Statistical Analysis And Variational Data Assimilation Experiments For Observation Errors Of FY-3 Microwave Observation

Together with the assigned background errors,the observation errors provide the weighting of observations in the data assimilation system.The correct specification of observation errors is essential to obtaining optimal analyses.FY-3 seriers are the second generation of Chinese polar-orbiting weather satellite.The data of FY-3 seriers is widely used in the major global numerical weather prediction center and play a more and more important role in the assimilation system.However,the specification of observation errors for FY-3 instruments are relatively simple.For weakening the defects,the research of improving the observation errors of FY-3A/B MicroWave Temperature Sounder(MWTS)and MicroWave Humidity Sounder(MWHS)and its numerical experiments with variational data assimilation are carried out to improve the usage of FY-3 satellite data in data assimilation system.The main content is as follows:Firstly,to correct the biases in satellite data,the paper presents a bias correction scheme based on cycle updating.In the scheme,the repeated data assimilation is started two days prior to the analysis of our actual case study to adaptively tune and spin-up the coefficients.The results show that this scheme can obtain optimal initial coefficients for bias correction and remove biases of FY-3A/B MWTS and MWHS effectively when they are assimilated in the regional model.Secondly,the Desroziers method is first used to evaluate the spatial correlation and inter-channel correlation of the observational errors of MWTS and MWHS on FY-3A and FY-3B in WRF Data Assimilation system(WRFDA).It was found that the error standard deviations of the MWTS and MWHS were less than that used in the WRFDA.MWTS had small inter-channel errors,while MWHS had significant inter-channel errors.The horizontal correlation length scales of MWTS and MWHS were about 120 and 60 km,respectively.A comparison between the diagnosis for instruments onboard two satellites showed that the observation-error characteristics of the MWTS or MWHS were different when they were onboard different satellites.In addition,it was found that the error statistics were dependent on latitudes and scan positions.Finally,to verify the influence of correlated errors of FY-3 satellite data on data assimilation analysis and weather forecast,the NWPSAF(EUMETSAT Satellite Application Facility on Numerical Weather Prediction)one dimensional variational data assimilation(1DVar)and WRFDA three dimensional variational data assimilation(3DVar)are used to conduct the data assimilation and forecast experiments,respectively.The diagnosed results of 1DVar using Desroziers method are similar with the results of WRFDA,but due to the different error characteristics of different error sources in two assimilation system there are still some differences.The influences of considering correlated observation errors on observations and analysis were studied.The results show that using correlated observation error covariance matrix could reduce the difference between observation and analysis,and increase the difference between background analysis,especially for humidity.The study on observation impact shows that accounting correlated errors could increase the degree of freedom of signal(DFS)and the reduction of error variances.These results showed that considering the correlated errors made the analysis value closer to the observation,so that the observation could provide more impact on the analysis.Comparing the analysis with global position system radio occultation(GPSRO)data,we could find that accounting for correlated observation errors reduced the root mean standard error between analysis and GPSRO data.That is to say,considering the correlated observation errors in assimilation could obtain a more accurate analysis.For optimal assimilation of MWTS and MWHS data,A scheme is proposed to consider inter-channel correlated observation errors in WRFDA.This scheme uses the diagnosis results obtained by the Desroziers method as updated observation error covariance matrix.the Cholesky decomposition method is used to compute the inverse of the observation error covariance matrix,further adjustment for the error standard deviations is conducted.The added computational cost of this scheme is little.Accounting for error correlations significantly changes the weighting of observations in the analysis.The use of updated observation-error covariance matrix leads to improvement on forecast skill,especially when the standard deviations of observation error are inflated.Introducing inter-channel observation error correlation also significantly improve track forecasts and slightly improve intensity forecasts.