Biometrical Models Based Assessment of Genotype x Environment Interaction of Regional Cotton Yield Trails

Document Type : Original Article

Abstract

THE OBJECTIVES of this study were to use different biometrical models in assessing the genotype by environment interaction (GE) in cotton yield trials, determine the relationship among different stability statistics, and compare the relative efficiency of these models in explaining the GE effects. Variation in lint cotton yield was evaluated in fifteen extra long stable genotypes across 10 environments (location-year combinations) in 2010 and 2011. The combined analysis
of variation showed that the main effects were highly significant and sum of squares proportions (remaining after removing the sums of squares due to error and replications) were 65.30%, 10.6%, and 24.1% for environments, genotypes and interaction, respectively. Pattern analysis split each of environments and genotypes into different lineages of homogenous clusters. This was reflecting the tremendous effects of environments, seasonal variation and genotypic
differences and emphasizing the importance of deep investigation of GE interaction. Joint regression model revealed that the proportion due to regression line was 7.65%. The greater part of GE interaction was due to deviation from regression line. Meanwhile, the proportions of the first two principal components in GE interaction were 36.45% and 19.5%, respectively, with the first I PCA being significant. This reflects the importance of AMMI model in isolating the relevant parts of GE interaction and excluding the irrelevant parts. AMMI-1 and AMMI-2 models were high informative in describing
the main effects and their interaction. AMMI model was superior to joint regression model in terms of its predictive ability
and efficiency in explaining the pattern of GE sum of squares. Moreover, AMMI determined the genotypes with specific stability as well as the discriminative environments. Ranks of stable genotype and magnitude of stability measurement varied with each model. Neither coefficient of regression nor the coefficient of deviation significantly correlated with the mean performance. IPCA1 significantly correlated with the trait mean performance. AMMI stability value (ASV) was highly correlated with the deviation from regression and with Tai coefficients. As expected, both of regression coefficient and α, the deviations from regression and λ, were positively correlated. AMMI model assembled each group of (E2 and E6), (E9 and E10) and (E4 and E7) to establish a mega environment for breeding the associated genotypes. When contrasting stability measurement, genotypes G88, G93, G84×PimaS6,F 81338/08 and F7 1310/08 were commonly exhibited average stability, therefore they could targeted for the simultaneous improvement of yield and stability. At the level of specific stability and adaptability, however, AMMI model dominated other models. It is important to take into consideration the results of specific stability and adaptability, especially when the component of interaction within environments being higher, rather than among environments as it was evident here.

Keywords