I think coupling two phenomena need to be under the general laws: conservation of mass, energy and momentum. In writing coupled equations, you consider these laws and follow the equations and then simply add the derivatives.
the answer can be found in using some kind of interconnection and exchange protocol like HLA (High Level Architecture), providing you exactly the mechanism you're searching for. Look for Simulation and HLA and search for Prof. Bruzzone and Revetria papers in order to have information about its implementation.
Create an output file from one model and so the the other model can read it or as an alternative although very much the same is to build a black box that will represent one of the models in the other.
In doing this you should consider the effects each model may have on each other and you may want to take the above actions both ways.
There are various solutions to your question. The two models can be synthesized using the low-level techniques, such as RPC and shared memory, combined under the general framework, such as DEVS, interacted via an interoperation middleware, such as HLA/RTI.
These solutions are general approaches so that you can just choose one of them considering your conditions and model objectives.
I am afraid that if the two models are based on discrete event systems, there are no obstacles to apply these solutions. However, one of them is a continuous model, the problem is not simply solved.