Genotype × environment interactions for grain yield of upland rice backcross lines in diverse hydrological environments

Genotype by environment (G × E) interactions were investigated in Vandana and a subset of 13 BC2 and BC3 lines of an improved indica upland rice cultivar, Vandana, backcrossed with a drought-tolerant traditional japonica cultivar, Moroberekan, which has a thick and extensive root system, in response to eight hydrological field environments conducted at Los Baños, in the Philippines, between 2001 and 2003. The G × E interaction accounted for 13% of the total sum of squares with environment and genotype responsible for 84 and 3%, respectively. Cluster analysis identified four environment and six genotype groups, which accounted for 70% of the G × E sums of squares. Of this, AX1, AX2 and AX3 accounted for 27, 22 and 21% of the G × E-SS, respectively. AX1 represented yield potential; AX2 was related to soil conditions, aerobic status and possibly VPD; and AX3 to change in phenology (days to flowering) with stress. The four environment groups were considered as broadly representative of contrasting rice production environments, including lowland-type, upland-wet season and upland-aerobic environments that experienced vegetative- or anthesis-stage drought stress. Genotype groups differed in adaptation to these diverse environments. For genotype groups G1–G6, G3 (VM150) had stable yields across environments; G1 (VM134) had the greatest grain yield in lowland-type environments (E2); G5 (VM135) in wet season environments (E3); G6 (VM168) in anthesis-stage drought (E4); G2 (Vandana and VM26) in vegetative- and anthesis-stage drought (E1 and E4); G4 had average yields across environments. Implications for breeding of rice adapted to contrasting hydrological environments are discussed, with the caution that adaptation to more than one environment type is desirable, because, as is demonstrated in this paper, an untimely climatic event can transform one environment type into another. Our results suggest that selection in one environment type may not give benefit in other environment types, so testing in more than one environment type is essential.