The PRISM project has defined six subsystems of the earth's climate system which define six categories of component models. These are models of the general circulation of the atmosphere (AGCMs) and of atmospheric chemistry (ACMs), models of the ocean general circulation (OGCMs) and of the marine bio-geo-chemistry (BGCMs), sea ice models (SIMs), and models for land surface processes (LSMs).
The most recent AGCM of the MPI-Met is ECHAM5. It runs in a coupled configuration with the OGCM MPI-OM, also recently developed at the MPI-Met. These two component models define the base of the MPI-Met earth system model and also of the COSMOS earth system model, which is the German community physical climate model. Optionally, the climate model is comprising the BGCM HAMOCC, also developed at the MPI-Met. A sea ice model is embedded in the ocean model code and will therefore not be addressed as a separate component model in the present report. For a precise definition of component models read the PRISM handbook on the Standard Compilation Environment (Legutke and Gayler (2004)). The inclusion of a comprehensive land surface and chemistry component are present research activities at the institute. The adaptation of these models to the PRISM system is in work and will be documented in a later edition of this report.
In the following sections each of the MPI-Met component models used within the PRISM system is briefly described. Special emphasis is put on those aspects that are directly related to the coupling physics.
ECHAM5 is the 5th generation and most recent model of the ECHAM AGCM series developed at the MPI-Met. Depending on the configuration, the model resolves the atmosphere up to 10 hPa for tropospheric studies or up to 0.01 hPa for middle atmosphere studies (often referred to as MAECHAM5). The middle-atmosphere version is not yet used in a PRISM coupled model.
Prognostic variables are vorticity, divergence, temperature, surface pressure, specific moisture, cloud water, cloud ice, and width and skewness of the cloudiness distribution. The prognostic equations for vorticity, divergence, temperature, and surface pressure are solved by the spectral transform method. The prognostic equations for specific moisture and cloud variables are solved in grid point space using wind fields derived from vorticity and divergence.
ECHAM5 includes a land surface scheme which is implicitly coupled to the atmosphere (Schulz et al. (2001)). As mentioned above, a sea ice model is part of the ocean model MPI-OM.
Grid cells of ECHAM5 can be partially covered by land, ocean, and sea ice. Surface fluxes of reflected solar and outgoing long wave radiation, turbulent fluxes of sensible and latent heat, moisture and momentum are calculated separately for the land, the sea ice, and the open ocean fraction of each grid cell using the different conditions and characteristics (e.g. skin temperature, roughness parameters) defined for each of the surfaces. At a specific height above the ground, the blending height (Groetzner et al. (1996)), all conditions are assumed to be horizontally homogeneous. The blending height is related to the advective space scale. The concept is based on the underlying assumption that the scales of the surface conditions are smaller than the grid scale. This is also an assumption underlying the sea ice rheology (Hibler (1979)). The spatial scales of the sea ice surface (ice floes) vary between some centimetres and about 10 km. In its default version, the model is formulated on 19 or 31 layers defined in hybrid sigma-pressure vertical coordinates with the top level at 10hPa. For both discretisations the lowest computation level (blending height) is about 30m.
Depending on the area fraction of land in a grid cell, the atmosphere uses either the land flux (if the land fraction is 
50 %), or the ocean/sea ice flux if the ocean fraction is the larger. However, all fluxes calculated over sea water or sea ice are used for the transformation of the fluxes from the atmosphere to the ocean model grid in order to better represent the coastal gradients and also to improve the flux conservation. The latter has also motivated the use of the partial land/sea cell partitioning for the atmosphere grid. The atmosphere coastal cells are locally adapted so that the coast lines of both grids define the same global ocean and continental surface areas. The concept is illustrated in Figure 2.1 sketching the surface layers of the atmosphere and the ocean with the heat and freshwater flux components and their interpolation to the ocean grid. The check pattern arrows are ignored for the budget calculations in the atmosphere, however they are used for the interpolations between the grids. The ocean does not have partial cells. The transformation of the surface conditions from the ocean to the atmosphere grid is based on a surface averaging method (compare section 3.2.1).
 |
The surface albedo on land depends on the specified background albedo, the forest fraction of the cell, the snow depth on the ground and canopy, and of the surface slope and temperature (Röske (2001)). On sea ice, it is a function of skin temperature, while on sea water it is constant. The function over sea ice depends on the existence of snow.
The model contains a scheme to simulate the lateral water flows on the land and discharge into the oceans (Dümenil and Todini (1992) and Hagemann and Dümenil (1996)). Precipitation onto grid cells being part of the continental ice sheets (Greenland and Antarctica) normally accumulate over time since sublimation is usually much smaller than precipitation.
An ice sheet model which simulates meltwater surface runoff and discharge or mass discharge by ice streams or glacier calving is not yet coupled to the MPI-Met PRISM model. In order to close the freshwater budget in the coupled system, the precipitation over the ice sheets is accumulated during each coupled time step and is then distributed to the coastal ocean cells and passed to the ocean together with the river and continental boundary runoff. The latent heat of fusion corresponding to the discharged water volume is passed with the net heat flux, assuming that all discharge from the ice sheets happens as frozen fresh water at freezing point.
