Reliable (regional) climate change predictions cannot be achieved without enhanced European collaboration. PRISM has been set up to foster this collaboration, and to provide a common model infrastructure. A first step requires an assessment of the science of climate simulation in Europe. It was with this aim in mind that a meeting was organized in June 2001 in Les Diablerets in Switzerland. The workshop was attend by about 50 researchers and students from all over Europa..

For the preparation of the meeting an organizing committee was formed consisting of Guy Brasseur (MPIM Hamburg, co-chair), Gerbrand Komen (KNMI, De Bilt, co-chair), Hervé Le Treut (IPSL), Alan O'Neill (Un. Reading), Markku Rummukainen (SMHI) and Sophie Valcke (CERFACS).

This committee defined the following objectives:

• To review the status of Climate and Earth System Modelling in Europe: What models are available in Europe? What are the scientific requirements?
• For each sub-component - also for the coupled models - what are the arguments in favour of preserving or not a diversity of models; what is an adequate validation; what can we expect in terms of new development; what may differentiate the existing models, and the models to be, in terms of coupling with other sub-components?
• To present the PRISM project
• To formulate scientific requirements for the PRISM model infrastructure
• Foster the interaction between PRISM and external groups/users

• What are the scientific problems to be addressed with the coupled models that the PRISM modelling structure will allow us to assemble?
• Does the PRISM concept have all relevant components to address these scientific questions?
• What is the definition of a "component"? Is the partitioning of the Global Climate System into components as done in PRISM the relevant one?

On behalf of the organising committee,

Guy Brasseur, MPIM & Gerbrand Komen, KNMI
 
 

2. Discussion and summary reports

A significant amount of time was reserved for discussion. Part of the discussion will be continued elsewhere. In a number of case the discussion have been summarized. These summary reports are provided in this section.

2.1 The Atmosphere Component

Summary by Serge Planton

The different models applied to climate studies in Europe are ARPEGE-Climat (ECMWF+Météo-France/CNRM), ECHAM (MPI), LMDZ (IPSL/LMD), the Unified model (UKMO + UGAMP), the DNM and MGO Russian models and the ECMWF atmospheric model used for seasonal forecasting. The four first mentioned models are involved in PRISM. The scientific requirements concern the use of the models in stand-alone or coupled simulations to study climate variability, climate predictability and the impact of human activities on climate. Some of these models, or at least models with the same dynamical cores, are also applied for weather forecasting or pollution studies. Model diversity is useful since multi-model ensembles allow the investigation of this part of the uncertainty of simulated climate (mean, variability, sensitivity to external forcings) which is due to the representation of physical processes. A reduction of the ensemble to only the "best" models is not easy since the identification of these models depends on the choice of the evaluation criteria (mean or variability, spatio-temporal domain, ….). An adequate validation of models supposes the application of a wide range of methodologies: direct comparison to observations in stand-alone and coupled simulations, process studies, intercomparison exercises, use of a hierarchy of models to interpret the more complex model results by theoretical considerations. The future developments of models imply improved physical parameterisations and new numerical approximations to reduce systematic errors, to include new processes, to improve accuracy and efficiency of climate simulations. Higher horizontal and vertical resolution (tightly linked to the evolution of parameterisations), and higher upper level, will be used to improve the representation of climate and climate variability, including extreme events. The implication of preceding remarks for coupling is that the atmosphere component needs also to be coupled to simplified versions of the other components (soil water and energy sub-component, over-simplified chemistry, mixed-layer ocean, thermodynamic sea-ice, ….). The main scientific requirements for PRISM infrastructure concern the need to keep stand-alone simulations easily performed, to allow multi-model simulations through easy exchange of components and to preserve fast communication between fast interacting components. This last technical issue has to be checked at the very beginning of the PRISM project implementation.

2.2 Ocean Models, Sea-Ice Models and Keynote Lecture 1

Summary by Alan O’Neill

In this session Dr Claus Böning discussed the state of ocean and sea-ice modelling, and outlined the areas where scientific and technical developments were needed. Ocean models in Europe fall into a few broad classes, based on choices of horizontal and vertical grids. The scientific requirements include flexibility in the choice of physical parametrization, resolution, and modularity dictated by the science needs. Co-ordinated sensitivity studies of are also needed. Model diversity is not large at the moment, and further small sub-divisions of model types should be discouraged. To have community models, core groups should be in place to maintain and document them, so that unnecessary duplication can be avoided. Concerning model validation, one should distinguish between validation of equilibrium and transient properties, the latter being much more difficult. There are no good data sets to test eddy parametrizations, so that high-resolution models are needed as benchmarks. Priorities for model developments are: improved physical parametrizations, the use of hybrid vertical co-ordinates, and higher horizontal resolution in certain regions. To enable coupling of models in the PRISM project, the coupler must be able to handle any horizontal grid, and an initialisation strategy must be developed for high-resolution experiments (for which long-duration spin ups may not be feasible). Core groups should be established to co-ordinate model development and the exchange of results. It is recommended that PRISM should submit the results of its demonstration projects to the ocean model intercomparison project, OMIP.

