FE 13 - COSMO-Model System: Interpolation Program INT2LM

Documentation of the Changes in INT2LM 1.7

26.11.2007

Version 1.7 of the interpolation program INT2LM is available since end of November 2007. This short guide gives an overview on the main modifications, the changes in the Namelists and the results. The main subjects of the new version are the developments for very high-resolution runs with the COSMO-Model and the extensions for the climate mode. While all modifications for the high-resolution runs have been intensively tested at DWD and other COSMO centers for about a year now, the extensions for the climate mode are still in the testing phase. Therefore it is not recommended to use this version as preprocessor for the COSMO-CLM, the climate mode of the COSMO-Model. Development and testing for this mode is going on and an updated version will be delivered as soon as possible.

1. Modifications since Version 1.4

Developments for Runge-Kutta high-resolution runs

Treatment of qr, qs, qg:
The interpolation of the additional humidity variables for rain, snow and graupel from a coarse grid model to the COSMO-Model grid has been introduced. It can be controlled by the new Namelist variables:

Name	Meaning	Default
`lprog_qrqs`	Interpolate `qr`, `qs` to the fine grid	`.FALSE.`
`lprog_qg`	Interpolate `qg` to the fine grid	`.FALSE.`

Additional smoothing of the orography:
The high-resolution model had some problems with the roughness of higher mountains. Therefore, an additional smoothing of the orography can be done. This is controlled by the new Namelist variables:

Name	Meaning	Default
`lxso_first`	Extra smoothing of orography before low-pass filtering	`.FALSE.`
`ilow_pass_xso`	Type of low-pass filter for extra smoothing of steep orography	`0`
`numfilt_xso`	Number of sequential applications of xso-filter	`1`
`ilow_pass_oro`	Type of low-pass filter for orography filtering	`1`
`numfilt_oro`	Number of sequential applications of orography-filter	`1`
`ifill_valley`	Type of valley filling `0`: no filling `1`: MIN (before smoothing the orography) `2`: MIN (after smoothing the orography)	`0`
`rfill_valley`	Mask for valley filling: if the height distance to the four direct neighbor grid points is larger than `rfill_valley`, this valley is filled up and set to the minimum height value of the four neighbors	`0.0_ireals`

Additional humidity splitting after balancing and / or filtering of pressure deviation:
A balancing and / or filtering of the pressure deviation had been introduced for high-resolution runs in one of the last versions. This is done after splitting the relative humidity to qv and qc. But after changing the values for the pressure deviation, the variables for qv and qc do not fit to pp any more, which makes an additional splitting of the relative humidity in module src_vert_inter_lm.f90 necessary. Without the additional splitting, wrong (even negative) values occured for variable qv_s in the initial conditions. But the problem with negative qv_s was also reported even if the additional splitting was used, but less severe. Therefore, a clipping of qv_s_lm has been introduced in the module src_2d_fields. This changes the results for the moisture variables qv and qc in case of lm2lm. No negative values can now occur for qv_s.

Back to Contents

Introduction of Ensemble Mode

The ensemble mode has been introduced into INT2LM to provide (different) boundary conditions for the ensemble members of the COSMO-Model. The ensemble member ID, the ensemble type and the total number of ensemble members are set and written to the corresponding grib headers.

The ensemble mode is controlled by a new Namelist switch leps_bc in the group /CONTRL/. Default is .FALSE. If leps_bc = .TRUE., a new Namelist group /EPSCTL/ is read which contains the following namelist parameters:

Name	Meaning	Default
`iepsmem_bc`	ID of the member in the ensemble of boundary conditions (meaningful value `≥ 0)`	`-1`
`iepstyp_bc`	ID of the boundary ensemble generation type (meaningful value `≥ 0)`	`-1`
`iepstot_bc`	total number of boundary ensemble members (meaningful value `≥ 0)`	`0`
`lchk_bc_typ`	if `.TRUE.`, check member ID of input data (only if `leps_bc=.TRUE.)`	`.FALSE.`

There is a new data module (data_epsctl.f90) and the following modules have been modified:

data_int2lm_control.f90: introduced new Namelist switch leps_bc
src_lm_output.f90
src_namelists.f90
src_read_coarse_grid.f90

Back to Contents

Introduction of a smooth transition of orography at the lateral boundaries

If the orographies of the coarse grid and the fine grid differ too much in the vicinity of the lateral boundaries, problems can occur during the simulation. To avoid these, a smooth transition of the fine orography to the coarse one has been implemented. This can be controlled by the new Namelist variables:

Name	Meaning	Default
`llbc_smooth`	If `.TRUE.`, run with a smooth orography transition at the lateral boundaries.	`.FALSE.`
`nlbc_smooth`	Number of grid points for smooth orography transition at the lateral boundaries.	`20`

Back to Contents

Extensions for climate mode

Another major upgrade for INT2LM is the implementation of the climate mode. The main subjects are the introduction of NetCDF as I/O format (described below) and the provision of the slowly varying parameters (like plant characteristics) also in the boundary values.

The use of the climate mode can be controlled by the following new Namelist variables.

Name	Meaning	Default
`lbdclim`	if `.TRUE.`, special data are written to the boundary files	`.FALSE.`
`lcm2lm`	if `.TRUE.`, data from a (common) coarse grid climate model are read.	`.FALSE.`

The switch lcm2lm has been implemented different compared to the former beta-version int2lm2_b2. In the beta-version this switch controlled both, reading data from a climate model and writing additional data to the boundary files. But this made it impossible to produce the additional boundary data in case of lgme2lm, lec2lm or llm2lm. Therefore we implemented another switch lbdclim, which controls the computation and writing of the additional boundary data.

There are a number of additional switches, which can not only be used in the climate mode:

Name	Meaning	Default
`llake`	if `.TRUE.`, read and process additional lake parameters	`.FALSE.`
`lyear_360`	if `.TRUE.`, use a climatological year with 360 days per year.	`.FALSE.`
`l_cressman`	if `.TRUE.`, use a Cressman scheme for the horizontal "Match"-interpolation (implemented for mass grid points only)	`.FALSE.`
`l_bicub_spl`	if `.TRUE.`, use a bicubic spline interpolation	`.FALSE.`

There are additional switches that control the treatment of the slowly varying fields. The computation of these fields has been moved from the subroutine external_data.f90 to the module src_2d_fields.f90 (new subroutine plant_characteristics in this module). In climate mode, these fields have to be computed also for producing boundary conditions.

Name Meaning Default

itype_t_cl

to choose, which deep soil (climatological) temperature shall be chosen:

0:	Take `t_cl` from coarse grid model
1:	Take `t_cl` from the external parameters for the COSMO-Model, if it is available there.

0

itype_rootdp

to choose, how the root depth shall be computed

0:	Input from external parameter for the COSMO-Model is taken and modified with an annual cycle.
1:	Input from external parameter for the COSMO-Model is taken without any modification, but limiting to a minimum value of 0.12.
2:	Input from external parameter for the COSMO-Model is taken and modified with an annual cycle. In addition, the values are adapted to ECOCLIMAP niveau.

0

itype_ndvi

to choose treatment of plant cover and leaf area index:

0:	Plant cover and leaf area index for the COSMO-Model are produced for a special day by using the data sets for vegetation and for rest and modifying these with an annual cycle.
1:	Plant cover and leaf area index for the COSMO-Model are produced for a special day 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 GME. This is only possible, when using GME data.

0

itype_w_so_rel

to choose type of relative soil moisture input:

0:	Artificial profile relative to pore volume.
1:	Relative to pore volume (read from coarse grid data).
2:	Relative to field capacity (read from coarse grid data).

0

If type_w_so_rel = 0, an artificial profile for the soil moisture is set, no matter which input model is used. The settings type_w_so_rel = 1/2 only work for lcm2lm = .TRUE..

NOTE: In Version 1.5 itype_w_so_rel has been introduced as iw_so_rel_type. We changed the name in order to be consistent with other names of that type in INT2LM and in the COSMO-Model.

Back to Contents

Extensions for chemistry

Another option that has been added to INT2LM is the possibility of interpolating additional fields for chemistry variables to the fine grid. This is only implemented for option llm2lm and is controlled with the Namelist variable:

Name	Meaning	Default
`l_chemistry`	if `.TRUE.`, interpolate additional chemistry fields	`.FALSE.`

The metadata of the additional fields are declared in Grib Tables 241 and 242. If l_chemistry=.TRUE., all fields from these 2 tables, which are in the input data set, are read, interpolated to the fine grid and written to the output data field. If fields from these tables are not in the input data set, only a warning is given, no other action is done.

Back to Contents

Changes for I/O and Introduction of NetCDF

With the introduction of the climate mode also the NetCDF-format for I/O has been implemented into the INT2LM. For every file that is read or written, the user can decide whether it should be done in Grib1 or NetCDF. This can be controlled by the Namelist variables listed in the following table. Possible values of the strings are 'grb1' or 'ncdf'.

Name	Meaning	Default
`ylmext_form_read`	Input format for the external parameters of the COSMO-Model grid	`'grb1'`
`yinext_form_read`	Input format for the external parameters of the coarse grid	`'grb1'`
`yin_form_read`	Input format of the coarse grid data	`'grb1'`
`ylm_form_write`	Output format for the COSMO-Model variables	`'grb1'`

To proper implement the NetCDF format, some more implementations of the generic routine distribute_values were necessary in the module parallel_utilities.f90.

In former versions of INT2LM, analysis or initialized analysis data could only be done file by file. This is now possible in one run of the INT2LM. The new Namelist variables to control this, are (in group /DATA/):

Name Meaning Default

yinput_type

to choose the type of the input data:

`'forecast':`	Coarse grid data are forecast data (second character in file name is `'f')`
`'analysis':`	Coarse grid data are analysis data (second character in file name is `'a')`
`'ana_init':`	Coarse grid data are forecast data (second character in file name is `'i')`

'forecast'

ytunit_in

Time unit for input data. This character determines, how the filename for the input data is set. It occurs as the fourth character in the filename.

`'f'`	Forecast format: Time is specified as `'ddhhmmss'` with 2 digits for the day, the hour, the minutes and the seconds.
`'d'`	Day format: Time is specified as `'yyyymmdd'` with 4 digits for the year and 2 digits for the month and the day. Note: With this format, I/O can only be specified with an increment of full hours (or multiples).

'f'

ytunit_out

Time unit for output data. This character determines, how the filename for the boundary (output) data is set. It occurs as the fourth character in the filename.

`'f'`	Forecast format: Time is specified as `'ddhhmmss'` with 2 digits for the day, the hour, the minutes and the seconds.
`'d'`	Day format: Time is specified as `'yyyymmdd'` with 4 digits for the year and 2 digits for the month and the day. Note: With this format, I/O can only be specified with an increment of full hours (or multiples).

'f'

The following 2 Namelist variables have been eliminated:

in group /CONTRL/: lanalysis
in group /DATA/: ytunitbd

The file YUCHKDAT is opened and closed in every step.

Back to Contents

Some (minor) modifications and extensions

There were more modifications and extensions, which are described briefly here.

Changes for the SLEVE coordinate:
When reading the external parameter for the coarse grid model, the subroutine sleve_split_oro is called by all MPI-tasks (while in the COSMO-model it is only called by MPI-task 0). Therefore, a new argument has been added to the parameter list for the MPI-task ID, in order to do the control output only by MPI-task 0.
Modifications for skipping levels at top of variables in ECMWF input files.
The ECMWF-model reaches much higher in the atmosphere than most COSMO-Model applications. To reduce the necessary memory, levels above the COSMO-Model can be skipped. This is controlled by the new Namelist variable:

Name Meaning Default

nlevskip number of levels to skip at top of input model 0
Modifications for improving snow interpolation from ECMWF
Modifications for accepting ECMWF orography on model levels
Elimination of some Namelist variables in group /CONTRL/
The variables nstart, nstop, nincbound, dt have been eliminated. Start, stop and boundary increment can now only be given in multiples of 0.25 hours with the variables hstart, hstop, hincbound. The timestep is now set as a constant dt=900s.

Name	Meaning	Default
`nlevskip`	number of levels to skip at top of input model	`0`

Debug output:
For controlling the verbosity of debug output similar to the COSMO Model the following Namelist variables have been introduced:

Name Meaning Default

idbg_level To control the verbosity of debug output: with a higher value, more debug output is printed to STDOUT. 2

lprintdeb_all

`.TRUE.:`	All tasks print the debug output
`.FALSE.:`	Only task 0 prints the debug output

.FALSE.

Back to Contents

Bug Corrections

As usual, there have been several bug fixes.

src_grids.f90 and src_decomposition.f90:
Adaptations for domains that are located around the date line
src_lm_output.f90, subroutine write_nc_vdefs:
The value of kedim has to be used as third dimension instead of ke_in.
src_namelists.f90:
Corrected internal setting of nstart, nstop and nincbound, which were not set, if the defaults for hstart, hstop and hincbound were not changed.
src_read_ext.f90 and external_data.f90:
Bug correction for orography filtering in sequential programs.
interp_utilities.f90:
The internal logical switch for the Cressman scheme was wrongly initialized, in Version 1.5, so that the Cressman scheme was always run.
src_memory.f90:
Dimensions of the new variables cgas_in, caero_in were wrong in Version 1.5.

Back to Contents

2. Modifications in Module `mpe_io.f90` for Little Endian Machines

There was a problem with writing the proper structure of GRIB records on little endian machines (most Linux machines). If the number of bytes for a GRIB record did not fit into a full integer word (which is 4 bytes long), the last bytes were lost, leading to a corrupt GRIB end section: only 77 instead of 7777 were present in the GRIB record.

This is due to the different byte ordering by little endian (counting from right to left, instead from left to right). On big endian machines, the representation of a GRIB record in the machine has the form (note especially the last 2 integer words):

| G | R | I | B | | x | x | x | x | .... | x | x | 7 | 7 | | 7 | 7 | y | y |

where x is any information belonging to the GRIB record and y is not belonging to the GRIB record any more.

On little endian machines, the representation of each integer word is just reversed:

| B | I | R | G | | x | x | x | x | .... | 7 | 7 | x | x | | y | y | 7 | 7 |

Within mpe_io.f90, the bytes are organized in an integer array, but they are send via Message Passing Interface using the MPI_BYTE data type and the correct length of the GRIB record. The result is, that from the last integer word, only the two bytes containing yy are send, but not the two necessary bytes containing 77. When writing the data to disk, the little endian byte numbering is reversed, so that the GRIB record really is identical to the big endian version. But the last two bytes are not correct.

This problem is now by-passed by just sending additional bytes, in order to transfer the whole last integer word of the GRIB record. A modified version of mpe_io.f90 is included in the new INT2LM version and also in the new version of the COSMO-Model.

Back to Contents

3. Changes of the Namelists

Changes in existing groups:

Group /CONTRL/

Name	Meaning	Default
`lprog_qrqs`	Interpolate `qr`, `qs` to the fine grid	`.FALSE.`
`lprog_qg`	Interpolate `qg` to the fine grid	`.FALSE.`
`lxso_first`	Extra smoothing of orography before low-pass filtering	`.FALSE.`
`ilow_pass_xso`	Type of low-pass filter for extra smoothing of steep orography	`0`
`numfilt_xso`	Number of sequential applications of xso-filter	`1`
`ilow_pass_oro`	Type of low-pass filter for orography filtering	`1`
`numfilt_oro`	Number of sequential applications of orography-filter	`1`
`ifill_valley`	Type of valley filling	`0`
`rfill_valley`	Mask for valley filling	`0.0_ireals`
`llbc_smooth`	to run with a smooth orography transition at lateral boundaries.	`.FALSE.`
`nlbc_smooth`	Number of grid points for smooth orography transition	`20`
`lbdclim`	to write special data to the boundary files	`.FALSE.`
`lcm2lm`	to read data from a (common) coarse grid climate model	`.FALSE.`
`llake`	to read and process additional lake parameters	`.FALSE.`
`lyear_360`	to use a climatological year with 360 days per year.	`.FALSE.`
`l_cressman`	to use a Cressman scheme for the horizontal "Match"-interpolation	`.FALSE.`
`l_bicub_spl`	to use a bicubic spline interpolation	`.FALSE.`
`l_chemistry`	to interpolate additional chemistry fields	`.FALSE.`
`itype_t_cl`	to choose, which deep soil (climatological) temperature shall be chosen:	`0`
`itype_rootdp`	to choose, how the root depth shall be computed	`0`
`itype_ndvi`	to choose treatment of plant cover and leaf area index:	`0`
`itype_w_so_rel`	to choose type of relative soil moisture input:	`0`
`leps_bc`	to switch on the ensemble mode	`.FALSE.`
`idbg_level`	to control the verbosity of debug output	`2`
`lprintdeb_all`	to print debub output on all processors	`.FALSE.`

Detailed explanations for the new variables can be found above. Note that the variable itype_w_so_rel has first been introduced as iw_so_rel_type. To be consistent with other names of that type, it has been adapted.

The following namelist variables have been deleted from the group /CONTRL/:
lanalysis, nstart, nstop, nincbound, dt, ke_soil, czml_soil, ldebug.

New Group /EPSCTL/

Name	Meaning	Default
`iepsmem_bc`	ID of the member in the ensemble of of boundary conditions	`-1`
`iepstyp_bc`	ID of the boundary ensemble generation type	`-1`
`iepstot_bc`	total number of boundary ensemble members	`0`
`lchk_bc_typ`	to check member ID of input data	`.FALSE.`

Group /GRID_IN/

Name	Meaning	Default
`polgam_in`	angle between the north poles of the systems	`0.0`
`ke_soil_in`	number of input levels in multi-layer soil model (without climatological layer)	`7`
`czml_soil_in`	depth of main level of soil layers (middle of layer) in meter	`0.005, 0.02, 0.06, 0.18, 0.54, 1.62, 4.86, 14.58`
`endlat_in_tot`	transformed latitude of the upper right grid point	`0.0`
`endlon_in_tot`	transformed longitude of the upper right grid point	`-40.0`
`lushift_in`	shift of u-velocity due to grid staggering	`.FALSE.`
`lvshift_in`	shift of v-velocity due to grid staggering	`.FALSE.`
`east_add_in`	add an extra column to the East	`0`
`west_add_in`	add an extra column to the West	`0`
`north_add_in`	add an extra column to the North	`0`
`south_add_in`	add an extra column to the South	`0`
`nlevskip`	number of levels to skip at top of input model	`0`

Group /LMGRID/

Name	Meaning	Default
`ke_soil_lm`	number of levels in multi-layer soil model in the COSMO-Model (without climatological layer)	`7`
`czml_soil_lm`	depth of main level of soil layers (middle of layer) in meter	`0.005, 0.02, 0.06, 0.18, 0.54, 1.62, 4.86, 14.58`
`cw_so_rel_lm`	artifical volumetric soil water content (0-1) profile	`0.75, 0.75, 0.75, 0.75, 0.75, 0.75, 0.75, 0.75`
`polgam`	angle between the north poles of the systems	`0.0`

Group /DATA/

Name	Meaning	Default
`yinput_type`	to choose the type of the input data	`'forecast'`
`ytunit_in`	Time unit for input data	`'f'`
`ytunit_out`	Time unit for output data	`'f'`
`yinext_form_read`	Input format for the external parameters of the coarse grid	`'grb1'`
`ylmext_form_read`	Input format for the external parameters of the COSMO-Model grid	`'grb1'`
`yin_form_read`	Input format of the coarse grid data	`'grb1'`
`ylm_form_write`	Output format for the COSMO-Model variables	`'grb1'`
`ldwd_grib_use`	use some DWD specific Grib settings	`.TRUE.`

Back to Contents

4. Changes of the Results

Because of some adaptations in the filtering of orography, there are quite some changes in the results for the option lgme2lm=.TRUE.. But these differences are small. Other changes are:

If negative values of qv_s occured before, these are now clipped to 0.
Only for lm2lm=.TRUE.:
The values of qv and qc are changed, if lbalance_pp=.TRUE. or lfilter_pp=.TRUE.

Back to Contents