Extended Abstract (1.3M)P3.11
Assimilation of water vapor radiances from geostationary imagers in 4DVAR
Christina B. Köpken, ECMWF, Reading, Berks., United Kingdom; and G. Kelly and J. N. Thépaut
The direct assimilation of water vapor (WV) clear-sky brightness temperatures (CSBT) from geostationary satellites became operational at ECMWF in April 2002 using the four-dimensional variational assimilation (4DVAR) system and data from METEOSAT-7. Currently, CSBT data from GOES-8 and GOES-10, processed by the Cooperative Institute for Meteorological Satellite Studies (CIMSS), are included in pre-operational tests. From both satellite series, data are used in the form of average radiances of the pixels diagnosed as cloud-free within areas of about 50·50 km2 for GOES and about 80·80 km2 for METEOSAT, and with hourly time resolution. The paper describes quality control issues specific to the geostationary imager data as well as assimilation results.
Operational monitoring of CSBT from METEOSAT and GOES shows that the GOES data are at 0.5-1 K bias in good agreement with brightness temperatures simulated from model first guess profiles. METEOSAT CSBT exhibit a significantly larger bias of about 3.5 K. For assimilation, the biases are removed through a statistical bias correction based on model predictors. The monitoring also helps identifying systematic quality problems. For METEOSAT satellites, it reveals image anomalies present around local midnight that are caused by solar stray light intruding into the radiometer. GOES-8 experiences systematically lower brightness temperatures around local midnight, this effect being possibly linked to the calibration. Additionally, CSBT data affected by remaining cloud contamination need to be excluded from assimilation. Here, particularly for the GOES satellites, usage is made of the available window channel data in order to define quality control thresholds.
When assimilated, the CSBT data correct the upper tropospheric humidity field in areas of known model deficiencies. The influence of both METEOSAT and GOES data show consistent patterns. With 4DVAR including the time dimension, it can exploit the high observation frequency of the geostationary imagers which provides information on the movement of humidity patterns. Thus, analysis increments are also derived for the wind field. While the analysis draws well to CSBT data, the fit to other conventional observations does not degrade. In some areas there are small improvements in the fit of the first guess to radiosonde humidity and tropical wind observations. A positive influence is seen particularly in the agreement of the model with other WV radiances from the High-resolution Infrared Sounder, HIRS-12, and the Advanced Microwave Sounding Unit, AMSUB-3. The impact on forecast quality is slightly positive to neutral for different areas of the globe, the positive impact being found on upper level winds and geopotential.
Poster Session 3, Data Assimilation
Tuesday, 11 February 2003, 3:30 PM-5:30 PM
Previous paper Next paper