Helge Drange reviewed the status of sea-ice modelling. At the time of the meeting, there was no complete overview available of sea-ice models in Europe. (A review is scheduled to start in September 2001.) Classes of models include those that treat sea-ice as a viscous plastic or an elastic viscous plastic medium. Model diversity is required because sea-ice models for coupled climate simulations are still in a developing phase. The imperative is to develop robust models in view of the possibility that sea-ice might reduce dramatically over the next 50 years. Greater collaboration with the oceanographic, atmospheric and biogeochemistry groups will be essential. Model developments should include higher horizontal and vertical resolution, inclusion of different ice types and sea-ice biota in biochemical models. Sea-ice models are validated with data from passive microwave sensors, upward-looking sonars and sea-ice stations. The dynamical aspects of sea-ice models have been compared in a sea-ice model intercomparison project (2000), and a corresponding project for the thermodynamical aspects has started recently. Sensitivity studies are needed in coupled models, especially for the ice-free conditions expected in the second half of this century.

The final talk in the session was a keynote address given by Taroh Matsuno, Head of the Frontier Research System for Global Change project in Japan. New high-resolution climate models are being developed to be run on the Earth Simulator, a very powerful NEC supercomputer scheduled for completion in March 2002. For instance, a non-hydrostatic atmospheric general circulation model is envisaged, with a horizontal resolution of around 5 km, in which meso-scale convective activity is treated explicitly. Collaboration with the PRISM project was encouraged.

2.3 Land/surface Schemes (LSS)

Summary by Markku Rummukainen

Why diversity and flexible coupling of LSSs?

It has been shown that the uncertainty in climate simulations, linked to LSSs is considerable, affecting the local-to-regional surface climate and even larger scales. A wide variety of LSSs, from the simple bucket-approach to very complex ones are used. The complexity of the modern schemes suggests that an LSS should be maintained as a separate model system component.

Models in Europe

Different schemes are in use, such as MOSES (UKMO), ISBA (Météo-France), ECHAM (MPIfM), SECHIBA (CNRS, LMD) and ORCHIDEE (IPSL).Depending on the particular implementation, river routing, carbon and vegetation components might be included.

Implications for coupling

• A general interface (not OASIS) already exists and is considered by modeling groups (ISPL/ORCHIDEE, HC/MOSES, MPI; DMI/ECHAM, Météo-France/ISBA.)
• The question was raised whether this general interface should be adopted in PRISM for the coupling between atmospheric model components and LSSs. If OASIS is preferred, its implementation of passing of fluxes needs to be made more efficient.
• There is a close coupling between the PBL and the LSS. If an implicit solver (e.g. LMD, UKMO, ECMWF, Météo-France) is used, the PBL needs to call the LSS. In the case of an explicit solver, LSS can be computed outside the PBL. A general interface from within PBL allows handling both of these solver alternatives.
• Concerning the coupling of an atmospheric model and LSS, it seems reasonable to have the same timestep and horizontal spacing in the PBL and in the LSS. The latter can, however, be finer than in the atmospheric component (cf. subgrid surface tiling). Multiple PBLs might be considered.
• From the atmosphere/PBL, a LSS can need/needs the following parameters: Net-SW(+radiation-related diagnostic quantities), downwelling LW, latent heats of sublimation and evaporation, rainfall, snowfall, subgridscale precipitation distribution, lowest atmospheric layer T, q, u, v, their sensitivity to sfc-fluxes and if implicit solver, sfc layer diffusivity (C_h).
• A LSS returns the following parameters to the atmospheric/PBL model: Sfc-albedo, radiative-T, sfc-?, E, LE, H, sfc roughness, displacement height, the ratio of actual-to-potential E, passive tracers and if explicit solver, C_h.
• LSS coupling in a climate system models implies also that there is a connection between LSS and ocean models via river routing and coastal flow, that there might be a connection between LSS and lake models via river flow and atmospheric chemistry and LSS via tracer fluxes.

Summary by Markku Rummukainen
 
 

Why obgc?

Modeling obgc addresses the fact that there is a coupling between climate (change) and the ocean natural C-cycle, through the physical and biological C-pumps in the ocean, leading to uptake, release and long-term storage of C in the world ocean. In addition, obgc is relevant to an end-user such as fisheries and in atmospheric chemistry in the form of DMS-fluxes to the atmosphere.

Models in Europe

Obgc models of the type of NPZD (nutrients, phytoplankton, zooplankton, detritus) are typically 1-D, simulate the biological production in the oceans, driven also by ocean transport in 3-D and solar/wind forcing from the atmosphere. Examples of models in use are HAMOCC, PISCES and HadOCC. The 3-D transport is derived from an OGCM.

