Reprocessing 12-yr Microwave Humidity Sounder Historical Data of Fengyun-3 Satellites

Space-borne microwave radiometers are effective instruments for monitoring global changes in the earth’s climate system, and their observations are widely used in weather and climate research. Water vapor is one of the climate variables with extensive changes in time and space; thus, consistent and stable long-term global observations are required to analyze the climatic characteristics of water vapor evolution in the earth–atmosphere system. High-frequency microwave observations from meteorological satellites were used to study the water vapor. The 183-GHz channels, which align adjacent to the water vapor absorption lines, are suitable for water vapor observations and provide the primary data sources for the research of atmospheric moisture in the troposphere, precipitable water, and atmospheric ice water paths (Ferraro et al., 2016), as well as assimilation in numerical weather prediction (NWP) and reanalysis (Rienecker et al., 2011).

The Fundamental Climate Data Record (FCDR), Climate Data Record (CDR), and Essential Climate Variable (ECV) of conical scanning microwave imagers and cross-track scanning microwave sounder have been constructed for demanding applications (Hans et al., 2019). Research on the FCDR at the microwave humidity channels (183 GHz) developed relatively late due to the complexity of water vapor variation in the earth–atmosphere system and observational uncertainties. The 16-yr FCDR based on Advanced Microwave Sounding Unit-B (AMSU-B) and MHS brightness temperature for hydrological research was released by the NOAA (Ferraro et al., 2016). In recent years, the European Meteorological Satellite (EUMETSAT) has constructed FCDR via the recalibration of historical data from meteorological satellites at microwave water vapor absorption channels (183 GHz), in the frame of the fidelity and uncertainty in climate data records from earth observation (FIDUCEO) program (Hans et al., 2019). In 2020, a new CDR was released for the upper tropospheric humidity (UTH) observed by the SSMT-2, AMSU-B, and MHS from 1994 to 2017 (Lang et al., 2020). Observations from the Suomi National Polar-orbiting Partnership (SNPP) and the Joint Polar Satellite System (JPSS) were recently reprocessed by the NOAA Center for Satellite Applications and Research (STAR; Zou et al., 2020). This reprocessing eliminated the temporal inconsistency of scientific data caused by differences in the processing techniques, and greatly improved its calibration accuracy. In this program, historical data from the Advanced Technology Microwave Sounder (ATMS) were recalibrated by using the latest version of the operational calibration algorithm as the baseline. This reprocessing reduced the long-term radiation bias between SNPP/ATMS and AQUA/AMSU-A Channel 7 to 0.003 K yr⁻¹, meeting the stability requirement of satellite measurements in climate trend detection (Zou et al., 2020).

Fengyun-3 (FY-3) is a second-generation polar-orbiting operational meteorological satellite in China. Four FY-3 satellites (FY-3A/B/C/D) were successfully launched carrying Microwave Humidity Sounder (MWHS) on board. MWHS is a cross-track scanning instrument with channels centered in the 183.31-GHz water vapor line for atmospheric moisture profiling under clear and cloudy conditions. From the successful launch of FY-3A in 2008 to FY-3D in 2017, MWHS (FY-3A/B MWHS and FY-3C/D MWHS-II) has been providing global observations for over 12 yr. The characteristics of the historical data were reviewed, and errors in the operational system were corrected in this study. The calibration framework was standardized for the four MWHS onboard the FY-3 satellites, FY-3A/B/C/D, in preparation for the homogenized global microwave humidity dataset (183 GHz) to construct the FCDR for NWP, reanalysis, and climate research in China.

The MWHS and MWTS, together with Infrared Atmospheric Sounder (IRAS), constitute the Vertical Atmospheric Sounding System (VASS) for Fengyun satellites. The channel characteristics of the MWHS are comparable to those of the AMSU-B and MHS NOAA Polar-orbiting Operational Environmental Satellites (POES).

Specifications of the MWHS onboard FY-3A/B/C/D are listed in Table 1. The FY-3A/B MWHS has three channels centered near the wing of the 183.31-GHz water vapor absorption line (183.31 ± 1, 183.31 ± 3, and 183.31 ± 7 GHz) and two channels at 150 GHz with dual polarization, making it a unique in-orbit instrument (Zhang et al., 2008; Gu et al., 2010, Gu et al., 2012; Lu, 2011; Wang et al., 2013). The follow-on satellites FY-3C/D were further improved for temperature and humidity sounding to embrace the successful operational application of the VASS aboard FY-3A/B. Different from that of FY-3A/B, the MWHS onboard FY-3C/D has five channels near the conventional water vapor absorption line of 183 GHz. The specifications of the water vapor channels are similar to those of the ATMS onboard the SNPP and JPSS-1. For FY-3C/D, the additional temperature sounding channels near the 118-GHz oxygen absorption line enhance the detection capability of FY-3 for atmospheric temperature profiling and enrich our knowledge of the atmospheric thermal structure, which further benefits the atmospheric detection at sub-millimeter wave bands in the geostationary orbit (Guo et al., 2014; He et al., 2015; Lawrence et al., 2015; Guo et al., 2020). Since the launch of FY-3B, MWHS observations have been operationally assimilated in the ECMWF, the Met Office, and the Numerical Weather Prediction Center of the China Meteorological Administration, and positive impacts have been shown in the assimilation (Chen et al., 2015, Chen et al., 2018; Li and Liu, 2016; Carminati et al., 2018; Lawrence et al., 2018).

After the launch of FY-3C/D in 2013 and 2017, respectively, MWHS attracted worldwide attention because of its special configuration. For the first time, channels at 118 GHz were used to observe the global temperature from space. Assimilating observations from these channels have the slightly positive impact on the forecast accuracy. This resulted in an improvement of the atmospheric motion vector (AMV) of up to 0.2%. In addition, the assimilation trial has shown that observations at 118 GHz are suitable for precipitation retrieval (Lawrence et al., 2018).

The noise equivalent delta temperature (NEDT) of the MWHS onboard FY-3 indicates that the performance of MWHS has been improved gradually from FY-3A to FY-3D (Fig. 1). FY-3A/C crosses the equator in the morning, and FY-3B/D cross the equator in the afternoon. In the frame of the FY-3 satellite series, another three satellites are planned to carry MWHS to provide global coverage with an optimal temporal resolution (Yang et al., 2019; Zhang et al., 2019). Although the design and architecture of the MWHS are the same as those of FY-3C/D, the performance will be significantly improved owing to advanced technology and manufacturing.

Figure 1.  The noise equivalent delta temperatures (NEDT) for FY-3A MWHS, FY-3B MWHS, FY-3C MWHS-II , and FY-3D MWHS-II.

Since the launch of FY-3A in 2008, more than 12-yr continuous MWHS observations have been archived. The observation periods of these MWHS observations are shown in Fig. 2. The MWHS onboard FY-3A and FY-3B completed their operational missions on 4 May 2014 and 29 September 2021, respectively. The MWHS onboard FY-3C/D are still in orbit and operational. Notably, the FY-3B MWHS has been providing global observations successfully for more than 10 yr.

Figure 2.  FY-3A/B/C/D MWHS data records during 2008–2021.

MWHS is a total power radiometer. Its observations of hot target, cold space, and earth scene involve the same optical and electrical signal paths, and the advantage of the MWHS in-orbit calibration algorithm has been previously ensured (Li et al., 1999). Given the nonlinearity of the transfer function, a quadratic term calculated from the thermal vacuum (T/V) tests in the prelaunch phase should be considered in the linear calibration of MWHS, and nonlinearity correction is considered in the calibration of MWHS (Gu et al., 2012, 2013; Weng et al., 2013). Thus, the radiometer calibration transfer function of the FY-3 MWHS can be described as JPL (2000):

where ${C}_{\mathrm{c}}$ and ${C}_{\mathrm{w}}$ are the cold space and internal blackbody calibration target radiometer output count, respectively, $ {R}_{\mathrm{c}} $ and $ {R}_{\mathrm{w}} $ are the corresponding radiances for the cold space and internal blackbody calibration target, respectively, $ {C}_{\mathrm{s}} $ and $ {R}_{\mathrm{s}} $ are earth scene measurements, $ g $ is the calibration gain, and $ u $ is the nonlinear correction coefficient determined in the prelaunch T/V test (Atkinson, 2001; Gu et al., 2010; Guo et al., 2015).

Reprocessing the FY-3 MWHS historical data can be useful in developing FCDR and improving the application of microwave observations. Following the latest operational calibration algorithm of the FY-3D MWHS, the calibration framework was standardized to reprocess the historical data. As shown in Fig. 3, the standardization of the reprocessing processes mainly includes the following four aspects.

Figure 3.  FY-3 MWHS radiometric recalibration flow chart. The shaded boxes indicate the optimized and unified parts in the reprocessing method.

1) For the FY-3A MWHS, the nonlinear parameter u, which represents the nonlinear characteristics of the instrument, was reconstructed. Instead of the nonlinear brightness temperature (Gu et al., 2010), the nonlinear parameter was used to characterize the nonlinearity of the transfer function. Observations of the FY-3A MWHS are thus recalibrated by using Eqs. (1)–(3).

2) The calibration process and static parameter formats were unified. The T/V tests in the prelaunch phase were reviewed to update the static parameters used in the calibration procedure. These parameters were derived only when the instrument reached a stable state in a vacuum chamber.

3) Quality control of the telemetry parameters was integrated into the reprocessing of the historical data. The telemetry parameters include the hot-load temperature, instrument temperature, counts of the hot-load and cosmic background, scan angles, and scan periods. They were re-analyzed, and their thresholds were set accordingly for quality control.

4) The checksum in the MWHS raw data was used to eliminate satellite transmission errors.

The major changes to the operational system of the MWHS, including the in-orbit status and calibration parameters, are shown in Fig. 4. Failure of the FY-3A MWHS Channel 2 occurred on 8 October 2009. Lunar contamination correction was implemented in the operational calibration system of the FY-3A/B MWHS starting in March 2011. Thus, additional data reprocessing were applied to earlier observations to remove lunar intrusion. The memory chip of the FY-3B MWHS was damaged on 26 November 2016, and the backup was activated on 8 December 2016. Therefore, the observations of FY-3B MWHS during this period are not available. The FY-3C MWHS-II powered on and off twice due to the adjustment of the satellite platform in-orbit, resulting in data gaps. On 2 February 2015, misuse of the blackbody temperature and the instrument temperature at Channels 3–9 was identified, and the corresponding parameters were updated in the calibration algorithm. Automatic Gain Control (AGC), which is related to the nonlinear coefficient, was initialized on 9 October 2015 and 10 April 2018. The antenna correction coefficients were updated on 17 March 2015 to improve the calibration accuracy of the FY-3C MWHS-II. The antenna pointing angle was corrected by shifting the offset of the scanning mechanism, which further improved the accuracy of the geolocation.

Figure 4.  FY-3 MWHS major instrument status and calibration algorithm changes since 2008.

The FY-3A/B/C MWHS observations have been reprocessed based on the algorithm presented in the previous section. To further validate the accuracy of the reprocessing algorithm, the simultaneous nadir overpass (SNO) of the FY-3 A/B/C MWHS and the reference instrument were compared. Because there is no counterpart instrument deployed with the 118-GHz channels, only observations at the 183-GHz channels are evaluated. Similar instruments and their data availabilities are presented in Table 2. Given the temporal coverage and consistency, the METOP-A MHS, whose accuracy was reported within 0.6 K compared to GPM (Global Precipitation Measurement) Microwave Imager (GMI), is taken as a reference sensor (Berg et al., 2016). Level 1C files provided by the GPM Intersatellite Calibration Working Group (XCAL team) were used for SNO comparisons in this study. The channel specifications of the MWHS and MHS are listed in Table 3. Note that the channel of the MWHS at the farthest wing of 183 GHz is a double-sideband, while the MHS is a single-sideband set at 190.3 GHz. Sondeur Atmospherique du Profil d’ Humidité Intropicale par Radiométrie (SAPHIR) onboard Megha-Tropiques also has water vapor channels at 183 GHz; however, SAPHIR center frequencies differ from those of MWHS, and thus, it was not appropriate for comparison in this study.

The observed brightness temperatures during the lifetimes of the FY-3A/B/C MWHS and METOP-A MHS were matched with the-183 GHz channels. The threshold for the matched data was 20 min and 3 km at scan angles less than 5° near the nadir. To avoid possible impacts from clouds and environments, observations with a standard deviation greater than 1 K in 3 $ \times $ 3 pixels were screened out (Yan and Weng, 2008; John et al., 2013).

The FY-3A/B/C MWHS operational and recalibration data were compared with the MHS using the SNO method described above. Theoretically, the difference in the brightness temperatures between the two sensors is negligible at the same channels, and only a constant and stable bias is allowed when the sensors observe the same target simultaneously. Figure 5 shows the bias of the monthly mean brightness temperatures between the two instruments at the corresponding channels. The left (right) panels of Fig. 5 show the SNO bias of the operational data (reprocessing dataset) and the MHS. The intersatellite bias between the Metop-A MHS and FY-3A/B/C MWHS is caused by the calibration error of MWHS, since Metop-A MHS has a bias of less than 0.6 K (Berg et al., 2016). For FY-3A, the improvement after reprocessing is attributed to nonlinearity correction, as mentioned in Section 3. The bias of FY-3B is significantly reduced after optimizing the nonlinearity in the T/V tests, where the new static parameters in the calibration algorithm are recalculated at a stable and reliable instrument state. Therefore, the new static parameters of the FY-3B MWHS better depict the operating status of the instrument in orbit. The difference between the operational calibration algorithm and reprocessing model of FY-3C is minimal, and the change in the bias is insignificant. Note that the spike in November 2016 was due to the damaged memory chip of FY-3B.

Figure 5.  SNO bias between FY-3A/B/C MWHS and METOP-A MHS.

After reprocessing, seasonal, and interannual variability are still found for the 183.31 ± 3-GHz channel. To better explain this phenomenon, the monthly mean instrument temperature of the FY-3 MWHS is shown in Fig. 6. The temporal variability at 183.31 ± 3 GHz is strongly correlated with the instrument temperature. Thus, we assume that the instrument temperature dependency of the nonlinearity is not fully removed at this channel, and further efforts are needed to correct the anomalous variation. Aside from the instrument temperature, contributions from the channel frequency difference between MWHS and MHS cannot be neglected at 183.31 ± 7 GHz, resulting in a bias up to −2 K.

Figure 6.  As in Fig. 5, but for instrument temperature.

The standard deviations of the SNO are shown in Fig. 7. The consistency of the standard deviation between the three datasets was improved. After reprocessing, the SNO standard deviation of FY-3B was reduced because of the updated static parameters. The optimization of the data quality control scheme and checksum effectively removes abnormal information, resulting in a relatively stable standard deviation. Note that data from November 2016, when the hardware was damaged, were not evaluated.

Figure 7.  As in Fig. 5, but for SNO standard deviation.

Figures 8–10 show further comparisons of the SNO results of MWHS and MHS at 183.31 ± 1, 183.31 ± 3, and 183.31 ± 7 GHz, respectively. The calibration bias of the operational data is found to be scene-temperature dependent, and the value is generally larger than that of the reprocessed data. The three channels subject to severe impacts of the instrument temperature exhibit strong correlations to the scene temperature (top panels of Figs. 8–10), consistent with Fig. 5. In addition to the nonlinearity, the contamination of the instrument temperature, which is not completely removed in the operational calibration, further impinges on the intersatellite bias to be scene-temperature dependent (Zou and Wang, 2011). Moreover, a bias of up to 3 K was found for FY-3B MWHS in the operational version, due to the damaged memory chip of FY-3B.

Figure 8.  SNO between FY-3A (left panel), FY-3B (middle panel), and FY-3C (right panel) MWHS and METOP-A MHS at 183.31 ± 1 GHz. The top panel shows scatterplots of the temperature dependence (TB) of SNO bias for the MWHS operational data, the middle panel shows this for the recalibrated MWHS data, and the bottom panel is the probability distribution of the bias from operational and recalibration.

Figure 9.  As in Figs. 8, but for the channel at 183.31 ± 3 GHz.

Figure 10.  As in Fig. 8, but for the channel at 183.31 ± 7 GHz.

For all channels, the reprocessing significantly improved the SNO bias, on the order of 1–2 K, indicating the effectiveness of the reprocessing method (bottom panels of Figs. 8–10). Especially for FY-3B MWHS, the bias is reduced from 3.2 to 0.8 K, and its distribution follows a Gaussian distribution at 183.31 ± 3 and 183.31 ± 7 GHz (Figs. 9–10).

After reprocessing, the scene-temperature dependency was mitigated for all channels because of the updated nonlinear coefficients used in the reprocessing. The changes at the 183.31 ± 7 GHz channel of the FY-3C MWHS are negligible compared to those of FY-A/B (Fig. 10). This is because the T/V test scheme of FY-3C was improved before launch, and accurate nonlinear parameters were expected in the operational calibration.

The reprocessing of historical data from FY-3 MWHS observations spanning more than 12 yr is presented in this study. The unified format of the FY-3A/B/C L1 reprocessing dataset includes the observed brightness temperature, geolocation information, and unified-format static parameters. After reprocessing, the bias was approximately 1 K in magnitude for most channels. For FY-3B MWHS, the bias between MWHS and MHS at Channel 4 reduced from 3.2 to 0.8 K. Meanwhile, the data consistency and stability of FY-3A/B/C also improved. This version has been released to users for atmospheric parameter retrieval and reanalysis assimilation.

The scene-temperature dependency in the calibration biases is significantly mitigated via the correction of nonlinearity parameters for the MWHS channels, and the correlation between the observations and the instrument temperature is alleviated. Based on this version, further refinement in the recalibration model will be conducted to remove or mitigate the intersatellite bias, following the inter-calibration algorithm developed by Zou et al. (2006) and Zou and Wang (2011). Subsequently, radiation traceability methods will be applied to achieve consistent and stable calibrations of MWHS onboard FY-3A/B/C and generate the FY-3 MWHS FCDR. The improvement in the data quality will not only benefit the retrievals of tropospheric water vapor and precipitable water in the climate system but also provide robust spaceborne observations for NWP and reanalysis.