Current systems for enacting scientific experiments, and in particular simulation workflows, do not support multi-scale and multi-field problems if they are not coupled on the level of the mathematical model. We present a life cycle that utilizes the notion of choreographies to enable the trialand-error modeling and execution of multi-scale and/or multifield simulations. The life cycle exhibits two views reflecting the characteristics of modeling and execution in a top-down and bottom-up manner. It defines techniques for composing dataintensive, scientific workflows in more complex simulations in a generic, domain-independent way, and thus provides scientists with means for collaborative and integrated data management based on the workflow paradigm.