Pre-processing and Bias Correction of AMSU-A Satellite Data

- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner
Monday, 3 February 2014
Hall C3 (The Georgia World Congress Center )
Sihye Lee, Korea Institute of Atmospheric Prediction Systems, Seoul, South Korea; and J. H. Kim, J. H. Kang, and H. W. Chun

As a part of the KIAPS Observation Processing System (KOPS) for data assimilation of numerical weather prediction model, we have developed the modules for satellite radiance data pre-processing and quality control, which include observation operators to interpolate model state variables into observation space radiance. AMSU-A (Advanced Microwave Sounding Unit-A) Level-1D radiance data was extracted using the ECMWF (European Centre for Medium-Range Weather Forecasts) BUFR (Binary Universal Form for the Representation of meteorological data) decoder and a first guess by model background was calculated with RTTOV10.2. For initial quality checks, the pixels contaminated by large amount of cloud liquid water, heavy precipitation, and sea ice were removed. Different channels for assimilation, rejection, or monitoring were selected for different surface types since the errors from the skin temperature were caused by inaccurate surface emissivity. In radiance data pre-processing, correcting the bias caused by the instruments and radiative transfer model errors is crucial. We have developed off-line bias correction modules in two steps, based on 30-day innovation statistics (observed radiance minus background; O-B). The scan bias correction was calculated individually for each channel, satellite, and scan position. Then a global multiple linear regression of the scan-corrected innovations against several predictors was employed to correct the airmass bias.