1. Introduction
After completing the launch of the first-generation polar-orbiting satellites in the Fengyun series (FY-1A–D) in 2007, China began its second-generation polar-orbiting satellite systems. Between 2008 and 2020, five satellites in the FY-3 series will be launched into orbit, with the first two labeled as experimental. On 27 May 2008, FY-3A was launched with 11 instruments on board. After the intensive calibration/validation period in its first 6 months in orbit, most of the instruments were declared suitable for user applications. The performance of atmospheric sounding instruments in particular meets or exceeds the specifications (Dong et al. 2009; Yang et al. 2009; Zhang et al. 2009). On 5 November 2010, the second satellite in the new series, FY-3B, was successfully launched, and all of the instruments on board are expected to be working after completing its in-orbit verification.
As part of the FY-3A/B calibration processes, it was deemed important to assess the instrument biases with respect to the simulations from numerical weather prediction models. Today, as the NWP analysis fields become more accurate than any single observing system and the atmospheric radiative transfer model is considered very accurate except for surface-sensitive channels and limb effects at large scan angles, the simulated satellite sounding measurements can be used as a “truth” for characterizing the performance of new instruments during their in-orbit verification period (Bell et al. 2007). The biases between simulated and observed measurements can also be properly quantified through careful investigation, and a data-friendly quality control scheme can subsequently be developed for applications of these observations in NWP data assimilation systems. An initial evaluation of the FY-3A Microwave Humidity Sounder (MWHS) against the National Oceanic and Atmospheric Administration (NOAA)-18 Microwave Humidity Sounder (MHS) revealed their similar bias distributions at two upper-level sounding channels (Guan et al. 2011). In this study, the measurements from the Microwave Temperature Sounder (MWTS) on board FY-3A will be further compared to those of NOAA-18’s Advanced Microwave Sounding Unit-A (AMSU-A). A similar comparison of four instruments on board FY-3A with those on the Meteorological Operation (MetOp)-A satellite can be found in Lu et al. (2010).
An improved characterization of FY-3A MWTS performance with respect to NOAA microwave sounding instruments would also benefit the climate research in the future. FY-3A MWTS is a four-channel cross-track scanning radiometer that is similar to the Microwave Sounding Unit (MSU) on board the early NOAA satellites from Television and Infrared Observation Satellite (TIROS)-N through NOAA-14 from 1979 to 2007. Since 1998, the AMSU-A, which has 11 more channels than MSU, replaced MSU and has since flown on NOAA-15, -16, -17, -18, and -19. Therefore, there are more than 30 yr of MSU global satellite data, which have been widely used by the community for climate studies. Using brightness temperature measurements of MSU/AMSU-A’ three channels, that is, MSU channels 2–4, which are similar to MWTS channels 2–4, and AMSU-A channels 5, 7, and 9, several research groups came up with various upper-air temperature climate data records (Christy et al. 1998, 2000, 2003; Mears et al. 2003; Mears and Wentz 2009; Vinnikov and Grody 2003; Zou et al. 2009). With MWTS on the FY-3 series, the MSU/AMSU time series can be directly extended well beyond 2020.
A single satellite has a limited 5–10-yr life span. Long-term climate data records can be obtained by stitching together measurements from overlapping satellite observations. However, requirements on satellite calibration for detecting global climate change are much higher than for capturing day-to-day weather fluctuations. The climate trend differences among different research groups still exceed the widely accepted accuracy requirement of 0.01–0.02 K decade−1 (Ohring 2005). A much more refined postlaunch, satellite-by-satellite calibration is required to deduce long-term climate trends with adequate accuracy, precision, stability, and consistency. Therefore, this study also contributes to future integration of FY-3A MWTS data into long-term climate data records.
This paper is organized as follows: Section 2 will present the major instrument characteristics of NOAA-18’s AMSU-A and FY-3A’s MWTS. Model simulations of global brightness temperatures are described in section 3. Numerical results on the evaluation of FY-3A’s MWTS against NOAA-18’s AMSU-A are provided in section 4, which consists of four subsections on global biases and standard deviations (see section 4a), the scan-angle dependence of biases (section 4b, the scene-temperature dependence of MWTS biases in polar regions (section 4c), and a root cause analysis (section 4d). The paper concludes in section 5.
2. AMSU-A and MTWS instrument characteristics
On 20 May 2005, the NOAA-18 satellite was successfully launched into a circular, near-polar, afternoon-configured (1400 LT) orbit with an altitude of 854 km above the earth and an inclination angle of 98.74° to the equator. The AMSU-A component on board NOAA-18 consists of two modules—AMSU-A1 and AMSU-A2—that provide a total of 15 channels. There are 12 channels in the frequency range from the 50.3- to 57.3-GHz for atmospheric temperature profiling from the earth’s surface to about 42 km (or 2 hPa). The other three channels are located at 89, 23.8, and 31.4 GHz. Measurements from these channels are sensitive to surface emissivity and temperature, atmospheric cloud liquid water, and water vapor and can be utilized to improve atmospheric sounding quality in lower-tropospheric conditions down to the surface (Goodrum et al. 2009).
