Climate Model Initialization and Improvement using Particle Filters CLIMIP

Lead Research Organisation: University of Reading
Department Name: Meteorology


Policymakers require predictions of regional and local climate for the near-term (the next 1-30 years) in order to plan effective adaptation strategies. Such predictions exist, but are subject to considerable uncertainties, firstly because our global climate models (GCMs) are imperfect, and because of the internal variability of climate, which can mask or enhance the changes due to external forcings, especially on regional scales (e.g. Hawkins and Sutton, 2009). Progress in climate science to improve GCMs and reduce uncertainty, and to use them more effectively to predict some of the internal variability component would therefore be extremely valuable. This project addresses both of these areas by providing a novel new method for initialising the GCMs, comparing that method with the state-of-the-art DePreSys system, and by identifying the processes responsible for model biases, and hence suggesting methods for GCM improvement.

Planned Impact

The outcome of this research would benefit the whole climate community. Furthermore, operational centers like the Met Office, ECMWF, and other meteorological centers, coastal management, air pollution management, oil industry, and insurance companies will all benefit from this research on the long run. All these communities use climate predictions and some form of data assimilation and encounter highly nonlinear problems that need sophisticated data-assimilation methods. Particle filters are among the few methodologies that can handle truly nonlinear problems, and have the ability to bring climate prediction forward.
Active 'workfloor' collaboration with the Met Office and ECMWF ensures a quick dissemination of the results in the meteorological and oceanographic communities.
Through the NCEO, of which the PI is co theme leader of the Data Assimilation theme, efficient exchange of results within the NCEO community, including the climate theme, is guaranteed. This is strongly enhanced by the contact with the climate community via the NCAS link. Furthermore, the NCEO Knowledge Exchange manager Andy Shaw will organize workshops to promote NCEO achievements outside NCEO, and results from this project will contribute to that activity.
All results will be published in scientific journals and discussed in scientific meetings all over the world to ensure maximal exposure to the scientific community. Finally, a workshop on 'Climate model improvement through data assimilation' will be organised half-way the project to bring in scientific expertise from elsewhere, and to expose our results and ideas to the wider community.


10 25 50

publication icon
Browne P (2015) Twin experiments with the equivalent weights particle filter and HadCM3 in Quarterly Journal of the Royal Meteorological Society

publication icon
Lang M (2016) A systematic method of parameterisation estimation using data assimilation in Tellus A: Dynamic Meteorology and Oceanography

publication icon
Lang M (2017) Data Assimilation in the Solar Wind: Challenges and First Results. in Space weather : the international journal of research & applications

publication icon
Van Leeuwen P (2014) Representation errors and retrievals in linear and nonlinear data assimilation in Quarterly Journal of the Royal Meteorological Society

Description A practicle problem we had to solve is to combine the climate model HadCM3 with our particle filter code to explore fully nonlinear data assimilation in climate models. The standard way of doing this turned out to be impractical, but we found a very efficient new method exploring the parallelisation codes of MPI. The new method is so efficient that it will be the basis of a new publication, now under review, and is expected to revolutionise the way ensemble data-assimilation codes and complex numerical models can be combined.
Another paper in which we use EMPIRE to perform the worlds first nonlinear data assimilation on climate model HadCM3 has been published.
Exploitation Route These developments have featured in SANGOMA, an EU FP7 project on efficient ensemble data-assimilation frameworks for marine social services.
Furthermore, EMPIRE turns out to be so successful that it will form the basis of the NCEO data-assimilation framework and will, as such, form part of a national infrastructure.
Sectors Aerospace, Defence and Marine,Digital/Communication/Information Technologies (including Software),Environment,Financial Services, and Management Consultancy,Healthcare,Retail,Transport

Title Ensemble data assimilation framework 
Description We have produced a very efficient way to connect any numerical geoscience model written in any language that supports MPI to our ensemble data-assimilation code. The main new feature is to communicate between the geoscience model executable and the ensemble model executable via MPI statements. This will revolutionise the speed and efficiency at which complex geoscience models can be implemented into a data-assimilation framework, making the data-assimilation systems much more flexible and user-friendly. 
Type Of Material Computer model/algorithm 
Year Produced 2013 
Provided To Others? Yes  
Impact This will revolutionise the speed and efficiency at which complex geoscience models can be implemented into a data-assimilation framework, making the data-assimilation systems much more flexible and user-friendly. 
Description EMPIRE is a data-assimilation software package that contains state-of-the-art ensemble data-assimilation methods and that can be combined very easily with any numerical model via MPI. 
Type Of Technology Software 
Year Produced 2015 
Open Source License? Yes  
Impact The user group of EMPIRE grows rapidly.