Research Of Data Processing Core Technology For Geosynchronous Interferometric Infrared Sounder(GIIRS)

The Geosynchronous Interferometric Infrared Sounder（GIIRS）on-board FY-4A satellite is the first infrared hyperspectral instrument operating in geostationary orbit in the world.This fourier transform spectrometer offers high spectral resolution,which can provide atmospheric temperature and humidity profile information in the vertical direction.Based on interference spectroscopy,the used GIIRS level 0 data preprocessing technology affects the accuracy and quality of level 1 result.The radiance received by detector pixel that deviates from the main optical axis,for the used plane-array detector,has spectral line shift on the basis of the off-axis effect.Due to the influence of discrete sampling,it is difficult to ensure that the actual initial sampling point starts exactly from the position where the optical path difference is 0,and the position often has a random sub-sampling error of up to 0.5 sampling points,which makes the spectrum have the imaginary part,and this is inconsistent with the theoretical one.Cloud detection is a key step before data assimilation,and the impact of cloud pollution needs to be reduced as much as possible to improve the GIIRS data utilization rate.At present,like the most of satellite observation data used in numerical weather forecast mode,the GIIRS cloud detection method only retains the data of clear IFOVs,and the whole channel data should be removed when the corresponding IFOV is contaminated by cloud,leading to the loss of some available channel information.The main content of this paper can be concentrated in two aspects about data processing core technology for GIIRS.The sensitivity analysis of spectrum result based on the ZPD shift is beneficial to research the more accurate ZPD correction scheme by adding phases,and the spectral resolution of all GIIRS’s detector pixels can be uniformly corrected by the relevant information such as the number of sampling points collected from interferograms,the reference laser frequency and the angle of off-axis pixel,and then this experiment compares and analyzes the adjusted spectrum with the initial one;Then,a cloud detection algorithm is set up with GIIRS observation and the RTTOV simulation by adopting the method given by Mc Nally,together with radiation characteristics.The main conclusions are summarized as follows:（1）The zero-adding before the FFT transform substantially increases the optical path difference and artificially improves the spectral resolution.Interpolation resampling causes the loss of spectral information and the spectral shift and radiation deviation are obvious.This paper proposes to truncate the interferogram by controlling the number of sampling points,using DFT transform without adding zeros,in principle to avoid the effect of traditional resampling in the spectral domain.The two methods of subtraction with the cold background have the same spectrum results,and the subtraction method in the interference domain has higher computing efficiency.The simulation results show that:1)When the cold background,blackbody and earth scene interferograms have the same ZPD offsets,the obtained spectrums are almost always consistent with the initial one,and the brightness temperature deviation is within 0.1 K when the offset is 1;2)The situation is different if only the ZPD position of cold background is shifted.In this case,the spectrum processed in the spectral domain has hardly changed,and the deviation is within 0.2 K when the offset is 1.The one processed in the interference domain is very sensitive to the ZPD offset of background interferogram,and the maximum value of the deviation can reach 60 K when the offset is also 1.At this time,it is necessary to correct the ZPD shift under sub-sampling.Based on information such as the reference laser wavelength and the off-axis angle of the probe,by controlling the number of sampling points,the spectral resolution of GIIRS’s off-axis pixel can be corrected.The off-axis effect makes the spectrum shift,and the greater the off-axis angle is,the more sampling points the interferogram requires.After adjustment,the spectral resolution is close to the theoretical value,which is 0.625 cm^-1,and the maximum deviation is within 20 ppm.This kind of rough calibration method is effective,and it has been applied to the latest on-orbit spectral calibration of GIIRS,and has achieved good application results.（2）The proposed cloud detection method selects channels based on the radiation characteristics such as GIIRS sensitivity,and finally obtains nine channels in the long-wave CO2 band,which are 701.25 cm^-1,702.5 cm^-1,703.125 cm^-1,703.75 cm^-1,715.625 cm^-1,740 cm^-1,748.125 cm^-1,746.875 cm^-1 and 750 cm^-1.The 42000 IFOVs used in the experiment are studied to obtain the radiance deviation threshold,the value at 06:00 during the day is 8 m W/m2.sr.cm^-1,and the value at 12:00 at night is 3 m W/m2.sr.cm^-1.The results of the proposed method are generally identical to those from the AGRI CLMs,and because the spatial resolution of GIIRS is significantly greater than AGRI instrument,some minor differences will occur for IFOVs with a certain of broken clouds.Using the derived the heights of cloud top for each IFOV,the clear channels with respect to the cloud IFOVs can be identified with the data utilization increased by around 13.76% in statistics.This proposed cloud detection algorithm can provide an important reference for GIIRS data assimilation.