Research In Bias Correction Of Data Assimilation For IASI Radiances

Hyperspectral infrared atmospheric sounding is advanced technology of satellite meteorology, which is a hot research area at home and abroad. Infrared Atmospheric Sounding Interferometer (IASI) as an infrared Fourier transform spectrometer was launched in2006. IASI observations are found to be of high quality, particularly in the important region of the15μm CO2temperature-sounding band. Study shows that the additional use of IASI data produces a statistically significant positive impact on forecast quality. Recently, IASI radiances haven’t been made full use of, while the assimilation of microwave radiances is more and more mature. In order to use radiances from IASI, biases between the observed radiances and those simulated from the model first guess must be corrected.In this article, taking into account both IASI instrument characteristics and the specific circumstance of GRAPES, a bias correction scheme for IASI radiances on MetOp-A satellite is developed based on the bias correction scheme for ATOVS radiances integrated in GRAPES and the bias correction scheme for IASI radiance at ECMWF. The scheme utilizes a scan correction and a air-mass correction. Scan bias, which is a relative bias between limb and nadir measurements, is caused by the variation of scan position. Statistics shows that the farther a scan position is away from the nadir point, the larger its scan bias is. And scan biases are almost symmetrical about the nadir point. Instead of latitudinal dependency like microwave radiances, there is a different smaller amplitude bias for each2x2FOR of IASI, that is currently not removed. So scan bias correction can only allow for the effect of scan angle, rather than the dependence of latitude. Air-mass bias is corrected according to the weather condition at that time and represented with predictors computed from the background field. According to the characteristics of the selected channels and data sets and the specific circumstance of GRAPES, four predictors are used,1000-300hPa thickness,200-50hPa thickness,50-20hPa thickness, model surface skin temperature. The reasonableness of the selected predictors is analyzed with canonical correlation analysis method. Then the coefficients of air-mass bias correction are computed by performing a least-squares fit on a large sample.After calculating the coefficients of scan bias correction and air-mass bias correction, and then applying them to the original bias, biases before correction and after are compared in terms of bias of FOV, global bias, bias with variation of latitude, the scatter of observed radiances and those simulated from the background field, PDF of the bias, time series of bias. Verification results show that scan bias correction removes the effect of scan position, and after the air-mass bias correction bias is greatly reduced. At the meanwhile its standard deviation is also reduced to a certain extent. And the distribution of the deviation is more close to the Gaussian form. With the evolution of time, the bias can be maintained at a relatively low level of stability. This work laid a good foundation for IASI data assimilation in GRAPES.