysis revealed highly significant Tukey's test, heterogeneity of environmental regressions and residual deviations. Mean grain yields, within-macroenvironment error mean squares, and stability parameters varied considerably among 16 macroenvironments. The highest grain yields were recorded in Osijek in both years and in Varaždin in 1996, with more than 11 t ha-1 . It seems that Feričanci would be optimum test site with relatively high and consistent yield and high values of entry mean squares indicating satisfactory differentiation among cultivars. However, four clustering methods generally did not correspond. According to three out of four clustering methods, two macroenvironments of Feričanci provide similar results. Employing other methods such as shifted multiplicative models, which effectively eliminate significant rank-change interaction, appears to be more reasonable.
显示更多 [+] 显示较少 [-]