Résumé
One of the main challenges of breeding programs is to identify superior
genotypes from a large number of candidates. By gradually increasing the
frequency of favorable alleles in the breeding population, recurrent
selection improves the population mean for target traits, increasing the
chance to identify promising genotypes. In rice, population improvement
through recurrent selection has been used very little to date, except in
Latin America. At Embrapa (Brazilian Agricultural Research Corporation),
the upland rice breeding program is conducted in two phases: population
improvement followed by product development. In this study, the CNA6
population, evaluated over five cycles (3 to 7) of selection, including 20
field trials, was used to assess the realized genetic gain. A high rate of
genetic gain was observed for grain yield, at 215 kg.ha-1 per cycle or
67.8 kg.ha-1 per year (3.08%). The CNA6 population outperformed the
controls only for the last cycle, with a yield difference of 1,128
kg.ha-1. An analysis of the product development pipeline, based on 29
advanced yield trials with lines derived from cycles 3 to 6, showed that
lines derived from the CNA6 population had high grain yield, but did not
outperform the controls. These results demonstrate that the application of
recurrent selection to a breeding population with sufficient genetic
variability can result in significant genetic gains for quantitative
traits, such as grain yield. The integration of this strategy into a
two-phase breeding program also makes it possible to increase quantitative
traits while selecting for other traits of interest.