CSDMS 2015 annual meeting poster HarutyunShahumyan

Presentation provided during CSDMS annual meeting 2015

Challenges and a solution in coupling dissimilar models for complex planning policy analysis

Harutyun Shahumyan, University of Maryland, Maryland, United States. harut@umd.edu
Rolf Moeckel, University of Maryland, Maryland, United States.

CSDMS2015 poster HarutyunShahumyan.png


Close model integration has become the mantra among model developers. New tools under development, such as CSDMS or OpenMI, promote tight integration of very different models and ease information transfer between the same. Continuously increasing computational capacities enable ever more comprehensive model integrations. From a technical perspective, the prospects of tight model integration are excellent. However, the research presented also exemplified limitations of model integration.

Python wrappers were developed to loosely couple land use, transportation and emission models developed in different environments. ArcGIS Model Builder was used to provide a graphical user interface and to present the models’ links and workflow. With the use of Python wrappers, the implementation of the coupler is separated from the models’ source codes. This gives an additional flexibility, which can help in terms of portability, performance and maintenance of the codes. Though some limitations, this system supports different types of components and links them under a single user interface without changing their original source codes. The integrated system calls automatically component models developed in Java, Cube script, MS Excel and C++ environments.

The suggested approach is especially efficient when the models are developed in different programming languages, their source code is not available or the licensing restrictions make other coupling approaches infeasible. Though this research focuses on dynamic spatial models that integrate transportation, land use and environmental impact models, the methodology is not limited to this type of systems. It can also be applied to other systems requiring consecutive implementation of standalone components including non-spatial models. Notwithstanding the ability to run complex model scenarios, the probably most important lesson learned of this research refers to the level of model integration. A key finding of this research is that model integration should depend on direction of information exchange and frequency of data flows. While this simple but robust loose integration method has satisfied the project’s initial goals, further tighter integration within the CSDMS is currently explored in the view of enhancing the models performance and data exchange speed as well as to widen their scope of applications.

* Please acknowledge the original contributors when you are using this material. If there are any copyright issues, please let us know and we will respond as soon as possible.