omsz
>>
  Hungarian Meteorological Service  founded: 1870
Research and development | Numerical Weather Prediction  | Analysis of the Atmospheric Environment | 
Felbontás: KicsiFelbontás: NormálFelbontás: KözepesFelbontás: NagyobbFelbontás: Nagy  

AROME

AROME is a non-hidrostatic mesoscale NWP model with sophysticated physical parametrization package. The AROME project was started in 2000 at Meteo-France when the need arised to develop a high resolution limited area model espcially for nowcasting purposes.

Finally the decision was made at 2002 to create AROME from the dynamical cernel of ALADIN and the physical parametrization package of MESO-NH research model. For assimilation purposes it was planned to use the 3dVar system that already exists in ALADIN. (It is possible that instead the more precise 3dFGAT assimilation procedure will be used.)

Schematic figures of the combination of the 2 models
Figure 1
Schematic figures of the combination of the 2 models

The first prototype of the model was created in 2003. A lot of technical and scientific problem was found since the two models differ from each other in many respects. In the combination of the two model the main aspect was to make the possible smallest change in the already existing code so the main goal was to create interface routines to connect the two models.

Currently AROME is developed in parallel with ARPEGE/ALADIN and MESO-NH. This means that the source code of AROME is incorporated into the source of ARPEGE/ALADIN so during the phasing - implementation of developments into export cycles - this model version is also updated. At the moment AROME is running day-by-day only at Meteo-France but other ALADIN countries also plan to participate in the development and future operativ use. It is expected that AROME becames operational at 2008.

Main characteristics of AROME

  • Limited area, spectral, non-hydrostatic model
  • Semi-Lagrange advection scheme is used
  • Two time level semi-implicit time scheme is used
  • Due to limited area it needs Lateral Boundary Conditions (LBC) for which Davies coupling scheme is used.
  • NH dynamic is needed due to high horizontal resolution since at this resolution the hydrostatic assumption (to neglect the time derivative of vertical velocity) is not valid any more. Because of the NH dynamics the introduction of 2 new prognostic variables is needed: w (vertical velocity) and p (NH pressure).
  • 4 parametrization schemes were adapted from MESO-NH model. (See more detail in MESO-NH documentation).
Microphysics
  • Uses 5 prognostic variables: vapor, cloud water, cloud ice, rain, snow, graupel (it is possible to introduce hail later) - Distinguishes warm-cloud and mixed-phase microphysic processes.
  • Distinguishes warm-cloud and mixed-phase microphysic processes.
  • Warm-cloud processes: autoconversion, accretion, evaporation, sedimentation.
  • Mixed-phase processes: nuleation, ice-crystal autoconversion, aggregation, raindrop contact freezing, riming, melting, deposition, Bergeron-Findeisen effect, ice-crystal sedimentation.
  • Adjustment to saturation: to avoid oversaturation, and to correct negative mixing ratios.
Schematic figures of the microphysics processes
Figure 2
Schematic figures of the microphysics processes

Surface processes

  • 4 surface types: sea, inland water, nature, town
  • Sea and inland water parametrization is not yet worked out (only Charnock formel is applied)
  • Over nature the 3 layer ISBA scheme is used (similar to what is applied in ALADIN, but more precise, e.g. 3 layer snow scheme).
  • Over town TEB (Town Energy Budget) scheme is used. Town consists of one road, two walls and roof.
  • The surface scheme is fully externalized that means that it is independent (structurally) of the atmospheric model it only means atmospheric forcing (radiation fluxes, T2m, RH2m, U10, ...).
Turbulence
  • In MESO-NH a 3D turbulence scheme is implemented but in AROME a simplified version is used with 1D (vertical) turbulent terms only, since over 1km one can assume that the gradient of the horizontal turbulent fluxes can be neglected.
  • Prognostic equation for TKE (turbulent kinetic energy) other (non-isotropic) second order moments are diagnosed
  • 1.5 order closure based on the computation of mixing length
Convection

At 2.5km resolution there is no need to parametrize convection since at this resolution one all the cloudy processes are described explicitely. However AROME should be able to run also on coarser resolution where there is already a need of convection parametrization.

  • Convection scheme is based on mass-flux approximation
  • Updraft, downdraft entrainment of environment, and detrainment into the environment is assumed.
  • It is based on CAPE closure (Fritsch Chapell closure) which is based on the assumption that CAPE in a grid element is removed within an adjustment period.
Radiation
  • Morcrette radiation scheme is used (as in the ECMWF model).
Chemistry
  • The newest AROME version already possesses chemistry scheme.

Create initial and lateral boundary conditions

To run the model one needs initial and latery boundary conditions (LBC). The LBC can be created from ARPEGE global model forecast or from an ALADIN forecast (that was integrated on a larger domain) with the help of ALADIN 927 configuration (this means basically an interpolation from input to target geometry).

The initial upper-air fields can be interpolated from an ALADIN or from ARPEGE run.

The externalized surface scheme on the other hand needs a separate initial file. The creation of this file is presently a bit complicated and can only be done on the machine of Meteo-France. The ALADIN (FA format) file containing the initial surface fields should be converted to grib format than with the help of a MESO-NH tool to interpolate from the GRIB file to the target AROME geometry. For the interpolation one also needs an other file containing physiographic fields (albedo, cover type, orography, etc.) for all months of the year. This can be created by an other MESO-NH procedure. These letter files should be created only once for a given domain, e.g. these are independent of the date of model integration.

Local application

We plan to use AROME at HMS in the future so we installed the code (version: cy29_t1) on IBM p655 cluster. Presently we test it mainly by analysing case studies.

Properties of the model version at HMS:

DomainHungary
Horizontal resolution2,5 km
Vertical number
of levels
49
LBCALADIN/HU, ARPEGE, ALADIN/3d-var
Graphical representationnetcdf,gmt,diaprog
Orografy of the AROME model

Post processing and graphical visualization

The visualization is a bit more complicated than in case of ALADIN since the model outputs are not stored in a single file but upper-air fileds and surface fields are stored in separate files (in different format).

The upper-air fields are stored in FA format (similar to ALADIN). These fields can be post-processed with ALADIN fullpos configuration (to convert from spectral to grid-point space). The post-processed files can than be converted to netcdf format and visualized in the HAWK system.

he files containing the surface fields are in the so called LFI format and can be visualized in different way:

  • With the diaprog program (it is based on NCARG library) (Figure 3).
  • One can also convert the LFI file to netcdf and than visualize in HAWK. The advantage of this procedure is that we can than compare the result with other model output (Figure 4).
  • With the diaprog program one can convert the file to ascii and then plot it with Gmt software
  • Finally there is a possibility to convert LFI format to GRIB and then visualize it with Metview. This procedure is however not yet working at HMS.
Surface fields visualization with diaprog program
Figure 3
Surface fields visualization with diaprog program

The result of an AROME case study: +24h accumulated total precipitation

AROME result visualization in HAWK
Figure 4
AROME result visualization in HAWK

Upper left: ALADIN forecast, upper right: AROME, bottom right: radar image