Providing a traceable, modular spectrum of models

We will construct a framework to investigate and improve traceability of simulated climate, ocean circulation, and other key variables, within a spectrum of GENIE fast ESMs and between that spectrum and an overlapping spectrum of higher complexity Hadley Centre models (the modularity of which is being tackled in the SIAMESE proposal) (see Figure). The two model spectra will overlap over a limited range of spatial resolutions, thus aiding inter-comparison. However, the constituent models still differ in underlying physics and process representation. Traceability across that gap in process representation will be tackled by assimilating the output of the slower models in the faster models.

GENIE-PRISM integration: Modularity is already central to GENIE but the ability of the framework to couple full complexity components will be enhanced by adoption of PRISM standards. Use of the PRISM environment (coupler, compiling and execution, GUIs) for GENIE components will be investigated. Particular attention will be paid to the extra overheads this adds to computations and recommendations to the PRISM development teams will be made accordingly. The potential for performing Grid runs (including ensembles and parallel processing) within the PRISM framework will also be explored using GENIE expertise and technologies.

Ocean: Horizontal model resolution has been argued to be important for the stability of the thermohaline circulation (THC). Furthermore, the fast GOLDSTEIN ocean model uses frictional geostrophic equations in contrast to the full primitive equations, and it is important to know how this affects its behaviour. Hence different resolutions of GOLDSTEIN with the following horizontal grids will be included in our extended GENIE framework:

• (i) 18x18 (original),
• (ii) 36x36 (current standard resolution in GENIE),
• (iii) 72x72 (current "high" resolution),
• (iv) 96x72 (as in FAMOUS),
• (v) 288x144 (as in HadCM3).

(i)-(ii) have a sine of latitude grid whereas (iii)-(v) use constant increments of latitude, as already developed for (iii) in an ongoing RAPID project. Vertical resolutions of GOLDSTEIN from 8 to 32 layers are already in use and the vertical structure can be matched to that of FAMOUS and HadCM3. The high resolutions will enable inter-comparison of (iv) with FAMOUS and (v) with HadCM3 on the same grid, and assimilation of output from FAMOUS and HadCM3 into the matching resolutions of GOLDSTEIN. Traceability through decreasing resolutions of GOLDSTEIN is more transparent because of the common physics. Retuning of parameters is required but data assimilation toward a common target can be used.

Atmosphere: The GENIE dynamic atmosphere model (the IGCM) is a particularly flexible tool for exploring traceability to the Hadley suite of models, and to fully investigate the role of spatial resolution. Our standard resolution (T21, L8) is higher horizontal resolution than the FAMOUS model, and it is straightforward to increase the IGCM to HadAM3 and even HadGEM resolution. For a range of resolutions, we will apply our optimal tuning techniques to ensure that, at each resolution, the model is equivalently tuned. We will thus be uniquely placed to clearly identify resolution effects from poorly tuned models. Moreover, we will also apply optimal tuning to alternative physical parameterisations, such as convection and cloud cover. In addition, a gap in dimensionality, complexity and speed exists between the simple 2D EMBM and the dynamical 3D atmosphere models used within GENIE and full complexity GCMs. We will use our fast IGCM to both tune the EMBM and improve the underling parameterisations. Hence we will generalise the framework to ensure that it can couple other types of atmosphere model based on fundamentally different physics.

Toward a common framework: By experimenting with use of the BFG to couple the natural system components of GENIE we will complement FLUME activity and the SIAMESE proposal in working towards a common community PRISM-based ESM framework.

Technological Challenges

Data Assimilation