We develop socio-environmental models that can be used indifferently for testing scientific hypotheses calling, for example, for sensitivity analyses, or for consultation support bringing into play participatory and prospective exploration of scenarios. This project defines a semantic method for specifying and implementing the initialization of simulation scenarios and the indicators to be observed. We advocate explaining separately the structure of model status (initial and during simulation), that of the input data and, particularly, dynamics parameterization and the structure of the indicators in order to be able to specify transformations between those structures. Thus, based on the structures, it will be possible to define both the possible scenarios and the indicators in the form of MDE (Model Driven Engineering) model transformations. This approach is generic for problems related to the initialization and observation of socio environmental models.