Documentation of the Changes in COSMO-Model 4.5 / 4.6

24.09.2008

The biggest change to the COSMO-Model has been in Version 4.5, where the input of observation data from NetCDF files has been implemented. More optimizations for vector processors have been performed, which improve the performance of the model on the NEC. There were other modifications, which are described in detail below. Version 4.6 just was a fix of a bug, introduced with 4.5.

1. Documentation of the Changes

Reading observation data from NetCDF files (by Christoph Schraff)

A new interface has been introduced to read and pre-process observation data from NetCDF files instead of an AOF file. These NetCDF files are obtained by simple conversion from BUFR files which contain BUFR reports according to Table Driven Code Formats (TDCF) of WMO (see http://www.wmo.int/pages/prog/www/WMOCodes/OperationalCodes.html, http://www.wmo.int/pages/prog/www/WMOCodes/TemplateExamples.html) for those data types where TDCF have been defined, namely for SYNOP, SHIP, PILOT and TEMP types, but also for BUOY and AMDAR. For ACARS, BUFR from ARINC Centre 56 (USA) and from UK Met Office can be read, or alternatively, a combined format. For wind profiler, RASS, and VAD radar wind, a format defined by DWD is read, since there is no standard format defined by WMO yet. The names of the input files, that can already be used, are:

cdfin_temp
cdfin_tempship
cdfin_pilot
cdfin_pilot_p
cdfin_amdar_ml
cdfin_amdar
cdfin_acars: ACARS combined format from DWD
cdfin_wprof
cdfin_rass
cdfin_radar_vad
cdfin_synop
cdfin_synop_mob
cdfin_ship
cdfin_buoy
cdfin_acars_uk: ACARS BUFR from UK Met Office
cdfin_acars_us: ACARS BUFR from ARINC Centre 56 (USA)

Other input files cannot yet be used. If a file is empty, it should be removed.

From which files the observations are read is controlled by the 2 Namelist variables (in the group NUDGING):

Name	Meaning	Default
`itype_obfile`	to determine, from which file(s) the observations are read `1`: read observations from AOF `2`: read observations from NetCDF files	`1`
`ycdfdir`	directory in which the NetCDF input observation files reside	`./`

Back to Contents

Further optimizations for vector processors (by Uli Schättler and Jens-Olaf Beismann, Christian Bollmann (NEC))

Some more routines have been optimized for vectorization:

Subroutines hydci_pp and hydci_pp_gr of module src_gscp (Microphysics)
In the routines for the microphysics there are several IF-clauses, to check, which conditions are valid. Up to now, all these IF-clauses have been in a big loop. For vector processors it is better, first to collect all indices, for which a certain condition holds, and then to loop only over these indices.
On scalar machines, this organization takes more or less the same time (perhaps slightly slower).
numeric_utilities_rk: (Runge-Kutta dynamics)
The coefficients for the Bott advection routines have been modified to save several divisions during the forecast.

Note: Although these changes are mathematically neutral, they change the results of the Runge-Kutta dynamics, because the numerical calculations are different.

Synthetic satellite images (Use of RTTOV7-library)
The dimensions of several variables from the RTTOV-library have been changed, to enable vectorization over the profiles. The routine org_sat_tbs (in the module src_sat_tbs) and the call in organize_satellites have been adapted accordingly. Now the RTTOV-library is called for a whole vector (a latitude) of grid points.

Note: Because of the changes in dimensions, the COSMO-model from Version 4.5 onwardsd definitely needs the modified RTTOV7-library, named RTTOV7_synsat_vector. It will not work with the old version any more.

src_lheat_nudge: (Latent Heat Nudging))
The definition and allocation of the global field blacklist_tot has been moved outside the subroutine lhn_obs_prep, because the values are also needed after this routine has finished.

Back to Contents

Modifications for Runge-Kutta Dynamics

Incorporation of modified Tiedtke scheme (by Michael Baldauf, Dmitrii Mironov):
This is the same as in Version 4.4 for the Leapfrog-scheme: The tendencies of qi and qc of the modified convection scheme are now also delivered to the Runge-Kutta dynamics.

An additional flux calculation method (Lax-Wendroff) has been implemented in src_integrals.f90 (by Ronny Petrik, MPI Hamburg).

Optimized horizontal advection routines are now also used for the field wcon in advection_pd in module src_advection_rk.f90. (by Michael Baldauf).

A more efficient implementation of a lower boundary condition for w has been implemented in fast_waves_rk.f90 (by Günther Zängl).

Note:

This changes the results of the Runge-Kutta applications slightly!

Back to Contents

Additional namelist parameter for SSO scheme (by Jan-Peter Schulz, Ulrich Schättler)

An additional namelist variable has been implemented in group /PHYCTL/ to modify the calling frequency of the SSO scheme (Introduction of SSO-scheme: see Version 4.4):

Name	Meaning	Default
`nincsso`	time step increment for running the SSO scheme	`5`

Back to Contents

New (alternative) reference atmosphere (by Günther Zängl)

A new (alternative) reference atmosphere has been introduced, which is based on the temperature profile

T0(z) = T00 + delta_t EXP(-z/h_scal),
with default values of T00 = 213.15 K, delta_t = 75 K and h_scal = 10 km; (in the model code, T00 = t0sl - delta_t).

Thus, the reference atmosphere approaches an isothermal profile in the stratosphere, whereas the existing reference profile has an increasingly negative vertical temperature gradient in the stratosphere. The vertical extent of the model domain is no longer limited with the new reference atmosphere.

Except for idealized simulations, the reference atmosphere can only be chosen in INT2LM (from Version 1.9 on). All parameters of the reference atmosphere are coded in the GRIB files, and the COSMO-Model determines the type of reference atmosphere by encoding the GRIBs.

The following holds in Version 4.5 and 4.6 (might be changed in the future to account for additional vertical coordinate types):

For the old reference atmosphere the vertical coordinate type (ivctype) in Grib is: 1 - 3.
The internal model variable irefatm is set to 1.
For the new reference atmosphere the vertical coordinate type (ivctype) in Grib is: 4 - 6.
The internal model variable irefatm is set to 2.
It is: ivctype = :
1, 4: Pressure-based vertical coordinate on input
2, 5: Heigt based verical coordinate on input
3, 6: SLEVE coordinate

The new reference atmosphere needs 2 additional parameters, which are also coded in the GRIBs:

delta_t: temperature difference between sea level and stratosphere
h_scal: scale height

For the old style of coding (ldwd_grib_use=.TRUE.), it is

delta_t = igds (29 + ke1 + 5)
h_scal = igds (29 + ke1 + 6)

For the new style of coding (ldwd_grib_use=.FALSE.), it is

delta_t = igds (31 + ke1 + 4)
h_scal = igds (31 + ke1 + 5)

Back to Contents

Modifications / Alternatives for idealized cases (by Günther Zängl)

Implementation of a new sponge layer near upper model boundary (according to Klemp et al., MWR 2008):
Modifications have been done in fast_waves_rk and organize_dynamics. A new value for the Namelist parameter itype_spubc = 3 in group /DYNCTL/ has been implemented.

Finish the implementation of the radiative lateral boundary condition. (Modifications have been done in fast_waves_rk, src_relaxation, src_setup and organize_data):
Two new Namelist parameters have been added to group /DYNCTL/:

Name	Meaning	Default
`lradlbc`	to switch on/off the radiative lateral boundary condition. This parameter can be used only in connection with idealized cases (`lgen=.TRUE.`)	`.FALSE.`
`relax_fac`	Reduction factor for the lateral relaxation zone (effective only if `lradlbc = .TRUE.`; it allows to combine the radiative lateral boundary condition with a weak Davies relaxation.	`0.01`

Possibility to switch off surface momentum fluxes in the Runge-Kutta dynamics for idealized cases, even if the turbulence scheme is used. A new Namelist parameter lfreeslip_sfc has been implemented. Default is .FALSE..

Back to Contents

More Namelist Input for Ensemble Mode (by Christoph Gebhardt)

Three more Namelist parameters have been introduced in the Namelist group /TUNING/:

Name	Meaning	Default
`entr_sc`	mean entrainment rate for shallow convection	`0.0003`
`mu_rain`	shape parameter of the rain drop size distribution	`0.5`
`cloud_num`	cloud droplet number concentration	`5.00E+08`

Before, these were parameters in the convection or the microphysics, resp. As namelist parameters they are used by the ensemble group, but are considered to be of general interest.

More Namelist parameters have been included in the Namelist group /EPSCTL/:

Name	Meaning	Default
`fac_lai`	a factor to modify values of the leaf area index. Values should stay in the interval `[rmin_lai, rmax_lai]` (see below).	`1.0`
`rmin_lai`	Minimum value for leaf area index modification value.	`0.0`
`rmax_lai`	Maximum value for leaf area index modification value.	`8.0`
`fac_plcov`	a factor to modify values of the plant cover. Values should stay in the interval `[rmin_plcov, rmax_plcov]` (see below).	`1.0`
`rmin_plcov`	Minimum value for plant cover modification value.	`0.0`
`rmax_plcov`	Maximum value for plant cover modification value.	`1.0`
`fac_rootdp`	a factor to modify values of the root depth. Values should stay in the interval `[rmin_rootdp, rmax_rootdp]` (see below).	`1.0`
`rmin_rootdp`	Minimum value for root depth modification value.	`0.0`
`rmax_rootdp`	Maximum value for root depth modification value.	`2.0`

These parameters are considered to be of interest only for ensemble runs and therefore are available only in the EPS namelist group.

Back to Contents

2. Summary of new and changed Namelist Variables

There are the following new Namelist variables:

Group	Name	Meaning	Default
`/NUDGING/`	`itype_obfile`	to determine, from which file(s) the observations are read read observations from AOF read observations from NetCDF files	`1`
`/NUDGING/`	`ycdfdir`	directory in which the NetCDF input observation files reside	`./`
`/PHYCTL/`	`nincsso`	time step increment for running the SSO scheme	`5`
`/PHYCTL/`	`lfreeslip_sfc`	switch on/off surface momentum fluxes even if turbulence scheme is used	`.FALSE.`
`/DYNCTL/`	`itype_spubc`	Type of Rayleigh damping in the upper levels. Implemented new option 3. chooses the damping against boundary fields chooses the damping against filtered forecast fields; can be chosen, if the boundary data is only available on frames (NEW) Sponge layer near upper model boundary according to Klemp et al. (MWR 2008)	`1`
	`lradlbc`	to switch on/off the radiative lateral boundary condition. This parameter can be used only in connection with idealized cases (`lgen=.TRUE.`)	`.FALSE.`
	`relax_fac`	Reduction factor for the lateral relaxation zone (effective only if `lradlbc = .TRUE.`; it allows to combine the radiative lateral boundary condition with a weak Davies relaxation.	`0.01`
`/TUNING/`	`entr_sc`	mean entrainment rate for shallow convection	`0.0003`
	`mu_rain`	shape parameter of the rain drop size distribution	`0.5`
	`cloud_num`	cloud droplet number concentration	`5.00E+08`
`/EPSCTL/`	`fac_lai`	a factor to modify values of the leaf area index. Values should stay in the interval `[rmin_lai, rmax_lai]` (see below).	`1.0`
	`rmin_lai`	Minimum value for leaf area index modification value.	`0.0`
	`rmax_lai`	Maximum value for leaf area index modification value.	`8.0`
	`fac_plcov`	a factor to modify values of the plant cover. Values should stay in the interval `[rmin_plcov, rmax_plcov]` (see below).	`1.0`
	`rmin_plcov`	Minimum value for plant cover modification value.	`0.0`
	`rmax_plcov`	Maximum value for plant cover modification value.	`1.0`
	`fac_rootdp`	a factor to modify values of the root depth. Values should stay in the interval `[rmin_rootdp, rmax_rootdp]` (see below).	`1.0`
	`rmin_rootdp`	Minimum value for root depth modification value.	`0.0`
	`rmax_rootdp`	Maximum value for root depth modification value.	`2.0`

Back to Contents

3. Changes of Results

The results of the Runge-Kutta applications change because of the following modifications:

Changes in the coefficients of the Bott advection routines (Only numerical changes).
Consideration of tendencies for cloud water and cloud ice from the Tiedtke convection.
Changes in the lower boundary condition for the vertical velocity.

Back to Contents