The model and the climate it simulates is described in Roeckner et al. (2003) and Roeckner et al. (2004).
Three dimensional spectral initial data for the prognostic variables vorticity, divergence, temperature, and specific humidity are read from file AresAlev_jan_spec.nc![[*]](file:/usr/lib/latex2html/icons/footnote.png){kind=icon}
. The data represent January values.
The data, and therefore the file name, depends on the horizontal grid (Ares) and the number of vertical levels (Alev).
January initial data for the soil scheme are read from file AresOres_jan_surf.nc. In addition, the file contains the grid definition, masks and orography as well as surface data such as background albedo, vegetation type, soil characteristics, and surface roughness. Since the model formulation allows for partial land coverage in a cell, which is determined in order to fit the coast lines of the ocean and the atmosphere grids, this file depends on both, the atmosphere's and the ocean's (Ores) horizontal resolutions.
 |
Climatological monthly land surface data of vegetation ratio (AresOres_VGRATCLIM.nc) and leaf area index (AresOres_VLTCLIM.nc) have been interpolated from a 1km data set (Hagemann (2002)) to the model grid and (partial) land/sea masks. The files therefore depend on the atmosphere and the ocean model grid. The climatological monthly land surface temperature data (Ares_TSLCLIM.nc) depend on the atmosphere grid only.
Most tracer gases are well mixed in the standard model version. They are prescribed with constant mixing ratios of 438 ppm CO
, 1.65 ppmv CH
, 305 ppbv NO, 0.280 ppbv CFC11 and 0.484ppbv CFC12. When the middle atmosphere version is used, CH and NO can be prescribed as horizontally constant vertical profiles with decreasing values in the middle atmosphere.
In the standard version with no comprehensive ACM included, ozone is specified from a climatology. The monthly and zonal mean ozone concentration profiles are conservatively interpolated to the model time and levels (Fortuin and Kelder (1998)). The data do not depend on the vertical resolution or on the land/sea mask. The file name is Ares_O3clim2.nc.
The hydrological discharge model runs on a 0.5 degree geographical grid independent of the atmosphere grid. It reads time independent parameters (e.g. river direction) and initial (January) data from hdpara.nc and hdstart.nc. Both files are therefore independent of the model grid.
The file surrta_data contains input data used by the radiation scheme.
The model initially reads sea surface temperatures (SST) from a file called Ares_amip2sst_clim.nc. In the coupled configuration the data is overwritten by values provided from the ocean model at the start of the simulation.
ECHAM5 generates one output file per output event (triggered by the namelist variable trigfiles). It contains model raw diagnostic data. The data are either actual values or time averages of the period specified by the namelist event-variable putdata. The file name is echamid_expid_yyyymm.dd[.nc]. The echamid is a number, that can be attributed to the model user running the experiment (see 4). The string expid represents the experiment ID, yyyy the year of the first time level in the file, mm the first month and dd the first day. The appendix .nc is used when the output format is netCDF, which is the PRISM standard format. Optionally, GRIB can be used as output file format.
|
In contrast to the ocean model MPI-OM, ECHAM5 does not use cpp flags to (de)activate optional compilation of physical parameterisations. Cpp flags are only used to configure the model for different platforms or for different coupled constellations (e.g. whether or not HAMOCC is part of the coupled model). These cpp flags are listed in table 5.1.
ECHAM5 reads five namelist files, named columnctl, dynctl, physctl, postctl, runctl and sdsctl. In the PRISM model constellation, only the namelists runctl and sdsctl are used to read parameters that differ from the default values set in the source code. These parameters are used for model controlling, not for the specification of model optional physics. They are listed in tabel 2.3.
|
MPI-OM is the global ocean general circulation model (OGCM) of the MPI in Hamburg. It is based on the primitive equations and utilises the usual assumptions for large scale ocean models (hydrostatic, Boussinesq). The model is formulated on an Arakawa C staggered (Arakawa and Lamb (1977)) horizontal curvilinear grid. The grid is generated by a conformal mapping of a global geographical grid which allows to place the poles at any point of the globe and can accommodate varying resolution in space (see figure 2.3). Other features are the use of geo-potential vertical coordinates, a free surface which allows to specify freshwater fluxes at the surface, partial(height) bottom grid cells which makes the resolution of the bathymetry independent of the number of layer or their vertical spacing, and an embedded dynamic-thermodynamic sea ice model with viscous-plastic rheology(Hibler (1979)) and snow cover.
The model formulation includes options for isopycnal diffusion, the Gent-McWilliams eddy parameterisation, a parameterisation of slope convection, and a choice of higher-order advection schemes, as well as a parameterisation of mixed layer deepening. The model and its climatology are described in Marsland et al. (2003).
 |
Geographical positions of the grid cells are read from the file anta_Ores.ext8 with Ores being the Ocean horizontal grid acronym. Other information on the model grids, such as grid point separation, masks, and bathymetry are stored in arcgri_Ores.ext8 and topo_Ores. BEK_Ores contains specifications on ocean basins used for diagnostics.
MPI-OM needs for its initialisation three dimensional data of potential temperature and salinity on the model grid. They are read from INITEM_Ores.Olev.ext8 and INISAL_Ores.Olev.ext8 (Levitus et al. (1998)). Olev reflects the number of vertical levels.
If the surface salinity is restored to climatological values, monthly mean data are read from SURSAL_Ores.ext8 (Levitus et al. (1998)).
PRISM provides an infrastructure for coupled models as well as for stand-alone component models. If MPI-OM runs in stand-alone mode, it does not receive forcing fields from the atmosphere model via OASIS. Instead, daily climatological near-surface conditions derived from 15 years of ECMWF re-analysis data in the OMIP project (Röske (2001)) are used to force the ocean. The input data comprise total cloud coverage, precipitation, solar radiation, surface temperature and dew point, 10m wind speed and wind stress ( files GICLOUD, GIPREC, GISWRAD, GITDEW, GITEM, GIU10, GIWIX, GIWIY). Surface fluxes of heat and evaporation are calculated with the help of bulk formulas. In the forced mode a 30 day calendar is used.
Climatological monthly runoff data are provided as well in files runoff_obs and runoff_pos. The latter contains latitude and longitude of the discharge positions. The corresponding grid cell indices are calculated in the model for the actual grid. The file therefore does not depend on the grid.
|
|
MPI-OM generates a large number of diagnostic output files. The time averaging is controlled by the namelist variable IMEAN. Each code is written into a separate file named fort.unit according to the unit the file is written to. The file format is EXTRA. At the end of each run a tar file is created from the averaged data files with is named fort.date_enddate.tar where date represents the date of the first and enddate of the last day of the run. In addition a file containing time series is generated from the data of each model time step. The content is described in detail in routine DIAGNOSIS.F90 of the source code. The file is saved as TIMESERdate_enddate.
The MPI-OM source code contains a number of cpp flags for conditional compilation of different parameterisations of the model physics. A list of those flags that are activated for the compilation of the PRISM model can be found in table 2.6. Cpp flags related to the configuration for different platforms or controlling of the coupled model run are given in table 5.1.
|
In addition to cpp flag (de)activation, it is possible to tune the model by modifying a number of namelist variables in OCECTL. These variables together with the values set for PRISM coarse resolution experiments are listed in table 2.7. Note that optimised values may change with resolution and forcing data as well as with the coupled configuration. Also included are namelist values which control the flow of the experiments.
|
HAMOCC5 is a model of the ocean carbon cycle of the NPZD class including a semi-labile pool of dissolved organic carbon. The basic version contains seven oceanic aqueous tracer which are transported by the ocean's advection and mixing routines and diffusively exchanged with the sediment scheme. Particulate matter (organic carbon, calcium carbonate, silicate) accumulates in the sediment layers. Dissolution is also possible. The biological model is based on a Redfield stoichiometry for organic material. Phytoplankton growth is described by a Michaelis-Menten kinetics with growth rates limited by temperature, wind speed, vertical mixing, and light intensity. The model is described in detail in Kriest et al. (2004) and in Maier-Reimer et al. (2005).
HAMOCC runs on the MPI-OM grid. It receives all information about the model geometry in a parameter list from the ocean model. It needs only one input file, INPDUST_Ores which contains monthly dust fluxes obtained by nudging the ECMWF ERA15 data set into the ECHAM4/T42/L19 model (Timmreck and Schulz (2004)).
|
The output files of HAMOCC are written in netCDF file format.
Monthly mean output data is saved in file bgcmean_date_enddate.nc. The naming rule concerning the time period the file spans is the same as that for the MPI-OM output files (section 2.2.1). The data is stored as n+1 dimensional arrays where n is the dimension of the variable, the additional dimension is for the months. The file also contains data averaged over the full run period.
|
The averaging period of the data in timeser_bgc_date_enddate.nc is set in the namelist by specifying parameter nfreqts1. The file contains station time series. The positions of the stations (grid cells) is also specified via namelist input.
HAMOCC is a submodel called from the MPI-OM ocean model (section 2.2). According to the PRISM coding rules for such models (which in a more general context are called packages), the model must get all the information it needs from the calling model through subroutine parameter lists. This rule is not completely kept for MPI-OM and HAMOCC. The module MO_COMMO1 from MPI-OM is used in two HAMOCC routines. Thus the specification of the grid defining cpp flag is necessary. Besides, the flag GMBOLUS has to be activated in order to define the same data blocks used in both models. This will be changed soon. All other cpp flags are used to control the model flow. They are described in 5.4.
The namelist of HAMOCC contains variables controlling the model flow. The variables and a short explanation are listed in table 2.10. It is possible to define stations (model grid cells) for which time series are printed.
|