E-Science applications are often interdisciplinary and require resources from different infrastructures; the interoperability between heterogeneous infrastructures is an important requirement for constructing large scale experiments. Analyzing interoperability issues by including both application and infrastructure aspects promotes a global view on interoperability of different layers, and we argue it can converge to an optimal reference model to guide the development of service layers in e-Science infrastructures. This paper proposes an Open Distributed Processing based interoperability reference model for e-Science.