Serial of Applications of Satellite Observations An Introduction to Hyper-spectral Infrared Sounders Onboard Polar-orbiting Meteorological Satellites

Yin Mengtao　Zou Xiaolei,(　Department of Earth, Ocean and Atmospheric Science, Florida State University, USA 　Center of Data Assimilation for Research and Application, Nanjing University of Information and Science & Technology, Nanjing 0044)

Serial of Applications of Satellite ObservationsAn Introduction to Hyper-spectral Infrared Sounders Onboard Polar-orbiting Meteorological Satellites

Yin Mengtao1Zou Xiaolei1,2
(1Department of Earth, Ocean and Atmospheric Science, Florida State University, USA 2Center of Data Assimilation for Research and Application, Nanjing University of Information and Science & Technology, Nanjing 210044)

Polar-orbiting meteorological satellites circulate above the Earth at about 800-km altitude, completing 14 orbits daily. A single orbit takes about 100 minutes. Each polar-orbiting satellite provides observations on the so-called descending (ascending) node when moving from north (south) to south (north). The local time for all the descending nodes to cross the equator remains constant for a fixed polar-orbiting satellite, although their longitudes are different. The same is true of ascending nodes. Different from a geostationary satellite that provides temporally continuous observations within a limited spatial and spectral domain[1], a polar-orbiting meteorological satellite can provide global coverage in multiple visible, infrared and microwave bands twice daily. Observations from polar-orbiting meteorological satellites have played important roles in numerical weather prediction (NWP), climate study and product retrieval of meteorological variables.

Polar-orbiting meteorological satellites with infrared sounders onboard are launched into early-morning, morning and afternoon orbits. The descending nodes of early-morning and morning orbits pass the equator at about 6:00 AM and 10:00 AM local equatorial crossing time (LECT), respectively. The LECT of ascending nodes of afternoon orbits is at about 1:00 PM local time①. National Oceanic and Atmospheric Administration (NOAA) started its Polar Orbiting Environmental Satellite (POES) series in 1978. NOAA-13 failed to operate in an afternoon orbit. NOAA-6/8/10/12/15 are earlymorning satellites. NOAA-17 is a morning satellite. The remaining NOAA POES, including NOAA-18/19 and Suomi NPP, are afternoon satellites. Other countries also operated polar-orbiting meteorological satellites. Other countries also operated polar-orbiting meteorological satellites. Two morning-orbiting satellites MetOp-A/ B has been launched by European Organization for the Exploitation of Meteorological Satellites (EUMETSAT) since 2006. The Chinese Fengyun-3 (FY-3) polar-orbiting meteorological satellite series started from 2008 on. FY-3A/B are experimental meteorological satellites, and FY-3C is an operational meteorological satellite. FY-3A/C are morning satellites and were launched in May 2008 and September 2013, respectively. FY-3B is an afternoon satellite and was launched in November 2010. China plans to launch an early-morning-orbiting satellite, FY-3E, in 2018. By then, the FY-3 satellites will provide global observations with three different orbits, i.e., early-morning, morning and afternoon orbits. Table 1 provides a list of the current operational polar-orbiting meteorological satellites with their launch dates, infrared sounders, status and agencies.

The first High-resolution Infrared Radiometer Sounder (HIRS) was onboard Nimbous-6 satellite, which was launched in 1975. HIRS had 16 infrared channels and one visible channel. The follow-up HIRS instruments, HIRS/2/3/4 onboard the NOAA-6 to 19 had 19 infrared and one visible channel. Table 2 lists the central wavenumbers and the bandwidth at each channel of the first HIRS and HIRS/2/3/4. It is seen that the 1219.51 cm-1channel of the first HIRS was removed from HIRS/2/3/4. Four new infrared channels were added to HIRS/2/3/4 with their central wavenumbers at801.92, 1029.87, 1364.26 and 2500.00 cm-1, respectively.

Hyper-spectral infrared sounders, include Atmospheric Infrared Sounder (AIRS) onboard the National Aeronautics and Space Administration (NASA) Aqua satellite since 2002, Infrared Atmospheric Sounding Interferometer (IASI) onboard MetOp-A/B satellite, and Cross-track Infrared Sounder (CrIS) onboard Suomi NPP satellite since 2010, represent the advancement in infrared sounding technology. The spectral resolutions of AIRS, IASI and CrIS are much higher than HIRS instruments. AIRS has 2378 channels covering a spectral range from 650 to 2700 cm-1. IASI provides radiance measurements with 8461 channels that are located in a spectral range from 600 to 2800cm-1. CrIS provides radiance measurements at a total of 1305 channels, which are divided into longwave (650 to 1095 cm-1), midwave(1210 to 1750 cm-1) and shortwave (2155 to 2550 cm-1) bands②. HIRS/2/3/4 only provide radiance measurements at 19 channels from 650 to 2700 cm-1. The above three hyper-spectral infrared sounders have different spectral resolutions. The spectral resolution (Δν) of AIRS increases with increasing central wavenumber (ν) and is inversely proportional to a constant spectral resolving power (R)[2], i.e.:

where R=1200. IASI has a constant spectral resolution of 0.25cm-1over its entire observing spectral range. CrIS has a constant spectral resolution at each of its three spectral bands. The spectral resolutions of longwave, midwave and shortwave bands are 0.625, 1.25and 2.5 cm-1, respectively. The full spectral resolution (FSR) mode allows CrIS to have a spectral resolution of 0.625 cm-1over the full spectral range of CrIS[3]. The spectral resolution of CrIS shortwave band is much coarser than that of the corresponding IASI band. It was found to be difficult to apply the absolute frequency calibration in the CrIS shortwave band by employing IASI shortwave observations during the post-launch period due to the differences of spectral resolutions between two instruments[4]. By utilizing the FSR mode, the absolute frequency calibration in the CrIS shortwave band becomes straightforward using IASI[5].

CrIS is the newest hyper-spectral infrared sounder and will be taken as an example for further discussions. CrIS is a cross-track scanning instrument. A single scanline of CrIS consists of 30 fields of regard (FORs), with each FOR consisting of nine fields of view (FOVs). As the satellite Suomi NPP moves in the along-track direction from South to North, the hyper-spectra infrared sounder CrIS observed 30 FORs in the cross-track direction from West to East. The horizontal resolution of CrIS observations is determined mainly by the beam width. The scan angle and the altitude of satellite also have an impact on CrIS data resolution. The beam width for CrIS is 0.963°, corresponding to an FOV with a 14-km diameter at nadir. The sizes and distributions of FOVs and FORs along a single scanline of CrIS near the equator are shown in Figure 1. The footprints of the FOV and the FOR in the figure were calculated based on the center longitude and latitude of a particular FOV, the beam width as well as the zenith angle, the azimuth angle and the altitude of Suomi NPP satellite. A detailed description of the mathematical formula for the calculation of FOR and FOV sizes can be found in the appendix. From Figure 1 it is seen that the sizes of the FOV and the FOR increase with scan angle, confirming that the horizontal resolution of CrIS observations is the highest at nadir and decreases with an increasing scan angle.An overlap is found for CrIS FOVs with large scan angles in the cross-track direction. An enlarged view of the nine FOVs for FORs 1, 15 and 30 in Figure 1 are displayed in Figure 2. FOV 5 isthe center FOV, FOVs 1, 3, 7 and 9 arecorner FOVs, and the FOVs 2, 4, 6 and 8 areside FOVs[3].The corner and side FOVs rotate around the center FOV counter-clockwise from the west to the east for a single scanline. It is reminded that there is no overlap between neighboring FOVs within a single FOR.

The cross-track and along-track diameters of the nine FOVs along the same scanline of CrIS in Figure 1 are provided in Figure 3. It is worth noticing that the crosstrack diameters of the nine FOVs increase with scan angle more greatly than the along-track diameters. The crosstrack diameters are slightly smaller than the along-track diameters at nadir due to a larger latitudinal distortion of the nine FOVs in the along-track direction. The latitudinal distortion is caused by the larger radius of the Earth at the equator (6378.1 km) than at the pole (6356.8 km). It leads to a higher altitude of the Suomi NPP satellite at higher latitudes than low latitudes such that the FOV observed by the CrIS instrument is larger at higher latitudes. At the largest scan angle, the minimum cross-track and maximum along-track diameter of the FOVs is about 39 and 25 km, respectively.

Similar to CrIS, all HIRS series and AIRS are crosstrack scanning instruments. There are 42 FOVs and 56 FOVs along a single scanline of the first HIRS and HIRS/2/3/4, respectively. The total number of FOVs for a single scanline of AIRS is 90. It is worth noticing that the horizontal resolution of each generation of HIRS series is different. The horizontal resolution of the first HIRS and HIRS/2 at nadir is 25 and 17.7 km, respectively. The horizontal resolution of visible and infrared shortwave channels of HIRS/3 at nadir is 20.3 km, and that of infrared longwave channels of HIRS/3 is 18.9 km. The nadir resolution of HIRS/4 is 10 km, nearly twice as high as that of the other HIRS instruments. The nadir resolution of AIRS is 13.5 km. A comparison of sizes and distributions of FOVs among AIRS, CrIS and infrared longwave channels of HIRS/3 near nadir is provided in Figure 4. At the same scan angle, the FOV size is the largest for the infrared longwave channels of HIRS/3, the smallest for AIRS, and moderate for CrIS. Differences in FOV sizes of the infrared longwave channels among HIRS/3, AIRS and CrIS arise mainly from differences in the beam widths of the three instruments as well as the altitudes of the corresponding satellite platforms. The beam widths for HIRS/3, AIRS and CrIS are 1.3, 1.1 and0.963°. The altitude of Aqua satellite with AIRS onboard is 705 km, while the altitude of Suomi NPP satellite with CrIS onboard is 834 km. Although the beam width for AIRS is larger than that for CrIS, the FOV size for AIRS is smaller than that for CrIS due to a lower altitude of Aqua than that of Suomi NPP. Near nadir, no overlaps occur between neighboring FOVs for the three instruments in both cross-track and along-track directions. A large space between neighboring FOVs for the infrared longwave channels of HIRS/3 exists in both the cross-track and alongtrack directions. A small space between neighboring FOVs is observed in cross-track directions for AIRS and both cross-track and along-track directions for CrIS.

Under clear-sky conditions, the measured infrared radiance comes from a specific volume of the atmosphere, which is determined by the beam width, the weighting function, and the observing time period. A single CrIS FOR consisting of nine FOVs takes about 0.2 s to observe[3]. As is mentioned above, CrIS provide radiance observations at 1305 channels in the spectral range of 655-2550 cm-1. The radiance observations may come from different atmospheric volumes with significant overlaps. Hence, the radiance observations of CrIS full spectral range contain significantly redundant and thus correlated information. In NWP, a channel selection becomes necessary for CrIS data assimilation in order to avoid error correlations between different channels and to reduce the computational expense. The channel selection for CrIS has two main principles: select channels with high sensitivity to a certain atmospheric species and high vertical resolution. The former is to effectively reduce the redundancy between different channels and the latter is to maximize the vertical resolution of the retrieval product[6]. The vertical resolution of CrIS observations is determined by the weighting function of each channel. The narrower the weighting function is, the higher the vertical resolution is for a specific channel. The atmosphere at the altitude of weighting function peak contributes most to the radiance observed by that channel[7]. The weighting functions of different channels reach the maximum at different altitudes, which is the basis for retrieving the vertical profiles of atmospheric species. In addition, the vertical observing range of channels is also considered in the channel selection for CrIS. Gambacorta et al.[6]select a total of 399 CrIS channels for applications in NWP data assimilation system. This subset of CrIS channels includes 24 surface temperature, 87 temperature, 62 water vapor, 53 ozone, 27 carbon monoxide, 54 methane, 52 carbon dioxide, 24 N2O, 28 HNO3and 24 SO2sounding channels. Figure 5 presents the weighting function profiles of CrIS longwave infrared, shortwave infrared, water vapor and surface temperature channels calculated by the Community Radiative Transfer Model (CRTM)[8]under the US standard atmosphere. The infrared longwave, midwave and shortwave channels are indicated in blue, green and red colors, respectively. Figure 6 provides the distributions of altitudes of weighting function peaks for the 399 CrIS channels. It is seen that temperature channels are distributed in longwave and shortwave bands. The infrared longwave temperature channels (660 to 750 cm-1) are arranged compactly from 1000 to 10 hPa, providing the vertical profile of atmospheric temperature with high vertical resolution. The infrared shortwave temperature channels (2200 to 2420 cm-1) are arranged in a similar pattern to infrared longwave temperature channels but more compactly in the vertical range of 60 to 10 hPa, which can provide more information about the upper atmospheric temperature. Ozone channels are distributed over the spectral range of 990 to 1070 cm-1. The strong vibrational absorption band of ozone is near1041.67 cm-1. About 90% ozone is concentrated in the stratosphere within the altitude range from 10 to 50 km, and the remaining 10% ozone is concentrated near the Earth’s surface③. Water vapor channels are distributed over the following two spectral ranges: 780-1210 cm-1and 1310-1750 cm-1. The longwave water vapor channels (780 to 1210 cm-1) can provide the water vapor information near the surface. The midwave water vapor channels (1310 to 1750 cm-1) are arranged compactly in the vertical range from 800to 200 hPa, enabling the vertical profiling of the atmospheric water vapor. Surface temperature channels are distributed over two spectral ranges of 770-1095 cm-1and 2460-2540 cm-1. It is worth mentioning that the infrared shortwave surface temperature channels (2460 to 2540 cm-1) are not used in the National Centers for Environmental Prediction (NCEP) NWP systems due to a potential contamination of sun glint[9].

Hurricane Sandy made landfall at Cuba at 0600 UTC October 25, 2012. The sea level pressure and sea surface temperature of NCEP Final (FNL) global analysis at the same time is presented in Figure 7a. The observed brightness temperature of CrIS infrared longwave surface temperature channel 79 from the descending node of Suomi NPP at the same time is provided in Figure 7b and 7c. It is found that Hurricane Sandy is located over a warm sea surface with a low-pressure center of less than 998 hPa (Figure 7a). Compared to microwave, the wavelength of infrared is shorter, implying that the infrared radiance is attenuated in clouds more quickly. If the cloud has a large optical depth, the radiance measuredby CrIS channel 79 mainly comes from the cloud top, otherwise from the Earth surface.The brightness temperatures over cloudy areas are as low as 195 K, while those over clear-sky areas can reach up to 295 K (Figure 7b). A warm anomaly is observed near the Sandy center. The brightness temperatures in Hurricane Sandy’s eye are as high as 260 K, in a great contrast to those in the neighboring environment of lower than 200 K. It reflects a typical warm core structure in the hurricane center with thick clouds within and outside the eye wall. Figure 8a presents the weighting function distributions of 11 CrIS infrared longwave temperature channels. The cross section of brightness temperatures for these 11 CrIS infrared longwave temperature channels through the hurricane center in the along-track direction from the ascending node of Suomi NPP at 0600 UTC October 25, 2012 is provided in Figure 8b. It is seen that the brightness temperature reaches the maximum at the surface within the eye. The brightness temperature difference between the hurricane center and the nearby environment is as high as 60 K. The horizontal and vertical structures of Hurricane Sandy are well captured by CrIS infrared longwave temperature channels.

The prior hyper-spectral infrared sounders including AIRS and IASI have been widely used in NWP data assimilation system. McNally et al.[10]designed two experiments to explore the impact of AIRS data assimilation using only clear-sky observations. One experiment was to assimilate the clear-sky radiance from a single instrument (AIRS, HIRS and AMSU-A) in the ECMWF four-dimensional variational data assimilation system. AIRS data assimilation was found to outperform the assimilation of data from other two instruments with lower spectral resolutions (HIRS and AMSU-A).Another experiment was to add AIRS clear-sky observations into the ECMWF operational data assimilation system. It was found that AIRS had a positive impact on ECMWF operational forecasts. Guidard et al.[11]studied the impact of IASI data assimilation using both clear-sky and cloudy observations. The IASI clear-sky measurements were found to improve the model forecasts, while the IASI cloudy measurements had a neutral influence on the model forecasts due to the shortage of an effective method which can retrieve cloud parameters of high precision. The applications of the newest hyper-spectral infrared sounder (CrIS) in NWP have not yet to be demonstrated. On the other hand, all three hyper-spectral infrared sounders have been applied in the retrieval of meteorological variables and climate research. The AIRS/Advanced Microwave Sounding Unit (AMSU) retrieval product processing system has been running since 2002. IASI, AMSU and Microwave Humidity Sounder (MHS) have constituted the trace gas product processing system since 2008. The CrIS/Advanced Technology Microwave Sounder (ATMS) processing system has been operational since 2013. Gambacorta et al.[12]compared the accuracy of retrieval products from AIRS/AMSU, IASI/AMSU/MHS and CrIS/ATMS systems using NOAA Center for Satellite Applications and Research (STAR) Operational Hyper Spectral Retrieval Algorithm. Their results showed that the CrIS/ATMS system could provide vertical profiles of atmospheric temperature and water vapor with the same accuracy as those from the other two retrieval systems, except for the temperature in the lower troposphere and the water vapor in the middle troposphere. Under the FSR mode, which enables high spectral resolution of 0.625 cm-1across the full spectral range of CrIS, the vertical profile of carbon monoxide provided by CrIS/ATMS system is comparable in accuracy to the existing carbon monoxide retrievals from AIRS/AMSU and IASI/AMSU/MHS systems. In summary, the CrIS/ATMS processing system already satisfies the requirements for meteorological product retrieval and climate research. The values of CrIS hyper-spectral infrared radiance measurements and their retrieval products in NWP and climate research could be fully realized only when significant improvements in bias correction, quality control and cloud detection and retrieval algorithm for CrIS measurements are made.

注释

① http://nsmc.cma.gov.cn/NewSite/NSMC/Channels/100351.html

② http://www.wmo-sat.info/oscar/instruments/view/93

③ http://www.ozonelayer.noaa.gov/science/basics.htm

④http://www.nasa.gov/mission_pages/hurricanes/archives/2012/ h2012_Sandy.html#4

[1]达成, 邹晓蕾. GOES成像仪资料简介. 气象科技进展, 2014, 4(4): 52-61.

[2]Aumann H H, Chahine M T, Gautier C, et al. AIRS/AMSU/HSB on the Aqua mission: Design, science objectives, data products, and processing systems. IEEE Trans Geosci Remote Sens, 2003, 41: 253-264.

[3]Han Y, Revercomb H, Cromp M, et al. Suomi NPP CrIS measurements, sensor data record algorithm, calibration and validation activities, and record data quality. J Geophys Res Atmos, 2013, 118: 12734-12748.

[4]Strow L L,Motteler H,Tobin D, et al. Spectral calibration and validation of the Cross-track Infrared Sounder (CrIS) on the Suomi NPP satellite. J Geophys Res Atmos, 2013, 118: 12486-12496.

[5]Esplin M, Strow L L, Bingham G, et al. CrIS full spectral resolution test results. 92nd AMS Annual Meeting, New Orleans, LA, January 2012.

[6]Gambacorta A, Barnet C D. Methodology and information content of the NOAA NESDIS operational channel selection for the Cross-Track Infrared Sounder (CrIS). IEEE Trans Geosci Remote Sens, 2013, 51: 3207-3216.

[7]Petty G W. A First Course in Atmospheric Radiation. Madison,Wisconsin: Sundog Publishing, 2006.

[8]Weng F. Advances in radiative transfer modeling in support of satellite data assimilation. J Atmos Sci, 2007, 64: 3799-3807.

[9]Chen Y, Han Y, Delst P V, et al. Assessment of shortwave infrared sea surface reflection and nonlocal thermodynamic equilibrium effects in the community radiative transfer model using IASI data. J Atmos Oceanic Technol, 2013, 30: 2152-2160.

[10]McNally A P, Watts P D, Smith J A, et al. The assimilation of AIRS radiance data at ECMWF. Q J R Meteorol Soc, 2006, 132: 935-957.

[11]Guidard V, Fourrié N, Brousseau P, et al. Impact of IASI assimilation at global and convective scales and challenges for the assimilation of cloudy scenes. Q J R Meteorol Soc, 2011, 137: 1975-1987.

[12]Gambacorta A, Barnet C D, Wolf W, et al. The NOAA Operational Hyper Spectral Retrieval Algorithm: A crosscomparison among the CrIS, IASI and AIRS processing systems. International TOVS Study Conference. Jeju Island, Korea, March 2014.

附录

已知CrIS瞬时视场中心（P）的经度λP、纬度φP、仪器波束宽度ω及Suomi NPP卫星（S）天顶角μs、方位角φs、与地球之间的距离h。将φP转换为地心纬度（geocentric latitude）γP：

式中，α是地球的扁率，它与地球赤道半径（ra）和极地半径（rb）的关系为：

地球中心由O表示，已知γP，可求出O与P之间的距离（dOP）：

进一步利用λP可求出P点在笛卡尔坐标系里的三维坐标即向量OP：

已知dOP、γP、μs、h，可利用正弦定理求出S与P之间的距离（dPS）：

rOP为向量OP的单位向量，还已知μs、φs，利用旋转矩阵可求出单位向量rPS：

已知向量OS、OP，可求出向量SP和单位向量rSP，还已知ω，利用旋转矩阵可求出单位向量rSF1，F1为瞬时视场轨迹上一点：

S点与F1点的距离为dSF1，则：

F1点在地球表面，满足椭球体公式：

整理（11），可得：

式（12）为dSF1的一元二次方程。若方程有两个不同实数解，取较小值；若方程有两个相同实数解，取该值；若方程无解，则向量OF1不与地球表面相交。进一步可求出F1的纬度φF1和经度λF1。利用旋转矩阵将rSF1绕SP逆时针旋转度=10i，i=1，2，3，…，36）可求出单位向量rSFi：

再根据式（9）－（12）可算出dSFi，进一步可求出Fi的纬度φFi和经度λFi。

每9个CrIS瞬时视场组成一个CrIS能视场。每个能视场内，瞬时视场5被称为中心瞬时视场，瞬时视场1、3、7、9被称为对角瞬时视场。已知OS、OP、OPj(j=1,3,7,9)，可求出SP5和SPj，进一步可求出能视场对应的波束宽度θ：

已知SP5和θ，可求出单位向量rSP5，再根据式（7）－（12）可求出能视场轨迹的纬度和经度