HTML
--> --> -->| Channel no. | Frequency (GHz) | Specification (K) | NEDT (K) | Beam width (°) | Peak Weighting Function (hPa) |
| 1 | 23.8 | 0.5 | 0.25 | 5.2 | Surface |
| 2 | 31.4 | 0.6 | 0.3 | 5.2 | Surface |
| 3 | 50.3 | 0.7 | 0.35 | 2.2 | Surface |
| 4 | 51.76 | 0.5 | 0.28 | 2.2 | 950 |
| 5 | 52.8 | 0.5 | 0.25 | 2.2 | 850 |
| 6 | 53.596 ± 0.115 | 0.5 | 0.27 | 2.2 | 700 |
| 7 | 54.4 | 0.5 | 0.25 | 2.2 | 400 |
| 8 | 54.94 | 0.5 | 0.25 | 2.2 | 250 |
| 9 | 55.5 | 0.5 | 0.28 | 2.2 | 200 |
| 10 | 57.29 | 0.75 | 0.4 | 2.2 | 100 |
| 11 | 57.29 ± 0.217 | 1 | 0.53 | 2.2 | 50 |
| 12 | 57.29 ± 0.322 ± 0.048 | 1 | 0.55 | 2.2 | 25 |
| 13 | 57.29 ± 0.322 ± 0.022 | 1.25 | 0.82 | 2.2 | 10 |
| 14 | 57.29 ± 0.322 ± 0.010 | 2.2 | 1.13 | 2.2 | 5 |
| 15 | 57.29 ± 0.322 ± 0.0045 | 3.6 | 1.8 | 2.2 | 2 |
| 16 | 88.2 | 0.3 | 0.27 | 2.2 | Surface |
| 17 | 165.5 | 0.6 | 0.39 | 1.1 | Surface |
| 18 | 183.31 ± 7.0 | 0.8 | 0.35 | 1.1 | 800 |
| 19 | 183.31 ± 4.5 | 0.8 | 0.41 | 1.1 | 700 |
| 20 | 183.31 ± 3.0 | 0.8 | 0.48 | 1.1 | 500 |
| 21 | 183.31 ± 1.8 | 0.8 | 0.53 | 1.1 | 400 |
| 22 | 183.31 ± 1.0 | 0.9 | 0.68 | 1.1 | 300 |
Table1. ATMS channel characteristics.
Spaceborne microwave remote sensing observations, such as ATMS, are a key data type for numerical weather prediction (NWP). Zou et al. (2013) demonstrated that the assimilation of ATMS radiances into the Hurricane Weather Research and Forecasting model helps to improve both the hurricane track and intensity forecast performance. Previous studies have also shown positive impacts on global weather forecast skill brought by the Advanced Microwave Sounding Unit-A (AMSU-A), which is the predecessor of ATMS (Eyre et al., 1993; Andersson et al., 1994; Derber and Wu, 1998; Qin et al., 2012). Besides being a part of observational inputs for NWP models, their applications also include retrieving surface temperatures, atmospheric temperatures, total precipitable water, liquid water paths, and ice water paths under almost all weather conditions except for heavy precipitation. Tian and Zou (2016) showed that the measurements from both AMSU-A and ATMS can be used to analyze the three-dimensional hurricane warm-core structures with a temperature profile retrieval algorithm they proposed. Tian and Zou (2018) combined the microwave temperature sounder instruments on multiple satellites to retrieve the three-dimensional warm-core structure temporal evolutions in Hurricanes Harvey, Irma, and Maria. Zou and Tian (2018) further refined the temperature retrieval algorithm by training the retrieval coefficients with Global Positions System (GPS) radio occultation (RO) temperature profiles for achieving better accuracies of the temperature retrieval products.
However, before any of these applications, the bias features of each channel have to be characterized. Any bias has to be properly quantified and then removed. Zou et al. (2014) characterized the noise and bias characteristics of the ATMS onboard S-NPP using NWP analysis/forecast fields. ATMS brightness temperatures (TBs) were simulated with atmospheric temperature and water vapor profiles from GPS RO observations as an input to the Community Radiative Transfer Model (CRTM). CRTM is known to be able to rapidly simulate radiances with an accuracy of less than 0.1 K for microwave sensors (Liu et al., 2013). It was shown that S-NPP ATMS biases for the temperature sounding channels 5–15 could be well characterized by GPS RO data. In this study, the in-orbit accuracy of the ATMS onboard both the recently launched NOAA-20 satellite and the S-NPP satellite were estimated using GPS RO level-2 retrieval profiles from the two Global Navigation Satellite System (GNSS) Receivers for Atmospheric Sounding (GRAS) onboard the Meteorological Operational (MetOp)-A and MetOp-B satellites (Gorbunov et al., 2011).
Since NOAA-20 operates in the same orbit as S-NPP but about 50 min ahead of it, NOAA-20 allows an important overlap in ATMS observational coverage. This gives meteorologists a new opportunity to obtain ATMS information at a half-hour interval, twice daily, for fast-evolving weather systems such as hurricanes. An example is shown in this regard for Hurricane Florence (2018).
2.1. Data description
Atmospheric temperature and humidity profiles retrieved from GPS RO observations of the GRAS instrument onboard both MetOp-A and -B serve as inputs for model simulations of ATMS antenna temperatures. With the aid of the Radio Occultation Processing Package, the raw RO measurements of excess Doppler shifts can be used to retrieve the atmospheric refractivity. The GPS RO level-2 atmospheric temperature and humidity profiles can then be retrieved from the refractivity with a one-dimensional variational data assimilation method with the 137-level ECMWF reanalysis data as first guess (Healy and Eyre, 2000; Culverwell et al., 2015). The horizontal resolution of each RO profile is about 270 km (Kursinski et al., 1997). The estimated errors for temperature profiles are less than 1 K from the surface to about 40 km and less than 0.4 K for the layer from 8 km to 25 km (Angling, 2016). The GPS RO level-2 retrieval profiles used in this study were provided by the Radio Occultation Meteorology Satellite Application Facility (ROM SAF) under the European Organisation for the Exploitation of Meteorological Satellites (EUMETSAT) (Nielsen et al., 2016). The MetOp-A and -B RO mission can generate approximately 1200 level-2 temperature/humidity profiles daily. In order to estimate the biases in ATMS measurements of antenna temperatures, the observed ATMS antenna temperatures (temperature data records) during the period from 23 January to 23 July 2018 were compared against the CRTM-simulated TBs generated with RO profile data during the same time period as input background information to the CRTM.2
2.2. Methodology
In practice, before any TB simulations, a bi-weight quality control (QC) needs to first be applied to the MetOp-A and -B RO profile data to ensure the validity and quality of RO profiles input into CRTM. The first step in the QC procedures is to ensure all refractivity values are physical, i.e., positive. The bi-weight mean and bi-weight standard deviations of the temperatures at different pressure levels are then calculated. In order to ensure that all RO profiles input into CRTM are of reasonable accuracy, outliers whose deviations from the bi-weight mean are 2.5 times larger than the bi-weight standard deviation are further excluded. More details regarding the formulation and the bi-weight QC method can be found in Zou and Zeng (2006).RO profiles that pass the abovementioned bi-weight QC procedures are collocated with ATMS observations from both S-NPP and NOAA-20 under the criteria of being within a 100-km spatial distance and 3-h time difference (Zou et al., 2014). The total numbers of GRAS RO profiles collocated within ATMS observations from S-NPP and NOAA-20 are 34 965 and 35 870 under clear-sky conditions, respectively, and 48 743 and 46 964 under cloudy conditions, respectively. Above 800 hPa, less than 10% of data are masked out by the bi-weight QC (Fig. 1). All of these collocated RO profiles are then used as inputs to the CRTM to generate ATMS simulations.
Figure1. Variations of data count removed in the bi-weight QC with respect to pressure (black curve) and the weighting functions of channels 8 (blue curve) and 13 (red curve).
Figure2. (a) Biases and (b) standard deviations of ATMS channels 8–13 calculated from the differences in TBs between the S-NPP (black bars) and NOAA-20 (blue bars) observations and RO/CRTM simulations at nadir FOVs during the 6-month period from 23 January to 23 July 2018. The red curve indicates the specification values of S-NPP ATMS channels.Figure 3 shows scatterplots of S-NPP and NOAA-20 ATMS O-B values at channel 9 versus those at channel 8. The distribution of the NOAA-20 (Fig. 3b) data points looks tighter than that of the S-NPP (Fig. 3a) data points, which is also reflected by a smaller standard deviation of the former than the latter. Figure 3 also shows the minimal impacts of clouds on these upper-level channels and the small differences of the bias and standard deviation between all-weather and clear-sky-only data. An inter-channel correlation of O-B between channels 8 and 9 is seen for NOAA-20 and S-NPP.
