| Home > Newsevents > Training > Rcourse_notes > PARAMETRIZATION > SURFACE_ASSIMILATION > |
Land surface assimilation
March 2001
By Jean-François Mahfouf and Pedro Viterbo
European Centre for Medium-Range Weather
Forecasts
Table of contents
Next Section
Previous Section
4 Simple land surface initialisation methods
Due to the difficulties presented above, soil variables are initialised empirically in most operational forecasts models.
A first method for initialising soil variables is to set the analysed soil variables to climatological values without using any model information (through the first-guess). In practice, this is done by adding a relaxation to climatology to the prognostic equations of the land surface scheme (Blondin 1991). The underlying idea of this term is that simple surface schemes can only describe correctly the variables having a short time scale evolution, like the surface soil temperature, and that variables having a time scale longer than a few days must be prescribed. Problems arise from the specification of climatological soil values which are based on indirect atmospheric observations and very simplifed transfer models. These products, like the soil moisture climatology of Mintz and Serafini (1992), have a high level of uncertainty as shown by Viterbo and Beljaars (1995). Another potential problem is that with a relaxation of deep variables towards climatology, seasonal anomalies cannot be forecasted. An example of such deficiency has been shown by Beljaars et al. (1996) with two versions of the ECMWF model for the 1993 US floods.
The other approach to initialise soil variables is to set the analysed values to the first-guess. In that context, an absolute confidence is given to the land surface scheme for evolving its prognostic variables and no control exists to prevent the land surface scheme from drifting to an unrealistic state. Such drifts can occur through positive feedbacks with the atmosphere, in situations where the scheme experiences systematic errors in the atmospheric forcing (too much radiation, too much rainfall, ...) or from a misrepresentation of some land surface processes. The second point can be checked with stand-alone simulations where the atmospheric forcing is prescribed. An example of a drift of the current ECMWF land surface scheme within the data assimilation system is described in the next section.
Training Course Notes Front Page
Table of contents
Next Section
Previous Section
 |
12.06.2002 |
 |
© ECMWF |
