Abstract
We developed rice varieties during a collaborative rice breeding research project using a wide diversity of genetic resources from Yunnan, China, and Japan. In 1999, the new varieties developed during the project covered more than 200, 000 ha in Yunnan. To develop higher yielding and widely adaptable varieties, we analyzed statistically the genotype-environment (GE) interaction using multilocation yield test data in Yunnan. Uniform rice nursery data from 12 locations and two years were used as a basis for examining variety adaptability in Yunnan. Variance components for varieties × locations and varieties × locations × years indicated that the interaction of varieties with environments was very important in determining relative yields. Progress in the development of high-yielding and adaptable varieties was apparent in the comparison of Hexi 34 and Hexi 35 with a standard variety, Yunjing 9, and other improved Yunnan varieties. Some statistical methods were used to group the varieties based on the analysis of the GE interaction. The sum of squares (SS) of GE interaction was partitioned by linear regression into joint, genotypic, and environmental regression, stability variance analysis, and by additive main effects and the multiplicative interaction (AIVIMI) model into three significant principal component axes. The AMMI model was most effective in partitioning the interaction SS among the three methods. The AMMI analysis suggested the presence of certain close relationships between yield stability and some agronomic characteristics of rice varieties, such as early maturing habit, panicle density, and seed fertility.
