The maturity of the GreenLab model at the single plant and uniform crop level has led us to consider its application to the landscape scale, where heterogenity in space and time becomes an issue but where applications are. Our focus remained on plants and the related biophysical processes but we aimed at ""functional landscapes"" that should eventually achieve some quantitative realism. We adopted a bottom-up approach, starting with what we knew how to do and expanding according to where our experiments showed potential improvements. The first prototypes had severe shortcomings in software architecture and led us to consider a new simulation formalism for our needs, as well as a new model for competition on resources. The synchronisation formalism and software architecture is a lot more flexible and generic than what we had originally, but at the same time it lays some drastic constraints on the models themselves and their capacities. It can be seen as a drawback, but we consider this to be a great modelisation tool: it forces the modeler to be perfectly clear about the assumptions of the model, specifically about the time behaviour. The reimplementation of the plant model in this architecture highlighted some interesting points that could bring advances in the modelisation of single plants.