Joint Poster Session JP1J.10 Towards a 1D+3DVar assimilation of radar reflectivities: Ongoing results

Monday, 24 October 2005
Alvarado F and Atria (Hotel Albuquerque at Old Town)
Olivier Caumont, CNRM, Toulouse, France; and É. Wattrelot, V. Ducrocq, F. Bouttier, C. Guéguen, and G. L'Hénaff

Handout (752.9 kB)

Radar data encounter increasing interest for Numerical Weather Prediction, and in particular for the next generation of NWP models. Indeed, most meteorological centers plan to run operational nonhydrostatic models with resolution 1-4 km before the end of this decade. Radar data will be well placed to provide high-resolution information about wind and precipitation which the verification and initialization of such high-resolution models require. Our aim is to prepare the use of radar data in the future high-resolution nonhydrostatic model (named AROME) of Météo-France.

At present, the emphasis is put on reflectivities, and we first intend to assimilate them as 1D retrievals of temperature and humidity into a 3DVar assimilation system. Within this framework, an observation operator for reflectivites was designed that will be used for both the verification and the assimilation of the AROME model. For this task, the research nonhydrostatic high-resolution MESO-NH model offered the possibility to mimic the under-development AROME model. In particular, the MESO-NH model has an advanced representation of the water cycle, with five hydrometeors (cloudwater, rainwater, primary ice, snow and graupel) governed by a bulk microphysics parameterization which will be implemented in the AROME model. In order to specify the observation operator, we therefore developed a radar simulator as a postprocessing component of MESO-NH, allowing to test and choose among the different possibilities for the observation operator.

Besides, a new processing chain was set up to handle raw reflectivity data and make them available to the assimilation system: these data are first pre-processed in order to remove artefacts and weight each pixel by a quality flag. After that, they are stored as vertical columns on a horizontal Cartesian grid in the Observational DataBase. Then, to be able to assimilate these data, a first mandatory check is made through a monitoring approach.

Currently, we are developing the 1D inversion through twin experiments, that differ from each other in their initial conditions. A first simulation provides a reference solution from which reflectivity observations are simulated. Then, the 1D inversion is applied and its results are assimilated using the second experiment as a background.

- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner