Genotype by environment interaction for grain yield and carbon isotope discrimination of barley in Mediterranean Spain

Carbon isotope discrimination (Δ) has been found to be either positively or negatively related to grain yield of small grain cereals when grown in contrasting environments. In order to clarify a possible association between grain yield of barley (Hordeum vulgare L.) and Δ of mature kernels, five 6-rowed and five 2-rowed barley cultivars were evaluated in 22 rainfed environments of northern Mediterranean Spain. Analyses of variance suggested that the genotypic Δ values were more consistent across environments than the genotypic yields. Genotype×environment (G×E) interaction for grain yield was further explored by fitting an AMMI (additive main effects and multiplicative interaction) model. The first 2 multiplicative axes were found significant. The AMMI2 model provided more accurate estimates of genotypic yields within environments than the conventional unadjusted means across replicates. AMMI2 estimates were used for input into cluster analysis, grouping environments that ranked genotypic yields similarly. Three major groups were obtained, with average yields of 2.42 t/ha (cluster I), 3.06 t/ha (cluster II), and 5.16 t/ha (cluster III). The genotypic ranking for Δ did not vary substantially across clusters, but it changed for grain yield. The average genotypic yields in the low-yielding cluster I ranked opposite to those in the high-yielding cluster III, suggesting the existence of a crossover point at an intermediate yield level. The association between grain yield and Δ for genotypic means within clusters was variable. In cluster I, yield and Δ tended to be negatively related, whereas they were positively related in clusters II and III. Genotypes with lower Δ, i.e. with higher transpiration efficiency, performed better in low-yielding environments (mostly those grouped in cluster I). On the contrary, a high genotypic Δ was of advantage in medium (cluster II) and high-yielding environments (cluster III). This observation supports the assumption that drought tolerance and high yield potential under non-limiting growing conditions may be antagonistic concepts in barley. Genotypic means for kernel number per m 2 and Δ were consistently and positively related within clusters, suggesting that a constitutively high Δ may have been driven by a large genotypic reproductive sink. The convenience of using Δ as a selection criterion in areas exhibiting a considerable G×E interaction for grain yield is discussed.