Résumé
Crop modeling in support of breeders’ decisions on selection criteria can benefit from the new global focus on phenomics because it provides new information on existing genetic diversity for useful traits. This study attempted an in silico prediction of margins for genetic improvements of early vigor (biomass produced during vegetative growth) and drought resistance combined, based on virtual recombination of several traits (here syn. model parameters) within ranges of trait variation observed in a panel of diverse rice genotypes. The Ecomeristem model was parameterized by multi-parameter optimization procedures applied to observed datasets for 136 rice genotypes. The traits within the observed ranges were then recombined in silico to generate a virtual population of 9000 individuals. Simulations of real and virtual phenotypes under three water treatments, using finite water resources during stress cycles, indicated strong and similar trade-offs between constitutive vigor and drought resistance in both real and virtual, recombinant populations. A substantial margin for potential genetic improvement of vigor with unchanged drought resistance was predicted, drawing chiefly from structural growth and development traits that would increase internal demand for assimilates (larger and thicker leaves, increased leaf appearance rates). Increased vigor would not necessarily require greater photosynthetic potential per se. However, improved drought resistance with unchanged constitutive vigor would require greater water economy (increased photosynthetic potential and limited water use, therefore higher transpiration efficiency) and greater tolerance of leaf extension and gas exchange rates to drought, while tillering ability should be limited in favor of larger and thicker leaves. These results carry significant uncertainty because they predict virtual genotypes and their phenotypes, based on simple assumptions in the model (namely on gas exchange) and in genetics (free, additive trait combinability). But the approach is innovative and may eventually help developing ideotypes drawing from information of existing diversity and integrative modeling of phenotypes.