By formalizing elemental traits, their compensations and their genetic variability, crop modelling can support the phenotypic and genetic dissection of complex agronomic traits. It can accordingly be used to predict optimal trait combinations and theoretical margin for genetic improvement for a given environment (Cooper et al. 2014). The present study aims to develop and provide a proof of concept of such modelling approach applied to the case of rice early vigor and its response to drought. The plant growth model Ecomeristem (Luquet et al. 2012) was heuristically parameterized for each of 136 accessions constituting a japonica rice diversity panel, using phenotypic data acquired in a greenhouse at IRRI under two water conditions (well-watered vs. dry-down, Rebolledo et al. 2013).Model parameters controlling leaf morphogenesis, transpiration rate, light conversion efficiency and their drought regulation, were in silico recombined within the range explored by the 136 accessions in order to generate a virtual population of 9000 individuals. Simulations of real and virtual phenotypes under three water treatments pointed out strong and similar tradeoffs between constitutive vigor and drought tolerance, estimated as shoot growth maintenance. A substantial margin for potential genetic improvement of vigor and drought resistance was however pointed out, depending respectively on parameter related to C sink strength and water or light use efficiency (Luquet et al., in press). These results are theoretical as they rely on the prediction of phenotypes from virtual genotypes, based on simple modelling (particularly on gas exchange) and genetic (free, additive trait combinability) assumptions. However they provide further insight into the way genetic and physiological information should be further combined into models toward ideotype exploration. This will be discussed with respect to the results of the Genome Wide Association Study recently performed using 12,221 single-nucleo