Integrated modelling

The assembling of multiple interacting model components of various domains such as environmental, social and economic systems is known as integrated modelling. The process of developing integrated models comprises of several phases; that are conceptual development, software development, model assessment and application as well as stakeholder participation. While software tools and frameworks in system programming languages exist to aid in the technical development of integrated models, less emphasis is on exploratory model analysis and long-lasting lifetime of process components.

This  project is aimed to aid in the phases software development and model assessment. We will develop a life cycle management system that allows the construction of generic spatio-temporal model components and their coupling. Next to the model construction, execution schemes for component execution are considered. As uncertainty is present in data and process descriptions, the framework will incorporate execution schemes allowing uncertainty estimation of components and integrated model.

Several issues are addressed within this project:

  • to provide a scheduling scheme to execute coupled components supporting fixed and variable time steps as well as accounting for individual events. Including event-based scheduling will ease the construction of multiple paradigm models such as field-based models interacting with agents.

  • to schedule coupled models in schemes such as Monte Carlo or Particle Filtering.

  • to formalise the concepts of model components by means of ontologies. Well described process components will ease the construction of integrated models and strengthen the reusability of components in the model development process. Formalised specifications of components will also ease the translation of information between domain boundaries such as passing geographic information to economic components.

  • quality aspects related to the component construction. Advocate processes such as unit and integration testing; sharing code, data and documentation; peer-review of components. Best practices that can strengthen confidence in components and lead to reusable model libraries.

  • to provide a programming interface suitable for scientists not being software development experts. By using scripting languages such as Python instead of system programming languages like Fortran or C, model development can focus on process descriptions and exploratory modelling.

The project is funded  by VITO (Flemish Institute for Technological Research, Belgium; in close cooperation with the Environmental Modelling Unit at VITO.

You can find information about a framework prototype here.