Implications for coupling

• Presently, 1-way coupling between an OGCM and obgc is used. The transport (nutrients, reacting species) calculated for obgc is obviously related to the calculations in the OGCM, but otherwise the obgc-schemes are not associated to a specific OGCM.
• 2-way coupling might be considered in the future, e.g. obgc returning ocean color to an OGCM.
• Coupling of obgc is also needed to sea ice, atmospheric radiation and wind and could be implemented for LSS and river flow (nutrient transport).
• Obgc might return information to the atmosphere/atmospheric chemistry (CO₂, DMS)
• OGCM and obgc timesteps can be different.
• Different horizontal grids (staggering) are used in implemented OGCM-obgc pairs.
• There might be a parallel to the question of how to couple between an AGCM and atmospheric chemistry model, even though the number of transported species in an obgc is likely less than in atmospheric chemistry models. Also, so far there is no return flow of information from obgc to the OGCM (but chlorophyll might be used in the future?).

Retain OGCMs as is and build a tracer transport scheme into obgc-schemes. The coupling is then needed between the OGCM and the transport-obgc-scheme. This has the following advantages: offline-use of an obgc is easy, the OGCMs do not need to be modified and no interpolation is required in the coupling procedure. However, one needs to construct a separate transport scheme (different for each OGCM) to be included to the obgc-scheme.

Alternatively, the OGCM could be split in physical and dynamical modules and the latter used for transporting the biogeochemical species, thus avoiding the need to develop a transport model.

An obgc needs the following physical variables from an AGCM+OGCM: surface wind, solar radiation, ocean temperatures and salinity, sea ice, 3-D ocean transport (advection, convection, mixing).

2.5 Regional Climate Models

Summary by Sophie Valcke

The RCMs involved in PRISM are:

1. The Rossby Centre RCA, RCO and RCAO: a regional atmospheric, ocean and coupled atmosphere-ocean-sea_ice-hydrology climate model. RCA has HIRLAM dynamics but improved mesoscale (climate) physics.
2. The HadRM3H having the Met Office Unified Model dynamics and physics.
3. The HIRHAM4 from the DMI, having the HIRLAM dynamics and the MPI ECHAM4 physics.

The validation of RCMs is through perfect boundary-condition experiments (reanalysis data, e.g. ERA, as lateral boundary conditions) and comparison with observational datasets, preferably high-resolution ones. When RCMs are driven by results from a GCM, e.g. in climate change experiments, it is important to understand the behaviour/biases in the global model and how it affects the regional model results/performance.

Further developments of RCMs include going to even higher resolution (5-10 kms), two-way RCM-GCM coupling (i.e. the information calculated by the RCM feeds back to the GCM), and further refinements in RCM-RCM coupling for different components of the climate system (atmosphere, ocean, sea-ice, land, etc.).

After the discussion, the conclusions concerning the technical requirements for including RCMs in PRISM are the following:

1) RCM-RCM coupling:

• The physical interfaces implied in the coupling of different climate components in RCMs should not be different than those implied in the coupling of climate system components in global models. Technically, the coupler should be designed to allow regional coupling which does not bring strong additional constraints.

• Both one-way and two-way RCM-GCM model coupling should be allowed in PRISM. One-way coupling could be done on-line or off-line (given that the appropriate information was saved during the GCM run) using the same interface.
• The appropriate physical interfaces between the different global and regional components need to be defined in PRISM (for example, what is the coupling information that needs to be exchanged between an atmospheric GCM and a atmospheric RCM?). This will imply an interaction between the PRISM WP3h and all other relevant component WP (WP3b to WP3g).
• Technically, RCM-GCM coupling will probably ask for some additional coupler functionalities, such as automatic extraction of coupling field sub-domains, time interpolation, or 3-D spatial interpolation.

Summary by Sophie Valcke

In this session two additional discussion papers were presented (see Annex - B1, "Proposal for a PRISM coupling infrastructure" by Jan Polcher, and "Additional remarks on the PRISM infrastructure" by Martin Stendel (Annex B2).

Jan Polcher presented two important coupling concepts. The first concept concerns the execution flow of the models being coupled: this flow can be either "fixed" or "data-driven". In the fixed-flow approach, the model execution is controlled by a pre-defined hard-coded sequence of launches/calls to the different parts of the different component models. In the data-driven approach, each component model is launched/called once and its execution is automatically controlled (i.e. suspended/pursued) by the inputs needed from the other models at the different points in the code.

The second concept concerns the exchange of data: this exchange can be either "point-to-point", either "one-sided". In the point-to-point case, a source model sending coupling information has to know the

corresponding target model and sends them directly. In the one-sided case, when a component is ready to deliver some data (send operation) or when it requires some input data (get operation), it will first notify a data-broker that centralizes the information on all posted send and get. The data-broker will perform the matching between the different send and get from the different models and will ensure

that the exchange is performed accordingly.

