status of the cosmo-software and documentation wg 6: reference version and implementation wg...

Status of the COSMO-Softwareand Documentation

WG 6: Reference Version and Implementation

WG Coordinator: Ulrich Schättler

07.-11.09.2009 11th COSMO General Meeting - Offenbach, Germany 2

Contents

• Source Code Administration: Developments in 2008/09

• DWD plans for the near future• Model Documentation• COSMO Newsletter• COSMO Technical Reports


Source Code Administration

• Starting point last autumn: – COSMO-Model 4.2 and INT2LM 1.7officially distributed– COSMO-Model 4.5 and INT2LM 1.8 as development

versions

• Plans for the year 2008/09:– Consolidate last development versions– Include COSMO-ART– Include 2-moment microphysics– Prepare the COSMO-Model for an easy use of test-cases– Prepare the COSMO-Model for OASIS coupling– Include optimizations by NEC and IBM– Hybrid parallelization



• INT2LM 1.9 (03.09.09)– Implementation of new reference atmosphere– Modifications in the initialization of perturbed pressure– Reading additional external parameters

• 12 monthly ratios of NDVI (DWD, Grib 1 only)• 12 monthly values of plcov, lai, z0 (CLM in NetCDF only)• 12 monthly values of aerosol types (DWD, Grib1 only)• Minimum stomata resistance of plants• Thermal radiative surface emissivity

– Modifications for the vertical interpolation in case of llm2lm– Alternative interpolation of soil humidity– Alternative treatment of IFS humidity– Additional options for itype_w_so_rel– Bug fixes and technical adaptations



• COSMO-Model 4.6/4.7: (12.12.08)

– Correction updates and bug fixes• Nudging: observation processing in NetCDF

• Dynamics: remove analytical computation of reference pressure, which must be done simultaneously in INT2LM

• Soil Model: split IF statements with llandmask and m_styp (which is only defined, if llandmask=.TRUE.)

• Initialize ldiabf_satad=.FALSE. for Leapfrog



• COSMO-Model 4.8: (13.02.09)– New reference atmosphere– Implemented stable integration of the Coriolis terms by

calculating their tendencies in every RK substep (optional)– Included (optional) computation of „old“ 2m temperature;

can be chosen by itype_diag_t2m– Splitted computation of dynamical and convective gusts

(vmax_10m still is the sum of both)– Included several options for computing dynamical gusts

with itype_diag_gusts– Technical adaptations and bug fixes– Call to SR global_values only, if number of

processors > 1



• COSMO-Model 4.9: (15.07.09)– Implementation of COSMO-ART

• This is an initial release which most probably needs further testing• Introduced ifdefs for a clean splitting of this component (-DCOSMOART,

-DPOLLEN); please note: lmorg.F90 lmorg.f90)• Source Code is not distributed with the COSMO-Software, but only by KIT

(the Karlsruhe Institute of Technology)

– Changes in the dynamics• More accurate (and expensive) discretization of metric terms for the

Runge-Kutta scheme• Implemented option for using potential temperature as advected variable• Introduced itype_lbc_qrsg to control the treatment of humidity

variables at the boundary zones

– Technical adaptations and bug fixes


• Some routines have been (partly) reorganized for better vectorization (e.g. synthetic satellite images, radiation on a coarser grid)

• Implemented 3D versions for diffusion and advection operators. This had negative effects on the IBM (destruction of cache re-use), therefore it is implemented with an ifdef NECSX (default: FALSE)

• In some routines, a loop collapsing has been implemented „by hand“ (using above ifdef)

Timings for a 12 hour COSMO-DE forecast (should take about 11 minutes on as few processors as possible)

NEC SX-9

112 PE

IBM pwr6

1616 PE

Time to Optimization

4.2 14951.5 720.5

4.4 2021.8 659.1 3 days

4.8 1024.2 678.5 3 weeks

4.9 634.2 679.8 3 months


Vector Optimizations:

The COSMO-Model still is a performance portable model!


INT2LM COSMO-ModelNew Reference Atmosphere done done

COSMO-ART, PollenInterpolation of Pollen still missing

done

New External Parameters doneUse of new parameters still missing

Optimizations by NEC, IBM –NEC: nearly done

IBM: still missing

2-Moment Microphysics –

Easy Use of Test Cases –

OASIS coupling –

Hybrid Parallelization

Summary of Software Status




• Actions in the next weeks– Testing of the new versions of INT2LM and the

COSMO-Model; additional work, if necessary– Distribution of the new versions to all partners– Preparation of new external parameter data sets

with additional fields– Every COSMO partner updates its operational

implementation– Every COSMO partner takes care of the update

of the corresponding COSMO web pages!

DWD Plans for the near future


COSMO-EU

Grid spacing: 7 km

Layers: 40

Forecast range:

78 h at 00 and 12 UTC


1 grid element: 49 km2

COSMO-DE

Grid spacing: 2.8 km

Layers: 50

Forecast range:

21 h at 00, 03, 06, 09,

12, 15, 18, 21 UTC


Global model GME

Grid spacing: 40 km

Layers: 40

Forecast range:




Model System at DWD


• Grid spacing reduced from 40 to 30 km.• Number of layers increased from 40 to 60.• Number of gridpoints increased from 14.7 to 39.3 Mill.• Time step decreased from 133.33 s to 100.00 s.• Cost of model increased by a factor of more than four.• Uppermost model layer moved from 10 hPa to 5 hPa.

• Prognostic precipitation without advection of rain (qr) and snow (qs).

• Tiedtke convection scheme replaced by Bechtold scheme.

• Variational soil moisture assimilation scheme (SMA).

GME 30 km / 60 Levels


30 km

Max: 2796 m

40 kmMax: 2551 m


GME L40GME L60ECMWF L91ECMWF L60


• GME data volume to be transferred will increase by a factor of at least 2.7 = (60/40) *(256/192)2.

• Including rain (qr) and snow (qs) in the initial and lateral boundary data will increase the factor to about 3.3.

• This might imply an increase of the Internet connection bandwidth at the NMHS receiving GME data!

• The improved quality of GME will lead to an improved quality of the COSMO-Model forecasts, especially for near surface weather conditions.

Impact on Users of GME Data

Model DocumentationCOSMO Newsletter

COSMO Technical Reports


Model Documentation

• Dynamics and Numerics, Physics, Assimilation: – no updates have been done (at least none have been

reported)

• User Guides for INT2LM, COSMO-Model: – Updates for the new / changed Namelist variables

• Documenting the procedures for INT2LM has started because of the Interoperability Programme


COSMO Newsletter

• COSMO Newsletter No. 9 has been published in December 2008


COSMO Technical Report

• TR No. 12: Adriano Raspanti: Final Report on Priority Project

VERSUS (Verification System Unified Survey)

• TR No. 13:Chiara Marsigli: Final Report on Priority Project

SREPS (Short Range Ensemble Prediction System)


THE END

Thank you for your Attention

status of the cosmo-software and documentation wg 6: reference version and implementation wg...

Documents

f90source code

int2lmsoil model

implementation of cosmo

reference version

only12 monthly values

dwd plans

bug fixescall

bug fixesnudging