Data assimilation
An important research topic at the NWP department is the development of a data assimilation system for the ALADIN (link) model. Data assimilation means the generation of initial conditions for the model using all available information about the current state of the atmosphere. Main information are meteorological observations, the background (which is usually a model forecast valid for the time we want to generate the initial conditions) and their reliability (error covariance matrices). In case of limited are models (LAM), data assimilation can be replaced by an interpolation of the initial conditions of a global or another LAM model. Most of the LAM models are initialised this way. The problem with this method is that the interpolated initial conditions, does not describe on the actual small (meso) scale meteorological systems, as the information comes from a poorer resolution. It is important thus, to create the initial conditions directly on the high-resolution grid, which can be done only by running a local data assimilation. Of course we obtain extra information running a high-resolution data assimilation, only if the resolution of the observation network is similarly high as the model grid resolution. The density of observations can be increased, by using new observation techniques (aircraft, satellite observations, wind profilers) in the data assimilation system.
The work on data assimilation started in June 2000 in the Hungarian Meteorological Service. The first experimental assimilation cycle was run during the summer 2001 on an SGI Origin 2000 (16 processors) computer. A more developed version of the system was installed on the presently operative machine of the service (IBM p690, 32 processors) as a quasi-operational application in November 2002. During 2003-2004 the development continued by including new observations (aircrafts, different satellite observations, wind profilers) to the system. The data assimilation system is used operationally since May 2005.
The main characteristics of the system:
| Scheme | Three dimensional variational (3DVAR) analysis |
| Background error covariances | NMC method |
| Background | 6 hour ALADIN forecast |
| Assimilation cycle | 6 hour (4 long cut-off + 2 short cut-off analyses per day at 0, 6, 12 and 18 UTC) |
| Lateral boundary conditions in the cycle | ARPEGE long cut-off analyses |
| Production forecast | 48 hour ALADIN forecasts starting from the 00 and 12 UTC analyses of the data assimilation system |
The data assimilation cycle
Used observations:
| Surface observations (SYNOP) | Surface pressure |
| Upper air radiosonde observations (TEMP) | Temperature, wind, geopotential, specific humidity |
| Aircraft observations (AMDAR) | Temperature, wind |
| Satellite observations (ATOVS/AMSU-A, MSG/AMV) | Temperature, wind |
Future developments:
The main development in the future will still focus on the use of new observation types. Also, it is planned, that the NMC background error covariance matrix will be replaced by a new one computed by an Ensemble method. For the longer term, the 3DVAR scheme is planned to be replaced by a 3D-FGAT method and a 4DVAR even later.
Documentation:
The following paper can be interesting for those who are interested in data assimilation and in the data assimilation system of the ALADIN model (in Hungarian):
A few documents about the data assimilation system of the ALADIN/HU model:
The following websites offer detailed information on data assimilation generally:
>
Main activities 
