One of the major developments in COSMO within the last years was the common COSMO-ICON Physics. The idea behind that development was to reduce the amount of maintenance and therefore to use the same source code packages in COSMO and in ICON. Besides the unification of the packages between COSMO and ICON, the technical modifications to use a blocked data structure were a big challenge.
With COSMO-Model Version 5.05 all physics packages are now in the blocked data format. The last two packages included from ICON, which are activated in COSMO Version 5.04e, are the turbulence and the surface schemes. The experiments done at DWD showed heterogeneous results, which might be due to the fact, that the turbulence and the surface scheme have significant differences to the schemes used up to now in COSMO. Alas the COSMO performance using the new ICON parts is worse than using the "old" schemes still in ijk-data format. One reason for this can be the fact that COSMO cannot take over all modifications from ICON, as for example the treatment of external parameters or the tile approach.
DWD therefore decided to choose a configuration, with which the behaviour of the former packages in ijk-data format can be reproduced rather closely. Of course, it is also possible to choose a configuration which is closer to the ICON like setup.
In the following we want to document these different setups: the conservative (or old-style COSMO physics) and the advanced (or new ICON style physics) approach. We start with an overview of the physical packages in COSMO and ICON, and which ones are shared. The main differences in these two setups are in the turbulence scheme and in the soil model TERRA. The different namelist settings for these schemes are discussed separately. Also in INT2LM there are necessary namelist settings to get the ICON-like behaviour of the physics.
The following table gives an overview on the different parameterizations and which packages are used in COSMO and in ICON.
| Scheme | COSMO | ICON | blocked format implemented in COSMO Version |
| Microphysics | Prognostic water vapour, cloud water, ice, rain, snow, graupel (Doms, 2004; Seifert 2010) | 5.01 | |
| Radiation | Ritter-Geleyn δ two-stream | RRTM | 5.03 |
| Subgrid Scale Orography | Lott and Miller (1997) | 5.04e | |
| Turbulence | prognostic TKE scheme (Raschendorfer) | 5.04a | |
| Surface Schemes |
TERRA (Heise and Schrodin, 2002) FLake (Mironov) SeaIce (Mironov) |
5.04e | |
| Convection | Tiedtke (deep or shallow) | 5.04b | |
| Tiedtke-Bechtold (only optional in COSMO) | |||
While the microphysics and the SSO scheme already were the same in ICON and in the old COSMO ijk-physics, there have been considerable modifications in ICON for the turbulence and the soil model TERRA. In the following we document how the namelist configuration can be chosen to either use a setup for reproducing the old ijk-schemes (as close as possible) or to use the new ICON modifications.
In the turbulence scheme most modifications in ICON have been implemented as options. Therefore it is possible to choose between the former COSMO or the new ICON implementation. Several switches have been implemented for that. But some of these switches are not namelist variables but are hard-coded internally. In order to avoid re-compilation, a new namelist variable loldtur has been implemented in the COSMO namelist group /PHYCTL/ to set these internal switches accordingly. In addition, some namelist variable have also to be chosen differently.
The following table lists the namelist variables in /PHYCTL/ that have to be set for choosing the "OLD" (COSMO-like) configuration or the "NEW" (ICON-like) setup of the turbulence scheme.
| /PHYCTL/ | OLD | NEW | Explanations | Default |
|---|---|---|---|---|
| loldtur | .TRUE. | .FALSE. | If .TRUE. internal switches are set to use former COSMO implementation of special algorithms in the turbulence scheme. | .FALSE. |
| itype_vdif | -1 | -1 | To choose the type of vertical diffusion. At the moment there are still problems using itype_vdif≥0, which has been implemented with the new ICON version. Therefore only itype_vdif=-1 can be recommended for both configurations (vertical diffusion done at the end of the dynamics). | 1 |
| ltkeshs | .FALSE. | .TRUE. | If .TRUE., a separate horizontal shear production for TKE is considered. This was not the case in the OLD COSMO configuration. | .TRUE. |
| itype_sher | 0 | 1 | In the new code, itype_sher=1 would also consider a horizontal shear correction. As no horizontal shear production for TKE is computed (see above), we have to set itype_sher=0 for the OLD COSMO configuration (only vertical shear of horizontal wind). | 0 |
| imode_tran | 1 | 0 | imode_tran=1 sets the mode of the TKE equation in the transfer scheme in the same way as it was done in the OLD COSMO implementation. ICON uses value 0. | 0 |
| icldm_tran | 0 | 2 | icldm_tran sets the mode of cloud representation in the transfer scheme. The old COSMO implementation did not consider clouds, this is chosen with value 0. In ICON (value 2) also a sub grid (turbulent) condensation is considered. | 2 |
| lemiss | .FALSE. | .TRUE. | icldm_tran Read an external parameter: a map for the thermal radiative surface emissivity. Also see INT2LM below. | .FALSE. |
| itype_aerosol | 1 | 2 | icldm_tran To choose Tanre (1) or Tegen (2) aerosol climatology. Also see INT2LM below. | 1 |
Also in the namelist group /TUNING/ there are some variables that are chosen differently in the ICON setup. At least compared to the settings that were used at DWD for COSMO-DE or for the former COSMO-EU. Of course the choice of these variables depend on special configurations and domains. The following table lists the variables that are now chosen at DWD for COSMO-D2 ("OLD") and for ICON ("NEW").
