Use this URL to cite or link to this record in EThOS: http://ethos.bl.uk/OrderDetails.do?uin=uk.bl.ethos.530274
Title: The formal generation of models for scientific simulations
Author: Tang, Daniel
Awarding Body: University of Leeds
Current Institution: University of Leeds
Date of Award: 2010
Availability of Full Text:
Access from EThOS:
Access from Institution:
Abstract:
It is now commonplace for complex physical systems such as the climate system to be studied indirectly via computer simulations. Often, the equations that govern the underlying physical system are known but detailed or highresolution computer models of these equations (“governing models”) are not practical because of limited computational resources; so the models are simplified or “parameterised”. However, if the output of a simplified model is to lead to conclusions about a physical system, we must prove that these outputs reflect reality and are not merely artifacts of the simplifications. At present, simplifications are usually based on informal, ad-hoc methods making it difficult or impossible to provide such a proof rigorously. Here we introduce a set of formal methods for generating computer models. We present a newly developed computer program, “iGen”, which syntactically analyses the computer code of a high-resolution, governing model and, without executing it, automatically produces a much faster, simplified model with provable bounds on error compared to the governing model. These bounds allow scientists to rigorously distinguish real world phenomena from artifact in subsequent numerical experiments using the simplified model. Using simple physical systems as examples, we illustrate that iGen produces simplified models that execute typically orders of magnitude faster than their governing models. Finally, iGen is used to generate a model of entrainment in marine stratocumulus. The resulting simplified model is appropriate for use as part of a parameterisation of marine stratocumulus in a Global Climate Model.
Supervisor: Not available Sponsor: Not available
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID: uk.bl.ethos.530274  DOI: Not available
Share: