Integrating statistical and ecophysiological analysis of genotype by environment interaction for grain filling of barley in Mediterranean areas. II. Grain growth

In Mediterranean areas, grain growth of temperate cereals often progresses under the harmful influence of drought and high temperature. Genotypic responses are mediated by the specific occurrence of these constraints, thus causing genotype by environment (G × E) interaction. Field experiments were carried out in 12 environments of northern Spain to characterize G × E on grain growth of five six-rowed and five two-rowed barley (Hordeum vulgare L.) cultivars. Grain growth was defined as the result of two components: grain-filling rate (GFR) and grain-filling duration (GFD). Genotypic and environmental descriptors were used as concomitant variables at the levels of the genotypic and environmental factor to partition G × E. For a first exploration of G × E, AMMI (additive main effects and multiplicative interaction) models were used. Subsequently, separate factorial regression models were fitted for GFR and GFD. G × E for GFR could be partially attributed to the joint effect of two pre-anthesis climatic variables (rainfall during heading, and average maximum temperature during jointing). The factorial regression model for GFR explained more than half of the analysis of variance G × E sum of squares with a quarter of its degrees of freedom (d.f.). Overall, six-rowed cultivars were more affected by low rainfall at heading and high temperature during jointing than two-rowed types. The inclusion of these pre-anthesis variables suggests that G × E for GFR could be related to large differences in source/sink balance between two- and six-rowed genotypes at anthesis. The factorial regression model for GFD contained the pre-anthesis climatic covariable ratio of rainfall to total evapotranspirative demand during heading, and the genotypic covariable anthesis date, suggesting that G × E for GFD was related to differences in phenology among genotypes. This model retained about 35% of the analysis of variance G × E sum of squares with fifth of its d.f. The success of the integration of statistical and ecophysiological tools for the explanation of grain filling in barley is discussed