Evaluating OpenMI as a model integration platform across disciplines

For decision makers in the domains of agriculture and environment, for instance in government agencies, farmers, environmental NGOs and farmers' unions, it is beneficial to evaluate ex-post or to asses ex-ante the impacts of their choices. To research these interdisciplinary relationships, models developed by different scientific disciplines and often operating at different scales can be integrated into model chains that cover processes across disciplines. In order to assemble models into an operational model chain conceptual, semantic and technical levels of integration have to be taken into account. The main focus of this paper is on technical integration to ensure repeatability and reproducibility of model chain runs and to optimize use of computer hardware for model simulations. Technical integration itself can be achieved by different approaches (i.e. manual, scripting, building or using a proprietary framework, using an open framework based on standards). From the many available modelling frameworks (e.g. OMS, TIME, KEPLER, FRAMES, MODCOM, OpenMI) the emphasis will be on OpenMI, the Open Modelling Interface and its use and usefulness as a readily available, generally accepted and open standards based framework. OpenMI is an open source software standard for dynamically linking models at runtime, which can potentially be used in many domains, but is currently mainly applied in the water and environmental domains. This paper describes and evaluates the use of OpenMI in several multi-disciplinary large projects that worked on integrated models. These projects operated in the disciplines of agriculture, land use, nitrogen cycling, forestry, hydrology and economics. To this end two workshops were organized to acquire feedback from both software developers and modellers that contributed to the aforementioned projects on the use of OpenMI. Perceived advantages and disadvantages of OpenMI differed between modellers and software engineers, although both identified the lack of standard functionality as a major disadvantage and the prescription of a way of working through OpenMI as a standard as a major advantage. In conclusion, OpenMI can be used as a standard for technical model integration across disciplines, and it is not limited to one particular discipline. (C) 2012 Elsevier Ltd. All rights reserved