This document of March 2001 is a tentative synthesis of previous contributions.
In dynamical adaptation mode, the initial conditions for ALADIN short-range forecasts are simply interpolated from those of the coupling model (ARPEGE hereafter). Whatever the quality of the interpolation method the resulting fields are smooth with respect to the higher resolution surface forcing in couples model. This leads at least to some spin-up during the early hours of forecast. Moreover, sometimes, some important small-scale features produced by the previous ALADIN forecast are lost and one may face situations when a 24h forecast is better than teh 12h forecast valid at the same time and based on more recent observations.
The idea of blending is to combine the "large scales" resolved by the ARPEGE analysis with the "meso-scale" features provided by the short-range ALADIN forecast. The underlying hypothesis is that the short-wave part of the guess is more realistic than the short-wave part of the interpolated ARPEGE analysis, thanks to a better balance with the high-resolution surface forcing (orography, soil and vegetation characteristics). This might be considered as a meso-scale data assimilation without (directly) suing observations.
The document presents :
- blending of spectral fields by digital filtering
- scale selection in the global assimilation of spectral fields
- principles of digital filter initialization (DFI) blending
- tunings
- blending of gridpoint fields
- gridpoint fields in ARPEGE/ALADIN
- standard initialization of gridpoint variables in ARPEGE
- blending of surface variables
- management of other gridpoint files
- the blending assimilation suite
- perspectives
- combination with 3d variational assimilation
- combination with soil/surface assimilation
- case of nested models
- references and appendix