Ecology, like many disciplines, commonly relies on simulation to provide insights into the dynamics of complex systems. Yet there are two unresolved problems for ecological studies relying on simulation. First, it is often the case that simulators representing the same system, designed for ostensibly the same purpose, differ in their results with the reasons buried deep within computer code. Second, ecology is a diverse discipline and each subdiscipline necessarily has its particular simulation methods. This raises a problem as to how models from these various fields can be coupled for transdisciplinary studies. We built a new simulation platform named 3Worlds, grounded on a concept familiar and common to all fields of ecology: the ecosystem. We defined the ecosystem for the purpose of simulation by a precise set of rules. The platform can implement models from fields as diverse as food web, population and landscape ecology, energy and material stocks and fluxes, and techniques such as agent-based, cellular automata and discrete-event simulation. In addition, we developed a dynamic graph to represent ecosystems as a set of interacting components. Our approach goes some way to unifying ecology for the purpose of simulation and reduces the problem of code comparison to a comparison of two graphs: (1) a specification graph that complies with the rules of what constitutes an ecosystem, and (2) the successive graph states of a particular simulation trajectory representing the ecosystem. Two applications constitute the core of 3Worlds. ModelMaker builds the ecosystem compliant model and ModelRunner executes the model represented as a dynamic graph. A library of ~24 models illustrates how 3Worlds can simulate very different systems, from simple 1-equation 1-variable models to individual-based systems with thousands of ecosystem components.