Most of the LACE partners were represented in Prague during this quarter (10 people from 5 countries) and worked on various subjects.
The idea (initially suggested by Vincent Cassé) of blending of long waves from the ARPEGE analysis with short-wave signal obtained by the ALADIN short-range forecast (the guess) using digital filter has been described in the previous ALADIN Newsletter.
The further research on blending has been carried on, focusing on the investigation whether the digital filter initialisation of the initial state of ALADIN (i.e. after blending) is still needed (so called external DFI). The results are not completely conclusive but suggest that certain initialisation will be still necessary in the production forecast due to presence of noise in the first few hours of the model integration. However, the filtering may be somehow weaker than it is in the case of usual dynamical adaptation because the small-scale features built by blending are already partly filtered out by the internal DFI. Another important result is that external DFI is not needed within the blending cycle for short range forecasts preparing the guess. A series of tests aiming to determine the weaker filter will be launched once the parameters of blending (lower truncation and properties of internal DFI) are tuned on several cases.
Treatment of the surface prognostic variables represents another problem in the blending. The suggested solution takes the difference of the ARPEGE guess and analysis projected to the ALADIN grid (via EE927) giving the ARPEGE analysis increment and adding this increment to the ALADIN guess.
The research on the blending is currently carried on by Dijana Klaric (HR), Stjepan Ivatek-Sahdan (HR), Radmila Bubnová (CZ) and Jean-François Geleyn (FR).
The 3D VAR development in ALADIN went on in RC LACE Prague Team.
Results of the summer ALADIN 3DVAR action at Toulouse was ported to ALADIN/LACE library. Meanwhile the necessary background error statistics for LACE domain (using the NCEP method) were calculated over 90 days by a common effort of Claude Fischer (FR), Mária Siroká (SK) and Radmila Bubnová (CZ). Currently the single observation experiments are working and experiments with the full observation dataset can be run now. Further research is oriented towards deeper diagnostics on the background error statistics.
Yong Wang (AT) started the study of the influence of the vertical interpolation procedure APACHE on the structure of the background errors of ALDIN/LACE. The results are not yet conclusive due to the lack of sufficient data series on the archive machine.
Couple of days in September two experts from Météo-France worked with RC LACE Prague Team with the aim to port nearly the complete ARPEGE data assimilation/forecast suite to the local system. Some work is still necessary to optimize the ARPEGE suite and to synchronize it with the ALADIN suite (intended implementation in the new, SMS-based operational suite), address some issues of shared memory (TOVS handling and minimization are not implemented for multitasking) but the essential part is done. It is estimated that the remaining work will be finished before the end of 1999 and that the next year RC LACE computing center would be able to cope with some exceptional perturbations in the ARPEGE production.
Jozef Vivoda (SK) concentrated in September on the resuscitation of the 2D vertical plane version of ALADIN to further enable the idealised experiments, especially focusing on the non-hydrostatic semi-Lagrangian scheme. The 2D version works now in AL11. He also introduced a new feature of the sponge on the model top and developed a new diagnostic tool computing the vertical velocity, surface drag and the vertical momentum flux.
David Dvorák (CZ) currently runs various configurations of the semi-Lagrangian scheme on the idealised case of the frontal collapse with the focus on the scheme stability.
Meriem Zitouni (HR) ported the last developments in the diagpack to the local library and carried on the work on the implementation of the land-sea contrast into the surface structure function allowing more appropriate analysis along the coastal lines.
Since diagpack is designed to provide information to forecasters some convection diagnostic indices calculated from the analyses state would be welcome. These indices are usually two-dimensional so their incorporation into FullPos would not be appropriate. Instead a tool for computation of some stability indices has been developed by Harald Seidel (AT). This tool works on the fullpos files and the results are added to the input file. The work will go on.
Some minor improvements were made in the verification package veral. An embryo of a precipitation subpackage of veral was developed but it still suffers by some missing data in the local CMAFOC observation files.
The verification generally suffered by the local lack of a steady archiving facility.
|
Home |