Proceedings - C-SRNWP Project

Figure 7: 3 hour aggregated precipitations (mm/3h) as deduced from radar data (left), and as
modelled by the Arome 2.5 km model (right). In the latter case, the initial state was obtained
by a preliminary 12 hour long RUC with a 3 hourly 3D-VAR data analysis of surface
observations, aircraft, radiosonde and satellite radiances (ATOVS, HIRS, MSG/SEVIRI).
Top: 0-3 hour lead time; bottom: 3-6 hour lead time.
5. Surface assimilation
There exist for the time being a variety of approaches for initializing surface fields. It is
planned to derive Sea Surface Temperature and possibly ice cover sheet information from upto-date
satellite observations (NESDIS and Eumetsat products). Furthermore, in France for
instance, land surface fields could be initialized straight away from the coupled
soil/surface/hydrology system “SAFRAN”. For a more genuine data analysis inside the LAM
NWP model, we still do have the CANARI Optimal Interpolation code, which actually now is
used in Czech Republic (in link with upper-air blending of Arpège fields) and Morocco. Most
likely, France and Hungary also are going to test its implementation in link with the 3D-VAR
upper-air analysis.
For the future developments, provisional plans are to integrate a specific surface field analysis
(first using an Optimal Interpolation approach, then using some Dynamically adaptive
algorithm derived from a 2D-VAR approach). This system would be coded inside the
externalized surface scheme SURFEX, which is to replace our present operational surface
model ISBA in the coming years. Additionally, the development of a spatialization tool for
2D horizontal fields (having strong heterogeneities) is planned. This tool would allow the
analysis of screen-level fields.
Further plans for surface assimilation will be discussed and decided within the frame of the
Aladin/Hirlam collaboration.
6. Plans for 2007 and outlook
►Background error covariances:
• Errors of the day for screening
• Gridpoint maps of Sigma_b’s for minimization
• Filtering of the ensemble background errors (low-pass)
• Native Aladin ensemble of analyses
• Derive from those an Arome ensemble forecast
• Wavelets (A. Deckmyn, Belgium; T. Landelius, Hirlam)
200