Résumé
A collaborative CIAT/Cirad project aims to create new improved upland rice germplasm for drought tolerance based on population improvement through recurrent selection (RS). In the framework of a multidisciplinary team (ecophysiology, molecular genetics and breeding), we seek to enhance this breeding strategy through the integration of marker-assisted breeding tools. This requires improving methods for high throughput phenotyping in the field. Four hundreds lines selected from diverse recurrent populations were screened under drought conditions in Villavicencio CIAT experimental station (Colombia) during the dry season 2008/2009. Drought stress was applied for two weeks at panicle initiation stage. Drought response under water deficit at flowering stage was evaluated based on soil moisture and canopy temperature at flowering with an infra-red thermographic camera and final grain yield. Leaf temperature at/around flowering stage exhibited strongly significant varietal differences that were negatively correlated with soil moisture content and yield. This phenotyping approach permitted identifying genotypes that display good maintenance of transpiration (and thus sustained growth) under drought stress, related mainly to limited water extraction and/or greater root depth; it also enabled identifying genotypes that have high transpirationrates (and thus higher potential growth rates) under irrigated control conditions. Selecting suitable combinations of both characteristics is expected to help identifying drought tolerant material with high yield potential. The methodology will be further improved by (1) including on each thermographic photo a wet/dry object to normalize canopy temperatures against weather fluctuations for situations where micro-meteorological backup is not available, and (2) sampling leaf material for complementary delta 13C measurements indicative of transpiration efficiency (TE). The integrated methodology is being implemented in RS breeding schemes and in parallel in a genetic association study on a japonica rice variety panel, in order to identify promising genomic regions and polymorphisms for marker development. This will eventually permit developing efficient marker-assisted recurrent selection (MARS) approaches. (Texte intégral)