Toward model-driven engineering principles and practices for model replicability and experiment reproducibility

Recent years have seen a proliferation of the use of simulation models in computational science. Most of these models have never been independently replicated by anyone but the original developer. Furthermore, there is a growing credibility gap due to widespread, relaxed attitudes in communication of experiments, models, and validation of simulations used in computational research. We examine various issues and challenges involved in model replication and simulation experiment reproducibility. Model-driven simulation engineering principles and model transformation concepts are adopted as solution strategies to improve replicability of models and reproducibility of experiments. A process model, an architectural framework, and an implementation strategy is introduced to address identified issues in simulation experiment management and model replication.