On 27 May 2008, the FY-3A satellite was successfully launched into a circular, near-polar, morning-configured (1000 LT) orbit with an altitude of 836 km above the earth and an inclination angle of 98.75° to the equator. The differences of the local equator crossing time (ECT) of the two closest swaths between FY-3A and NOAA-18 are about 3–4 h. Like AMSU-A, MWTS is also a cross-track-scanning radiometer but only has four subset channels from the AMSU-A set, which are similar to MSU on the early NOAA satellites.
Table 1 lists some of the channel characteristics of the NOAA-14 MSU, NOAA-18 AMSU-A, and FY-3A MWTS, including channel frequency, peak weighting function height, radiometric temperature sensitivity (NEΔT), the total number of fields of view (FOV) per scan line, the horizontal resolution of the nadir FOV, as well as the 3-dB radio frequency (RF) bandwidth (Mo 1996, 1999; Dong et al. 2009). The extreme scan position of the earth view to the beam center (48.3°) is the same for both the MWTS and AMSU-A instruments. It takes 16 s for the FY-3A MWTS to complete one scan line from left (the first FOV) to right (the last FOV), which is twice as long as that of NOAA-18’s AMSU-A. After each earth scan line, MWTS antenna then views the cold space (~2.73 K) and an onboard warm blackbody target (~290 K) for calibration. There are only 15 scene FOVs along each MWTS scan line, compared to the 30 FOVs of AMSU-A. Therefore, beam positions 15 and 16 are near the nadir direction of AMSU-A, and beam position 8 is the nadir direction of MWTS. The diameter of MWTS FOV at nadir is 62 km, which is larger than that of AMSU-A (48 km). The FY-3A MWTS swath width is 2250 km, which is slightly smaller than that of AMSU-A (2300 km).
Channel characteristics and specifications of MSU, MWTS, and AMSU-A.
Figure 1 displays weighting functions (WFs) calculated from a standard U.S. atmospheric profile for the four channels on the FY-3A MWTS and the corresponding four channels on the NOAA-18 AMSU-A with the same specified frequencies as MWTS. MWTS channel 2 has two bands located at both sides of the center frequency (i.e., double-sided bands), which do not cover the center frequency absorption. The bandwidth of any MWTS channel is consistently slightly broader than the corresponding AMSU-A channel. This is reflected in WF distributions shown in Fig. 1. The WF of the MWTS channel 4 is much broader and slightly higher than that of AMSU-A channel 9. The WFs for the other three channels, MWTS channels 1, 2, and 3, are nearly identical to (or only slightly broader than) AMSU-A channels 3, 5, and 7.
Weighting functions of the FY-3A MWTS channels 1–4 (solid lines) and NOAA-18 AMSU-A corresponding channels 3, 5, 7, and 9 (dash lines) calculated from the U.S. standard atmospheric profile.
Citation: Journal of Atmospheric and Oceanic Technology 28, 10; 10.1175/JTECH-D-11-00023.1
Radiance data in level 1B format from both FY-3A’s MWTS and NOAA-18’s AMSU-A during January 2010 are employed for this study. The same technique used for AMSU-A (Mo 1999) is used to calibrate MWTS (R. You 2009, personal communication). A global distribution of the total number of observations within each 1° × 1° grid during this month is shown in Fig. 2. Overall, the number of AMSU-A data is about 4 times more than the MWTS data. Such a difference in data amount mainly comes from a combined result of the time for MWTS to complete a scan line (16 s), which is 2 times longer than that of AMSU-A (8 s), as well as the total number of FOVs on a single MWTS scan line (15 FOVs), which is only half of those of AMSU-A (30 FOVs). The differences of spatial resolution between the two instrument measurements may also have contributed a little to the smaller amount of MWTS observations. Data comparisons between the two instruments are made only for MWTS channels 2–4 and AMSU-A channels 5, 7, and 9. These three oxygen channels contain important tropospheric and low-stratospheric temperature information that is much needed by NWP models. NOAA-18’s AMSU-A data are currently ingested and assimilated at major NWP operational centers around the world. FY-3A’s MWTS data will soon be incorporated into operational forecasts at the European Centre for Medium-Range Weather Forecasts (ECMWF) and China (Lu et al. 2010). Comparisons for MWTS channel 1 with AMSU-A channel 3 will be made in a future study because they are farther away from the O2 absorption line center and the results need to be interpreted and analyzed with many other auxiliary data and a more advanced radiative transfer capability.
Global distribution of the total number of observations within each 1° × 1° grid box.
Citation: Journal of Atmospheric and Oceanic Technology 28, 10; 10.1175/JTECH-D-11-00023.1
3. Global simulations of brightness temperatures
Absorption and emission of radiation at microwave frequencies by the atmospheric constituents, such as oxygen, is the physical basis for remotely sensing the atmospheric profiles, such as temperature. A radiative transfer model can simulate the microwave radiation at the top of the atmosphere, which is measured by a radiometer on board a satellite. The required inputs to radiative transfer model for simulation include atmospheric profile of temperature, water vapor content, variable gas concentrations (e.g., ozone), and cloud and surface properties. In this study, the radiative transfer for TIROS Operational Vertical Sounder (RTTOV; Saunders et al. 1999) is used to produce a set of global simulations of brightness temperatures that are measured by both MWTS and AMSU-A. The vertical profiles of temperature and specific humidity as well as surface pressure from the National Centers for Environmental Prediction (NCEP) Global Forecast System (GFS) 6-h forecasts are used as input to RTTOV. The NCEP GFS 6-h forecast fields have a horizontal resolution of 1° × 1° and 26 vertical levels. The highest vertical level is around 10 hPa.
A direct pixel-by-pixel comparison between FY-3A’s MWTS and NOAA-18’s AMSU-A is not possible over a global scale because brightness temperature measurements from these two satellites do not overlap except for the region over the high latitude where the two satellite orbits pass over. For example, only three simultaneous nadir overpasses are found during the entire month of January 2010 (see Table 2). Instead, global simulations of brightness temperature described above are used as a “reference” or “truth” for comparing the performance of the two instruments. Specifically, statistical features of the differences between the MWTS-observed and RTTOV-simulated brightness temperatures are examined and compared with those of the NOAA-18 AMSU-A.
Brightness temperatures at three simultaneous nadir overpass locations: Point A: 70.53°N, 6.96°W at 10 Jan 2010; point B: 71.07°N, 51.44°W at 10 Jan 2010; and point C: 70.55°S, 45.39°E at 22 January 2010. The collocation criteria are that the spatial and temporal separations between NOAA-18 and FY-3A measurements are less than 12.5 km and 60 s, respectively.
A simple quality control (QC) is applied to identify outliers defined as those measurements whose values deviate from model-simulated values by more than Zscore times the standard deviation. These outliers are most likely associated with those observations that are affected by clouds and precipitation and cannot be simulated from the radiative transfer model because the input profiles from NWP models lack reliable information about clouds. To minimize the impact of outliers on the calculated values of mean and standard deviation, the biweight mean and biweight standard deviation are used (Zou and Zeng 2006). We set Zscore = 2 for MWTS channels 2 and AMSU-A channel 5 and Zscore = 1.7 for other channels. On average, for two lower-level sounding channels, the percentage of outliers of MWTS and AMSU-A is similar (Fig. 3a) but the number at MWTS channel 4 is much larger than that at AMSU-A channel 9. About 18% (17%), 9% (9%), and 13% (9%) outliers are found for channel 2 (channel 5), channel 3 (channel 7), and channel 4 (channel 9) in MWTS and AMSU-A data (Fig. 3), respectively. More than twice as many outliers are identified over land than over ocean for lower-level sounding channels.
(a) Percentage of outliers in MWTS (solid bar) and AMSU-A (dashed bar) data over land (green) and ocean (red) in January 2010. (b) Total number (unit: 106) of MWTS (channels 2–4) and AMSU-A (channels 5, 7, and 9) data points after QC. The total number of all data points over ocean and over clear-sky oceanic data points is also indicated. The clear sky is defined by cloud fraction being less than 5% for MWTS and liquid water path less than 0.05 kg m−2 for AMSU-A.
Citation: Journal of Atmospheric and Oceanic Technology 28, 10; 10.1175/JTECH-D-11-00023.1
The total number of observations from MWTS channels 2–4 and AMSU-A channels 5, 7, and 9 that pass QC are shown in Fig. 3b. The total number of all data points over ocean and over ocean with clear sky is also indicated. An MWTS FOV with 5% of a cloud fraction is identified as a clear-sky microwave measurement. The cloud is derived from another visible/infrared instrument called the Medium Resolution Spectral Imager (MERSI) on board FY-3A. For AMUS-A, a liquid water path less than 0.05 kg m−2 (Weng and Grody 1994) is used to define a clear FOV. Similar to Fig. 2, the AMSU-A data are about 4 times more than MWTS data in all categories.
4. Numerical results
a. Biases and standard deviations
The global bias and standard deviation of brightness temperature differences between satellite observations and model simulations are shown in Fig. 4, both with and without removing the outliers, over ocean in clear-sky conditions (Figs. 4a,b). Compared with AMSU-A data, the global bias of the MWTS data is smaller but the standard deviation is larger, except for that of channel 5. At AMSU-A channel 5 and MWTS channel 2 the standard deviations are similar and mostly are due to data over land (figure omitted). Results using only nadir data (Figs. 4a,b) are similar to those using all scan-angle data (figure omitted).
(left) Global bias and (right) standard deviation of brightness temperature differences between FY-3A MWTS observations and NCEP GFS simulations for different datasets defined in Fig. 3b.
Citation: Journal of Atmospheric and Oceanic Technology 28, 10; 10.1175/JTECH-D-11-00023.1
To obtain further insights into the differences of global biases and standard deviations between the two instruments, a global distribution of the brightness temperature differences between satellite observations and model simulations on a typical day, 2 January 2010, is provided in Fig. 5. For brevity, the brightness temperature differences between MWTS observations and global simulations will be denoted as 
Global distribution of (left) 
Citation: Journal of Atmospheric and Oceanic Technology 28, 10; 10.1175/JTECH-D-11-00023.1
The latitudinal dependence of MWTS biases is observed throughout the month. Figures 6 and 7 shows the biases calculated within every 5° latitudinal bands for the entire month of January. For both channel 7 and channel 9, 
Latitudinal dependence of biases (left) 
Citation: Journal of Atmospheric and Oceanic Technology 28, 10; 10.1175/JTECH-D-11-00023.1
As in Fig. 6, but for standard deviations.
Citation: Journal of Atmospheric and Oceanic Technology 28, 10; 10.1175/JTECH-D-11-00023.1
Positive biases are found for both MWTS channel 2 and AMSU-A channel 5 north of 20°N (Figs. 6a,b). These two channels also have large standard deviations. It is found that these are mostly caused by data over land. Figure 8 shows that the averaged brightness temperature differences for all outliers within the 15°–60°N latitudinal band for these two channels. Data are grouped into five different surface pressure intervals. The total percentage of positive and negative outliers is also provided. Positive biases prevail, and biases are larger over areas with higher topography. This may be caused by smaller surface emissivity used in the forward model because these two low channels peak at 700 hPa, which is close to high terrain surfaces in the Northern Hemisphere.
Averaged brightness temperature differences for all outliers within the 15°–60°N latitudinal band for (left) MWTS channel 2 (
Citation: Journal of Atmospheric and Oceanic Technology 28, 10; 10.1175/JTECH-D-11-00023.1
b. Scan-angle dependence of biases
Both MWTS and AMSU-A are cross-track-scanning microwave radiometers. A feature of the cross-track sounder is that the optical pathlength varies with scan angle. For MTWS and AMSU-A upper-level sounding channels, this effect dominates the characteristics of microwave-observed brightness temperature. The optical pathlength is the shortest at the nadir (i.e., zero scan angle) and the longest at the limb scan position of the earth view, which is 48.3°. Correspondingly, the brightness temperature is highest at the nadir for tropospheric channels and lowest at the nadir for stratospheric channels. As the scan angle increases, the brightness temperature is expected to decrease for tropospheric channels and increase for stratospheric channels. This feature of cross-track sounder is called the limb effect.
The level 1B radiances or brightness temperatures used in this study are not limb-adjusted to nadir values. This is reflected in Figs. 9a–f in which the global monthly mean brightness temperatures from FY-3A’s MWTS (solid line, left panels) and NOAA-18’s AMSU-A (solid line, right panels) observations are presented. The MWTS channels 2 and 3 and AMSU-A channels 5 and 7 are in the troposphere in which the atmospheric temperature usually decreases with altitude. The global mean brightness temperature curves down as a function of beam position, a result of an upward shift of the weighting function peak with increasing scan angle. The brightness temperatures at nadir could be different from those at the largest scan angle by more than 10 K, which corresponds to an upward shift of the WF peak by about 1 km. MWTS channel 4 and AMSU-A channel 9 are in the stratosphere in which the atmospheric temperature usually increases with altitude. The global mean brightness temperature slightly curves up as a function of beam position, which is also a result of an upward shift of the weighting function peak with increasing scan angle.
(a)–(f) Global monthly mean brightness temperatures from the (left) FY-3A MWTS (solid line) and (right) NOAA-18 AMSU-A (solid line). The corresponding model simulations are shown (dashed line). Scan-angle dependence of (g) bias 
Citation: Journal of Atmospheric and Oceanic Technology 28, 10; 10.1175/JTECH-D-11-00023.1
For NWP applications, the limb effect of a cross-track sounder is modeled in forward radiative transfer models. As seen in Fig. 9 (dashed line), the limb effect of cross-track MWTS and AMSU-A on brightness temperature measurements is mostly taken into account in the radiative transfer model. In general, a symmetric distribution of model-simulated brightness temperature with respect to scan angle is expected. This is true for the tropospheric channels for both instruments (dashed line in Figs. 9c–f). An asymmetric distribution of model simulation is noticed for MWTS channel 4 and AMSU-A channel 9 (Figs. 9a,b). A detailed diagnosis for the cause of this asymmetry reveals it arises from the global mean input temperatures being asymmetric if high-latitude data (50° north and south) are included. Figure 10, which presents the scan-angle dependence of model-simulated AMSU-A channel 9 brightness temperatures within different latitudinal bands, further confirms this. It is thus suggested that the NWP analysis fields within (50°S, 50°N) are most suited for characterizing the performance of cross-tracking satellite instruments.
Scan-angle dependence of model-simulated AMSU-A channel 9 brightness temperatures in different latitudinal bands including all data in January 2010.
Citation: Journal of Atmospheric and Oceanic Technology 28, 10; 10.1175/JTECH-D-11-00023.1
The size of FOV increases with increasing scan angle: the larger the FOV, the higher the atmospheric inhomogeneity within the FOV. The atmospheric inhomogeneity within FOVs at larger scan angles cannot be explicitly simulated in radiative transfer models. Furthermore, satellite observations at large scan angles could also be obstructed by the spacecraft radiation, which is also difficult to take into account in the forward model and calibration process. Therefore, angular-dependent biases between the observed brightness temperatures and those simulated from radiative transfer models are anticipated. In many applications, such as NWP radiance assimilation and satellite remote sensing systems, the angular-dependent biases between the observed brightness temperatures and those simulated from radiative transfer models are removed based on some empirically fitted relations from Figs. 9g,h (see Harris and Kelly 2001; Weng et al. 2003). One way to examine the systematic errors in the radiative transfer models that arise from the limb effect is to calculate the bias 
Similar to global biases, a latitudinal dependence exists in scan biases. Figure 11 presents the variation of scan biases with latitudes, in which the latitudinally dependent nadir biases (see Figs. 6c,d) are already subtracted. The data are divided into each 5° latitudinal band. The scan biases vary significantly with latitude, especially at the edge of the scan where biases are often largest. A significant latitudinal dependence of scan biases was also found for MSU channel 2 (Harris and Kelly 2001). This finding raises a concern on the traditional global bias correction algorithm in the NWP data assimilation (i.e., the one-size-fits-all approach) and suggests that a latitudinally dependent bias correction scheme is required for microwave sounding data assimilation.
Latitudinal dependence of scan biases. The latitudinally dependent nadir biases (Fig. 6) are subtracted. (left) FY-3A and (right) NOAA-18.
Citation: Journal of Atmospheric and Oceanic Technology 28, 10; 10.1175/JTECH-D-11-00023.1
c. Scene-temperature dependence of MWTS biases in polar regions
Table 2 provides brightness temperatures at three simultaneous nadir overpasses (SNOs) found at 70.53°N, 6.96°W on 10 January, 71.07°N, 51.44°W on 10 January, and 70.55°S, 45.39°E on 22 January. The data location separation between NOAA-18 and FY-3A measurements is less than 12.5 km in space and 60 s in time. For channels 4 and 9, the MWTS-observed brightness temperatures are 1–1.5 K warmer than the AMSU-A measurements at all three SNOs. For channel 3 and 7, the MWTS-observed brightness temperatures are about 0.8 K colder than the AMSU-A measurements at the SNOs near the North Pole, but 1.1 K warmer at the SNO near the South Pole.
In fact, the brightness temperature differences between MWTS and AMSU-A for the three SNO measurements reflect a quite general situation. Figure 12 provides global distributions of 100-hPa temperatures at 1200 UTC 2 January and 30 January 2010, as well as the differences between 2 and 30 January in the Southern and Northern Hemispheres. Note that the temperature in the polar regions of the Southern Hemisphere is the warmest in this month and the polar region in the Northern Hemisphere is dominated by wavenumber-1 planetary scale feature. The polar region in the Northern Hemisphere experienced a much larger variability of brightness temperature than that in the Southern Hemisphere from the beginning to the end of January 2010. Keeping this in mind, we examine the daily variation of the MWTS- and AMSU-A-observed and RTTOV-simulated brightness temperatures at channel 4 within the latitudinal band 80° ± 1.5°N, as well as the differences between observations and simulations from 1 to 31 January 2010 (Fig. 13). The patterns of MWTS-observed brightness temperatures generally agree well with those from AMSU-A and the simulations. The cold episode in the middle of January in the 0°–90°E longitudinal zone and the warm episode in late January near the 30°E–180° longitudinal zone are well captured by both instruments (see Figs. 13a–d). The AMSU-A observations are colder than the model simulations, but the differences are less than −1 K (Fig. 13f) and are uniform throughout all of the longitudes and all times, suggesting the good design of the antenna subsystem and a robust calibration. Contrary to AMSU-A channel 9, the differences between MWTS measurements and model simulations at channel 4 are largely positive near 80°N and vary strongly with longitude and time (see Fig. 13e). The MWTS observations could be more than 2–5 K warmer than the model simulations. The causes of this scene-dependent bias remain unknown. However, from the discussions below [see Eq. (1)], the uncertainty in characterizing the nonlinearity parameter could result in such a distribution.
Global distribution of 100-hPa temperatures at 1200 UTC on (a),(b) 2 and (c),(d) 30 Jan 2010, as well as the differences between 2 and 30 Jan in the (e) Southern and (f) Northern Hemispheres.
Citation: Journal of Atmospheric and Oceanic Technology 28, 10; 10.1175/JTECH-D-11-00023.1
(a),(b) Satellite-observed and (c),(d) RTTOV-simulated daily averaged brightness temperatures at (left) MWTS channel 4 and (right) AMSU-A channel 9 within the 78.5°–81.5°N latitudinal band. (e),(f) Differences between observations and simulations.
Citation: Journal of Atmospheric and Oceanic Technology 28, 10; 10.1175/JTECH-D-11-00023.1
Figure 14 displays the brightness temperatures and bias of MWTS channel 4 and AMSU-A channel 9 within the latitude band 80° ± 1.5°S. A substantial difference is shown in brightness temperature pattern for two instruments near the South Pole because of their different observational times. In January when the Southern Hemisphere is experiencing summer, the strong solar heating in the stratosphere causes a large diurnal change in stratosphere temperature and is well captured by the two instruments, with the brightness temperatures from MWTS being 1–2 K warmer than AMSU-A measurements. The temperature variations within this latitudinal band are, in general, less than 10 K (i.e., from 228 to 237 K), which is much smaller than the range of brightness temperature variation in the Northern Hemisphere counterpart (see Fig. 13). The westward propagation of the warm and cold episodes in the first half of the month and the eastward propagation of the warm and cold episodes in the second half of the month (Figs. 14c,d) are well captured by both instruments (see Figs. 14a,b). The AMSU-A observations are colder than the model simulations and the differences are less than 0.8 K (Fig. 14f) and are uniform throughout all the longitudes. The MWTS observations are close to model simulations and the differences are less than 0.4 K (Fig. 14e) and are very similar to those from AMSU-A.
As in Fig. 13, but for the Southern Hemisphere within 78.8°–81.5°S.
Citation: Journal of Atmospheric and Oceanic Technology 28, 10; 10.1175/JTECH-D-11-00023.1
Figures 15 and 16 present satellite-observed and RTTOV-simulated brightness temperatures at MWTS channel 3 and AMSU-A channel 7 within the latitudinal bands of 78.5°–81.5°N and 78.5°–81.5°S, respectively. A scene temperature-dependent bias is also observed for MWTS channel 3. However, the MWTS bias in the polar region of Northern Hemisphere is positive over warmer scene temperatures and negative over colder scene temperatures (Fig. 15), which is opposite to the sign of biases in the stratospheric MWTS channel 4 (Fig. 13). The MWTS bias in the Southern Hemispheric polar region is negative over warmer scene temperatures and positive over colder scene conditions. The differences between MWTS observations and simulations at channel 3 can reach ±1.5 K. The AMSU-A channel 7 observations are consistently colder than model simulations in the polar region of both hemispheres, with a difference being less than −1 K.
As in Fig. 13, but for (left) MWTS channel 3 and (right) AMSU-A channel 7.
Citation: Journal of Atmospheric and Oceanic Technology 28, 10; 10.1175/JTECH-D-11-00023.1
As in Fig. 15, but for the Southern Hemisphere within 78.8°–81.5°S.
Citation: Journal of Atmospheric and Oceanic Technology 28, 10; 10.1175/JTECH-D-11-00023.1
d. Root cause analysis of MWTS global biases
In general, Cc and Cw from several scans are averaged to avoid the spikes from random noise in count. For AMSU-A calibration, n = 3 in Eq. (2), t0 represents the current scan line, ti = t0 + iΔt, and Δt = 8; and for MWTS, Δt = 16. The values of the parameter μ in (1) can be derived from the prelaunch thermal vacuum data but could be different on orbit if the instrument temperature changes. The nonlinear coefficient μ after satellite launch was obtained for both MSU and AMSU instruments using the simultaneous nadir overpass method (Zou et al. 2006, 2009). The updated μ values from the SNO calibration can improve the calibration accuracies for MSU and AMSU instruments.
Based on Eq. (1), it is clear that observation errors of satellite brightness temperatures depend on the earth scene temperature as well as instrument temperature. The closer the blackbody temperature is to the earth scene temperature, the more accurate the two-point calibration is. Because the nonlinearity term is estimated at three instrument temperatures, the nonlinear correction is also more accurate when the earth scene temperatures are close to these three instrument temperature points specified in prelaunch calibration. A nonlinearity term for a typical instrument can be problematic for many applications if the parameter μ is not accurately quantified in the prelaunch calibration process. The evidence for MWTS nonlinearity issue will be highlighted below.
The bias distribution with latitude is another important area to be further investigated. Figures 17 and 18 present the scatterplots of the global brightness temperature biases for MWTS channel 4 and AMSU-A channel 9 measurements against the simulated brightness temperatures with latitude indicated by colored dots. For AMSU-A channel 9, the bias depends neither on latitude, nor on temperature. For MWTS channel 4, the bias distribution in the Southern Hemisphere is similar to that of AMSU-A (see Figs. 17a,b), but the bias increases with latitude in the Northern Hemisphere and reaches a maximum of 2–3 K (Figs. 15c–f). Similar characteristics are observed for the entire month of January 2010 (Fig. 18). Similar patterns are seen or data over ocean only or over ocean with clear sky only.
Scatterplots of the brightness temperature biases for (left) MWTS and (right) AMSU-A against the simulated brightness temperatures during 1–5 Jan 2010. (a),(b) 0°–90°S, (c),(d) 0°–50°N, and (e),(f) 50°–90°N.
Citation: Journal of Atmospheric and Oceanic Technology 28, 10; 10.1175/JTECH-D-11-00023.1
Scatterplots of the brightness temperature biases for (left) MWTS and (right) AMSU-A against the simulated brightness temperatures during 1–15 Jan 2010. (top) All data, (middle) ocean only, and (bottom) clear sky over ocean only.
Citation: Journal of Atmospheric and Oceanic Technology 28, 10; 10.1175/JTECH-D-11-00023.1
It is also interesting to compare the bias variation of MWTS channel 3 with that of AMSU-A channel 7. Figure 19 shows the scatterplots of the brightness temperature biases for MWTS channel 3 and AMSU-A channel 7 measurements during January 2010 against the simulated brightness temperatures. Because this MWTS channel measures the radiation from the troposphere and peaks near 300 hPa, the global brightness temperatures vary within a smaller range (about 24 K) and are also warmer than those at channel 4. For MWTS channel 3, the bias distribution in the Northern Hemisphere is similar to that of AMSU-A (with larger standard deviations), but is much larger in the Southern Hemisphere. The differences between MWTS observations and model simulation can exceed ±1 K, whereas the AMSU-A measurements compare much more favorably with model simulations. Similar to channel 9, the biases of AMSU-A channel 7 are also independent of latitude and temperature, suggesting a very successful calibration on nonlinearity.
As in Fig. 18, but for (left) MWTS channel 3 and (right) AMSU-A channel 7.
Citation: Journal of Atmospheric and Oceanic Technology 28, 10; 10.1175/JTECH-D-11-00023.1
The bias characteristics of MWTS measurements indicate that the nonlinearity term may not be well characterized in the calibration algorithm. However, this is not the only cause for the MWTS biases for the following reasons: If nonlinearity is the sole cause for the MWTS biases, the biases in the Southern Hemisphere of channel 4 would also be scene dependent because the scene temperature varies substantially in the area, that is, from 190 to 236 K. Thus, it is hypothesized that some other factors may have also contributed to the MWTS biases.
A possible cause for the latitudinal dependence of the MTWS bias may be the solar contamination on MWTS on board the calibration target. This event can often happen at the high latitudes in the Northern Hemisphere for morning-orbit satellites such as FY-3A. In January when the FY-3A satellite passes over the equator at 1000 LT in the ascending node, the instrument calibration target can be illuminated by the direct and stray solar radiation after the satellite is out of the earth’s shadow near the latitude of 30°N. If the warm count from the blackbody is affected by the extra solar radiation and the platinum resistance temperature (PRT) does not respond fast enough and the PRT temperature variation is not consistent with that of the warm count, then the calibration gain in Eq. (1) would be unstable, which can also cause some latitudinally dependent biases.
The drift in the MTWS receiver center frequency at channel 4 is another factor that was questioned as the root cause for MWTS bias (Lu et al. 2010). However, the center-frequency drift has yet to be confirmed by the MWTS instrument vendor. The actual sources of MWTS biases may be a combination of all of the above factors and would require further research on the root cause analysis.
A quantitative analysis of the biases at MWTS channel 2 versus AMSU-A channel 5 would be rather difficult because the biases can be also strongly contributed by the uncertainty in the forward model. For example, a large negative bias at a cold scene temperature in the Southern Hemisphere (Fig. 20) is more related to the region over the Antarctic continent where surface snow prevails and its emissivity is not well characterized in the RTTOV. For a warm scene region near 240–260 K, the large negative bias can be more related to scattering from precipitating clouds, which is also not included in the forward model. Throughout all of the scene temperatures, the large positive biases can be more related to the surface and atmospheric inhomogeneity within the satellite field of view, which has not been considered in the forward model. The biases of AMSU-A channel 5 over ocean are nearly independent of latitude and temperature, suggesting a very successful calibration of AMSU-A data over ocean.
As in Fig. 18, but for (left) MWTS channel 2 and (right) AMSU-A channel 5.
Citation: Journal of Atmospheric and Oceanic Technology 28, 10; 10.1175/JTECH-D-11-00023.1
5. Summary and conclusions
For a satellite sounding instrument, the bias between observed and simulated brightness temperatures can be used to conduct the root cause analysis of the calibration error sources and instrument performance anomaly. In this study, we derived the bias characteristics of the FY-3A MWTS and NOAA-18 AMSU-A and compared their global and regional distributions for the month of January 2010. The simulations are derived using NCEP GFS outputs as inputs to RTTOV. It is noted that the MWTS channels 3 and 4 show more biases than the corresponding AMSU-A channels, especially in high latitudes. Nonlinear biases for MWTS channel 4 brightness temperature measurements are found mainly north of 30°N and increase in magnitude toward higher latitudes in the Northern Hemisphere. Nonlinearity for MWTS channel 3 brightness temperature measurements is found mainly south of 30°S, with biases increasing in magnitude toward higher latitudes in the Southern Hemisphere. These bias results suggest that a more careful postlaunch calibration analysis is required for the MWTS data prior to its operational uses in NWP systems.
Although it is not possible to find the exact root causes for MWTS biases through comparisons between measurements with NWP model simulations alone, several possible factors are explained from the satellite calibration algorithm and the antenna subsystem. Because NOAA-18 is an afternoon orbit satellite, the AMSU-A calibration system is less vulnerable to the stray light contamination and has less calibration uncertainty, and therefore its spatial and temporal distribution can be used as a reference. It is suggested that the MWTS bias can be contributed by several possible causes, including the contamination of its calibration target by stray lights, detector nonlinearity, antenna reflector emission, and/or center frequency uncertainties. More studies are required to identify the exact root cause and propose the algorithms for correction so that the MWTS data quality can be further improved.
This study further demonstrates the usefulness of the NWP analysis fields for generating global simulations for characterizing the performance of new instruments during and after their in-orbit verification period. With the upcoming launch of FY-3C by China, we plan to extend this work to the postlaunch calibration of both FY-3A/B MWTS measurements. The methodology used in this study will be used to create an automated real-time monitoring system for comparing satellite measurements with NWP models, which will be used as a tool to monitor the long-term performance of satellite observing systems and to effectively incorporate satellite data in NWP modeling systems. Characterization of MWTS bias is also a critical step for linking the FY-3 data to NOAA MSU/AMSU time series and for creating a long-term fundamental climate data record (CDRs) in climate monitoring, reanalyses, and forecasts.
Acknowledgments
This work was jointly supported by Chinese Ministry of Science and Technology Project 2010CB951600, and Chinese Ministry of Finance Project GYHY200906006. The authors would like to express our sincere thanks to Dr. Sid Boukabara for his reviews on this manuscript.
REFERENCES
Bell, W., English S. , Candy B. , Hilton F. , and Atkinson N. , 2007: The assimilation of SSMIS radiances in numerical weather prediction models. IEEE Trans. Geosci. Remote Sens., 46, 884–900.
Christy, J. R., Spencer R. W. , and Lobl E. S. , 1998: Analysis of the merging procedure for the MSU daily temperature time series. J. Climate, 11, 2016–2041.
Christy, J. R., Spencer R. W. , and Braswell W. D. , 2000: MSU tropospheric temperatures: Dataset construction and radiosonde comparisons. J. Atmos. Oceanic Technol., 17, 1153–1170.
Christy, J. R., Spencer R. W. , Norris W. B. , Braswell W. D. , and Parker D. E. , 2003: Error estimates of version 5.0 of MSU–AMSU bulk atmospheric temperatures. J. Atmos. Oceanic Technol., 20, 613–629.
Dong, C., and Coauthors, 2009: An overview of a new Chinese weather satellite FY-3A. Bull. Amer. Meteor. Soc., 90, 1531–1544.
Goodrum, G., Kidwell K. , and Winston W. , 2009: Advanced Microwave Sounding Unit-A. NOAA User’s Guide with NOAA-N, -N′, NOAA, 35–44. [Available online at http://www2.ncdc.noaa.gov/docs/klm/cover.htm.]
Guan, L., Zou X. , Weng F. , and Li G. , 2011: Assessments of FY-3A Microwave Humidity Sounder measurements using NOAA-18 microwave humidity sounder. J. Geophys. Res., 116, D10106, doi:10.1029/2010JD015412.
Harris, B. A., and Kelly G. , 2001: A satellite radiance-bias correction scheme for data assimilation. Quart. J. Roy. Meteor. Soc., 127, 1453–1468.
Lu, Q., Bell W. , Bauer P. , Bormann N. , and Peubey C. , 2010: An initial evaluation of FY-3A satellite data. ECMWF Tech. Memo. 631, 58 pp.
Mears, C. A., and Wentz F. J. , 2009: Construction of the Remote Sensing Systems V3.2 atmospheric temperature records from the MSU and AMSU Microwave Sounders. J. Atmos. Oceanic Technol., 26, 1040–1056.
Mears, C. A., Schabel M. C. , and Wentz F. J. , 2003: A reanalysis of the MSU channel 2 tropospheric temperature record. J. Climate, 16, 3650–3664.
Mo, T., 1996: Prelaunch calibration of the Advanced Microwave Sounding Unit-A for NOAA-K. IEEE Trans. Microwave Theory Tech., 44, 1460–1469.
Mo, T., 1999: Calibration of the advanced microwave sounding unit-A radiometers for NOAA-L and NOAA-M. NOAA Tech. Rep. NESDIS 92, 53 pp.
Ohring G., Wielicki B. , Spencer R. , Emery B. , and Datla R. , 2005: Satellite instrument calibration for measuring global climate change. Bull. Amer. Meteor. Soc., 86, 1303–1313.
Saunders, R. W., Matricardi M. , and Brunel P. , 1999: An improved fast radiative transfer model for assimilation of satellite radiance observations. Quart. J. Roy. Meteor. Soc., 125, 1407–1425.
Vinnikov, K. Y., and Grody N. C. , 2003: Global warming trend of mean tropospheric temperature observed by satellites. Science, 302, 269–272.
Weng, F., and Grody N. C. , 1994: Retrieval of cloud liquid water over oceans using Special Sensor Microwave Imager (SSM/I). J. Geophys. Res., 99, 25 535–25 551.
Weng, F., Zhao L. , Ferraro R. , Poe G. , Li X. , and Grody N. , 2003: Advanced Microwave Sounding Unit cloud and precipitation algorithms. Radio Sci., 38, 8086–8096.
Yang, J., Dong C. , Lu N. , Yang Z. , Shi J. , Zhang P. , Liu Y. , and Cai B. , 2009: Chinese new generation of polar-orbiting satellites FY-3. Acta Meteor. Sin., 67, 501–509.
Zhang, P., Yang J. , Dong C. , Lu N. , Yang Z. , and Shi J. , 2009: General introduction on payloads, ground segment and data application of Fengyun 3A. Front. Earth Sci. China, 3, 367–373.
Zou, X., and Zeng Z. , 2006: A quality control procedure for GPS radio occultation data. J. Geophys. Res., 111, D02112, doi:10.1029/2005JD005846.
Zou, C.-Z., Goldberg M. D. , Cheng Z. , Grody N. C. , Sullivan J. T. , Cao C. , and Tarpley D. , 2006: Recalibration of microwave sounding unit for climate studies using simultaneous nadir overpasses. J. Geophys. Res., 111, D19114, doi:10.1029/2005JD006798.
Zou, C.-Z., Gao M. , and Goldberg M. D. , 2009: Error structure and atmospheric temperature trends in observations from the Microwave Sounding Unit. J. Climate, 22, 1661–1681.
