In 2012, ENES published its infrastructure strategy for the decade to come. The document was elaborated through two workshops supported by the FP7 IS-ENES project: a scoping workshop (Montvillargennes, France, March 2010) which resulted in a first version of the foresight. The text was then discussed extensively by the community at the subsequent workshop (Hamburg, Germany, February 2011).
Discussions have led to key conclusions and recommendations regarding:
- the need to access world-class computers,
- the need to further integrate the ENES community and to prepare the next generation of climate models for future massively parallel computers,
- the need to further develop the data infrastructure and its access for a larger range of users.
A comprehensive version of the document has been edited by and circulated among the various contributors.
To read the ENES infrastructure strategy document 2012-2022, click on the image below:
Mitchell J, Budich R, Joussaume S, Lawrence B and Marotzke J (2012), “Infrastructure strategy for the European Earth System Modelling community 2012-2022”, ENES Report Series 1, 33 pp.
URL https://verc.enes.org/community/about-enes/the-future-of-enes/ENES foresight.pdf
The document explores the infrastructure needed during the next decade to support European climate research for seasonal to centennial climate predictions. This research will be integral to providing the scientific basis for climate services. It is highly relevant to the objectives of the Joint Programming Initiative on Climate “Connecting Climate Knowledge for Europe”.
The scientific community working on climate modelling is organized within the European Network for Earth System modelling (ENES). It has outlined major scientific issues both related to the reliability of climate change predictions and to the scientific understanding of climate natural variability. The document presents a vision of what could be available in 10 years time. It is envisaged that by the end of the decade, convective scales will be fully featured in climate models. Initialization and ensemble techniques will be well developed.
Uncertainty will be well characterised ensuring appropriate diversity in both regional predictions and longer paleoclimatic simulations. It will be possible to initialise models from real data using full data assimilation techniques. It will be easier to evaluate models and their shorter range projections using hindcasts applied to real data, and such evaluation will be harnessed in a cycle of continuous model improvement.
The implications of this vision for infrastructure are set out. The most demanding goal is to ensure by the end of the decade convective scales are resolved in European climate models of the Earth system with the objective to obtain regional climate predictions for next few decades which are more reliable. This in turn will require intensive and adapted access to exascale computing (1018 operations per second), co-located to an unusually large data archive, which needs to be connected to national archives by networks transferring data at rates faster than Tera bits/second. European models will benefit from being more modern in terms of flexibility and usability, and having a designed diversity, with the number of model «families» possibly reduced and commensurate with the resources available for their development and use.They should scale very well on high performance computers, but also allow usage across the whole computing pyramid. This will mandate a better connected and organized European community, agreeing and working together on common goals.
The requirements to ensure the availability of appropriately skilled experts is outlined, and possible funding and governance are discussed.
ENES recommends the following action items for the climate modelling research infrastructure:
1. Provide a blend of high-performance computing facilities ranging from national machines to a world-class computing facility suitable for climate applications, which, given the workload anticipated, may well have to be dedicated to climate simulations.
2. Accelerate the preparation for exascale computing, e.g. by establishing closer links to PRACE and by developing new algorithms for massively parallel many-core computing.
3. Ensure data from climate simulations are easily available and well documented, especially for the climate impacts community.
4. Build a physical network connecting national archives with transfer capacities exceeding Tbits/sec.
5. Strengthen the European expertise in climate science and computing to enable the long term vision to be realized.
Strengthening the European climate modelling infrastructure will provide Europe with the necessary evidence and expertise for its mitigation and adaptation policies.