AbstractScientific community is trying to create climate models as realistic as possible to simulate interactions and to analyse causal relationships in our environment. With these many simulations, researchers are trying to establish the most likely situations, and aim to predict future behaviours. These simulation models have a high rate of variability, in particular the choice of various elements that influence a particular simulation like the CO2 cycle or the horizontal grid resolution. This variability is usually managed by a configurator. But the knowledge base used to develop these models is constantly evolving,
which means a constant review of models and so the configurator.
In order to best manage the variability and the scalability supported by the configurator, we suggest to rely on a feature model that provides a reasoning and verification support that is easier to understand and edit than source code. The goal of our work is to apply this suggestion to a real case. This project uses a feature modelling language called Textual Variability Language (TVL) and implements a configurator based on this language for the configuration of atmospheric models by the Community Atmospheric Model (CAM).
Moreover, the confrontation to the considered real case shows the importance of default values to simplify the configuration of a model by the users. However these values are not part of feature models and require an extension of feature modelling languages. So we'll suggest a theoretical extension for TVL.
|Date of Award
|12 Sept 2012
|Patrick Heymans (Supervisor), Arnaud Hubaux (Co-Supervisor) & Germain Saval (Co-Supervisor)
- Feature Modelling
- default values