The magnitude and nature of genotype-by-environment (G x E) interactions for grain yield, days-to-flower and plant height of rainfed lowland rice in Northeast Thailand were examined using random F-7 lines from seven crosses developed by the Thai breeding program. A total of 1116 lines and checks were evaluated in a multi-environment trial conducted across three years (1995-1997) and eight sites. The G x E interaction was partitioned into components attributed to genotype-by-site (G x S), genotype-by-year (G x Y) and genotype-by-site-by-year (G x S x Y) interactions. The G x S x Y interaction was the largest G x E interaction component of variance for all three traits. There was little G x S interaction for grain yield and days-to-flower. The G x S interaction was significant for plant height, but was the smallest component of variance for that trait. The G x Y interaction component of variance was significant for all three traits, but was small relative to the genotypic component for days-to-flower and plant height. For grain yield the G x Y interaction component was comparable in size to the genotypic component. Partitioning the genotypic and G x E interaction components of variance into among-cross and within-cross components indicated that there was significant variation both among and within the crosses for each trait. The relationships between the three traits differed among the crosses and the environments. A major factor contributing to the large G x S x Y interactions for grain yield was the genotypic variation for days-to-flower in combination with environmental variation for the timing and intensity of drought. Some of the interactions associated with timing of drought were repeatable across the environments sampled in the multi-environment trial, and to some extent the environments were characterised on the basis of whether there was pre-flowering, intermittent, or terminal drought. There was genotypic variation for grain yield after taking into consideration the influences of timing of drought in relation to plant development, Three of the seven crosses involved the Thai cultivars KDML105 and RD6 as parents. These crosses produced an array of progeny with lower yield than the Thai cultivars, suggesting it would be difficult to improve on the yield of these cultivars. In contrast three of the remaining four crosses, which did not have a Thai cultivar as a parent, produced progeny that had higher yield than KDML105 and RD6 indicating that yield of rainfed lowland rice could be improved above that of the these popular cultivars. (C) 1999 Elsevier Science B.V. All rights reserved.