Figure3. Differences in the observed and simulated TB (K), i.e. O-B, of channel 9 with respect to channel 8 for the ATMS onboard (a) S-NPP and (b) NOAA-20 under clear-sky (green dots) and cloudy (red dots) conditions. The mean values of clear-sky (cyan) and cloudy (black) conditions are indicated by the black cross, and standard deviations by the dashed box. Only nadir observations were included when calculating the mean and standard deviation values.The ATMS is a cross-track scanning radiometer, meaning the optical path varies with scan angle. Figure 4a shows the number of RO profiles collocated with ATMS observation pixels at each field of view (FOV) position from 23 January to 23 July 2018. The FOV denotes the angle of view at which the ATMS can effectively detect the radiation. About 600 RO profiles are collocated near nadir ATMS FOVs. The numbers of collocated GRAS RO profiles rapidly increase at larger scan-angle positions, resulting from larger FOV sizes and larger spatial separations between the centers of two neighboring FOVs at larger scan angles. Figure 4b shows the means (dashed) and standard deviations (solid) of the O-B of channel 8 at the 96 scan positions for the ATMS onboard NOAA-20 (blue) and S-NPP (black). The ATMS scan biases, with the biases calculated at nadir subtracted, are asymmetric with respect to scan angle for both NOAA-20 and S-NPP, with much smaller scan variation on the side of FOVs 1–48 than the side of FOVs 49–96. This asymmetry was found to be caused by the antenna sidelobe intercepts with the spacecraft, as ATMS is mounted on the side of the spacecrafts (Kim et al., 2014). It is also apparent that the NOAA-20 ATMS scan biases are smaller in magnitude than those of S-NPP on the side of FOVs 49–96. The standard deviations for both ATMS instruments show no significant scan-dependent features. Similar to the results in Fig. 2b, the NOAA-20 ATMS standard deviations are generally smaller than those of the S-NPP ATMS at most FOV positions. Figure 5 shows the O-B biases at channels 8–13 as functions of the 96 scan angles for the ATMS onboard NOAA-20 and S-NPP. Similar to channel 8 (Fig. 4b), the scan-dependent biases at channels 9–13 for the NOAA-20 ATMS are generally smaller than those for the S-NPP ATMS (Fig. 5). The patterns for NOAA-20 ATMS channels 12 and 13 appear less symmetric than those for the corresponding S-NPP ATMS channels. Scan biases for channels 9–11 display similar asymmetries with respect to FOV positions as that of channel 8.
Figure4. (a) Data count and (b) biases (dashed curves) with standard deviations (solid curves) of channel 8 of ATMS onboard S-NPP (black) and NOAA-20 (blue) calculated with simulated TBs with RO as backgrounds from 23 January to 23 July 2018. Profiles under all-sky conditions were included in the calculations.
Figure5. Biases of channels (a) 8–10 and (b) 11–13 of the ATMS onboard NOAA-20 (solid curves) and S-NPP (dashed curves) with respect to scan positions under all-sky conditions.Previous studies, including Zou et al. (2014) and Weng et al. (2013), have reported latitudinal dependences of microwave temperature sounder data biases. The latitudinal features of the NOAA-20 ATMS were examined by grouping the RO profiles collocated with ATMS near nadir pixels in every 10° latitudinal band. Figure 6a shows the data counts of RO profiles collocated with ATMS observations at nadir (FOV positions 47 and 48) within each latitudinal band. Figure 6b shows the means (dashed) and standard deviations (solid) of O-B at channel 8 as functions of latitudinal bands for the ATMS onboard NOAA-20 and S-NPP. Consistent with previous results, the NOAA-20 ATMS biases are greater in both magnitude and latitudinal variation than those of the S-NPP ATMS. The NOAA-20 ATMS biases are largest at high latitudes. The standard deviations show a similar pattern, with larger magnitudes at higher latitudes. Figure 7 shows the latitudinal bias distributions for ATMS channels 8–13. There is a strong latitudinal dependence of the ATMS biases from both S-NPP (Fig. 7a) and NOAA-20 (Fig. 7b). These latitudinal dependences are possibly due to the opposite sign of O-B in the background, i.e., RO profiles (Wang and Zou, 2012). The magnitudes of the biases of channels 8 and 9 are greater at low latitudes than at high latitudes, and vice versa for channels 10–13. Since S-NPP and NOAA-20 ATMSs use the same radiance-based calibration algorithm, the ATMS bias differences between S-NPP and NOAA-20 could arise from some coefficients in the processing coefficient tables, which are sensor-dependent and different for these two satellites (Ninghai SUN, personal communication, 2018).
Figure6. (a) Data count and (b) the mean (dashed curves) and standard deviations (solid curves) of O-B with respect to latitudinal bands of every 10° for the ATMS onboard S-NPP (black) and NOAA-20 (blue) for observations from 23 January to 23 July 2018. Profiles under all-sky conditions were included in the calculations.
Figure7. The mean differences of O-B at channels 8–13 with respect to latitudinal bands at every 10° calculated from 23 January to 23 July 2018 for the ATMS onboard (a) S-NPP and (b) NOAA-20. Profiles under all-sky conditions were included in the calculations.As the NOAA-20 ATMS has larger negative biases and smaller standard deviations than those of the S-NPP ATMS, the updated bias correction and error variance estimation for the NOAA-20 ATMS is required in order to assimilate NOAA-20 ATMS data in NWP and in linking ATMS observations from NOAA-20 to those of S-NPP and AMSU-A for climate studies (Tian and Zou, 2019).
Figure8. TB (K) observations at ATMS (a, b) channel 6 and (c, d) channel 8 from (a, c) S-NPP ATMS at 1629 UTC and (b, d) NOAA-20 ATMS at 1718 UTC on 10 September 2018 within and around Hurricane Florence. The hurricane center from the best-track data are indicated by a black cross symbol.ATMS, as mentioned in section 1, is a cross-track scanning radiometer with constantly changing scan angles in one scan line. The TB measurements thus have scan dependence, or limb effects, that can conceal much of the hurricane’s features. These limb effects can be removed by using the method described and applied in Zou and Tian (2018) and Tian et al. (2018). One may also use it to remove the prevailing scan variations to reveal the storm structures hidden in the TB measurements. NOAA-20 ATMS observations from 1 to 31 August 2018 were used to train the coefficients of limb correction. Figure 9 shows the mean scan variations in the TBs of ATMS channels 5–12 onboard S-NPP (solid) and NOAA-20 (dashed) as well as their differences (dotted). The scan patterns of the two ATMS instruments generally agree well with each other; both have some minor asymmetric distributions. The limb-corrected TB observations are given in Fig. 10. The storm’s rain-band features, which are vaguely visible in Fig. 8, can be clearly seen in the limb-corrected TB observations from both NOAA-20 and S-NPP, regardless of the storm’s relative location within a swath. Although Hurricane Florence was observed near the edge of an S-NPP ATMS swath, the warm center and cold rain bands (Figs. 10a and c) are as successfully recovered as those located near the nadir of a NOAA-20 ATMS swath (Figs. 10b and d) by the limb correction. In this instance, the center of Hurricane Florence was covered by NOAA-20 ATMS 50 min after the S-NPP ATMS; the limb-corrected ATMS measurements also enable the capture of the evolutions of the warm center during the 50 min.
Figure9. Scan variations in global mean TBs of ATMS channels 5–12 (solid lines for S-NPP and dashed lines for NOAA-20) and differences in mean TBs between NOAA-20 and S-NPP (dotted lines). Results shown in all panels were calculated using data from 1 to 31 of August 2018.
Figure10. TBs (K) after limb corrections at (a, b) channel 6 and (c, d) channel 8 observed by (a, c) S-NPP ATMS at 1629 UTC and (b, d) NOAA-20 ATMS at 1718 UTC on 10 September 2018 within and around Hurricane Florence. The center of Hurricane Florence from the best-track data is indicated by a black cross symbolThe observations of ATMS onboard both satellites in the case of Hurricane Florence (2018) were examined in this study. It was found that the hurricane can be covered by ATMS measurements that are 50 min apart from both satellites, giving up to four ATMS coverages daily. This back-to-back orbiting setup enables a better temporal resolvability with cloud penetrating microwave observations over extreme weather events like Hurricane Florence. The limb-effect correction algorithm was applied to ATMS antenna temperature observations to reveal the structures of Hurricane Florence embedded but not visible in the antenna temperature observations. Future studies should include better quantitative comparisons of NOAA-20 ATMS instrument biases and other features, such as seasonal and annual variations, with ATMS and AMSU-A instrument biases.
Acknowledgements. This research was supported by NOAA (Grant No. NA14NES4320003). The authors are grateful to those who contributed to the Comprehensive Large Array-data Stewardship System for providing the ATMS observation data, and to ROM SAF for providing the RO profile data. The software developed to perform the calculations in this study is available by contacting the Corresponding Author, Xiaolei ZOU, at xzou1@umd.edu.
