| 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
8 Conclusions
During the last five years, the initialisation of prognostic and diagnostic surface variables has been recognised as an important issue for numerical weather prediction. Weaknesses of initialisations based either on first-guess only or climatology have been clearly identified. As a consequence, soil moisture is currently initialised in various operational weather centres using sub-optimal analysis methods and observations from SYNOP reports (ECMWF, UKMO, CMC, Météo-France, HIRLAM). These techniques could be be improved by using analysis methods already tested for atmospheric variables like optimum interpolation in the sequential framework and 4D-Var (which is rather appealing by better accounting for the non-linearities of the problem as well as the temporal distribution of observations).
The other land surface prognostic variables are still crudely initialised. Concerning snow mass analysis at ECMWF, the current scheme could be improved by using a more recent climatology and a more realistic observation operator if snow density can be computed by the land surface parametrization.
Areas not yet explored concern the initialisation of deep soil temperatures for which the time scale of evolution is also much longer than short range forecasts, and the specification of vegetation properties having a seasonal cycle (such as the albedo, the vegetation cover or the leaf area index). Remote sense data that could be useful in that context include quantities such as satellite skin temperatures or the normalized difference vegetation indices (NDVI).
Finally, the examples shown in the paper illustrate the fact that the initialization and parametrization methods have to be developed together. Realistic models are needed to support indirect measurements techniques, e.g., forecast snow density allows a better use of snow depth measurements to initialise snow mass. On the other hand, refinements in the land surface parametrization used by numerical models might be needed in order to extract from new measurement techniques its full information content.
Training Course Notes Front Page
Table of contents
Next Section
Previous Section
 |
12.06.2002 |
 |
© ECMWF |