From the discussion, it appeared that one-sided exchange of data should be preferred. This allows the models to remain independent from one another and gives more flexibility in the coupling exchanges. Concerning the model execution flow, the "data-driven" approach seemed to be conceptually preferable as each model remains a self-consistent code with as little modifications as possible compared to the off-line (non-coupled) mode.

Another point that came out in the discussion is the efficiency of the coupling, especially between two heavily coupled components, such as an atmospheric model and an atmospheric chemistry model between which a high number of 3D coupling fields should be exchanged at high frequency. One way to ensure a efficient coupling is to perform memory exchange rather than exchange through message passing. This type of exchange is naturally easier to implement if both models form only one executable. However to judge the importance of considering memory exchange rather than message passing exchange, a first evaluation of the overhead associated with message passing should be done for a typical heavy coupling.

There was also a discussion on having a global coupled system formed by a reduced number of different executables. One argument was that a reduced number of coupled executables is easier to manage for the operating system. That argument was however questioned by the constructors and has to be investigated furthermore.

3. Conclusion

The meeting provided an overview of the current state of climate modelling in Europe, and it identified a number of issues that have to be resolved before work on a joint model infrastructure can start. There is a particular urgent need for an inventory of variables that each model component needs as input, and the required frequency of refreshment.

Annex A Abstracts of Presentations

A1. The Atmosphere Models

Serge Planton, Météo-France, CNRM, Toulouse, France

Since the first simulation over one month performed with an atmospheric quasi-geostrophic model at GFDL by Phillips (1956), several tens of Atmospheric General Circulation Models have been developed for climate applications. Today, about 30 AGCM are involved in the Atmospheric Model Intercomparison Project (AMIP) and 4 in PRISM: ARPEGE-Climat (ECMWF + Météo-France/CNRM), ECHAM (MPI); LMDZ (IPSL/LMD); Unified Model (UKMO). Each model includes a dynamical core (grid point or spectral, eulerian or semi-lagrangian) and a physical parameterization package (radiation, clouds and liquid-water, large-scale condensation, deep and shallow convection, boundary layer processes, gravity-wave drag, land surface processes and hydrology, atmospheric chemistry).

Within the context of WCRP and IGBP scientific programmes, European and national projects, AGCMs are used to study climate variability, climate predictability and impact of human activities on climate. Some models are also used for other applications like weather forecasting or pollution transport simulation and forecasting. In stand-alone mode, the models are applied for the analysis of climate states at equilibrium and for sensitivity analysis. Coupled to other components, they are mainly applied for the analysis of transient climate, for long-range forecasting and for process studies. For research applications, the tendency is now for the investigation of a wider range of times scales (from intra-seasonal to multi-centennial) and of climate processes with lower signal to noise ratio.

A complete model evaluation implies a great variety of methods. The most classical consists in a direct comparison of AGCMs outputs to observations (climatologies, reanalyses, surface networks, satellites, …). But the AGCMs are also indirectly evaluated through the comparison of their parameterisations used in single-column models to data from observed field experiments or from Large Eddy Simulation or mesoscale models. An other common way of evaluation consists in intercomparing the different models integrated under the same protocol and using common diagnostics like in AMIP. Recently, new datasets, new diagnostics and new methodologies have been developed, increasing the spectrum of the models validation process.

Among the main questions arising from research in climate modelling, systematic errors is of main concern. A complete list is difficult to establish but some of them are widely spread among the ensemble of climate AGCMs. The main error sources are the parameterizations of sub-grid scale processes, orography, resolution and numerics. These errors imply the development of new schemes particularly in the field of clouds and radiative interactions (boundary layer clouds and cirrus), convection and gravity wave drag representation.

Increase model resolution is of great value for a better representation of climate and particularly climate variability and extreme events, as to improve the link between the resolved processes and the climate impact studies. But improving the model resolution implies parallel improvement of the representation of the climate feedbacks representation as demonstrated by recent research results. For regional studies, variable resolution models realise a good compromise between improved resolution and moderate computing time increase allowing ensemble simulations (but limited area models remain needed for very high resolution).

Model ensemble with a single model is useful to investigate uncertainty due to chaotic climate behaviour, when multi-model ensemble helps at investigating uncertainty due to the representation of physical processes. The analysis of AMIP simulations and the results of a few European projects show that multi-model ensemble have greater performances than individual model in term of mean, variability and forecast skill. They also show that dedicated multi-model ensemble simulations allow evaluating these parts of climate variability or climate sensitivity that are due to specific physical processes.

A2. Atmospheric Chemistry Models

Guy Brasseur, MPIM, Hamburg, Germany

Still to be inserted

A3. Developments in Ocean Climate Modelling

Claus Böning, Institut für Meereskunde, Kiel, Germany

Over the last decade or so the status of ocean modelling has changed quite dramatically. While the pioneering era of the 1960s-1980s was largely restricted to a few select laboratories possessing large computers, with global model development almost exclusively driven by the need for incorporating the ocean's role in climate, today's field has become quite diverse: a main reason being the much more prominent role of modelling in modern oceanographic research, due to a more widespread realisation of the increased skill of model solutions in mimicing observed oceanic features. Thanks to the efforts of more numerous groups and individuals, ocean model development has accelerated considerably in recent years, as reflected in an increasing number of sophisticated numerical methods, algorithmic choices and parameterisations for sub-grid scale physics: combined with better data sets for model initialisation and forcing, and an exploitation of the increases in computing power, several of these refinements have contributed to improve the models' abilities to simulate certain observed behaviours. Going along with these developments is a substantial increase in model diversity and available options for algorithms and parameterisation, and correspondingly, a growing need for closer examination of interdependencies between alternative sets of choices and more systematic inter-calibration of different model concepts.

Understanding the relative impacts of resolution, parameterisation and numerics on the performance of ocean circulation models is still in its infancy, particularly concerning the requirements for faithfully capturing the ocean's role in climate. The lecture will touch upon these issues, including the questions of the key processes that need to be simulated and of the critical modelling factors that currently limit their realistic representation. The perhaps most basic model consideration is about the compromise necessary between a coverage of the time and space scales involved in the physics of the global thermohaline circulation and its response to changing atmospheric conditions: bridging the gap between the two different model classes - coarse resolution ocean climate models covering the secular aspects of ocean-atmosphere interaction and eddy-resolving models aiming at capturing the transient response on decadal-interdecadal time scales - is not achievable in the near future. The implications for model development needs will be illustrated based on examples of model behaviours in different dynamical regimes. The discussion will also draw from the recommendations of the recent WOCE/CLIVAR Workshop on Ocean Modelling for Climate Studies (held at NCAR, 1998; ICPO Puvblication Series No. 28), and subsequent deliberations of the WCRP Working Group on Ocean Model Development (www.ifremer.fr/lpo/WGOMD); it is felt that particularly its recent move towards the definition of an Ocean Model Intercomparison Project (OMIP) - a 'Pilot Phase' has been launched recently - should be of some relevance to the coordinated experimentation being planned in PRISM.

A4. Sea-Ice Models

Helge Drange, NERSC, Bergen, Norway

Still to be inserted

A5. Present status of model development at the Frontier Research System for Global Change

Taroh Matsuno, FRSGC, Yokohama, Kanagawa, Japan

The Frontier Research System for Global Change (FRSGC) is a joint research project of the National Space Development Agency and Japan Marine Science and Technology Center, with objectives to found the basis for global change prediction. The FRSGC consists of six programs and in one of them (Integrated Modeling Research program), new high-resolution climate models are being developed to be run on the earth simulator which is now under construction as a sister program of the FRSGC. It will be completed by March 2002.

The models now being developed at FRSGC are:

1. T213/L50 spectral AGCM with 0.1° _0.1° /L(40) OGCM and a coupled model of the two. The ocean is expected to represent baroclinic eddies and resulting water transport explicitly.
2. Non-hydrostatic AGCM with the horizontal resolution of around 5 km by which meso-scale convective systems are treated explicitly.

A6. Land/Surface models

Jan Polcher

Full lecture available in pdf.

A7. Biogeochemistry models

Olivier Aumont

Still to be inserted

A8. Regional Climate Models

Markku Rummukainen, Swedish regional climate modeling program (SWECLIM) and Swedish Meteorological and Hydrological Institute (SMHI), SE-601 76 Norrköping, Sweden

Compared to the established role of general circulation models (GCM) in climate modeling, regional climate models (RCM) are a more recent development. Regional climate system models are now being worked on, with coupled atmosphere, ocean, lake and hydrology components. Along-held view of regional climate modeling has been that they can be used to further interpret global simulation results. A reduced geographical domain allows for a higher resolution and thus, for a more detailed representation of topographical features (mountains, lakes and lake systems, regional oceans and coastlines). Simulation of sub-GCM-scale variability in parameters such as snow cover and physiography also becomes feasible.

Regional climate modeling relies on global simulations, from which time-dependent lateral boundary conditions, including the large-scale sea surface, can be gained. However, as the climate solution is calculated anew within the regional domain, using a higher resolution and possibly different physical parameterizations than in thedriving global model, the possibility of the regional solution diverging from the global one arises. Such a divergence might either represent a worsening of biases already in the global simulation or an improvement in the simulation skill. In the first case, one reason for this might be the difficulty in "correctly" imposing the boundary forcing from a GCM on an RCM. Alternatively, there might be cases where(systematic) errors in the forcing GCM are ("correctly") amplified by stronger higher resolution forcing of the error structures, within the RCM. Even in the case where the simulation skill improves when put through an RCM, it is not evident that this is desirable as it makes the interpretation of the simulation chain more difficult.

In this talk, usefulness and problems associated with regional climate modeling are discussed. Examples are then given of two applications of regional climate modeling:

1. as a test bench for parameterization schemes in (all) climate models, using analyzed/re-analyzed lateral boundary forcing. Examples of the simulation of the diurnal cycle in cloudiness and convection in different regions are shown. How these in turn affect the coupling of atmospheric and ocean models is also discussed.
2. to incorporate high-resolution detail in GCM-simulations, in a physically-consistent manner. Examples of regional simulations, illustrating results appearing on sub-GCM-scale, are given for the North European region.

A9. The PRISM project

Gerbrand Komen, KNMI, De Bilt The Netherlands

PRISM ( PRogramme for Integrated earth System Modelling) is a project to be funded under the Fifth framework Programme of the European Union.

Following the recommendations of the European climate research community (Euroclivar, November 1998), PRISM will be carried out as a pilot infrastructure project for the establishment of a climate research network. The workplan foresees

1. the creation of a European management structure for developing, coordinating and executing a long-term programme of European-wide, multi-institutional climate simulations;
2. the development of a set of portable climate community models and associated diagnostic software under standardised coding conventions that can be accessed by all European scientists;
3. the execution of a first suite of joint simulations.

PRISM has about 20 participants among which leading climate centres, universities, meteorological services and computer vendors. The project will cover a period of three years, and consists of four phases: a definition phase, a development phase, an assembly phase and a demonstration phase.
 
 

A10. Assessment and review of 10 years of modular coupling in France

Eric Guilyardi, Center for Global Atmospheric Research, University of Reading, United Kingdom

The French community was the last major European group to tackle coupled climate modelling. Both based on other groups experimentation and specific goals and constraints, several choices were made at the beginning of the 1990s. A review of these scientific, technical and human choices is presented from pionneering experiments to todays mature programs. The development and use of the now popular OASIS coupler helped maintain the existing diversity of models and eased the set up of many different configurations of coupled ocean-atmosphere models.

An assessment of these choices after 10 years is given, including scientific strategy, assessment and exploitation, and the related technical set-up, production, development and maintenance issues, all of relevance for PRISM. The coupled european GCMs SINTEX (ECHAM/OPA) and HadOPA (HadAM/OPA) are presented and discussed within this assessment. This leads to a summary of future challenges for climate GCMs. Some key scientific issues and several recommendations for the PRISM project are proposed.

A11 The Development and Application of Climate Scenarios For Impact, Adaptation and Integrated Assessments

Mike Hulme, Tyndall Centre for Climate Change Research, UK

Results from climate model experiments have been used as the basis for climate scenario construction since the early 1980s. The earliest experiments used for this purpose were no more than so-called "2xCO2" sensitivity experiments (although some of these experiments are more than 12 years old now, their results are still used in some scenario development). The development of coupled ocean-atmosphere models and their use in transient climate change experiments in the early 1990s altered the way in which climate scenarios were developed and applied. More recently, multiply-forced experiments and ensemble simulations have again allowed a more sophisticated approach to be taken in the construction of climate scenarios. Regional climate model experiments have also been conducted since the early 1990s, but results from such experiments have rarely been used for scenario construction for a number of reasons which will be elaborated.

It is becoming increasingly important for climate scenarios to be developed as part of a broader activity of developing integrated scenarios of the future, encompassing social, technological, economic and environmental change. The work published by the IPCC in the Special Report on Emissions Scenarios has been particularly important in this regard. The talk will illustrate a number of conceptual frameworks that have been developed to allow integrated scenarios, including climate scenarios, to be developed and applied.

Three specific challenges facing climate scenario developers are those of (i) combining model results with observed climate data, (ii) the process of "downscaling" from coarse to fine resolutions, and (iii) the adequate representation or communication of uncertainty about future climates. Some important principles relating to each of these points will be described.

Finally, some attention will be given to the development of future climate scenarios in Integrated Assessment Models of climate change. A large number of such models now exist, yet there is a diversity of approaches taken by such models to the representation of future climate. All such methods, to some degree, rely upon the results of more specialist climate models.
 

A12. Coupled Ocean/Atmosphere Modelling: Key Physical Processes

Herve Le Treut, IPSL, Paris, France

The recent  IPCC report has emphasized the fact that the spread between climate models in terms of sensitivity is basically what it was 20 years ago: the global warming in response to a CO2 warming is still within an approximate range from 2 to 4.5 degrees. There is necessarily a strong link between this fact, and the continuing difficulty of the present generation of coupled models to maintain equilibrium: although flux corrections are less and less used, this is often obtained through an adequate tuning of some “harmless” parameters in the atmospheric codes (single scattering albedo, cloud radius droplet). But, of course, many of the feedback processes which maintain the equilibrium of the climate system are the same  which determine its sensitivity, and there is no reason to think that such a tuning is really “harmless”: it may only hide model errors in the representation of key processes. If there is no convergence in the quantitative results of the models, there has been over the years a blossoming of new parameterizations. Cloud representation has evolved from a situation where cloud fraction only was predicted, very often from simple statistical relations, to a situation where many models also handle consistently cloud water content, cloud optical properties and their modifications by the aerosol indirect effects. There is still some way to go in order to have a consistent representation of boundary-layer clouds, shallow convection clouds, and convective clouds, because the K-diffusion representations used in the former, the mass-flux approach used in the latter, are still difficult to reconcile, but this is also clearly the direction taken by the community – and this is also essential for an adequate representations of feedbacks in the intertropical area.

