The Cirad/CIAT rice collaborative project focuses on the development and enhancement of rice synthetic populations through recurrent selection (RS) with the goal to develop and diffuse improved material, populations or lines, for various rainfed ecosystems in Latin America and the Caribbean (LAC). Our breeding strategy is based on the development of broad-base populations, their improvement thought recurrent cycles of selection and recombination, and the exploitation of their genetic wealth for line development through conventional breeding. Within our network of rice breeders from LAC, we released improved material as part of the breeding activities for upland rice in Colombia and shared germplasm of interest to LAC collaborators as part of prebreeding activities. We are developing large scale phenotyping methods adapted to field conditions to evaluate response to drought among synthetic populations. The thermographic infra-red technology was assessed for screening large number of families, as well as for single-plant evaluation within large populations. This evaluation method allowed us to identify interesting progenitors to develop new population with increased resilience to drought. Allelic variability measured in several synthetic populations revealed high level of neutral diversity. These activities are part of a much broader mission on the management of synthetic populations for more efficient use of RS breeding and for the development of a marker assisted recurrent selection program. (Resume d'auteur)