Résumé
Lowlands constitute one of the largest and most appropriate environments for rainfed rice cultivation in West Africa. In the context of climatic change, drought is among the most important abiotic constraints in this environment. The Generation Challenge Program (GCP) funds an innovative breeding program based on marker-assisted recurrent selection (MARS) for developing drought-tolerant cultivars with high yield under normal water regime, and good yield under drought stress. The detection of relevant association between molecular markers and phenotypes depends on improved methods for high-quality high-throughput phenotyping in the field. The project uses two phenotyping methods: (1) yield and yield components are evaluated under two hydric conditions (irrigated and stressed), and (2) during the drought period, the plant transpiration capacity is estimated through measurement of the canopy temperature. Plant canopy temperature acts as a good indicator of the plant water status as, when plants are undergoing water stress, their temperature increases. These changes in canopy temperature, relative to ambient temperature, result from the modification of plant transpiration through stomatal conductance control. A bi-parental indica population of 230 F3:5 lines derived from a cross between IR64 and B6144F, and 10 control varieties, were phenotyped under drought conditions at the International Center for Tropical Agriculture (CIAT) Villavicencio experimental station (Colombia) during the dry season 2012/13. Irrigation was suspended for 2 weeks at reproductive stage (45-60 days after sowing). Drought response was evaluated on the basis of canopy temperature obtained with a numerical infrared thermographic camera. The methodology was improved by introducing the crop water stress index (CWSI) calculated using climatic data collected from an on-site weather station to normalize canopy temperatures against micro-meteorological weather fluctuations. Leaf canopy temperature at reproductive stage exhibited strong and significant genotypic differences that were negatively correlated with soil moisture content. This phenotyping approach allowed us to identify genotypes with good maintenance of transpiration capacity (and thus sustained growth) under drought stress. This ability can potentially be related to two main causes: either a lower volume of water necessary for plant functioning (thus reducing the need for water extraction from the soil) and/or a greater root system depth, accessing a larger amount of water. Independently, the two variables (canopy temperature during stress and yield) showed wide diversity: canopy temperature during the drought period displayed values differing by more than 5°C between sensitive and tolerant lines. Similarly, yield differed strongly among genotypes. A negative correlation was observed between canopy temperature and yield, thus highlighting the importance of water availability during the reproductive period. This phenotyping approach also permitted us to identify genotypes able to sustain adequate growth under drought stress. The integration of phenotypic data obtained through the high-throughput phenotyping with molecular data (generated by AfricaRice with 484 single nucleotide polymorphism [SNP] markers) will constitutes a significant advance for the identification of genomic regions associated to this trait. If these results are confirmed, they will allow the development of efficient MARS approaches for breeding rice for drought tolerance. (Résumé d'auteur)