| /TUNING/ | OLD | NEW | Explanation | Default |
|---|---|---|---|---|
| tkhmin | 0.4 | 0.75 | Minimal diffusion coefficients [in m2/s] for vertical scalar (heat) transport. | 0.75 |
| tkmmin | 0.4 | 0.75 | Minimal diffusion coefficients [in m2/s] for vertical momentum transport. | 0.75 |
| rat_sea | 20.0 | 7.0 | Ratio of laminar scaling factors for heat over sea and land. | 10.0 |
| pat_len | 500.0 | 750.0 | Effective length scale of subscale surface patterns over land [in m]. | 100.0 |
| tur_len | 150.0 | 500.0 | Asymptotic maximal turbulent distance [in m]. | 500.0 |
| a_hshr | 1.0 | 2.0 | Length scale factor for separate horizontal shear production. | 1.0 |
| c_soil | 1.0 | 1.75 | Surface area density of the (evaporative) soil surface. | 1.0 |
Also in TERRA there have been considerable changes in the ICON version compared to the OLD COSMO implementation. Most of these modifications are not optional, so there is no way to mimic the OLD behaviour. Only for some algorithms, which already had several options, new ways have been implemented, which could be chosen now. Note that not all ICON options can be used in COSMO due to a different treatment of external parameters and the ICON tile approach.
The following table lists the namelist variables with which new ICON options could be chosen. The column "OLD" again lists the COSMO-D2 settings from DWD, which correspond to the former configuration of TERRA, while the column "NEW" lists the ICON settings. DWD did not make enough tests (independent from the turbulence settings) to judge the performance of the ICON settings in COSMO. But everybody is free to do so!
| /PHYCTL/ | OLD | NEW | Explanation | Default |
|---|---|---|---|---|
| itype_evsl | 2 | 4 | Type of parameterization of bare soil evaporation. Option 4 is the new implementation by Jan-Peter Schulz. Note that option 1 is no more available. | 2 |
| itpye_heatcond | 1 | 3 | Type of soil heat conductivity: Option 1 uses an average soil moisture. New option 3 uses a special (hardcoded) tuning factor to indirectly account for the impact of vegetation, which does not depend on soil moisture. | 1 |
| itype_root | 1 | 2 | Type of root distribution: Option 1 assumes a uniform distribution used in the OLD COSMO configuration. Option 2, which assumes an exponential distribution, is used in ICON. Please note, that if you want to use option 2, you also have to use itype_rootdp=4 in the INT2LM! | 1 |
| cwimax_ml | 1.0E-6 | 0.0005 | cwimax_ml is used as a scaling parameter for maximum interception storage. In the OLD COSMO implementation this was a constant with value 1.0E-6. Now it is a namelist variable. ICON uses a value of 5.0E-4 = 0.0005. | 1.0E-6 |
| idiag_snowfrac | 1 | 20 | Method for diagnosis of snow-cover fraction. (Further explanation necessary). | 1 |
| lstomata | .FALSE. | .TRUE. | Minimum stomata resistance: if lstomata=.TRUE., an external parameter is read for the minimum stomata resistance. Note that you also have to run INT2LM then with lstomata=.TRUE.! | .FALSE. |
Some of the different settings in COSMO and ICON come from different treatment of external parameters. While ICON computes the external parameters within the model, COSMO gets its external parameters from INT2LM. For a comparable treatment in COSMO and ICON the following namelist variabes in INT2LM could be changed (in the group /CONTRL/.
The following table lists the external parameters, the namelist variable to control their behaviour and the values for the OLD COSMO and the new ICON settings together with some short explanations.
| Parameter | Controlled by | OLD | NEW | ||
|---|---|---|---|---|---|
| PLCOV | itype_ndvi | 0 | Plant cover and leaf area index for the COSMO-Model and for a special day are produced by using the data sets for vegetation and for rest and modify these with an annual cycle. | 1 | Plant cover and leaf area index for the COSMO-Model and for a special day are produced by using only the data set for vegetation and an averaged ndvi ratio. This ndvi ratio is computed as a weighted mean between monthly mean values, which are taken from the external parameter data set for the COSMO-Model. |
| LAI | |||||
| ROOTDP | itype_rootdp | 0 | Input from external parameter for the COSMO-Model is taken and modified with an annual cycle. | 4 | Input from external parameter for the COSMO-Model is taken without modifications. |
| AER_SO4 | itype_aerosol | 1 | No additional Fields for aerosol types are read from the external parameters. The COSMO-Model has to run with constant values for the aerosol distribution on rural areas, urban areas, desert areas or the sea (Tanre aerosol climatology). | 2 | Additional external parameters for the aerosol types of sulfate, mineral dust, organic, black carbon and sea salt are read as monthly mean values. Actual values for the current day are computed and given to the COSMO-Model (Tegen aerosol climatology). |
| AER_DUST | |||||
| AER_ORG | |||||
| AER_BC | |||||
| AER_SS | |||||
| RSMIN | lstomata | .FALSE. | No external parameter is taken. A constant value for all grid points is determined in the COSMO-Model. | .TRUE. | Take a map from the external parameters for the minimum stomata resistance of plants. |
| EMIS_RAD | lemiss | .FALSE. | No external parameter is taken. A constant value for all grid points is determined in the COSMO-Model. | .TRUE. | Take a map from the external parameters for the thermal radiative surface emissivity. |
| Note: Usage of SSO_STDH and SSO_SIGMA |
| The new turbulence scheme now needs the external parameter SSO_STDH and the new scheme for TERRA needs SSO_SIGMA. Therefore the namelist variable lsso should be set to .TRUE. in any case when running INT2LM. |