Documentation of the Changes in the COSMO-Model
Version 5.10

21.07.2021

1. Implemented Modifications for TERRA-URB

(by Ulrich Schättler)

The urban-canopy land-surface scheme TERRA-URB has been implemented in the COSMO-Model. The TERRA_URB version used is 2.3, which is taken from the CCLM version COSMO-5.0-clm9. It has been implemented already in COSMO-Version 5.05 as a test version. This is now implemented into the official version. The tile-approach necessary for TERRA-URB has already been implemented in COSMO-Model version 5.09.

TERRA_URB is documented in the user guide: User_guide_for_TERRA_URB

In addition there are two scientific papers:

by Wouters et al.: The impact of impervious water-storage parametrization on urban climate modelling, 2015; http://dx.doi.org/10.1016/j.uclim.2014.11.005
by Wouters et al.: The efficient urban canopy dependency parametrization (SURY) v1.0 for atmospheric modelling: description and application with the COSMO-CLM model for a Belgian summer, 2016; http://www.geosci-model-dev-discuss.net/gmd-2016-58
by Garbero et al.: Evaluating the Urban Canopy Scheme TERRA_URB in the COSMO Model for Selected European Cities; https://www.mdpi.com/2073-4433/12/2/237

Control of TERRA_URB

Running TERRA_URB can be controlled by the following namelist variables in /PHYCTL/:

General control:

lterra_urb: to switch on / off the urban-canopy land-surface scheme (default: .FALSE.)

In addition there are some TERRA_URB internal control variables, which are only active, if lterra_urb is set to .TRUE.:

lurbfab: to switch on/off the urban fabric (in a bulk approach) (default: .TRUE.)
itype_ahf: to choose the type of anthropogenic heat
1. no anthropogenic heat flux
2. anthropogenic heat according to Flanner (2009) (default)
itype_kbmo_uf: to choose the bluff-body thermal roughness length parametrization
1. standard SAI-based turbtran;
2. Brutsaert/Kanda; (default)
3. Zilitinkevich
itype_eisa: to choose the type of evaporation from the impervious surface:
1. evaporation just like bare soil;
2. no evaporation;
3. PDF-based puddle evaporation (Wouters et al., 2015) (default)

The following real namelist variables can be used as tuning parameters. Their value is applied to every grid point of the domain. If the value -1 is chosen, an external parameter field is read from the initial data set.
Research is underway to provide 2D external parameters for these values.

`curb_fr_bld`	(default: 0.67)	Building area fraction
`curb_h_bld`	(default: 15.0)	Building height
`curb_h2w`	(default: 1.5)	Street canyon H/W ratio
`curb_ahf`	(default: -1.0)	default annual-mean AHF (-1 means that external data is used)
`curb_alb_th`	(default: 0.14)	default effective thermal albedo of building/pavement substrate
`curb_alb_so`	(default: 0.10)	default short-wave albedo of building/pavement substrate
`curb_hcap`	(default: 1.25E6)	default specific heat times density of building
`curb_hcon`	(default: 0.777)	default building material heat conductivity (former c_ala_bm)
`cskinc_urb`	(default: 1000.0)	urban vegetation conductivity factor.

Back to Contents

2. Modifications for NetCDF I/O

Online compression with Zlib for NetCDF: (by Jonas Jucker)

Integration of the built-in compression functionality of the NetCDF-library into COSMO. The online compression reduces the output size about half, although increasing time to write output significantly. Therefore only recommended with asynchronous output. The compression can be switched ON/OFF for each individual /GRIBOUT/ list.

In order to define the total memory available for buffers on the compute PE's, the hardcoded 1GB were replaced with the namelist parameter asynio_host_mem.

The following two new namelist-parameters were introduced:

Group /GRIBOUT/: lcompress_netcdf: determine whether compression is applied or not for a specific file (default is .FALSE.).
Group /RUNCTL/: asynio_host_mem: total memory available for compute PE to store buffers (default 1 GB).

Modifications to enable NetCDF Restart in Single Precision: (by Ulrich Schättler)

Adaptations for reading and writing restart files in NetCDF have only been done for double precision yet. Following modifications were necessary to also do it in single precision:

Up to now the subroutines write_nc_gdefs and write_nc_vdefs checked for a restart file by testing NF90_DOUBLE (standard output is written in single). Now we test for yextension=='o', which is true for a restart file.
Add and check definitions of T_SO_SAVE in the same way as T_SO.
Add 'soil2' to allowed names for NetCDF 3rd dimension.
The subroutine sfc_restart_copy_toblock in sfc_interface.f90 must not be ported to GPUs, it is only running on CPUs at the beginning of the program.
The xxx_save fields (which are in blocked format) have to be written to and read from restart files. For that the corresponding fields xxx_sijk are used. But when reading these fields from restart files, they were not updated on the device. This is done now.

Back to Contents

3. Technical Changes and Bug Fixes

acc_global_data.f90:
The calls to deallocate the GPU memory have been activated in subroutine acc_deallocate_data_fieldsnow. These have been skipped before because of a former compiler bug.
radiation_interface.f90:
The three radiation variables sobs, pabs and thbs have been added to the copy-out mechanism for the radiation and solar radiation variables, to make all copies complete. In principle it is not necessary, because these variables are also computed in ij-data format (a left over from the coarse-radiation grid).
But when doing this, the variables sobs_b and pabs_b have to be initialized to 0.0, when no solar radiation is active, otherwise we would keep a solar value at least for one time step too much. This has been corrected now and changes the results slightly.

Back to Contents

4. Changes to the Namelists

Group	Name		Meaning	Default
`/GRIBOUT/`	`lcompress_netcdf`	NEW	Determine whether compression is applied or not for a specific file.	`.FALSE.`
`/RUNCTL/`	`asynio_host_mem`	NEW	Total memory available for compute PE to store buffers (in units of GB).	`1`
`/PHYCTL/`	`lterra_urb`	NEW	To switch on/off the urban parameterization.	`.FALSE.`
	`lurbfab`	NEW	To switch on/off the urban fabric (in a bulk approach).	`.TRUE.`
	`itype_ahf`	NEW	Type of anthropogenic heat no anthropogenic heat flux (default) anthropogenic heat according to Flanner (2009)	`1`
	`itype_kbmo_uf`	NEW	Bluff-body thermal roughness length parametrization standard SAI-based turbtran Brutsaert/Kanda Zilitinkevich	`1`
	`itype_eisa`	NEW	Type of evaporation from the impervious surface evaporation just like bare soil no evaporation PDF-based puddle evaporation (Wouters et al., 2015)	`2`
	If the following real namelist variables are set to -1.0, an external parameter field is read from the initial data set. Otherwise, the value given is taken for every grid point in the domain.
	`curb_fr_bld`	NEW	Building area fraction.	`0.67`
	`curb_h_bld`	NEW	Building height.	`15.0`
	`curb_fr_bld`	NEW	Street canyon height/width ratio.	`1.5`
	`curb_ahf`	NEW	Default annual-mean anthropogenic heat flux.	`-1.0`
	`curb_alb_th`	NEW	Default effective thermal albedo of building/pavement substrate.	`0.14`
	`curb_alb_so`	NEW	Default short-wave albedo of building/pavement substrate.	`0.10`
	`curb_hcap`	NEW	Default specific heat times density of building.	`1.25E6`
	`curb_hcon`	NEW	Default building material heat conductivity (former c_ala_bm).	`0.777`
	`cskinc_urb`	NEW	Urban vegetation conductivity factor.	`1000.0`

Back to Contents

5. Changes of Results

Due to the fix for sobs and pabs in radiation_interface (setting to 0.0 before computing solar values), the results are slightly changed: at sun set, we kept a non-zero value for one time step too long in former versions.

Back to Contents

Documentation of the Changes in the COSMO-Model Version 5.10

21.07.2021

Contents:

1. Implemented Modifications for TERRA-URB

(by Ulrich Schättler)

2. Modifications for NetCDF I/O

3. Technical Changes and Bug Fixes

4. Changes to the Namelists

5. Changes of Results

Documentation of the Changes in the COSMO-Model
Version 5.10