Since Version 4.0, some new options have been added to the COSMO-Model and some more bugs could be fixed.
Lately we detected some problems with the reproducibility in the COSMO-DE. There were several reasons:
Problems in the Runge-Kutta dynamical core:
These problems were related to cases with high vertical velocity and also due to the lateral boundary relaxation. Fixes have been implemented in the modules:
Problems in the Data Assimilation:
There were differences in the mapping of an observation to a special grid point. It turned out, that the computation of the rotated latitude for a grid point can differ, when using different decompositions. This was due to the fact, that the "start latitude" of the subdomain was used for the computation and different rounding then lead to different results.
Now, the value for startlat_tot is used to compute the rotated coordinates of a special grid point.
Extensions for Reproducibility of Semi-Lagrange Advection:
To ensure that the summation of REAL variables is done in a reproducible manner requires that the whole field is summed up in one processor. If this is not done, summing up a field on distributed computers can deliver different results, depending on the decomposition.
Another method has been implemented now in the COSMO-Model, which is used in the Semi-Lagrange Advection, which transforms the REAL numbers to long INTEGERS first and then does the summation.
Back to Contents
Back to Contents
Back to Contents
Back to Contents
Back to Contents
There have been some more editorial changes and adaptations to the INT2LM:
There are the following new Namelist variables: Back to Contents
The clipping of variables to 0 affect some variables, e.g. the rain rates
and some moisture variables. But note, that only the values in the output
are affected. The COSMO-Model itself still works with the original values.
The changes in the gust parameterization do change the values for the maximal
wind gust VMAX_10M.
Technical Bug Fixes
Changed downward and upward component in call to subroutine fesft.
Initialize t_so, w_so in case of multi-layer soil model.Calculation of Supercell Detection Indices
Two Supercell Detection Indices can be calculated in the postprocessing.
There are two new output fields SDI1 and SDI2, which can
be specified for output.
Changes in the Gust Parameterization
A re-tuning of gusts has been implemented by using wind speeds at 10 meters
above ground, instead of 30 meters.
Brightband detection in Radar Data
The possibility to use radar beam height as an additional information has
been implemented. Therefore, an additional GRIB ("height_file") file is
required. This is controlled by the Namelist variable lhn_height
Also, a brightband detection algorithm has been implemented, which can be
chosen optionally. This is controlled by the Namelist variable lhn_bright.
Additional Technical Changes
2. Summary of new Namelist Variables
Namelist Group /RUNCTL/
Group Name Description Default
/NUDGING/
lhn_height
use height infos for radar data
.TRUE.
lhn_bright
apply bright band detection
.TRUE.
height_file
name of dxheight file
height_dx.grib1
3. Changes of Results