These are just examples, but they show that there is a clear need to preserve model diversity, and this is one important aim of PRISM. In later stages of PRISM development, however, one may question whereas maintaining this diversity is of equal importance for all the components of the atmospheric models. If we distinguish in the present codes a dynamical component, a radiative component, a cloud/convection/boundary-layer component, a chemical-transport component, and a surface component (which is now already a stand-alone component in the proposed architecture), we may then have a level of modularity which could be sufficient to organize and rationalize the diversity of the models in our community.
 

A13. Towards a coupled Global Environmental Model System

Tim Johns, Met Office, Hadley Centre for Climate Change Prediction, Bracknell, UK

Over the last 5 years, use of coupled climate models has proliferated amongst climate research worldwide. Such models are recognized as vital tools in the mission to advance understanding of the climate system and climate change. Following a brief review of scientific and political drivers behind their development and use, some examples will be given illustrating how coupled models have been applied to address topical issues relevant to the climate change debate and to help improve understanding. Building on recent model improvements and a range of such studies at the Met Office, a new 'Global Environmental Model' is currently being designed and built for future climate research. The design of this model and its role in the Hadley Centre's research plans will be discussed. Lastly, current Met Office thinking on requirements and technical approaches for achieving more flexible and efficient coupling model infrastructures (via 'FLUME' and 'PRISM') will be reviewed, with some speculation on the possible implications of these initiatives.

A14 (Keynote lecture) The role of Models in Climate Research
Brian Hoskins, Un. Reading, United Kingdom

It is clear that experimentation, simulation and prediction with climate models plays a central and integrating role in climate research. This talk emphasises the necessity for the close interaction between modelling, theory and observations for optimum progress in this research.

Observational data are used in the context of models and theory to give a view of the atmosphere. At one end of the scale this could be a prediction model and a 4D-Var analysis system and at the other end it could be a conceptual model of tropical convection and the deduction of precipitation from OLR. Models are used also to help design the observational system. Models can be used along with observations to extend the theory. These models may vary in the number of earth systems described, the complexity of the processes represented, and the extent and complexity of the domain. Theoretical understanding also plays a crucial role in the optimal use of models. It can give confidence in predictions. It can help experimental design. Finally it can give a diagnostic framework for viewing model output, for comparing models & observations and for the guidance of the improvement of models.

Annex B - Discussion papers

B1. A proposal for a PRISM coupling infrastructure

Jan Polcher

Available in pdf

B2. PRISM System Specification document Proposal Notes

Martin Stendel

Available in pdf

Annex C - Meeting programme

Monday 25/06 Arrival

Tuesday 26/06

09.00 Opening + short introduction to PRISM (Guy Brasseur)

Session 1 [Marie-Alice Foujols, chair]: Atmosphere

09.15 - 10.15 The atmosphere models (Serge Planton)

10.15 - 10.45 Discussion

11.00 - 12.00 Atmospheric Chemistry models (Guy Brasseur)

12.00 - 12.30 Discussion

Session 2 [Alan O'Neill, chair]: Ocean models

14.00 - 15.00 Ocean models (Claus Boening)

15.00 - 15.30 Discussion

15.45 - 16.45 Sea-ice models (Helge Drange)

16.45 - 17.15 Discussion

17.30 - 18.30 Keynote 1: Present status of model development at the Frontier Research System for Global Change (Taroh Matsuno)

Wednesday 27/06

Session 3 [Markku Rummukainen, chair]: Land/Surface and (Ocean) Biogeochemistry

09.00 - 10.00 Land/Surface models (Jan Polcher)

10.00 - 10.30 Discussion

11.00 - 12.00 Biogeochemistry models (Olivier Aumont)

12.00 - 12.30 Discussion

Session 4 [Sophie Valcke, chair]

Session 4.1: Regional Climate Models

14.00 - 15.00 Regional Climate models (Markku Rummukainen)

15.00 - 15.30 Discussion

Session 4.2 Other topics

15.45 - 16.15 The PRISM project (Gerbrand Komen)

16.15 - 16.45 Why and how to build climate coupled GCMs: assessment and review of 10 years of modular coupling (Eric Guilyardi)

17.45 - 17.15 Discussion

17.30 - 18.30 Keynote 2: Application of climate scenario's in impact modelling (Mike Hulme)

21.00 - 21.30 After DinnerTalk: The Earth Simulator, (Keiji Tani)

Thursday 28/06

Session 5 [Gerbrand Komen, chair]: Present and future state of coupled climate modelling:

09.00 - 10.00 Coupled ocean/atmosphere modelling: key physical processes (Herve Le Treut)

10.00 - 10.30 Discussion

11.00 - 12.00 Towards a coupled Global Environmental Model system (Tim Johns)

12.00 - 12.30 Discussion

Afternoon: Conference hike and discussion

Friday 29/06

Session 6 [Guy Brasseur, chair]: Modelling the Global Climate System in PRISM:

09.00 - 10.00 Keynote 3: The role of models in climate research (Brian Hoskins)

10.00 - 11.00 Session Summaries (session chairs)

11.00 - 12.30 Discussion (Guy Brasseur)
 
 

Annex D - List of participants
 

Aldrian, E.
Max-Planck-Institut für Meteorologie
Hamburg
Germany

Aumont, O.
LODYC/IPSL/UPMC
Paris
France

Böning, C.
Universität Kiel
Kiel
Germany

Böttinger, M.
Deutsches Klimarechenzentrum GmbH
Hamburg
Germany

Brasseur, G.
Max-Planck-Institut für Meteorologie
Hamburg
Germany

Cuijpers, H.
KNMI
De Bilt
The Netherlands

De Montety, A.
Université catholique de Louvain
Louvain-la-Neuve
Belgium

Declat, D.
CERFACS
Toulouse
France

Dijkstra, H.
IMAU
Utrecht
The Netherlands

Döscher, R.
Swedish Meteorological and Hydrological Institute (SMHI), Rossby Centre, Norrköping
Sweden

Drange, H.
Nansen Environmental and Remote Sensing Center, Bergen
Norway

Fairhead, L.
Laboratoire de Météorologie Dynamique du CNRS, Paris
France

Fichefet, T.
Université catholique de Louvain
Louvain-la-Neuve
Belgium

Foujols, M-A.
Pôle de Modélisation de l'IPSL
Paris
France

Gayler, V.
Max-Planck-Institut für Meteorologie
Hamburg
Germany

Gordon, C.
Hadley Centre H011
Bracknell, Berkshire
United Kingdom

Griggs, D.
Hadley Center for Climate Prediction and Research,
Bracknell, Berkshire
United Kingdom

Guilyardi, E.
CGAM
Reading, Berkshire
United Kingdom

Hagedorn, R.
ECMWF
Reading, Berkshire
United Kingdom

Haywood, A.
University of Reading
Reading, Berkshire
United Kingdom

Hoskins, B.
University of Reading
Reading, Berkshire
United Kingdom

Hulme, M.
University of East Anglia
Norwich
United Kingdom

Johns, T.C.
Hadley Center for Climate Prediction and Research,
Bracknell, Berkshire
United Kingdom

Komen, G.J.
KNMI
De Bilt
The Netherlands

Latour, J.
Fujitsu European Centre for Information Technology, Toulouse
France

Le Treut, H.
Laboratoire de Météorologie Dynamique du CNRS, Paris
France

Legutke, S.
Max-Planck-Institut für Meteorologie
Hamburg
Germany

Linstead, C.
Potsdam-Institut für Klimafolgenforschung (PIK), Potsdam
Germany

Lorenz, P.
MPIfM
Berlin
Germany

Luthardt, H.
Max-Planck-Institut für Meteorologie
Hamburg
Germany

Matsuno, T.
Frontier Research System for Global Change (FRSGC), Yokohama, Kanagawa
Japan

Meier, H.E.M.
Swedish Meteorological and Hydrological Institute (SMHI), Norrköping
Sweden

O'Neill, A.
University of Reading
Reading, Berkshire
United Kingdom

Planton, S.
Météo-France
CNRM, Toulouse
France

Pohlmann, H.
Max-Planck-Institut für Meteorologie
Hamburg
Germany

Polcher, J.
Laboratoire de Météorologie Dynamique du CNRS, Paris
France

Redler, R.
NEC Europe Ltd.
Sankt Augustin
Germany

Ritzdorf, H.
NEC Europe Ltd
Sankt Augustin
Germany

Rummukainen, M.
Swedish Meteorological and Hydrological Institute (SMHI),
Norrköping
Sweden

Salas y Melia, D.
Météo-France
CNRM, Toulouse
France

Semeena, V.
Max-Planck-Institut für Meteorologie
Hamburg
Germany

Semmler, T.
Max-Planck-Institut für Meteorologie
Hamburg
Germany

Stendel, M.
Danish Meteorological Institute
Copenhagen
Denmark

Tani, K.
Earth Simulator Research and Development Center, Yokohama, Kanagawa
Japan

Thorsen, S.V.
Danish Meteorological Institute
Copenhagen
Denmark

Valcke, S.
CERFACS
Toulouse
France

Van Velthoven, P.F.J.
KNMI
De Bilt
The Netherlands

Wegner, J.
Max-Planck-Institut für Meteorologie
Hamburg
Germany