Documentation of the Changes in the COSMO-Model
Version 5.04g1

As testing with the new version is still going on without real satisfying results, also the development goes on. Here is the next development version.

Contents:

1. Reproducing the Behaviour of the old COSMO Physics
2. Updated TERRA to the latest ICON Version
3. GPU Port for Tiedtke Convection Scheme
4. Blocked Version of Herzog Turbulence Scheme
5. Modified Computation of HPBL
6. Modifications to SPPT
7. Technical Changes and Bug Fixes
8. Changes to the Namelists
9. Changes of Results

1. Reproducing the Behaviour of the old COSMO Physics

(by Matthias Raschendorfer, Ulrich Schättler, DWD)

Testing the new COSMO-ICON physics (namely the modified prognostic TKE turbulence and the modified version of TERRA) at DWD was not yet successful. We experienced several problems that could not be clarified up to now, for example:

• drying out of the soil
• numerical instabilities when running COSMO-D2

Therefore DWD decided to run the new code in a setting which comes rather close to the old ijk-version of the turbulence and the soil model. Especially for the turbulence scheme some hard-coded values have to be modified for that. In version 5.04f this could be realized by compiling with a pragma -DCOSMO_OLD. For ease-of-use this has now been implemented with a new namelist variable: loldtur

Default is: loldtur = .FALSE.

This means that the default setting is in a way that the new code is running in an ICON-like setting. Only if loldtur is set to .TRUE., several internal turbulence switches are modified to reproduce the old ijk-scheme as close as possible (but it is not bit identical to the old scheme).

Back to Contents

2. Updated TERRA to the latest ICON Version

(Modifications in TERRA by Günther Zängl, Jürgen Helmert, DWD)

Updates from latest ICON Version (2017-11-08)

• Soil ice parameterization according to K. Schaefer and Jafarov, E.,2016, (doi:10.5194/bg-13-1991-2016) (JH)
• Modifications for calculations of runoff and infiltration to avoid loosing too much soil moisture (GZ, JH)
• Introduced option itype_trvg==3:
  This option would need two additional input fields
  • z0: vegetation roughness length ( m )
  • plevap: function of accumulated plant evaporation (kg/m2)
  which are not available in COSMO, therefore this option is not chooseable in COSMO!
• Introduced the snow temperature offset for calculating the evaporation as additional input field (tsnred). This field is also not available for COSMO and is always given with all values = 0.0. In this way the computations for COSMO are not changed.

Back to Contents

3. GPU Port for Tiedtke Convection Scheme

(by Philippe Marti, MCH)

The Tiedtke convection scheme has been ported to GPU. Corresponding modifications have been done in conv_data.f90 and conv_interface.f90.

Back to Contents

4. Blocked Version of Herzog Turbulence Scheme

(by Ulrich Blahak, DWD)

The Herzog turbulence scheme, which can be switched on by itype_turb = [5,6,7,8], has been ported to the blocked data structure. New subroutines prankolmo_rk_blocked and horizontal_diffcoeffs_blocked have been implemented for that purpose.

Note that the other subroutines (partura, parturs) from src_turbulence have not yet been ported.

Back to Contents

5. Modified Computation of HPBL

(by Guy DeMorsier, MCH)

Measurements in Payerne have shown that the height of the planetary boundary layer (HPBL), as it is computed in SR calc_pbl_brn from pp_utilities.f90, is too high. The results were obtained from COSMO-2 data but they are also valid for all COSMO models. The subroutine calc_pbl_brn first determines a k-index (k_pbl) which represents the top of the PBL. The corresponding level of HHL then has been taken as HPBL. Now HPBL is computed by a linear interpolation between this and the next lower level. In this way the height of the PBL is reduced by about 50 to 100 meter.

Back to Contents

6. Modifications to SPPT

(by Andre Walser, MCH)

Revision of SPPT:

• changed calculation of rlatwei_rn and rlonwei_rn in src_stoch_physics.f90 to ensure same results with different compilers
• use tracer attributes for qx fields to define whether SPPT is applied for this field: new subroutine set_sppt_trcr_meta in src_stoch_physics.f90
• new namelist switch ltargetdiff_mask to switch off SPPT locally if targeted diffusion is active
• added output field RAPA: SPPT random pattern

Back to Contents

7. Technical Changes and Bug Fixes

• conv_interface.f90: (by Lucio Torrisi)
  
• sfc_interface.f90:
  Added new local variables dum1fl_s_t and dum2fl_s_t to acc present_or_create list.
• data_io.f90:
  
  • Set nrbit_phhl to 24 (was not initialized before)
  • Removed variable nvar
• src_input.f90, src_output.f90:
  Removed variable nvar
• turb_interface.f90: (by Uli Blahak)
  
  • Corrected start index of a loop in turb_prepare for itype_sher=2
  • Herzog turbulence scheme (itype_turb=5,6,7,8) in block data structure
  • Moved call to diag_lh_surfflux() after the idealized surface fluxes presetting
• src_obs_cdfin_comhead.f90:
  Fix to read reports with section1_data_category=6 and section1_int_data_sub_category=10 (which is not officially intended).
• src_artifdata.f90:
  
  • Bugfix in set_idealized_surffluxes_block(): removed factor 0.5 in computation of t_g, which is unnecessary here because 0.5*(u_i + u_i-1) has been replaced by u_m
  • Fix for t_ice in case of lseaice
• radiation_interface.f90:
  Fix for swdir_cor out-of-bounds

Back to Contents

8. Changes to the Namelists

Back to Contents

9. Changes of Results

• The modifications in TERRA do change the results.
• The modification in pp_utilities.f90 leads to different values for HPBL.
• The modifications for SPPT are only in affect, if running with lsppt=.TRUE.

Back to Contents