Poster Session P6.2 Validation of precipitable water from ECMWF model with GPS data during the MAP SOP

Monday, 21 June 2004
Olivier M. Bock, IPSL/CNRS, Paris, France; and C. Flamant, E. Richard, C. Keil, and M. N. Bouin

Handout (294.2 kB)

Precipitable water (PW) estimates from GPS data are used to validate operational analyses and re-analyses from European Centre for Medium-Range Weather Forecasts (ECMWF) during the MAP-SOP (Mesoscale Alpine Programme - Special Observing Period). GPS data are analysed with GAMIT GPS software for the period 7 September 1999 - 16 November 1999. The accuracy of GPS derived PW is first assessed through a comparison with high-resolution radiosonde data from the MAP database ( The agreement between both techniques is about 2 kg/m2, on average over 10 radiosonde stations throughout the Alps, but is shown to vary strongly from one site to another. Errors from surface pressure measurements, used to convert GPS delay retrievals to PW, are highlighted. The MAP re-analyses were produced with the T511/159L60 version of ECMWF model, using a 4D-VAR global assimilation system, with special MAP observations (windprofilers, high-resolution radiosondes, and surface observations). Compared to GPS data, both analyses and re-analyses are globally too dry over the MAP domain. Biases, evaluated at the sample GPS network (20 stations throughout the Alps), are about -3 kg/m2 for analyses and -1.7 kg/m2 for re-analyses. These biases are shown to be correlated with differences in the model’s orography and real topography (up to 1000 m at some GPS stations). Statistics in PW biases are presented at a number of GPS stations, with emphasis on special events encountered during the SOP such as heavy precipitation and foehn. GPS data are shown to provide useful information on rapid spatial and temporal evolution of PW content in the lower troposphere which might successfully improve analyses in future operational assimilation procedures.
- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner