A Study On Assimilation Of Radiance From FY3's Atmospheric Vertical Sounder System

In the past 2 decades,more and more satellite data have been used in numerical weather model.Great progress has been made about how to use these satellite data in NWP model correctly in recent years,especially in USA,English etc..A research scheme has been developed by NOAA and NASA in the end of 1980s.Up to now,an operational advanced data assimilation/satellite data retrieval/numerical prediction system has been set up successfully.A research from NCEP shows that there is a evident improvement to anomaly correlation from 5 day's numerical prediction while satellite clear radiance has been assimilated into numerical model directly.The 5 day's predicted anomaly correlation increased from 69%to 73%in the northern hemisphere and from 64%to 72%in the southern hemisphere.So,these results have brought a new successful method to introduce satellite data into numerical prediction.A global/regional three dimensional variational assimilation system has been build to assimilate satellite radiance directly in CMA.Results from batch experimentation indicate that great improvement is available to extended predictable forecast of global medium and short numerical prediction while ATOVS radiance is assimilated.Global 3DVAR system(SSI,GSI) of NCEP has been fetched in NMC to assimilate global ATOVS radiance.Analysis of smei-operational examination in 8 months shows that predictable forecast has been extended half a day to global medium numerical prediction in the northern hemisphere in NMC with the assimilation of ATOVS radiance.At the same time,the primary satellite data used in CMA's NWP model are from NOAA ATOVS and cloud track wind of foreign geostationary satellites,and only less than 10%satellite data that are received by CMA have been used in NWP model whereas the rate is up to 85%in foreign model.The 2^nd generation polar orbit meteorological satellite of CMA- FY3 was launched in 26 May,2008.Because capabilities of atmospheric vertical sounder(VASS) in FY3 are very similar to NOAA ATOVS,VASS radiance of FY3 is possible to be used in global/regional numerical model.But it is a flinty problem to us- how to assimilate these data correctly in our assimilations.In this paper,based on theoretic analysis to variational assimilation and radiative transfer in satellite remote sensing,radiative transfer function was reanalyzed to descript a mapping relation from model variables to channel brightness temperature observed by satellite in infrared and microwave.While the function(model) is semi-linearity and precise to the simulated radiance is corresponded to the one calculated by a line-by-line model,it is possible to generate a tangent linear model and a adjiont model corresponding to the radiative transfer model that is going to be used to introduce satellite observed radiance in data assimilation.A research on conventional radiative transfer,numerical approach and deviation analysis shows that,in case of parallel and plane atmosphere,homochromous transmittance could be computed in form of evolved Taylor series with a satisfied precise when pressure level is used during the calculation.Resolved transmittance is decomposed in a component depending on real atmospheric state and the other related to spectral characteristic of satellite sensor.The first part is so-called optical depth predictor and the latter is fast transmittance coefficients of sensor channel.In general, taking a set of selected typical atmospheric profiles as primary input,level-by-level transmittances in high spectral resolution are calculated in case of multi-zenith angles by a line-by-line model.Then a linear regression is used to yield an exclusive set of fast transmittances coefficients from the definition of optical depth predictors in any zenith angle the satellite sensor could arrive.While fundamentals are analysed to set up a fast radiative model,experiments to level-by-level channel optical depth of AIRS are performed.Predictors of mixed gas could provide better channel optical depth in 15μm because peak of simulated channel optical depth locates the adjacent pressure level to the peak of weighting function.Whereas the predicted channel optical depth in 4.3μm is not good as the value in 15μm.To water vapor predictors,great sensitivity,sometimes the simulated channel optical depth is even larger to the values generated by mixed gas predictors, has been found to such lower atmospheric sounding channels that are located in the far wing of CO₂ absorbing line.Also,analysis to the predicted optical depth to channels located in 6.7μm H₂O absorbing line shows that peak values to the optical depth computed by water vapor predictors are normally lower than the maximum channel weighting.So,a conclusion could be get that more corrections should be done to channel optical depth predictors of water vapor in upper atmosphere.When a small disturb is occurred in the primary input profiles of atmospheric temperature,humidity and O₃,deviations will be found to simulated channel brightness temperature.The largest sensitivity is found in middle and upper atmosphere from the predicted channel brightness temperature to absorbers.The lower atmosphere the channel observes,the more affect from water vapor could be found in simulated radiance to each channel.To channel of O₃,the maximum sensitivity comes from O₃ itself.Comparing to simulated channel radiance of IRAS by fat radiative transfer model,a convoluted channel transmittance with IRAS channel spectral response is performed to homochromous transmittance from GENLN2 to generate precise channel radiance. 43 profiles in TIGR-43 are used to perform dependent precision check.Standard deviations are less than 0.1K to air temperature sounding channels(channel 1-7, 14-19) in 15μm and 4.3μm.The maximum values increase to 0.3K to water vapor sounding channels(channel 11-13) in 6.7μm.34 profiles of NESDIS are designed to finish independent precision test.Evident decrease to STD,from 0.3K to 0.1K,is found in channel 10 in 9.6μm.And the STDs are decrease to 0.05K to channels in 15μm and 4.3μm,0.2K to the one in 6.7μm and 0.35 for window channels.Same results could be found to MWTS and MWHS.As a result,they are right to fast transmittance coefficients of FY3 VASS to the fast radiative transfer model.In another way,the fast forward model with coefficients of FY3 VASS could be used in data assimilation that will be done latter.Observations of NOAA16 ATOVS could be used to simulate radiance of FY3 VASS due to the similar parameters in channels'spectral,sensor's scan and NEâ–³T of channel in infrared and microwave between the two satellite sounding packages.All observations of NOAA 16 HIRS/3 could be used to set up observations of FY3 IRAS except the data of two channels not available in HIRS/3.Part of channel observations of AMSU-A could be used to simulate data of MWTS but with different resolution (15 foot points per scan line of MWTS vs 30 foot point of AMSU-A).To the dummy MWHS radiance,observations of all channels for each foot point on a scan line could be found in data of AMSU-B whereas the 8 foot points at the edge of per scan line of MWHS must be ignored.Static analysis is performed to deviations from computed channel brightness temperature of fast forward model to the observations of NOAA16 ATOVS in 6 satellite zenith angles.Systematic errors that are various with satellite zenith angle could not be introduced into the simulated FY3 VASS observations because of the parameter's gap between VASS and ATOVS.Based on the frame of GRAPES regional 3DVAR that has been thrown into operation in CMA,research has been developed to assimilation of radiance of satellite sounding sensors,especially to brightness temperature of FY3 VASS channels. Experiences from foreign operational NWP center show that it is the assimilation of microwave radiance provides the largest improvement to the numerical forecast,in particular to the track prediction of typhoon.After assimilation modules to radiance of FY3 VASS are finished,in a typhoon case,NOAA16 ATOVS radiances are used to build simulated observations of FY3 VASS.Evaluation to impact of assimilating'FY3 VASS radiance'shows that much smaller positive contribution to the numerical track forecast of typhoon mainly because great predigestion to microwave temperature sounding in FY3 MWTS(4 channels of MWTS vs 15 channels of AMSU-A).