The Kyoto protocol has clearly demonstrated the will of nations to take remedial action to mitigate the impact of anthropogenic climate change. This places a heavy responsibility on the climate research community to provide more reliable computations of the anticipated climate change for alternative scenarios of greenhouse gas emissions. Significantly improved assessments of the details of climate change and its impacts will be required, particularly at the regional level relevant for nations and economic regions such as Europe. This in turn will require an improved understanding of the climate system and its interactions with the socio-economic system. As we are unable to experiment with the single system Earth, modelling is the only analytical tool available for understanding the dynamics of the climate system and predicting the future evolution of climate, either under natural conditions or under the influence of anthropogenic impacts. As the climate system is highly complex, development and use of realistic climate models require a sophisticated software infrastructure and access to powerful supercomputers dedicated to climate research.

Recognising this need, the US has initiated an ambitious programme in climate research, the Accelerated Climate PRediction Initiative (ACPRI), for which multi-Tera-scale Supercomputer resources will be made available within a few years. A similarly ambitious programme based on a dedicated multi-Tera-scale Supercomputer is being pursued in the Japanese Earth Simulator Project. Although in the last years Europe has acquired a world leadership role in climate research (as acknowledged in the US National Research Council Report "Capacity of US Climate Modeling", 1998), it will not be able to compete with these initiatives by simply relying on past achievements. National plans for expanding existing national supercomputer facilities for climate research fall seriously behind the US and Japanese projections (see Figure 1). It is clear that Europe will not be able to match the US and Japanese initiatives on an individual national level. To remain competitive in international climate research, and to bear its fair share as a major economic player in addressing the global problem of climate change, Europe will need to coordinate, pool and enhance its efforts in climate modelling. This will require both organisational restructuring and significantly increased investments in supercomputer hardware.

The real and potential expertise in climate modelling in Europe is widely distributed. This is basically healthy, but implies that a successful European climate modelling initiative must develop mechanisms for effectively combining this expertise in a coordinated scientific framework. This is required already today within the framework of the present, nationally based climate supercomputing system in Europe. The present European climate research programme suffers from a basic dichotomy: while the EU Framework Programmes have clearly been successful in promoting cooperative research in general, access to the basic tools for climate modelling, namely supercomputers, has been restricted through local national or even institutional control. This implies that it has not been possible for consortia of European scientists to independently develop joint climate simulation projects that are approved and funded by the EU unless the projects are accepted within the specific programmes of the nations or institutions that operate the supercomputers. This places many highly qualified European scientists at a distinct disadvantage relevant to their more fortunate colleagues who happen to be affiliated with a dedicated national or institutional climate supercomputer centre. As a consequence, experience in the development and application of sophisticated climate models within Europe has been mainly localised in the few national climate research centres with dedicated supercomputer facilities, and major potential resources of climate modelling expertise in Europe have remained untapped. In the US, major initiatives have recently been taken to make climate models more modular, portable and independent on computer platforms. This allows easy exchange of physical, chemical and biological modules between climate models and model integrations with different model configurations.

The implications of the current European situation have been discussed in detail in the Euroclivar Report "Climate Variability and Predictability Research in Europe, 1999-2004" (November 1998). Euroclivar was a concerted action under the Fourth Framework Programme. It organised eight specialised workshops and formulated priorities for the European contribution to the international CLIVAR (Climate Variability and Predictability) Programme. These consist of specific recommendations in five interrelated research areas, European/Atlantic Climate Variability, Global Teleconnections, Anthropogenic Climate Change, Climate Observations (Past and Present) and Climate Modelling. The Euroclivar report summarises the consensus of the European scientific community. For easy reference the executive summary of this report as well as its chapter on Climate Modelling are reproduced in the Annex (section B7) of the present proposal. PRISM will build on the achievements of Euroclivar, in particular on the recommendations for integration of the European climate modelling effort (section B7.2.3). These recommendations can be summarised as follows:

A central objective in PRISM is the development of a flexible model structure (see figure 2) with interchangable model components exchanging information through standard interfaces with a universal coupler or directly with the other model components. This requires a considerable amount of standardisation, which however will be beneficial for further model development. A prototype of the coupler is already available, as are prototype versions of the component models, which are all state-of-the-art and already used by different groups in Europe. Once established, this will allow easy numerical experimentations and the easy exchange of different model components. It will also establish a standard for further model development. The standardisation will also allow the development of a set of common diagnostic tools, which will greatly facilitate the efficiency of climate research.

From the overall objectives of PRISM, a number of intermediate aims follow:

The first task is the detailed definition, for each component of the PRISM system (atmosphere, atmospheric chemistry, land-surface, ocean, sea-ice, ocean bio-geochemistry, or regional models) of a standard physical interface to the other components, i.e. the definition of the nature of information to be exchanged. The second task is the technical development of the coupler and model interface library that will allow an efficient transfer of the standard information defined in the first task.

The third task is the physical and technical interfacing of existing state-of-art models developed by the different partners with the PRISM system; this will be done following coding rules and portability criteria defined in the project. This step will involve adjustment, testing and documentation of the models. The candidate models and components modules are described in more detail in Part-C of the proposal so that the anonymity required in Part-B is not violated. The fourth aim is the development of common diagnostic and visualization tools. The fifth task is the development of a web-based interface that will allow the assembling of global climate models from the different model components, the submission of these climate models on the different computer platforms, the monitoring of the climate model integration, the control of the database archiving of the climate model data, and the link between the climate model outputs and the diagnostic and visualization tools.

Finally, project applications will be performed to demonstrate the usefulness of the system. The process of selecting a small number of demonstration runs from several possible numerical experiments already formulated is also part of the PRISM project, because it will pave the way for the creation of a European management structure.

Controlled experiments with the global earth system cannot be carried out. The only possibility for understanding the many interactions within the earth system and predicting future climate change is through computer simulations. Because of the inherent complexity of the earth system, realistic climate models are necessarily also highly complex and require significant computer resources. The reliability of climate models and predictions is therefore basically limited by the power of supercomputers. Process studies, parameterisation development, sensitivity investigations, model intercomparisons, climate change detection and attribution analysis, data assimilation techniques, etc. are all strongly dependent on the power of supercomputers. The available supercomputing power determines the form of the unavoidable trade-offs between model resolution, process complexity and the number and duration of climate simulations, and thereby ultimately defines the cutting edge of climate research.

Thus, the basic goals of the key action on Climate and the Environment as well as other components of the Fifth Framework Programme related to Global Change are critically dependent on the availability of efficient climate models and the effective use of supercomputers for climate simulations. This includes a number of projects that have been approved in the first call for the key action on Climate and Environment, and which can be conducted with greater efficiency with the tools to be developed as part of the present PRISM project. Of particular relevance for Europe is the increased model resolution needed to provide reliable simulations of climate change on regional scales, which requires considerable computer resources. Another challenge is the needed extension of the present generation of primarily physical climate models to earth system models, including the bio-geochemical cycles and the socio-economic system.

Climate modelling is one of the key applications used on the most powerful scientific computers available world-wide. Some of the largest computer systems in the world were specifically acquired for climate modelling, other scientific computer centres dedicate a significant part of their computer resources to climate research projects. Although all actual computer hardware used in high performance computing today is manufactured outside Europe in the United States or in Japan, the European Community has recognised the need for Europe to be at the leading edge for the applications of high performance computing in science, engineering and other fields. Since large-scale climate modelling is a fairly new and quickly evolving discipline there is no standard or obviously dominating modelling code world-wide. The European research community has the opportunity through the PRISM project to fill this gap and develop the dominant modelling framework for climate research world-wide. The leading hardware and software companies in the high performance computing market typically strive to cooperate very closely with the leading research groups in the important fields of high performance computing. This typically results in inclusion of these groups in design reviews, early access to new hardware and software, preferential pricing, collaborations to develop new software tools, libraries and subsystems. Also the leading vendors have often placed new scientific support centres near the leading research centres in the field of high performance computing to participate in the know-how but also in the publicity attracted by the leaders in the field.

