The average yield of rainfed lowland rice in Cambodia is the second lowest in Asia. The major constraints that cause this low yield are drought and poor soil fertility, and to date few well adapted high yielding genotypes have been developed. The development of drought tolerant and adapted genotypes to the poorer soils in the target population of environments is the major goal of the rice breeding program of Cambodia. Therefore, the original objective of this study was to examine genotypic variation for grain yield (GY) under late season drought and poor soil fertility and their interactions. However, the results of the first two years of experiments indicated that yield potential was an important character across water and fertiliser conditions and some of the interactions were associated with photoperiod sensitivity. Based on this finding, the study further addressed the following objectives: (i) the mechanisms associated with high yield potential,
(ii) the genotypic consistency under different environments and (iii) the effect of photoperiod sensitivity under different water conditions.
The work involved a total of 18 site-year field experiments, which formed a set of nine experiments (Experiment I-VIII). The first two experimental sets (Experiments I-IIA) were used to evaluate genotypic variation and their interactions for GY and putative drought tolerant traits across two water and fertiliser conditions. Two experimental sets (Experiments lIB-III) were used to examine the mechanisms associated with high yield potential. The effect of photoperiod sensitivity on GY under two water conditions with two seeding times and ages of seedlings at transplanting was studied in three sets of experiments (Experiments IV-VI). The series of Experiments VII and VIII were used to investigate the consistency of GY and drought response index (DRI) under 14 water-site-year environments and for GY at 12 site-year
environments, respectively. Randomly sampled sets of 6 to 80 breeding lines and local varieties originated from the Cambodian germplasm and breeding materials from the Thai-ACIAR project were used for all experiments.
To increase the chance of the late season drought stress, a technique that drained water from the 'fields was modified for the local conditions. It was effective in developing the drought stress during the target flowering through grain filling stage, with GY reduction varying from 8% to 58%. An application of inorganic fertiliser at the rate of 60-30-30 kg/ha of N-P2O5-K2O in Experiments I and II increased GY from 14-29%.
The genotype-by-environment (GxE) interaction component of variance for GY was small (σ2GXE/σ2g= 0.8-2.0) relative to the
genotype component. The phenotypic and genetic correlations between GY under the water-stress and well-watered treatments varied from mild (0.35) to strong (1.28). These small magnitudes of the GxE interaction component of variance, and the strong correlation between environments suggest that it is feasible to develop broadly adapted varieties for the rainfed lowland of Cambodia. High yield potential was found to be an important character in selection for such genotypes.
The high yield potential was found mostly in intermediate flowering genotypes, as they were able to produce intermediate to high total dry matter and harvest index. High yielding genotypes generally had an ability to produce higher dry matter during grain filling stage and took up more nitrogen and phosphorus and used them more effectively. High yield potential genotypes were also found to have a higher sink capacity (number of panicles per square meter), and had an ability to respond to
improved assimilate supply by increasing the grain size, or when source was reduced, to maintain the number of filled grains per panicle and grain size. Three genotypes; IR66327-KKN-8-P1-3R-0, IR66327-KKN-25-P1-3R-0 and IR66327-KKN-54P1- 3R-0 were found to have high yield potential and were well adapted across all experimental conditions.
Although the magnitude of GEls component of variance was small there were some effects of genotype-by-water and genotype-by-water-by-site-year interactions. These effects were found to be associated with days-toflower and DR!. Early flowering was found as a good mechanism to escape late season drought. The pattern analysis indicated that there was some level of genotypic consistency in DRI across seven drought environments. IR46331-PMI-32-2-1-1 and CAR4 had low and consistent values of DRI across all drought environments. On the other hand, IR57514-PMI-5-B-1-2 and CAR3 had the highest mean values of DRI, suggesting a level
of drought tolerance.
Drought stress, late seeding and the transplanting of old seedlings delayed flowering time of rainfed lowland rice and this delay was more pronounced in the photoperiod insensitive genotypes. When flowering was delayed and late season drought developed, the photoperiod insensitive genotypes were disadvantaged, particularly in the late maturity group as they had a longer delay in flowering. However, in the early flowering group, the photoperiod sensitive genotypes had low GY as they nowered too early when seeded late or transplanted with old seedlings but water condition was favourable late in the season.
The findings in this study were considered in terms of their implication to the breeding program for the rainfed lowland rice of Cambodia. The suggestions for modification of the breeding program focused on 1) the initial selection of donor lines with high yield potential and drought tolerance by screening of
local germplasm under irrigated and under the managed drought screen developed in the study. The donor lines, selected independent of their quality would be crossed to recipients lines which have good quality characteristics to develop new populations, 2) advancing the new populations to the F7-Fa generations and then selecting the lines for high yield potential, drought tolerance (using the draining technique) and other desired traits such as grain quality, 3) developing a multi-environment testing system based on the GxE interactions to identify the appropriate location and number of sites to represent the target population of environments, 4) initiating a selection program to identify experimental lines that provided reliable yields under drought and that responded with high yields in those sites (and years) with good moisture, and with good grain quality, 5) validating the predictive performance from the yield potential and imposed drought nursery in the target population of
environments, and 6) to characterising the target population of environments in order to exploit broad and specific adaptive mechanisms in the breeding program. Lastly there was a need to characterise the target population of environments for the exploitation of broad and specific adaptive mechanisms in the breeding program.