In many practical applications of porous media flow simulators, the most complex processes are confined to a small part of the model domain. The demands of a simulator on computational resources increase with the physical complexity of the model. Thus, a trade-off between physical accuracy and the computational demands of a model has to be made. Either great complexities are neglected in favor of a lean model or all processes are captured with a complex model which is superfluous in large parts of the domain. As a compromise between these options, a consistent transfer concept is introduced. It couples simple and complex models and adapts the resulting multiscale model to the physical processes actually occurring. As a basis for this, a decoupled formulation for non-isothermal compositional multiphase flow is discussed. It has the advantage that the size of the linear system of equations does not grow with the number of phases or components involved. The presentation reviews common concepts for the description of multiphase flow in porous media and provides a consistent derivation of the conservation equations of non-isothermal compositional flow and transport processes. Based on these equations, decoupled formulations for isothermal and non-isothermal compositional flow are derived, using the concept of the local conservation of total fluid volume. An isothermal and a non-isothermal multiphysics concept for the transition of complexity within a porous media domain are presented. Ideas for the extension of the multiphysics towards more complex systems and possible interfaces with multiscale methods are discussed. In the presentation, we want to introduce and distinguish two types of multi-scale approaches which have proven to model this simplified system very efficiently and accurately. We combine an adaptive grid method based on multi-point flux approximation with various local up-scaling techniques. The full abstract of this talk is available here.