The PRISM project has the opportunity to develop the dominant modelling framework for climate modelling world-wide, thereby establishing leadership for Europe in a key application of high performance computing and attracting the key companies in the high performance computing market to base a significant part of their research and development efforts in Europe.

Recognising all of these needs, the Fifth Framework Programme on Energy, Environment and Sustainable Development recommends in its infrastructure programme the provision of "Supercomputer facilities for the development of high-resolution climate and earth system models, allowing global and regional scale simulations" (p.31 of the EU document "Energy, Environment and Sustainable Development, Work Programme, document d_wp_en_199901.pdf). We propose PRISM as an essential contribution to the achievement of these objectives.

The PRISM proposal has evolved from the experience of the Euroclivar Concerted Action project and an earlier European Climate Computing Network (ECCN) project within the Human Capital and Mobility programme. Euroclivar succeeded in coordinating many climate modelling initiatives, but lacked the means for implementation of the proposed actions. The ECCN offered only a minor enhancement of the existing supercomputer resources of less than 1% and did not address the central organisational and technical support issues that will be the focus of PRISM. Thus it had only limited success in motivating scientists not already affiliated with the existing modelling centres to engage more actively in climate modelling. The PRISM proposal intends to overcome these shortcomings by adopting a new and innovative approach.

It is proposed to implement the Euroclivar recommendations by establishing groups with adequate technical support staff to develop the basic tools needed to conduct an integrated European modelling programme. If funded PRISM will greatly facilitate the exchange of models and model components. This exchangeability will contribute to the efficiency of climate research and will also remedy to the present unwanted situation in which it is virtually impossible for one group to reproduce results of another group. The implementation of an open and easily accessible European-wide structure for climate model simulations is an essential innovation needed to realise a coordinated European climate modelling programme and prepare for a later enhancement of the European supercomputer network for climate research.

The PRISM project represents an innovative approach to the development and enhancement of a very large scientific code designed to utilise the most powerful computer resources available in scientific computer centres in Europe. The PRISM project takes a full solution approach to the key computer aspects of climate modelling. It takes into account from the very beginning issues associated with portability and performance, and will adapt existing codes accordingly to prevalent high performance computer architectures (vector parallel and RISC-based massively parallel systems) as well as state of the art web-based networking and pre- and postprocessing capabilities. However, rather than developing a single very large code PRISM defines a system architecture and strict rules for code development within this architecture. It also includes in the project the necessary work to actually test the developed codes against the architecture and the defined rules and interfaces.

Climate modelling codes are among the most challenging applications for high performance computers today. To make significant steps forward with the next generation of climate models we need to be able to get the utmost performance out of the largest high performance computer systems available in Europe requiring special considerations to utilise modern high performance computing architectures. On the other hand the codes need to run very well on the various, different high performance computing systems available in the large scientific computer centres in Europe and therefore have to be quite portable. To achieve both goals - portability and highest possible performance - the PRISM project includes participation from computer manufacturers in high performance computing.

Climate simulations are CPU-intensive, producing large amounts of and generating heavy transfers of information through the network. Therefore, PRISM will also keep close contact with innovative projects in the field of data networking, such as Géant (a network infrastructure project following TEN-155) and Data-Grid, which is a European data grid intensive application around High Energy Physics.

PRISM will be a three-year project with different phases: system specification, technical development and demonstration. At the start of the project a short kick-off meeting will be held to review the workplan and relevant developments since the proposal was written. Then, the first half year will be dedicated to the system specification. This will comprise a review of existing platforms and existing models as well as the specification of a standard for I/O and archiving, a model environment and diagnostic tools. Decisions about the selection of model components will be made at the first project meeting. The specification phase will be followed by a period of development. During this period, the scientific demonstration project will be further defined in interaction with potential users through a user consultation meeting. In the last year a number of scientific demonstration integrations will be carried out. Major project milestones will be marked by the presentation of a number of documents, namely: