J. Meteor. Res.  2017, Vol. 31 Issue (4): 708-719   PDF    
The Chinese Meteorological Society

Article Information

Min MIN, Chunqiang WU, Chuan LI, Hui LIU, Na XU, Xiao WU, Lin CHEN, Fu WANG, Fenglin SUN, Danyu QIN, Xi WANG, Bo LI, Zhaojun ZHENG, Guangzhen CAO, Lixin DONG . 2017.
Developing the Science Product Algorithm Testbed for Chinese Next-Generation Geostationary Meteorological Satellites: Fengyun-4 Series. 2017.
J. Meteor. Res., 31(4): 708-719

Article History

Received October 8, 2016
in final form November 28, 2016
Developing the Science Product Algorithm Testbed for Chinese Next-Generation Geostationary Meteorological Satellites: Fengyun-4 Series
Min MIN, Chunqiang WU, Chuan LI, Hui LIU, Na XU, Xiao WU, Lin CHEN, Fu WANG, Fenglin SUN, Danyu QIN, Xi WANG, Bo LI, Zhaojun ZHENG, Guangzhen CAO, Lixin DONG     
1. Key Laboratory of Radiometric Calibration and Validation for Environmental Satellites, National Satellite Meteorological Center, China Meteorological Administration, Beijing 100081
ABSTRACT: Fengyun-4A (FY-4A), the first of the Chinese next-generation geostationary meteorological satellites, launched in 2016, offers several advances over the FY-2: more spectral bands, faster imaging, and infrared hyperspectral measurements. To support the major objective of developing the prototypes of FY-4 science algorithms, two science product algorithm testbeds for imagers and sounders have been developed by the scientists in the FY-4 Algorithm Working Group (AWG). Both testbeds, written in FORTRAN and C programming languages for Linux or UNIX systems, have been tested successfully by using Intel/g compilers. Some important FY-4 science products, including cloud mask, cloud properties, and temperature profiles, have been retrieved successfully through using a proxy imager, Himawari-8/Advanced Himawari Imager (AHI), and sounder data, obtained from the Atmospheric InfraRed Sounder, thus demonstrating their robustness. In addition, in early 2016, the FY-4 AWG was developed based on the imager testbed—a near real-time processing system for Himawari-8/AHI data for use by Chinese weather forecasters. Consequently, robust and flexible science product algorithm testbeds have provided essential and productive tools for popularizing FY-4 data and developing substantial improvements in FY-4 products.
Key words: geostationary meteorological satellite     FY-4     algorithm testbed     cloud properties    
1 Introduction

Since the launch of the first meteorological satellite TIROS-1 (Television and Infrared Observational Satellite) on 1 April 1960, satellites have become indispensable for studying the earth's atmosphere, ocean, and land. With high temporal resolution, meteorological data obtained from satellites in geostationary (GEO) orbit are greatly needed for monitoring tropical cyclones, severe weather, lightning, and air pollution development, and for supplying data for assimilation by numerical weather prediction (NWP) models (Schmit et al., 2005, 2008; Stuhlmann et al., 2005; Yu et al., 2009; Goodman et al., 2013; Greenwald et al., 2016). The Fengyun-4 (FY-4) series, the second generation of Chinese GEO-orbit meteorological satellites, was first proposed, and its schedule drawn up, around 2000 (Cao et al., 2014; Yang et al., 2017). The series is currently under development by the China Academy of Space Technology and will be operated by the National Satellite Meteorological Center of the China Meteorological Administration (NSMC/CMA).

The FY-4 series of GEO satellites, the successor to the Fengyun-2 (FY-2) series (Hu et al., 2013), is three-axis stabilized. In contrast, the current operational FY-2 series is spin-stabilized (Lu et al., 2008). The first satellite (FY-4A) is scheduled for launch in December 2016 (Yang et al., 2017). The FY-4 series of GEO satellites will further boost our capabilities for monitoring cloud and haze, atmospheric temperature, and atmospheric humidity, for example. This emerging GEO satellite system will carry much more advanced instruments that offer both improved temporal, spatial, and spectral resolutions and regional fast mobile-detection capability. The primary payloads of the FY-4 series will consist of four optical instruments: the Advanced Geostationary Radiation Imager (AGRI), Geostationary Interferometric InfraRed Sounder (GIIRS), Lightning Mapping Imager (LMI) (Cao et al., 2014), and Solar X-EUV Imaging Telescope (SXEIT). The latter instrument will not be available for the first (FY-4A) satellite (Yang et al., 2017). With improved imaging device performance, AGRI and LMI will further strengthen the ability for monitoring small-and medium-scale weather systems and lightning. GIIRS, from the perspective of GEO orbit, will resolve three-dimensional atmospheric temperature and moisture structures and their evolution. Development of extreme ultraviolet and X-ray solar observations will enhance space weather monitoring and warnings. In addition, Chinese scientists and engineers are planning to develop an instrument called MDAS (Microwave Detection Atmospheric Sounder) in the next FY-4 series GEO satellite. To coordinate with the FY-4 GEO satellites, CMA has already developed the Chinese next generation polar-orbiting satellite system in FY-3 (Yang et al., 2012; Min et al., 2014) for obtaining constellation observations and providing better meteorological services.

In contrast with the Chinese FY-4 meteorological and weather satellites, which remain under development, Himawari-8, the next generation GEO satellite operated by the Japan Meteorological Agency (JMA), was successfully launched on 7 October 2014. Its installed systems include a 16-band Advanced Himawari Imager (AHI) with a spatial resolution from 0.5 km (visible band at 0.64 μm) to 2.0 km (infrared band) and a full-disk observation frequency of 10 min. The AHI near real-time (NRT) observation data were released freely by the JMA on 7 July 2015 (http://www.jma-net.go.jp/msc/en/). The next-generation series of Geostationary Operational Environmental Satellites (GOES-R) operated by the National Oceanic and Atmospheric Administration (NOAA) of the United States, will be launched in 2016, following the existing GOES constellation of satellites (Schmit et al., 2008; Goodman et al., 2013). Two important payloads installed on GOES-R are the Geostationary Lightning Mapper and an improved 16-band Advanced Baseline Imager (ABI) (Schmit et al., 2005, 2008; Goodman et al., 2013). The European Organization for the Exploitation of Meteorological Satellites and the European Space Agency are cooperating in the development of the Meteosat Third Generation (MTG) European GEO satellite system, which is scheduled for launch in 2018. Based on the twin-satellite concept and the three-axis platform, this GEO satellite system consists of an imaging satellite (MTG-I) and a sounding satellite (MTG-S). The main payloads on MTG-I are the Flexible Combined Imager, the Lightning Imager, the Data Collection System, and Search and Rescue. MTG-S includes the Infrared Sounder and the Ultra-violet, Visible, and Near-infrared Sounder (Stuhlmann et al., 2005). This brief summary indicates that we are entering a globally dispersed and rapidly developing era in the production and application of GEO meteorological weather satellites.

To promote better applications and popularization of FY-4 series GEO satellite data, the NSMC/CMA formed an Algorithm Working Group (AWG) in early 2011 (Yang et al., 2017), aiming at developing various new science products, such as cloud height, cloud phase, cloud type, cloud optical and microphysical properties (daytime and nighttime), aerosol properties, land and sea surface temperature, snow/ice cover, and high temporal resolution indicators like moisture and temperature profiles. In the process of developing prototypes of FY-4 science algorithms, two innovative science product algorithm testbeds were initially proposed and developed by the scientists in the FY-4 AWG to serve as flexible tools for FY-4A AGRI and GIIRS algorithms. These testbeds are together known as the Fengyun Geostationary Algorithm Testbed-Imager/Sounder (FYGAT-I/S). The original and primary goal of the FYGAT-I/S was to help the scientists in the FY-4 AWG to develop, improve, and validate their prototypes of science product algorithms. In addition, with its capabilities and utilization convenience, the FYGAT-I/S can easily be popularized to make the best use of FY-4 or other GEO satellite data. Any authorized FY-4 users, based on their specific demands, can also freely select whichever science products generated by the FYGAT-I/S. The main sets of the FYGAT-I/S include the product algorithm and the temporal and spatial coverage. In short, the current main purposes of FYGAT-I/S can be summarized as: 1) serve as a research testbed for developing algorithms for quantitative products from GEO satellite-based measurements; 2) use as the prototype algorithm software for FY-4 operational science products; 3) provide algorithm software to process FY-4 Direct Broadcast data for regional real-time applications; and 4) serve as a testbed for reprocessing GEO satellite data for climate research and applications. Therefore, it is our hope that this robust and flexible science product algorithm testbed will play an important role in the development of applications of the FY-4 satellite data.

In Section 2, this paper briefly introduces the specifications and key applications of the FY-4A imager and sounder sensors. Section 3 illustrates the FYGAT-I/S in detail, including its functions, structures, and some initial product results produced by the FYGAT-I/S by using proxy Himawari-8/AHI and Atmospheric InfRared Sounder (AIRS) data. Finally, Section 4 provides a summary.

2 FY-4A imager and sounder 2.1 AGRI

AGRI is one of the most important payloads on the FY-4A GEO meteorological satellite platform (Yang et al., 2017); its prototype is shown in Fig. 1. It has two independent scanning mirrors for performing north–south and east–west scans, respectively. Generally, it scans a full-disk earth-view image every 15 min within a nominal nadir resolution of 1 km (for 0.64 μm visible bands) or 4 km (for other bands). For the first test satellites, FY-4A, AGRI will include 216 detectors for 14 bands covering the range of wavelengths from 0.45 to 13.8 μm, the visible to longwave infrared wavelengths (Yang et al., 2017). Table 1 lists the FY-4A/AGRI specifications and main applications. Bands 1–3, the reflective solar bands (RSBs) for retrieving cloud, aerosol, and haze properties, are designed to measure earth-view surface-reflected solar radiation during daylight hours. The installation of RSB bands will significantly enhance the quantitative application value of FY-4A imager data, which, in turn, are better than data obtained from the current FY-2 series imager. Bands 7–14, referred to as the thermal emissive bands (TEBs), make observations of the thermal emissions from earth targets (daytime and nighttime). In addition to the traditional TEBs of GEO satellite imagers, channel 14, centered at 13.5 μm in the CO2 absorption band, is new to the design of the Fengyun GEO satellites. This band is expected to improve the current retrieval accuracy of cloud mask and cloud top properties (Ackerman et al., 2008; Menzel et al., 2008). However, for the next FY-4B/AGRI, the first operational FY-4 satellite, the plan is to increase the temporal resolution of full-disk earth-view images to 15 min, enhance the spatial resolutions of infrared bands to 2.0 km, and add visible and infrared water vapor bands at 0.55 μm (for true color synthesis) and 6.90 μm (for layer water-vapor retrieval), respectively.

Figure 1 Prototype of FY-4A/AGRI.
Table 1 FY-4A/AGRI specifications
Channel No. Band (μm) Spatial resolution (km) Detection sensitivity Primary application
Visible & near-infrared 1 0.45–0.49 1 S/N ≥ 70 (ρ = 100%) Aerosol
2 0.55–0.75 0.5 S/N ≥ 200 (ρ = 100%) S/N ≥ 5 (ρ = 1%) @ 0.5 km Fog, cloud
3 0.75–0.90 1 Vegetation
4 1.36–1.39 2 S/N ≥ 200 (ρ = 100%) S/N ≥ 5 (ρ = 1%) Cirrus
5 1.58–1.64 2 Cloud, snow
6 2.1–2.35 2 Cloud, aerosol
Shortwave infrared 7 3.5–4.0 (high) 2 NEΔT ≤ 0.7 K (300 K) Fire, land, and surface
8 3.5–4.0 (low) 4 NEΔT ≤ 0.2 K (300 K)
Water vapor 9 5.8–6.7 4 NEΔT ≤ 0.3 K (260 K) WV
10 6.9–7.3 4 NEΔT ≤ 0. 3K (260 K) WV
Longwave infrared 11 8.0–9.0 4 NEΔT = 0.2 K (300 K) WV, cloud
12 10.3–11.3 4 NEΔT = 0.2 K (300 K) SST, cloud
13 11.5–12.5 4 NEΔT = 0.2 K (300 K) SST, cloud
14 13.2–13.8 4 NEΔT = 0.5 K (300 K) Cloud
Note: WV, water vapor; SST, sea surface temperature; S/N, signal-to-noise ratio; ρ, reflectivity; No., number; NEΔT, temperature resolution.

It is well known that the key applications of AGRI multiple spectral bands include cloud and aerosol properties, weather information, land and sea surface temperatures, and atmospheric water vapor, as documented in Table 1. To ensure and maintain the scientific data quality of AGRI, high-precision radiometric calibrations for all bands are required. Based on the three-axis stabilized approach, the real-time full-path in-flight radiometric calibration of sensors for all bands can be conducted by using onboard blackbody and visible radiometers.


Another significant instrument aboard FY-4A is GIIRS, which is the world's first high-spectral-resolution IR sounder to be installed on a GEO-orbit satellite platform (Yang et al., 2017). Figure 2 shows the prototype of FY-4A/GIIRS. This hyperspectral sounding instrument is a Fourier transform spectrometer; in this instrument, a Fourier transform is required to turn raw data into an actual spectrum (radiance versus wave number). Generally, the device samples the area of China every 67 min and the mesoscale regions every 35 min with a nominal nadir resolution of 16 km. For the sake of increasing the number of available pixels of GIIRS, the FY-4 leading group is planning, in the configuration of FY-4B or FY-4C, to enhance GIIRS' spatial resolution to fall within the range of 8–10 km. The research and development (or engineering) specifications of spectral resolution and spectral range are given in Table 2. The instrument provides 913 channels within 2 separate IR spectral regions; namely, 535 channels for the longwave infrared band (from 700 to 1150 cm–1) and 378 channels for the shortwave band (from 1650 to 2250 cm–1). The spectral range for the FY-4A sounder includes the thermal infrared window region (8–12 μm), parts of the two CO2 bands (centered at 15.5 and 4.3 μm), part of the strong water vapor (H2O) absorption band (5–8 μm, with a center at 6.3 μm), and the ozone (O3) band at 9.6 μm. H2O also absorbs continuously within the window region, at wavelengths larger than 10 μm. Furthermore, there are minor contributions from carbon monoxide, nitrous oxide, and the chlorofluorocarbons CFC11 and CFC12. The key applications of GIIRS data will focus on retrieving profiles of atmospheric H2O and O3 and performing data assimilation in numerical weather forecast models, and retrieving atmospheric temperature and water vapor profiles (Liu and Li, 2011). The designed radiometric calibration accuracy is within 1.0 K; spectral calibration accuracy is about 10 ppm.

Figure 2 Prototype of FY-4A/GIIRS.
Table 2 FY-4A/GIIRS longwave (LW) and shortwave (SW) specifications
Band Resolution (cm–1) Spectral range (nm) NEΔT Points
LW 0.8 700–1130 0.5 538
SW 1.6 1650–2250 0.1 375
3 Science product algorithm testbed 3.1 FYGAT-I

FYGAT-I was initially proposed and developed by the FY-4 AWG cloud and aerosol algorithm and application team to serve as a flexible retrieval algorithm development tool for FY-4A/AGRI. Because of the increase in its integration capability, almost all of the AGRI science product algorithms, such as cloud, aerosol, snow, land and sea surface temperature, outgoing longwave and shortwave radiation, and fire detection, can be accommodated at present. All of these improved FY-4A science product algorithms have partly inherited the algorithms of FY-2/Visible Infrared Spin Scanning Radiometer (VISSR) and partly referred to the GOES-R algorithms (www.GOES-R.gov; some GOES-R Algorithm Theoretical Basis Documents). In addition, Table 3 lists the current four types of NWP data and GEO imaging sensors integrated with the FYGAT-I system. NWP data include the NCEP reanalysis data (Kalnay et al., 1996), the Global Forecast System (GFS) data (Kanamitsu, 1989), and the T639 data developed by the CMA (http://nwpc.cma.gov.cn/sites/main/index.htm). GEO imaging sensors include FY-2/VISSR (Hu et al., 2013), FY-4A/AGRI (Yang et al., 2017), Meteosat Second Generation/Spinning Enhanced Visible and Infrared Imager (MSG/SEVIRI) (Schmetz et al., 2002), and Himawari-8/AHI. Note that FYGAT-I currently offers only cloud mask and cloud phase products for FY-2/VISSR. This is attributable to the fact that FY-2/VISSR data have only five bands (Hu et al., 2013).

Table 3 NWP data and GEO imaging sensor integrated with the FYGAT-I system
Category Temporal resolution Spatial resolution
NWP NCEP 6 h 1° × 1°
GFS1p00 3 h 1° × 1°
GFS0p50 3 h 0.5° × 0.5°
T639 3 h 0.125° × 0.125°
Imaging sensor FY-2/VISSR 0.5/1 h 5.0 km
FY-4A/AGRI 15 min 4.0 km
MSG/SEVIRI 15 min 4.0 km
Himawari-8/AHI 10 min 2.0 km

Figure 3 shows the functional structure of FYGAT-I. First, the testbed will initialize the platform parameters from user-set options for different science product algorithms, NWP data, and satellite imaging sensors. After initialization, the calculation of Planck thermal radiation lookup-tables (LUTs) and the input of satellite navigation data are conducted. As the priority data, the fixed satellite navigation data are made pixel-by-pixel for different GEO imaging sensors, including longitude, latitude, sensor zenith/azimuth angles, land and sea masks, ecosystem type, and elevation. These fixed parameters will be used in several science product retrieval calculations. The coefficients of Planck thermal radiation LUTs are variable, depending upon different imaging sensors' spectral response functions. These values are computed before integration with FYGAT-I. Afterwards, the measured values of radiance, albedo, or brightness temperature (BRT) of all bands stored in the GEO imaging sensor Level-1B (L1B) data are read. After validating the correctness of every band value in the L1B data, the FYGAT-I system will insert some climatological surface ancillary data, such as the global infrared-band emissivity, the visible band albedo, and the snow and ice mask, which are imported in connection with the observation times of the L1B data. The surface ancillary data are derived through the statistical analysis of multiyear data from the Moderate Resolution Imaging Spectroradiometer (MODIS) and National Snow and Ice Data Center (Comiso et al., 2003; Liang, 2003; Wan, 2008). The next module will process the NWP data with different temporal and spatial resolutions, including data input, temporal/spatial linear interpolation (only in the vertical direction), and spatial collocation. The interpolated and collocated NWP data will be used in atmospheric corrections (Wang and King, 1997) and the fast radiative transfer model (RTM). The testbed will call the Pressure layer Fast Algorithm for Atmospheric Transmittances (PFAAST) model (Eyre and Woolf, 1988) for the specific imager (Fengyun RTM for FY-4A/AGRI) to compute clear-sky atmospheric radiance and transmittance for IR bands using the interpolated NWP data. This fast RTM of GEO imaging sensors, developed by the Cooperative Institute for Meteorological Satellite Studies (CIMSS) of the University of Wisconsin–Madison (UW) (Eyre and Woolf, 1988), has been coupled into the FYGAT-I system. The total processing time of the PFAAST model for L1B data without parallelization (a serial code) ranges from 20 s (NWP with a horizontal resolution of 1° × 1°) to 80 s (NWP with a horizontal resolution of 0.125° × 0.125°) on a Dell Power-Edge M610 Server; this arrangement will be able to completely meet the efficiency requirements of future FY-4A operational application systems. Near real-time clear-sky radiances simulated by the fast RTM are always used in the retrievals of cloud properties such as cloud phase and type, cloud-top, cloud optical features, and so on (Menzel et al., 2008; Ackerman et al., 2010; Heidinger et al., 2012; Liu et al., 2014). Subsequently, the FYGAT-I system will call the user-defined science product algorithms one-by-one. Finally, the science product output module is executed to generate several sets of product data using the hierarchical data format 5 (HDF5).

Figure 3 Functional structure of FYGAT-I.

FYGAT-I is distributed as the science product testbed source code, a set of independent AGRI science product algorithms with the corresponding algorithm LUTs, user's option list file, and ancillary data. The primary testbed source code and AGRI product algorithm codes were developed in FORTRAN and C programming languages. The developed code has been compiled successfully and executed on Linux/ UNIX systems with Intel/g compilers and an IBM/Dell/Huawei server system. Before compiling FYGAT-I, we installed HDF 4/5 and wgrib2 (http://www.cpc.ncep.noaa.gov/products/wesley/wgrib2) tools and libraries, to read and write satellite and NWP data. Any authorized testbed user can freely select which FY-4/AGRI science product algorithms to incorporate into a given build of FYGAT-I and to run in a given program invocation. This robust and flexible design is intended to facilitate FY-4/AGRI algorithm development, integration, and validation. It is worth noting that, in FYGAT-I, the science product algorithm calling order must be correct, otherwise the testbed will terminate automatically or shut down. For instance, the cloud-mask algorithm must be called before other science product algorithms. This priority arises from the fact that the cloud-mask algorithm is required for determining cloudy or cloud-free pixels for other science products. Moreover, the scientists at CIMSS/UW have also developed a more complicated GEO imager observation data simulator and a GEO stationary Cloud Algorithm Testbed for future GOES-R/ABI applications (http://cimss.ssec.wisc.edu/geocat) (Greenwald et al., 2016).

As shown in Fig. 3, we divide science products into six categories: cloud mask, cloud, weather, aerosol, radiation, and surface; each product is stored in a separate HDF5 data file. At first, the independent cloud-mask product will be generated by FYGAT-I, because the important inputs of other science products, such as cloud-motion wind, are not integrated with FYGAT-I. In addition, the cloud-mask product is stored as one of the cloud products in the cloud HDF5 output file. Aside from the cloud-mask product, the list of cloud products includes cloud phase, type, top properties, microphysical and optical properties, and type-Ⅱ. The type-Ⅱ cloud product, for the benefit of weather forecasters, includes various cloud classifications, such as cirrus, cumulus, and altostratus. The weather product includes quantitative precipitation estimation, convective initiation (CI) index, and layer/total perceptible water (TPW), which is not currently available. The aerosol product consists of the dust aerosol test, land/ocean aerosol optical depth (in integration), and fog/low-cloud mask. The radiation product includes outgoing shortwave and longwave radiation, surface downward/upward longwave radiation, and downward shortwave radiation (in integration). The surface product contains land and sea surface temperature, fire-point detection, snow and ice mask, and albedo (not currently available). To date, most of the science products have been validated by using MODIS and ground-based instrument data.

Here, we use JMA Himawari-8/AHI data (http://www.jma-net.go.jp/msc/en/) as the proxy FY-4A data to validate our science product algorithms integrated with FYGAT-I. In addition, based on this robust and beneficial system, we have also made an NRT Himawari-8/AHI data processing system to provide better satellite data service in the CMA, beginning in 2016. Figures 4 and 5 show, at 1200 BT (Beijing Time; 0400 UTC) 19 August 2015, the primary cloud products generated by the FYGAT-I NRT processing system, including cloud mask, cloud phase, cloud-top height, cloud-top pressure, cloud daytime optical thickness, cloud daytime effective radius, cloud daytime ice-water path, and cloud type-Ⅱ products. Figures 4 and 5 depict a case of twin typhoons, named Goni (west) and Atsani (east), which we found at that time. In this case, the cloud-top heights of the two typhoons were close to 18.0 km; the typhoons were surrounded by cirrus clouds (obtained from the cloud type-Ⅱ product) and had ice-phase canopies (from the cloud phase product). The cloud optical thickness of about 50 around Goni's eyewall was greater than the cloud optical thickness of Atsani. From this finding we deduced that Goni was stronger and thicker than Atsani. Figure 6 shows the outgoing longwave radiation, outgoing shortwave radiation, land and sea surface temperature, and snow and ice mask products generated by FYGAT-I at 1200 BT (0400 UTC) on 12 May 2016. This case in the spring season showed that the averaged land surface temperature (LST) was about 28℃ at noon, local time, over the middle of China, and quite a bit of permanent snow coverage in the Tibetan Plateau region. The FY-4 AWG plans further developments of the FY-4/AGRI science product algorithms, based on the FYGAT-I system. In particular, these developments will be designed to improve and validate some key science products, such as the CI index and cloud mask.

Figure 4 Himawari-8/AHI products generated by FYGAT-I at 1200 BT (0400 UTC) 19 August 2015: (a) cloud mask; (b) cloud phase; (c) cloud-top height, and (d) cloud-top pressure. BT indicates local Beijing time. The titles above every sub-figure represent science product names, units, and observation times. "AHI08" means Advanced Himawari Imager-08.
Figure 5 Himawari-8/AHI products generated by FYGAT-I at 1200 BT (0400 UTC) 19 August 2015: (a) cloud daytime optical thickness; (b) cloud daytime effective radius; (c) cloud daytime ice water path; and (d) cloud type-Ⅱ. The titles are the same as in Fig. 4.
Figure 6 Himawari-8/AHI products generated by FYGAT-I at 1200 BT (0400 UTC) 12 May 2016: (a) outgoing longwave radiation; (b) outgoing shortwave radiation; (c) land surface temperature; and (d) snow and ice mask. The titles are the same as in Fig. 4.

FYGAT-S was also initially developed by the FY-4 AWG sounding algorithm team to serve as a flexible atmospheric and surface parameters retrieval tool for FY-4A/GIIRS under both clear and cloudy conditions. The algorithm retrieves temperature, moisture, and ozone profiles, as well as surface properties, such as surface temperature and emissivity (not currently available), cloud-top pressure, cloud optical depth and effective radius (not currently available) under cloudy conditions. In addition, the derived products, such as TPW, layer perceptible water, lifted index, convective available potential energy (CAPE), total totals index, Showalter index, and K-index, are retrieved. Just as in the case of FYGAT-I, the NWP data, as the priority data that includes the NCEP, GFS, and T639 data, are successfully integrated with the FYGAT-S system.

FYGAT-S, designed for Linux or UNIX systems, could be made suitable for Windows systems with minor modifications. This software, programmed in FORTRAN and C, has been tested successfully on IBM/Dell systems by using Intel/g compilers. To serve as a retrieval testbed for most on-orbit sounder systems, FYGAT-S includes several integrated components: the FY/RTM, based on the PFAAST model (Eyre and Woolf, 1988), and two widely used RTMs, the Radiative Transfer for Television and Infrared Observation Satellite (TIROS) Operational Vertical Sounder (TOVS) (RTTOV) (Matricardi, 2010) and the community Radiative Transfer Model (CRTM) (Liu and Boukabara, 2014). The statistical regression retrieval algorithm and the nonlinear physical retrieval algorithm, developed by Ma et al. (1999) and Li et al. (2000), are implemented in FYGAT-S; they differ from one another slightly in the solution each offers. In addition, the retrieval packages also contain the functions for bias correction, channel selection, observation system characterization, error analysis, and so on. Any authorized users may freely select science product algorithms incorporated with FYGAT-S and run those selections in a given program invocation.

Figure 7 shows the functional structure of FYGAT-S. First, the FYGAT-S configurations, such as channel selection, retrieval method, and output list, as well as the RTMs, are initialized. The first step of the testbed main body is to collocate data file names, such as L1B data. After that, the measured radiances of all bands stored in GIIRS L1B data are read in and converted to BRTs using the Planck function. Based on the BRTs, a cloud-detection process is conducted. There are two choices for cloud detection: one is the same as that used by the ECMWF, and the other is based on cloudy sky regression. For the clear-sky condition, the retrieval algorithm comprises two parts: the regression approach and the physical retrieval methodology. In future configurations of FY-4/GIIRS, FYGAT-S will directly use and collocate higher resolution cloud mask products generated by AGRI for sounder observations. For the regression stage, the training set is classified based on the surface type and the sensor scan angles. After that, taking the regression output as the first approximation, a non-linear iteration based on a one-dimensional variational physical algorithm is conducted. For the cloudy condition, the dual regression method is used. In this algorithm, the training set is classified based on cloud phase and sensor-scan angle. The final retrieval in the cloudy sky condition is obtained by using iterations generated by varying the regression coefficient according to the cloud properties inferred from the previous step. Finally, the science product outputs are archived in an HDF5 data file.

Figure 7 Functional structure of FYGAT-S. For the clear sky condition, the processes of "surface type classification", "scan angle classification", and "apply corresponding coefficient for clear sky retrieval" together are called the process of the single multi-variable regression algorithm, while the corresponding three processes for the cloudy condition are the parts of the dual regression algorithm.

As mentioned above, FYGAT-S is suitable for both low-resolution sounders, such as the High-resolution Infrared Radiation Sounder, and hyperspectral sounders, such as AIRS and the Infrared Atmospheric Sounding Interferometer. AIRS data were taken as the proxy FY-4A/GIIRS data in the development of FYGAT-S because they have been used widely in the science community. In addition to the atmospheric products from the AIRS science team (Susskind et al., 2003), the reanalysis of ERA-Interim (Dee et al., 2011) was also used for validation. The gridded ERA-Interim was interpolated to AIRS pixels by using bilinear interpolation spatially and linear interpolation temporally. The primary outputs of cloudy sky temperature and water vapor, retrieved with a dual-regression method, were validated at 1330 BT (0530 UTC) 20 September 2010, as shown in Fig. 8. The spatial distribution of retrieved cloud-top pressure (Fig. 8a) is consistent with the visible images of clouds obtained by FY-2E. In addition, the statistics indicating the bias between the retrieved temperature and water vapor and the reference data are shown in Figs. 8c, d. For the clear-sky algorithm, we take a granule of the Level 2 (L2) products of AIRS cloud-cleared radiances as input, and the corresponding temperature and humidity profiles are used for validation (Susskind et al., 2003). Figure 9 shows the geographical distributions of 500-and 850-hPa temperature of the AIRS L2 product and the retrieval, as well as their differences. It also demonstrates that the retrievals can capture the main temperature feature of the AIRS L2 product, both at 500 and 850 hPa. The discrepancies between the AIRS science team L2 product and the retrievals of FYGAT-S, for most areas, are within 2.0 K. Figure 10 shows the results of the statistical analysis within this region. In addition, the error analysis compared to ERA-Interim is presented. The difference between the AIRS L2 product and the retrievals is within or close to that of the AIRS L2 product and reanalysis. This confirms that the products retrieved by FYGAT-S are within an error level similar to the error presented by the AIRS science team product. The algorithms for derived products, such as TPW and CAPE, will be conducted in the next round of research.

Figure 8 (a) Visible imagery of cloud from FY-2E on 20 September 2010. (b) The spatial distribution of cloud-top pressure (hPa) retrieved by using dual regression from AIRS data. (c) The profiles of standard deviations of difference (RMSE) between retrieved AIRS temperature and ERA-Interim. (d) The profiles of standard deviations of difference between retrieved AIRS relative humidity (%) and ERA-Interim at 1330 BT (0530 UTC) 20 September 2010.
Figure 9 Spatial distributions of (a–c) 500-hPa and (b–f) 850-hPa temperature for (a, d) the AIRS L2 product, (b, e) the retrieval of FYGAT-S, and (c, f) their difference (K) at 1330 BT (0530 UTC) 20 September 2010.
Figure 10 Profiles of standard deviations of difference between retrieval and AIRS (black curve), AIRS and ERA-Interim (red curve), and retrieval and ERA-Interim (blue curve) for (a) temperature, (b) logarithmic specific humidity, (c) ozone, and (d) relative humidity, at 1330 BT (0530 UTC) 20 September 2010.
4 Summary

We introduce the current development of a science product algorithm testbed for the next generation of Chinese GEO meteorological satellites, FY-4, which will be operated by the NSMC, CMA. The scientists in the FY-4 AWG have already developed two innovative algorithm testbeds of FYGAT-I/S for the FY-4 imager and sounder. Both testbeds are programmed in FORTRAN and C for Linux or UNIX systems and have been tested successfully on an IBM/DELL server by using Intel/g compilers. The original motivation for developing two testbeds was to exploit the prototypes of science product algorithms for the FY-4 imager and sounder. The robust and flexible functions, designed for inclusion in the FYGAT-I/S, are used for selecting science product algorithms, NWP data, and sensors. These operations facilitate the work of authorized users and the developers of FY-4.

The main science products of the FY-4 imager and sounder have been processed successfully and demonstrated by using the proxy Himawari-8/AHI and AIRS data. Figures 4–10 show some initial science product results generated by FYGAT, such as cloud mask, cloud-top height, and atmospheric temperature profile. In addition, based on the robust FYGAT-I system, an NRT processing system of Himawari-8/AHI was made available at the beginning of 2016 at NSMC/CMA for Chinese weather forecasters. The original goal of FYGAT was to develop and validate science product algorithms for FY-4. We are confident that the robust and flexible FYGAT-I/S testbeds are essential and rewarding tools for popularizing FY-4 data and attaining substantial improvements in FY-4 science products in future.

Acknowledgments. The authors are particularly grateful to all the scientists of the FY-4 Algorithm Working Group at NSMC/CMA for their efforts in developing the FYGAT-I/S and FY-4 science product algorithms. The Himawari-8/AHI data provided by the Japan Meteorological Agency are freely accessible from the China National Meteorological Information Center. AIRS data are freely disseminated by one of the NASA's data archive centers. NOAA and the data center at ECMWF freely disseminate NCEP, GFS, and ERA-Interim data. We acknowledge their efforts in providing the high-quality data. Chinese FY-4 algorithm development has benefited from GOES-R algorithm development. We are grateful to the GOES-R algorithm working group for making GOES-R Algorithm Theoretical Basis Documents (ATBDs) available through www.GOES-R.gov.

Ackerman S. A., Holz R. E., Frey R., et al., 2008: Cloud detection with MODIS. Part Ⅱ: Validation. J. Atmos. Ocean. Technol., 25, 1073–1086. DOI:10.1175/2007JTECHA1053.1
Ackerman, S. , R. Frey, K. Strabala, et al. , 2010: Discriminating Clear-Sky from Cloud with MODIS—Algorithm Theoretical Basis Document (MOD35) Version 6. 1, Tech. rep. , MODIS Cloud Mask Team, Cooperative Institute for Meteorological Satellite Studies, University of Wisconsin, 117 pp. Available online at https: //modis-atmos. gsfc. nasa. gov/_docs/MOD35_ATBD_Collection6. pdf (accessed on 12 May 2017).
Cao, D. J. , F. X. Huang, and X. S. Qie, 2014: Development and evaluation of detection algorithm for FY-4 geostationary lightning imager (GLI) measurement. Proceedings of XV International Conference on Atmospheric Electricity, Norman, Oklahoma, U. S. A.
Comiso C., J. J. Cavalieri, D. Markus, 2003: Sea ice concentration, ice temperature, and snow depth using AMSR-E data. IEEE Trans. Geosci. Remote Sens., 41, 243–252. DOI:10.1109/TGRS.2002.808317
Dee D. P., Uppala S. M., Simmons A. J., et al., 2011: The ERA-Interim reanalysis: Configuration and performance of the data assimilation system. Quart. J. Roy. Meteor. Soc., 137, 553–597. DOI:10.1002/qj.v137.656
Eyre R., J. M. Woolf, 1988: Transmittance of atmospheric gases in the microwave region: A fast model. Appl. Opt., 27, 3244–3249. DOI:10.1364/AO.27.003244
Goodman S. J., Blakeslee R. J., Koshak W. J., et al., 2013: The GOES-R geostationary lightning mapper (GLM). Atmos. Res., 125–126, 34–39. DOI:10.1016/j.atmosres.2013.01.006
Greenwald T. J., Pierce R. B., Schaack T., et al., 2016: Real-time simulation of the GOES-R ABI for user readiness and product evaluation. Bull. Amer. Meteor. Soc., 97, 245–261. DOI:10.1175/BAMS-D-14-00007.1
Heidinger A. K., Evan A. T., Foster M. J., et al., 2012: A naive Bayesian cloud-detection scheme derived from CALIPSO and applied within PATMOS-x. J. Appl. Meteor. Climate, 51, 1129–1144. DOI:10.1175/JAMC-D-11-02.1
Hu X. Q., Xu N., Weng F. Z., et al., 2013: Long-term monitoring and correction of FY-2 infrared channel calibration using AIRS and IASI. IEEE Trans. Geosci. Remote Sens., 51, 5008–5018. DOI:10.1109/TGRS.2013.2275871
Kalnay E., Kanamitsu M., Kistler R., et al., 1996: The NCEP/NCAR 40-year reanalysis project. Bull. Amer. Meteor. Soc., 77, 437–471. DOI:10.1175/1520-0477(1996)077 < 0437:TNYRP > 2.0.CO; 2
Kanamitsu M., 1989: Description of the NMC global data assimilation and forecast system. Wea. Forecasting, 4, 335–342. DOI:10.1175/1520-0434(1989)004 < 0335:DOTNGD > 2.0.CO; 2
Li J., Wolf W. W., Menzel W. P., et al., 2000: Global soundings of the atmosphere from ATOVS measurements: The algorithm and validation. J. Appl. Meteor., 39, 1248–1268. DOI:10.1175/1520-0450(2000)039 < 1248:GSOTAF > 2.0.CO; 2
Liang S. L., 2003: A direct algorithm for estimating land surface broadband albedos from MODIS imagery. IEEE Trans. Geosci. Remote Sens., 41, 136–145. DOI:10.1109/TGRS.2002.807751
Liu C., Yang P., Minnis P., et al., 2014: A two-habit model for the microphysical and optical properties of ice clouds. Atmos. Chem. Phys., 14, 13719–13737. DOI:10.5194/acp-14-13719-2014
Liu Li, 2011: An improvement in forecasting rapid intensification of Typhoon Sinlaku (2008) using clear-sky full spatial resolution advanced IR soundings. J. Appl. Meteor. Climate, 49, 821–827.
Liu H., Q. Boukabara, 2014: Community Radiative Transfer Model (CRTM) applications in supporting the Suomi National Polar-orbiting Partnership (SNPP) mission validation and verification. Remote Sens. Environ., 140, 744–754. DOI:10.1016/j.rse.2013.10.011
Lu H. Zhang, F. M. Xu, 2008: Image navigation for the FY2 geosynchronous meteorological satellite. J. Atmos. Ocean. Technol., 25, 1149–1165. DOI:10.1175/2007JTECHA964.1
Ma L., X. J. Schmit, T. A. Smith, 1999: A nonlinear physical retrieval algorithm—Its application to the GOES-8/9 sounder. J. Appl. Meteor., 38, 501–513. DOI:10.1175/1520-0450(1999)038 < 0501:ANPRAI > 2.0.CO; 2
Matricardi M., 2010: A principal component based version of the RTTOV fast radiative transfer model. Quart. J. Roy. Meteor. Soc., 136, 1823–1835. DOI:10.1002/qj.v136:652
Menzel W. P., Frey R. A., Zhang H., et al., 2008: MODIS global cloud-top pressure and amount estimation: Algorithm description and results. J. Appl. Meteor. Climate, 47, 1175–1198. DOI:10.1175/2007JAMC1705.1
Min M., Zhang Y., Rong Z. G., et al., 2014: A method for monitoring the on-orbit performance of a satellite sensor infrared window band using oceanic drifters. Int. J. Remote Sens., 35, 382–400. DOI:10.1080/01431161.2013.871393
Schmetz J., Pili P., Tjemkes S., et al., 2002: An introduction to Meteosat Second Generation (MSG). Bull. Amer. Meteor. Soc., 83, 977–992. DOI:10.1175/1520-0477(2002)083 < 0977:AITMSG > 2.3.CO; 2
Schmit T. J., Gunshor M. M., Menzel W. P., et al., 2005: Introducing the next-generation advanced baseline imager on GOES-R. Bull. Amer. Meteor. Soc., 86, 1079–1096. DOI:10.1175/BAMS-86-8-1079
Schmit T. J., Li J., Li J. L., et al., 2008: The GOES-R advanced baseline imager and the continuation of current sounder products. J. Appl. Meteor. Climate, 47, 2696–2711. DOI:10.1175/2008JAMC1858.1
Stuhlmann R., Rodriguez A., Tjemkes S., et al., 2005: Plans for EUMETSAT's Third Generation Meteosat geostationary satellite programme. Adv. Space Res., 36, 975–981. DOI:10.1016/j.asr.2005.03.091
Susskind D. Barnet, J. M. Blaisdell, 2003: Retrieval of atmospheric and surface parameters from AIRS/AMSU/HSB data in the presence of clouds. IEEE Trans. Geosci. Remote Sens., 41, 390–409. DOI:10.1109/TGRS.2002.808236
Wan Z. M., 2008: New refinements and validation of the MODIS land–surface temperature/emissivity products. Remote Sens. Environ., 112, 59–74. DOI:10.1016/j.rse.2006.06.026
Wang H., M. D. King, 1997: Correction of Rayleigh scattering effects in cloud optical thickness retrievals. J. Geophys. Res., 102, 25915–25926. DOI:10.1029/97JD02225
Yang J., Zhang P., Lu N. M., et al., 2012: Improvements on global meteorological observations from the current Fengyun 3 satellites and beyond. Int. J. Digit. Earth, 5, 251–265. DOI:10.1080/17538947.2012.658666
Yang, J. , Z. Zhang, C. Wei, 2016: Introducing the new generation of Chinese geostationary weather satellites-Fengyun 4 (FY-4). Bull. Amer. Meteor. Soc. , doi: 10. 1175/BAMS-D-16-0065. 1, in press.
Yu Y. Y., Tarpley D., Privette J. L., et al., 2009: Developing algorithm for operational GOES-R land surface temperature product. IEEE Trans. Geosci. Remote Sens., 47, 936–951. DOI:10.1109/TGRS.2008.2006180