Documentation of the Changes in the COSMO-Model
Versions 4.15/4.16

07.12.2010

This short report documents the changes in the COSMO-Model versions 4.15 (implemented into the VCS on 2010-11-19) and 4.16 (implemented into the VCS on 2010-12-07). Version 4.16 is running operational at DWD since 2010-12-15. Version 4.16 only was a small technical update. The modifications are marked below.

Contents:

  1. Changes in the Dynamics
  2. Changes in the Physics
  3. Changes in the Assimilation
  4. Technical Changes
  5. Summary of Namelist Variables Changes
  6. Changes of Results


1. Changes in the Dynamics

Implicit vertical advection for the Runge-Kutta dynamics
(by Michael Baldauf)

A new option for 3rd order implicit vertical advection has been implemented in the Runge-Kutta scheme. At the same time the namelist control of vertical advection has been changed (in the group /DYNCTL/

New subroutines have been implemented for the new scheme in src_runge_kutta.f90 (calc_wcon_sqrtg), in src_slow_tendencies.f90 (complete_tend_uvwtpp_CN3Crow) and in numeric_utilities.f90 (solve_5banddiag, bandec, banbks).

Changes in fast_waves_rk
(by Günther Zängl)

A new subroutine weighting_factors_full2half has been implemented in fast_waves_rk.f90 to summarize the calculation of weighting coefficients for vertical interpolation.

Back to Contents


2. Changes in the Physics

Limited range of Namelist variable securi
(by Oliver Fuhrer)

The value of the /TUNING/ Namelist variable securi has been limited. According to numerical stability tests, the range from 0 to 1 is too big. The default value and the allowed range therefore have been reduced:

Default: securi_d = 0.5 and the range: 0 <= securi <= 0.5

This will change the results, if a value greater than 0.5 has been chosen before with limpltkediff=.FALSE. For limpltkediff=.TRUE., this security value has no effect any more.

Back to Contents


3. Changes in the Assimilation

Changes in the Latent Heat Nudging
(by Klaus Stephan)

There were bug fixes in src_lheat_nudge to correct bugs when applying bright band detection. This also includes a tuning of the parameter used for the detection. These fixes will change the results for LHN analysis.

The default value of namelist parameter nradar has been changed to 33 to consider the additional radar station "MEM" which will be operational soon.

Changes in the Nudging
(by Christoph Schraff and Klaus Stephan)

Introduction of a new subroutine to correct the dry bias of Vaisalla RS92 radiosonde measurements. This includes changes in data_nudge_all.f90, organize_assimilation.f90, src_obs_proc_cdf.f90 and src_obs_processing.f90. A new namelist parameter mqcorr92 has been introduced to control the bias correction
Group Name Meaning Default
/NUDGING/ mqcorr92 switch for bias correction for Vaisala RS92
  • = 0: no correction for humidity (default)
  • = 1: correct only solar radiation bias
  • = 2: correct total bias (incl. nighttime bias)
0
Also this bias correction causes a change in results. The analysis will become wetter during day time and drier during night time (if mqcorr92 set to 2). In case of mqcorr92=0 results do not change.

Back to Contents


4. Technical Changes

  • src_input.f90:
    Eliminated use of T_S_LAKE again, because the SST analysis now provides T_S as it is wanted by FLake.
  • numeric_utilities.f90:
    In the subroutine curl the dimension of tgrlat and acrlat was erroneously done with ie instead of je. In the subroutine potential_vorticity_rho (from module pp_utilities.f90) these parameters were not used at all and have therefore been eliminated.
  • src_output.f90
    Use time level nnow instead of preset variable itl for the variables VORTIC_U, VORTIC_V, VORTIC_W, POT_VORTIC, T_WATER and also in the calls to the subroutines caliq, calztd and calomega.
  • src_output.f90
    Increase value of lbm to 25000 for handling bigger frames (by Lucio Torrisi).
  • src_soil_multlay.f90
    Introduced ibot_w_so and snow_melt (from CLM Version):
    The soil model uses a local index i250, which corresponds to the last soil layer completely above 2.5 meter soil depth. This index is used to compute the subsoil runoff (runoff_g) as drainage flux through bottom of this layer and to control switching off the soil moisture gradient related flux below. For climate applications this index should be variable and there is a Namelist variable czbot_w_so, with which the desired soil depth for this layer can be given. The variable ibot_w_so has been introduced to replace i250. The default is chosen, so that the old settings are reproduced.
  • From Version 4.16: src_flake.f90
    Adaptation to the two time level scheme, update of I/O (in the subroutine flake_init).
  • From Version 4.16: src_input.f90
    Initialization of t_ice for all time levels (according to t_so).

Back to Contents


5. Summary of Namelist Variables Changes

There were the following changes for the Namelist variables:

Group Name Meaning Default
/DYNCTL/ lva_impl_dyn eliminated -
y_vert_adv_dyn string to specify the vertical advection scheme in the Runge-Kutta dynamics:
  • 'expl': explicit (as was lva_impl_dyn=.FALSE. before)
  • 'impl2': 2nd order implicit (as was lva_impl_dyn=.TRUE. before)
  • 'impl3': 3rd order implicit (this is new). before)
'impl2'
/NUDGING/ mqcorr92 switch for bias correction for Vaisala RS92
  • = 0: no correction for humidity (default)
  • = 1: correct only solar radiation bias
  • = 2: correct total bias (incl. nighttime bias)
1

Back to Contents


6. Changes of Results

Due to the bug fixes in the latent heat nudging the result will change for LHN analysis (for COSMO-DE). The bias correction also causes a change in results. The analysis will become wetter during day time and drier during night time (if mqcorr92 set to 2). In case of mqcorr92=0 results do not change.

Back to Contents