FY3B/IRAS Data Bias Correction,Cloud Detection,Quality Control And Assimilation Test

The thesis firstly reviews method of satellite data bias correction at home and abroad. On the basis, applying the idea of Harris and Kelly’s offline bias correction to correct the Infrared Atmospheric Sounder20channels brightness temperature bias of Feng-Yun the3rd (B) weather satellite in our country, which play key role to the numerical weather prediction. Bias correction contains scan bias and air-mass bias correction in two steps. Scan bias is attributed to instrument observed value varies with the instrument scan position. Statistics show that the scan position the farther away from the nadir, the greater the scan bias, and about nadir symmetry. Air-mass bias is corrected according the prevailing weather condition, and it is quantitatively represented with so-called "predictors", In this study, predictors are used:(1)1000-300hPa thickness;(2)200-50hPa thickness;(3)model surface temperature;(4)total perceptible water. Statistics sample data from18UTC,24th Dec,2012to00UTC,8th Jan,2013. Preliminary quality control is included the process of sample statistics, obtained the statistical bias correction coefficients is applied to correct IRAS channel brightness temperature of the next time. Through comparing the bias probability density function and bias stability change over time, which before and after bias correction, the correction is reasonable. FY3B/IRAS channel brightness temperature bias more Gaussian after bias correction, and meet the requirements of the variational assimilation.Secondly, IRAS data cloud detection is considered. Different from the commonly used channel brightness temperature threshold value method, this thesis adopts a smooth gradient method coupling channel SNR to study cloud detection. Adopting cloud imagery information of FY-2E geostationary satellite to carry on the mechanism analysis of brightness temperature bias threshold and smooth gradient method. Statistics data is between00UTC,9th Jan,2013and00UTC,14TH Jan,2013. IRAS channel brightness temperature bias obey Gaussian distribution after cloud detection. Further, minimum residual method is adopted to obtain the cloud parameter (effective cloud fraction) of FOV, late for Globe and Regional Assimilation and Prediction System, which assimilated IRAS cloud radiation to lay the foundation. We also study the IRAS data quality control. Adopting double weight method and principal component analysis (PCA) to perform data quality control. Double weight method can effectively eliminate outliers, who larger leave brightness temperature bias mean. During the process of adopting double weight method, statistics sample data, which from OOUTC,26th Dec,2012to18UTC,4th Jan,2013. From each time the channel brightness temperature bias mean, standard deviation, the percentage of outlier data, and skewness and kurtosis, the conclusion is that using double weight method to carry on quality control is feasible. PCA could grasp the structural relationships between data variables. Based on the PCA, using Hoaglin-T parameters to measure the sample outlier value, which could effectively eliminate the "outlier line" of ERAS channel brightness temperature. Considering the timely of GRAPES assimilated the IRAS data, different from GRAPES conventional hops, jumper and "box" method, this study adopts principal component analysis for IRAS data sparse.Finally, we discuss GRAPES assimilation IRAS data test. Including IRAS channel observation error re-estimated and diagnoses the influence of IRAS data to the GRAPES analysis field based on the information entropy and degrees of freedom for signal, which numerical approximation idea is adopted during the specific implementation process. Precipitation case of assimilation and controlled trials (combination of conventional data, microwave temperature data of Metop and FY3B/IRAS data) are carried on, show that assimilation IRAS data is feasible. This process of assimilation test results about quality control of IRAS data, and the analysis incremental of temperature, humidity, U and V wind filed are given, further given the mode precipitation forecast of the assimilation and control trials of GRAPES. Further, the paper has carried on the IRAS ideal data (IRAS channel simulated brightness temperature through adding the non-Gaussian random disturbance) experiment, whose errors follow non-Gaussian, theory of generalized variational assimilation is explored and researched. During specific implementation process, selecting the M-estimators to improve the objective functional of the classical variational assimilation, during the variational minimization process coupled with quality control, dynamic adjustment the contribution rate of IRAS data to objective functional. In the future, we intend to add the idea of the generalized variational to the GRAPES-3DVar assimilation module to study actual case.