J. Meteor. Res.  2018, Vol. 32 Issue (5): 804-818   PDF    
The Chinese Meteorological Society

Article Information

HAN, Yang, and Fuzhong WENG, 2018.
Remote Sensing of Tropical Cyclone Thermal Structure from Satellite Microwave Sounding Instruments: Impacts of Optimal Channel Selection on Retrievals. 2018.
J. Meteor. Res., 32(5): 804-818

Article History

Received January 22, 2018
in final form July 11, 2018
Remote Sensing of Tropical Cyclone Thermal Structure from Satellite Microwave Sounding Instruments: Impacts of Optimal Channel Selection on Retrievals
Yang HAN1, Fuzhong WENG2     
1. Nanjing University of Information Science & Technology, Nanjing 210044;
2. State Key Laboratory of Severe Weather, Chinese Academy of Meteorological Sciences, Beijing 100081
ABSTRACT: Accurate information on atmospheric temperature of tropical cyclones (TCs) is important for monitoring and prediction of their developments and evolution. For hurricanes, temperature anomaly in the upper troposphere can be derived from Advanced Microwave Sounding Unit (AMSU) and Advanced Technology Microwave Sounder (ATMS) through either regression-based or variational retrieval algorithms. This study investigates the dependency of TC warm core structure on emission and scattering processes in the forward operator used for radiance computations in temperature retrievals. In particular, the precipitation scattering at ATMS high-frequency channels can significantly change the retrieval outcomes. The simulation results in this study reveal that the brightness temperatures at 183 GHz could be depressed by 30–50 K under cloud ice water path of 1.5 mm, and thus, the temperature structure in hurricane atmosphere could be distorted if the ice cloud scattering was inaccurately characterized in the retrieval system. It is found that for Hurricanes Irma, Maria, and Harvey that occurred in 2017, their warm core anomalies retrieved from ATMS temperature sounding channels 4–15 were more reasonable and realistic, compared with the retrievals from all other channel combinations and earlier hurricane simulation results.
Key words: Advanced Technology Microwave Sounder (ATMS)     Microwave Retrieval Testbed (MRT)     warm core     hurricane     Irma     Maria     Harvey    
1 Introduction

Tropical cyclone is one of the most catastrophic weather systems, which usually causes severe damages to properties and lives. It forms with several environmental conditions: pre-existing tropical disturbance, warm water (> 26.5°C) with sufficient layer, at least 5° from the equator to provide the significant Coriolis force, low vertical wind shear, moisture in the lower-to-mid troposphere, and conditional instability through the atmosphere (Gray, 1968). Warm core is a vital feature for ensuring a consistent dynamic and thermodynamic tropical cyclone. Liu et al. (1999) simulated both the inner core and intensification of Hurricane Andrew (1992) and found that the magnitude of warm anomaly increased from 5 to 16 K with the height of warm anomaly dropping from 11 to 6.5 km as the storm deepened. Through model simulations, a warm core structure with 4.5 K at 400 hPa was well defined by Wang et al. (2010), and the sensitivity to different model physics options and initial conditions was demonstrated during pre-genesis evolution of Hurricane Felix (2007). Chen and Zhang (2013) emphasized the important relationship between upper-level warm core and rapid intensification (RI) in Hurricane Wilma (2005). They concluded that the significant convective bursts activity in the inner-core regions acted as an important ingredient in forming the upper-level warm core. Vigh and Schubert (2009) presented the theoretical argument of the condition that a tropical cyclone could rapidly develop a warm core structure and subsequently approach a steady state; to be specific, generating a warm core structure depended on the diabatic heating inside the radius of maximum wind of the tropi-cal cyclone/hurricane.

Earlier observational studies also confirmed the consistent warm core structure. The hurricane case with weak (La Seur and Hawkins, 1963), moderate (Hawkins and Rubsam, 1968), and strong (Hawkins and Imbembo, 1976) intensity was observed respectively by aircraft reconnaissance in support of NOAA hurricane research mission. Both horizontal and vertical structures of wind, temperature, humidity, and other parameters were analyzed. It is found that a maximum anomaly of greater than 16 K was located at heights varying from 600 to 200 hPa and the strongest gradients were within the eyewall region from the vertical cross-section of the temperature field. According to the Fourth Convection and Moisture Experiment (CAMEX-4) program, the warm core of Hurricane Erin was with a maximum perturbation temperature of 11 K at near 500 hPa, being broadly consistent with earlier finding (Halverson et al., 2006). By using the GPS dropsonde observation, the warm core structure of Hurricane Humberto (2001) was revealed and it exhibited a maximum of 6–7 K located at about 2-km height, which was weaker than that of other storms (Dolling and Barnes, 2014). From NASA’s Genesis and Rapid Intensification Processes (GRIP) project during Hurricane Earl (2010), Stern and Zhang (2016) concluded that the relationship between the height of warm core and hurricane intensity and its rate of change varies substantially from storm to storm. According to the NASA Hurricane and Severe Storm Sentinel (HS3) field campaign, the height of maximum warm anomalies in most cases occurs near 300 hPa (Komaromi and Doyle, 2017). In NOAA’s Sensing Hazards with Operational Unmanned Technology (SHOUT) field experiment, the warm core and convective structure evolve rapidly during the lifetime of Hurricane Matthew (Brown et al., 2017).

Satellite microwave observations are widely used for retrieving hurricane temperatures. Kidder et al. (2000) retrieved the atmospheric temperature profiles after removing the limb effect from the Advanced Microwave Sounding Unit (AMSU) on NOAA-15 satellite, and found a warm core centered at about 10.5 km. Also, a comprehensive link was created between tropical cyclone intensity and maximum temperature anomaly. Liu and Weng (2006) detected the warm cores of Hurricanes Katrina and Rita (2005) using the Special Sensor Microwave Imager Sounder (SSMIS) on the U.S. Defense Meteorological Satellite Program (DMSP) F16 satellite. The observed brightness temperatures with frequency of 54.4 GHz were taken as proxy data to objectively analyze the warm core of hurricanes. With the linear relationship between microwave radiance and atmospheric temperature and stable weighting function of sounding channels, Zhu and Weng (2013) developed a retrieval algorithm to translate the brightness temperature of the Advanced Temperature Microwave Sounder (ATMS) into the atmospheric temperature. The warm core of Hurricane Sandy was interpreted and it was with a low height and a large size. Based on that, the algorithm was modified to have a more homogeneous warm core structure and less-affected features from low troposphere precipitation (Tian and Zou, 2016). Liu and Weng (2006) used one dimensional variational (1DVAR) method to sequentially retrieve the profiles of atmospheric temperature, water vapor, and cloud water from AMSU measurement. With the AMSU-A measurement at frequencies less than 60 GHz, the profiles of temperature, water vapor, and cloud liquid water were derived firstly; then rain and ice water profiles were derived by adding high-frequency data at AMSU-B channels and finally all AMSU-A/B channels were used to refine the profiles. Besides atmospheric temperature, the Microwave Integrated Retrieval System (MIRS) can be utilized to retrieve the profiles of moisture, cloud liquid water, cloud ice water, total precipitable water (TPW), snow water equivalent (SWE), sea ice concentration (SIC), surface emissivity, and surface skin temperature over all surface types under all-weather conditions simultaneously (Boukabara et al., 2011). Theoreti-cally, it has the capability of applying to all the existing and future microwave sounding and imaging sensors.

This study focuses on the atmospheric temperature retrieval with the ATMS observations to derive the warm core structures of tropical cyclones, which will lay the firm foundation for intensity estimation of tropical cyclones. This paper is structured as follows. Section 2 will discuss the configuration of the Microwave Retrieval Testbed (MRT). Each component of the MRT will be described specifically, along with the ATMS instrumental characteristics and the 1DVAR mathematics basis. The idealized sensitivity of brightness temperature with respect to clouds and precipitation will be demonstrated in Section 3. The designed experiments in the MRT will be shown in Section 4, and a short introduction to Hurricane Irma will be provided in next section. Finally, the results and conclusions will be summarized in the last two sections.

2 Configuration of microwave retrieval testbed (MRT)

Satellite measurements at microwave frequencies play a crucial role in the earth observation system. The thermal radiation received by satellite instruments is emanated from various atmospheric layers and carries rich information regarding clouds and precipitation in addition to temperature and moisture. Through the simultaneous nadir overpass (SNO) measurements, ATMS data can be directly compared with AMSU-A data and a long-time series of satellite microwave sounding data can be established for climate applications (Weng and Zou, 2014; Zou et al., 2014a). In this study, we develop a retrieval testbed to derive a suite of atmospheric products from microwave data for aboard applications.

The Microwave Retrieval Testbed (MRT) is designed applicable for all microwave instruments with flexible modules as shown in Fig. 1. The inputs to MRT contains microwave observed brightness temperatures or radiances (e.g., ATMS or AMSU observations) and the background data (e.g., ECMWF-interim, GFS or GDAS data), which will serve as a-priori for the retrieval algorithms. For the retrieval algorithms, they can be the sequential 1DVAR (Liu and Weng, 2006) or simultaneous 1DVAR scheme (Boukabara et al., 2011) and the collections of the heritage microwave retrieval algorithms. Two commonly used fast radiative transfer models, the Community Radiative Transfer Model (CRTM: Weng, 2007) and the Radiative Transfer for TIROS Operational Vertical sounder (RTTOV: Saunder et al, 2007), can be alternatively used as part of forward operators. Since the CRTM and the RTTOV use the Advanced Adding and Doubling (Liu and Weng, 2006) and Delta-Eddington approximation (Bauer, 2006) as their radiative transfer solvers, respectively, they can be employed to test their sensitivities to the accuracy of the retrieval products. In both CRTM and RTTOV, the Fast Emissivity Model (FASTEM) is the key component to produce the emissivity over ocean, which is critical for ocean retrievals. If the priority is given to more accurate radiative transfer simulations, the line-by-line radiative transfer models such as MonoRTM (Clough et al., 2005) or Rosenkranz (Rosenkranz and Barnet, 2006) models are also optional in the MRT. Atmospheric profiles of temperature, moisture, and hydrometeors are retrieved as the standard outputs.

Figure 1 Microwave Retrieval Testbed (MRT) components and its major inputs and outputs.

As a first MRT experiment, the ATMS observations are used to retrieve the atmospheric temperature. Serving as a pathfinder of the Joint Polar Satellite System (JPSS), the first ATMS was carried onboard Suomi NPPsatellite which was launched successfully on 28 October 2011 (Weng et al., 2013). ATMS is cross-track scanning instrument with 22 sounding channels at frequencies ranging from 23 to 183 GHz. Table 1 gives the instrumental characteristics of the ATMS and shows a combined sounding capability of its ancestors (AMSU-A and MHS), thus providing higher spatial resolution and wider swath. Along each scan line, ATMS observes 96 field-of-view (FOV) samples within scan angle of ± 52.725° around the nadir direction. The ATMS channels 1–3, 16, and 17 (23.8, 31.4, 50.3, 88.2, and 165.5 GHz) are the atmospheric window channels, while the frequencies of channels 4–15 (50–60 GHz) and 18–22 (183.31 GHz) are located at oxygen absorption band and water vapor absorption band, respectively. The K/Ka bands have the same beam width of 5.2º, and the V and W bands have a beam width of 2.2º, whereas the G band beam width is 1.1º (JPSS ATMS ATBD, 2013). In general, ATMS can probe the thermal and moisture structure vertically from surface to 0.1 hPa under all-weather conditions. Since 2012, the ATMS antenna temperature data record (TDR) and sensor data record (SDR) has been released to user community for real-time applications after intensive calibration/validation (Cal/Val) and geolocation assessments (Weng et al., 2013; Kim et al., 2014; Zou et al., 2014b; Han et al., 2016; Weng and Yang, 2016; Yang and Weng, 2016). The data are also available for post-processing from the NOAA’s Comprehensive Large Array-Data Stewardship Systems (CLASS).

Table 1 The ATMS channel characteristics and their specifications of noise and calibration accuracy
CH Center frequency (MHz) Polorization Bandwidth
Freq. Stability
Accuracy (K)
max (K)
NEDT 3-db
bandwidth (Deg)
WF peak
1 23800 QV 270 10 1.0 0.3 0.7 5.2 Window
2 31400 QV 180 10 1.0 0.4 0.8 5.2 Window
3 50300 QH 180 10 0.75 0.4 0.9 2.2 Window
4 51760 QH 400 5 0.75 0.4 0.7 2.2 950
5 52800 QH 400 5 0.75 0.4 0.7 2.2 850
6 53596 ± 115 QH 170 5 0.75 0.4 0.7 2.2 700
7 54400 QH 400 5 0.75 0.4 0.7 2.2 400
8 54940 QH 400 10 0.75 0.4 0.7 2.2 250
9 55500 QH 330 10 0.75 0.4 0.7 2.2 200
10 57290.344[f0] QH 330 0.5 0.75 0.4 0.75 2.2 100
11 f0 ± 217 QH 78 0.5 0.75 0.4 1.2 2.2 50
12 f0 ± 322.2 ± 48 QH 36 1.2 0.75 0.4 1.2 2.2 25
13 f0 ± 322.2 ± 22 QH 16 1.6 0.75 0.4 1.5 2.2 10
14 f0 ± 322.2 ± 10 QH 8 0.5 0.75 0.4 2.4 2.2 5
15 f0 ± 322.2 ± 4.5 QH 3 0.5 0.75 0.4 3.6 2.2 2
16 88200 QV 2000 200 1.0 0.4 0.5 2.2 Window
17 165500 QH 3000 200 1.0 0.4 0.6 1.1 Window
18 183310 ± 7000 QH 2000 30 1.0 0.4 0.8 1.1 800
19 183310 ± 4500 QH 2000 30 1.0 0.4 0.8 1.1 700
20 183310 ± 3000 QH 1000 30 1.0 0.4 0.8 1.1 500
21 183310 ± 1800 QH 1000 30 1.0 0.4 0.8 1.1 400
22 183310 ± 1000 QH 500 30 1.0 0.4 0.9 1.1 300

Before ingesting the observations into the MRT, it is required to preprocess the ATMS TDR. To be specific, radiance processing consists of antenna pattern correction, footprint matching, and bias removal. Antenna pattern correction is a procedure to remove the effect of sidelobe and polarization cross-coupling. Footprint matching is a procedure to make the observations from two instruments into the same resolution sharing the same geolocation information and same viewing angle. For the beam width of ATMS channels 1 and 2, it is resampled to the equivalent beam width of 3.3º by using the Backus–Gilbert method. It will enhance the spatial resolution of the instrument and also increase the noise equivalent differential temperature (NEDT). The bias removal is to remove the inconsistencies between the ATMS observations and simulations from the forward model.

The forward model in this study is the CRTM developed at the US Joint Center for Satellite Data Assimilation (JCSDA). The CRTM is utilized to compute both the simulated ATMS radiances and the Jacobian with respect to the atmospheric state variables. The simultaneous 1DVAR algorithm is chosen to derive atmospheric temperature for its convenience. It focuses on minimizing the cost function J(X), which can be expressed as

$J({ X}) = \underbrace {\frac{1}{2}{{({ X} - {{ X}_0})}^{\rm{T}}} \times {{ B}^{ - 1}} \times ({ X} - {{ { X}}_0})}_{{J_{\rm b}}} + \underbrace {\frac{1}{2}{{({{ Y}^{\rm{m}}} - Y({ X}))}^{\rm{T}}} \times {{ E}^{ - 1}} \times ({{ Y}^{\rm{m}}} - Y({ X}))}_{J{\rm r}}, $ (1)

where X and X0 represent the retrieved and background state vectors (a set of geophysical parameters), and Y(X) and Ym are the forward operators using X and the measured vectors (radiances or brightness temperatures) as the input, respectively. B is the background error covariance matrix and E is the error covariance matrix for the forward calculation and/or instrument noise. The superscript T indicates a transpose of the matrix. The first term Jb in the right of Eq. (1) represents the penalty in departing from the background (a-priori) information, while the second term Jr indicates that from the measurements.

Assuming Y(X) is locally linear around X, the cost function can be minimized by

$\Delta {{ X}_{n + 1}} = [{({{ B}^{ - 1}} + K_n^{\rm{T}}{{ E}^{ - 1}}{K_n})^{ - 1}}K_n^{\rm{T}}{{ E}^{ - 1}}][{{ Y}^{\rm{m}}} - Y({{ X}_n}) + {K_n}\Delta {{ X}_n}],$ (2)

where the subscript n is the iteration index, $\Delta { X}$ is the departure from the background, and K is the Jacobian of Y(X) with respect to X. Equation (2) is preferred where the number of channels is relatively smaller. The iteration will start with the first guess, getting a new departure from the background at each iteration and be ended when the convergence achieves.

3 Sensitivity of brightness temperature to cloud and precipitation

In this sensitivity study, CRTM is applied to compute the optical depths and brightness temperatures at each of ATMS channels. Figure 2 shows the ATMS weighting functions and a schematic diagram of the cloud layer. Weighting functions of the ATMS channels 1–22 are calculated by using CRTM with the US standard atmosphe-ric profile. A similar cloud distribution is assumed as that in Liu and Curry (1993). Two cloud layers are simulated separately here: (1) a liquid phase cloud layer and (2) an ice cloud layer. The hydrometeor types correspond to two (rain and graupel) of six CRTM clouds, respectively. Liquid water cloud is specified at 700–800 hPa while the ice water cloud is located within 300–400 hPa. It should be noted that there are no mixed-phase layer and no liquid water droplets in the high-level cloud layer. The mean effective radius of liquid and ice cloud particles is set as 500 and 300 μm, respectively. Specifically, these cloud properties are combined with a special atmospheric profile with water vapor, temperature, and pressure of 100 layers, and the surface emissivity is generated from the FASTEM.

Figure 2 Weighting functions of the ATMS channels 1–22 calculated by using the US standard atmosphere profile. Two cloud layers are schematically shown for the sensitivity studies. A liquid phase cloud layer (cyan shaded) in case 1 is placed between 700 and 800 hPa while for case 2, an ice cloud layer (blue shaded) is located within 300–400 hPa.

For the atmospheric condition specified above, the brightness temperatures at different frequencies vary dramatically due to the effect of scattering and emission from clouds and precipitation. Figure 3 details the sensitivity of brightness temperature at ATMS channels to cloud water path for case one. At window channels 1, 2, 3, and 16, the brightness temperatures increase as the cloud liquid water increases. The initial low brightness temperature is a result of lower ocean emissivity. As cloud or rain water path increases, the cold oceanic scene in terms of brightness temperature is smeared off and the warmer brightness temperature arises from an increasing of the emission of cloud and raining liquid droplets. A similar situation exists in channels 4 and 5 in the oxygen absorption region. Meanwhile, the brightness temperatures of the rest of the oxygen sounding channels stay nearly constant, since the cloud layer locates below the weighting function peaks of the relevant channels. Figure 3bshows brightness temperature at channel 6 as an example. In Fig. 3c, brightness temperature at channel 18 decreases as the cloud liquid water increases, whereas those at channel 19 to 21 stay constant. Again, the weighting function peaks are all higher than the cloud top and thus they are not affected by cloud microphysics. For the ice cloud layer located between 300 and 400 hPa, the brightness temperatures at ATMS channels 1 and 2 as shown in Fig. 3d are not affected by ice clouds. For those oxygen sounding channels (see Figs. 3d, e), brightness temperatures are more affected by the ice cloud layer if their weighting function peaks below the cloud base (Figs. 3, 4). As shown in Fig. 3f, brightness temperatures at channels 18 to 21 decrease dramatically as the ice cloud path increases. The slope of decrease is larger for those lower heights of the weighting function.

Figure 3 Variations of brightness temperature of ATMS channels with respect to the (a–c) cloud liquid water path and (d–f) ice water path simulated over a water surface using CRTM. The effective radius is set to be 500 and 300 µm for liquid phase and ice cloud, respectively.

The sensitivity of brightness temperature at ATMS water vapor sounding channels to particle size is shown in Fig. 4. Here, the effective radii range from 100 to 400 μm. For liquid phase clouds, brightness temperatures at channels 18, 20, and 22 are shown in Figs. 4ac respectively and they monotonically decrease with cloud liquid water since these channels are less sensitive to surfaces and merely respond to the cloud scattering. Note that the variation for upper-level water sounding channel has much smaller magnitudes. However, for ice clouds (Figs. 4df), the larger the effective radius of the ice cloud particle, the stronger the effect of scattering on brightness temperature. For an ice water path ranging from 0 to 1.5 mm, brightness temperatures for a small size of 100 μm can decrease within a magnitude of 10–20 K but those for a large size of 500 μm can change within 100 K.

Figure 4 Variations of brightness temperature of all ATMS channels (a, d) 18, (b, e) 20, and (c, f) 22, with respect to the (a–c) cloud liquid water path and (d–f) ice water path. Different curves represent different effective radius (µm).

To understand the impacts of cloud vertical extension on ATMS sounding channels, we further simulate the responses of brightness temperature to ice cloud top. Here, the base of ice cloud is set as 500 hPa and the ice cloud top pressure ranges from 200 to 500 hPa. As shown in Fig. 5, the variation of brightness temperature with respect to the cloud top pressure is driven by the weighting function peak of a channel. For ATMS channel 18, which has a weighting function peaked at a lower level near 850 hPa, the brightness temperature is sensitive to ice cloud tops for all specified ice water paths.

Figure 5 Variations of brightness temperature of all ATMS channels (a) 18, (b) 20, and (c) 22, with respect to the cloud top pressure. The ice cloud base is fixed at 500 hPa. Cloud ice water path varies from 0.6 to 1.0 kg m–2.

Overall, brightness temperatures at water vapor sounding channels (183 GHz) are highly sensitive to cloud ice water path, particle ice, and cloud vertical extent. It is important to take into account of ice cloud physics when the high-frequency channels are included in the retrieval process.

4 Retrieval experiments in MRT

Currently, microwave retrieval testbed (MRT) is designed flexibly for all-weather remote sensing and has a capability of deriving a suite of atmospheric and surface parameters from microwave measurements. It is set up to retrieve as many parameters as possible from ATMS data. However, for hurricane conditions, MRT allows us for testing an optimal channel and deriving a best possible warm core anomaly. Three experiments are proposed in the study and are summarized in Table 2. Exp. 1 is proposed as our baseline approach. As shown earlier, ATMS brightness temperatures at channels 1 to 3 are all sensitive to surface emission and lower-level liquid phase clouds. Also, ATMS high frequencies at W/G bands are very sensitive to the clouds, especially for those peaking sounding channels. Exp. 1 excludes these channels. In Exp. 2, we use ATMS channels 1 to 15, which include surface sensitive channels at low frequencies; whereas Exp. 3 uses all the channels and is also the same as the setup in the operational products.

Table 2 Three experiments corresponding to uses of various ATMS channels
Exp # Channel usage Summary of purposes
1 4–15 Avoiding surface sensitive channels and high scattering effects at WG bands
2 1–15 Avoiding high scattering effects at WG bands
3 1–22 Including all processes

ATMS observations are collected during 2–6 September 2017 when Hurricane Irma was marching toward a landfall. In this study, all the channels are foot-print matched to the same spatial resolution. After running several steps of the MRT, ATMS data can be interpolated into the thermal structures of Hurricane Irma with the three different channel combinations above.

Hurricane Irma was one of the most catastrophic Saffir–Simpson scale Category 5 hurricane in the 2017 Atlantic hurricane season (Fig. 6). It was initially a tropical wave near the Cape Verde Islands on 26 August 2017 and developed as a tropical depression over the next day. During a series of eyewall replacement cycling, it fluctuated between Category 2 and 3. On 4 September, it rapidly intensified into a Category 5 hurricane under favorable conditions. At 0000 UTC 6 September, Irma reached its peak intensity with 185 mph (295 km h–1) winds and a minimum pressure of 914 hPa, making it the second most intense tropical cyclone in 2017. After the landfall on the Bahamian Island Little Inagua, its eyewall weakened back into a Category 4 hurricane shortly before 0600 UTC 8 September. But it regained Category 5 after 18 h and dropped as Category 3 after making landfall on Cuba. When it crossed from the warm water between Cube and Florida, the storm re-intensified from Category 3 to 4. Again making landfall on Cudjoe Key over one week, it finally dissipated off the coast of New England on 16 September.

Figure 6 The track of Hurricane Irma as reported from National Hurricane Center. The four selected centers of Hurricane Irma located at (18.50°N, 44.60°W), (18.00°N, 47.50°W), (17.10°N, 59.80°W), and (17.90°N, 62.60°W) and indicated by white circles (from right to left). The Saffir–Simpson scale of Hurricane Irma was Category 2, 3, 4, and 5 on 2, 3, 5, and 6 September 2017, respectively.

Figures 79 indicate horizontal distributions of the brightness temperature for ATMS channels 16–18 at 1616 UTC 2, 0448 UTC 3, 1704 UTC 5, and 0520 UTC 6 September 2017, respectively. Since brightness temperatures at these three channels are sensitive to scattering of cloud particles, they can be used to indicate the intensity of precipitation. In Fig. 7a, the brightness temperatures at channel 16 are both low near hurricane eye and in the surrounding areas. Since Irma is now a Category 2 storm, a widespread area of lower brightness temperatures in a range of 220–240 K indicates low lying clouds and clear ocean conditions. This is confirmed byFig. 8a at channel 17, which has a higher surface emissivity. As Irma evolves toward a more mature hurricane at Category 4 and 5, the lower brightness temperature at channel 16 remains near the hurricane eye whereas those at hurricane spiral rainfall bands evolve to warmer temperatures as shown in Figs. 7b, c, d. Since ATMS channel 17 is more sensitive to cloud scattering, brightness temperatures over oceans are more direct indicators of precipitating areas. The lower brightness temperature is a result of stronger precipitation. A cold brightness temperature distribution is surrounding the hurricane eyewall region at the ATMS channel 18. Brightness temperatures at other water vapor channels are also strongly correlated with the spiral rainfall bands (figure omitted).

Figure 7 Spatial distributions of brightness temperature for the ATMS channel 16 at (a) 1616 UTC 2 September, (b) 0448 UTC 3 September, (c) 1704 UTC 5 September, and (d) 0520 UTC 6 September, 2017. The center of Hurricane Irma is indicated by the cross symbol.
Figure 8 As in Fig. 7, but for the ATMS channel 17.
Figure 9 As in Fig. 7, but for the ATMS channel 18.

In order to run the MRT in summer 2017, a hurricane watch box is chosen within a latitudinal range of 5°N–40°N and a longitudinal range of 80°W–20°W. Therefore, the ATMS observations input into the MRT are fixed within the hurricane watch box. Four specific ATMS orbital data are selected, corresponding to Hurricane Irma when it is centered at (18.50°N, 44.60°W), (18.00°N, 47.50°W), (17.10°N, 59.80°W), and (17.90°N, 62.60°W) as shown inFig. 6.

In MRT, several modules such as bias correction, footprint matching, and 1DVAR are inherited from the MIRS (Liu and Weng, 2005, Boukabara et al., 2011). For a gi-ven geographical region within 20° × 20° centered at the eye of Hurricane Irma, the environmental mean temperature at each pressure level is computed by using the retrieved atmospheric temperatures under clear-sky conditions. The clear condition is defined by using a criterion where cloud LWP is less than 0.01 kg m2 (Weng et al., 2012). Eventually, the atmospheric temperature perturbation around the hurricane can be derived by subtracting the area mean from the retrieved temperature.

5 Analysis of MRT retrieval results

For ATMS, the along-track data for above specific times are directly synthesized through a hurricane cross-section of brightness temperature (Fig. 10). In doing so, ATMS data at channels 3–10 across the hurricane eye in the along-track direction is plotted against their respective position near the center. The pressure corresponding to the channel weighting function peak height is also labeled for a reference to the warm anomaly location. Figures 10ad display the evolution of hurricane from 2 to 6 September. Initially on 2 September, brightness temperature near the center is generally lower than those of the surrounding areas. At 0520 UTC 6 September when Irma reached Category 5, a warmer core brightness temperature is formed throughout the vertical column. At channel 8, which has a weighting peak at 250 hPa, the brightness temperature near the center is more than 10 K warmer than that outside of the storm. Thus, the brightness temperature cross-section from microwave sounding data is a very useful tool for monitoring the hurricane intensity.

Figure 10 Vertical cross-sections of the brightness temperature for the ATMS channels 3–10 along the black line in Fig. 6 at (a) 1616 UTC 2, (b) 0448 UTC 3, (c) 1704 UTC 5, and (d) 0520 UTC 6 September 2017. The cross symbols indicate the center of Hurricane Irma.

Figure 11 compares the vertical cross-sections of retrieved atmospheric temperature anomalies from 3 experiments on 2 September when Irma is a Category 2 storm. In Fig. 11a, an initial temperature anomaly appears at 200 hPa and extended to the near-surface. It seems to be a good reflection of a hurricane temperature in the initial stage. This structure is extremely similar to that of the previous study from Kidder et al. (2000). After adding the surface sensitive channels to Exp. 1, the warm core on the top is getting stronger but the cold anomaly (less than –10 K) seems unrealistic and this magnitude of cold anomaly has not been reported everywhere for Category 2 storm. When using all ATMS channels, the situation is very different: atmospheric temperature anomalies are positive and with a smaller magnitude. The thermal structure of Hurricane Irma at other times such as 0448 UTC 3 September and 1704 UTC 5 September 2017 are also obtained (figures omitted).

Figure 11 Vertical cross-sections of atmospheric temperature anomaly of Hurricane Irma along the black line in Fig. 6 at 1616 UTC 2 September 2017 from (a) Exp. 1, (b) Exp. 2, and (c) Exp. 3 retrieved from the MRT, respectively. The cross symbols locate the center of Hurricane Irma.

Figures 12ac compare the vertical cross-sections of retrieved atmospheric temperature anomalies from three experiments on 6 September when Irma is a Category 5 storm. It is seen that a warm core is as high as 18 K near 250 hPa when Hurricane Irma is in the mature stage. The structure and the magnitude are very similar above 200 hPa from all three experiments. The cold anomalies below 400 hPa in Fig. 12a might indicate cooling due to evaporation after precipitation falls below the freezing level; while for Exp. 2, only one small cold anomaly as low as about –12 K appears near the surface.

Figure 12 As in Fig. 11, but for the cross-section at 0520 UTC 6 September 2017.

When all the water vapor channels are included in the retrievals, warm temperature anomalies occur everywhere (Fig. 12c). The structure in Fig. 12c is unrealistic in most of the domain, except for the warm core region aloft. Note from our sensitivity studies that water vapor channels near 183 GHz are sensitive to atmospheric precipitation. Inaccurate forward models can lead to very large retrieval errors in temperature fields. It is shown that the magnitude of warm temperature anomaly from ATMS channels 4–15 is in overall agreement with the aircraft observations as in Hawkins and Imbembo (1976). While the magnitude of the cold anomaly may be a little high, the overall distribution seems also reasonable.

Hurricane Harvey is one of the strongest storms in record occurring in the Gulf region. It developed from a tropical wave on 17 August 2017 and rapidly intensified to be a Category 4 hurricane from 24 to 25 August after two landfalls. Maria is the Category 5 hurricane and the most intense tropical cyclone in 2017: the storm reached its peak intensity with a maximum sustainable wind of 280 km h–1 and a pressure of 908 hPa on 18 September. Figure 13 shows the vertical cross-sections of the two hurricanes at 0502 UTC 24 August and 1912 UTC 25 September 2017, respectively. From the cross-section of Harvey temperature anomaly retrieved in Exp. 3, the warm anomaly is located from 300 hPa to the surface, which is unrealistic when the storm is in a mature stage. However, the warm cores from Exps. 1 and 2 agree well with each other above 400 hPa. In Exp. 1, the warm anomaly is present but spreads across the hurricane center. Also, due to the strong scattering effect of rain bands, two cold anomalies are also present at the eyewall regions, which is similar to Hurricane Irma’s structure. For both Maria and Harvey, the warm anomalies near surface seem unrealistic and their reasons are under our further investigation.

Figure 13 Vertical cross-sections of atmospheric temperature anomaly of Hurricanes (a–c) Harvey at 1910 UTC 24 August and (e–f) Maria at 1725 UTC 25 September 2017, respectively. The left, middle, and right panels show the results from Exps. 1–3 retrieved from the MRT. The cross symbols on the bottom axis locate the center of the hurricanes.
6 Conclusions

A Microwave Retrieval Testbed (MRT) is developed for remote sensing of geophysical parameters from microwave observations. It can be used to explore an optimal channel combination to derive a realistic structure of atmospheric temperatures and other geophysical parameters. The sensitivity of different microwave channels to clouds and precipitation is fully investigated through radiative transfer simulations. Therefore, the idealized sensitivity experiment of brightness temperature with respect to clouds and precipitation was conducted in this study. It is shown that under the precipitation conditions, ATMS channels near 183 GHz is very sensitive to ice water clouds. When the ice cloud effective radius increases, the effect of scattering increases accordingly. Thus, uses of ATMS WG bands in retrievals will require highly accurate forward simulations.

MRT is configured with various selections of ATMS channels. By turning on or off the window channels or water vapor channels into the MRT, three experiments were conducted to produce different atmospheric temperatures during the evolution of tropical cyclones. Atmospheric temperature anomaly is computed from its retrievals subtracted by the environmental temperature around the hurricane. For Hurricane Irma during its life cycle, the hurricane has a warm core formed much larger in coverage and weaker in magnitude. As Irma intensified, the warm core becomes more compact and the magnitude is getting larger. The experiment with channels 4–15 produces a most reasonable and realistic anomaly distribution in all the hurricane phases and is consistent with the results of earlier regression-type approach (Kidder et al., 2000; Liu and Weng, 2005) and dropsonde observations (Hawkins and Imbembo, 1975). The rest of the ATMS channels are either sensitive to surface and/or water vapor, when they are included in the experiments, and can lead to unrealistic temperature structure if the forward operators have large biases in precipitation conditions. The uncertainty in radiative transfer models and lack of information (e.g., surface temperature and wind speed) as input to the model can all lead to large errors in temperature and water fields.

Since 1DVAR is a core module in MRT, background, first guess, and error covariances in both observation and forward models are all playing important roles for the final retrievals. Further investigations are ongoing and include (1) characterization of error covariances using Global Precipitation Measurement (GPM) and the Constellation Observing System for Meteorology, Ionosphere, and Climate (COSMIC) profiles as the first guess to the MRT, (2) understanding the impacts from updated scattering tables in latest CRTM and RTTOV, and (3) using regression-based retrievals as the first guess.

Bauer, P., E. Moreau, F. Chevallier, et al., 2006: Multiple-scattering microwave radiative transfer for data assimilation applications. Quart. J. Roy. Meteor. Soc., 132, 1259–1281. DOI:10.1256/qj.05.153
Boukabara, S. A., K. Garrett, W. C. Chen, et al., 2011: MiRS: An all-weather 1DVAR satellite data assimilation and retrieval system. IEEE Trans. Geosci. Remote Sens., 49, 3249–3272. DOI:10.1109/tgrs.2011.2158438
Brown, S., B. Lambrigtsen, B. Lim, et al, 2017: Demonstrating the impact of rapid repeat passive microwave observations from the global hawk: Implications for future smallsat or GEO missions. Proceedings of 2017 IEEE International Geoscience and Remote Sensing Symposium, IEEE, Fort Worth, TX, USA, 5938–5941, doi: 10.1109/IGARSS.2017.8128361.
Chen, H., and D. L. Zhang, 2013: On the rapid intensification of Hurricane Wilma (2005). Part II: Convective bursts and the upper-level warm core. J. Atmos. Sci., 70, 146–162. DOI:10.1175/JAS-D-12-062.1
Clough, S. A., M. W. Shephard, E. J. Mlawer, et al., 2005: Atmospheric radiative transfer modeling: A summary of the AER codes. J. Quantit. Spectros. Radia. Trans., 91, 233–244. DOI:10.1016/j.jqsrt.2004.05.058
Dolling, K., and G. M. Barnes, 2014: The evolution of Hurricane Humberto (2001). J. Atmos. Sci., 71, 1276–1291. DOI:10.1175/JAS-D-13-0164.1
Gray, W. M., 1968: Global view of the origin of tropical disturbances and storms. Mon. Wea. Rev., 96, 669–700. DOI:10.1175/1520-0493(1968)096<0669:GVOTOO>2.0.CO;2
Halverson, J. B., J. Simpson, G. Heymsfield, et al., 2006: Warm core structure of Hurricane Erin diagnosed from high altitude dropsondes during CAMEX-4. J. Atmos. Sci., 63, 309–324. DOI:10.1175/JAS3596.1
Han, Y., F. Z. Weng, X. L. Zou, et al., 2016: Characterization of geolocation accuracy of Suomi NPP Advanced Technology Microwave Sounder measurements. J. Geophys. Res. Atmos., 121, 4933–4950. DOI:10.1002/2015JD024278
Hawkins, H. F., and D. T. Rubsam, 1968: Hurricane Hilda, 1964. Mon. Wea. Rev, 96, 617–636. DOI:10.1175/1520-0493(1968)096<0701:HH>2.0.CO;2
Hawkins, H. F., and S. M. Imbembo, 1976: The structure of a small, intense hurricane—Inez 1966. Mon. Wea. Rev., 104, 418–442. DOI:10.1175/1520-0493(1976)104<0418:TSOASI>2.0.CO;2
JPSS ATMS ATBD, 2013: Joint Polar Satellite System (JPSS) Advanced Technology Microwave Sounder (ATMS) SDR Calibration Algorithm Theoretical Basic Document (ATBD). [Available online at https://www.star.nesdis.noaa.gov/smcd/spb/nsun/snpp/ATMS/ATMS-SDR-ATBD-2013-final.pdf]. Accessed online on 5 September 2018.
Kidder, S. Q., M. D. Goldberg, R. M. Zehr, et al., 2000: Satellite analysis of tropical cyclones using the Advanced Microwave Sounding Unit (AMSU). Bull. Amer. Meteor. Soc., 81, 1241–1259. DOI:10.1175/1520-0477(2000)081<1241:SAOTCU>2.3.CO;2
Kim, E., C.-H. J. Lyu, K. Anderson, et al., 2014: S-NPP ATMS instrument prelaunch and on-orbit performance evaluation. J. Geophys. Res. Atmos., 119, 5653–5670. DOI:10.1002/2013JD020483
Komaromi, W. A., and J. D. Doyle, 2017: Tropical cyclone outflow and warm core structure as revealed by HS3 dropsonde data. Mon. Wea. Rev., 145, 1339–1359. DOI:10.1175/MWR-D-16-0172.1
La Seur, N. E., and H. F. Hawkins, 1963: An analysis of Hurricane Cleo (1958) based on data from research reconnaissance aircraft. Mon. Wea. Rev., 91, 694–709. DOI:10.1175/1520-0493(1963)091<0694:AAOHCB>2.3.CO;2
Liu, G. S., and J. A. Curry, 1993: Determination of characteristic features of cloud liquid water from satellite microwave measurements. J. Geophys. Res. Atmos., 98, 5069–5092. DOI:10.1029/92JD02888
Liu, Q. H., and F. Z. Weng, 2005: One-dimensional variational retrieval algorithm of temperature, water vapor, and cloud water profiles from Advanced Microwave Sounding Unit (AMSU). IEEE Trans. Geosci. Remote Sens., 43, 1087–1095. DOI:10.1109/TGRS.2004.843211
Liu, Q. H., and F. Z. Weng, 2006: Radiance assimilation in studying Hurricane Katrina. Geophys. Res. Lett., 33, L22811. DOI:10.1029/2006GL027543
Liu, Y. B., D. L. Zhang, and M. K. Yau, 1999: A multiscale numerical study of Hurricane Andrew (1992). Part II: Kinematics and inner-core structures. Mon. Wea. Rev., 127, 2597–2616. DOI:10.1175/1520-0493(1999)127<2597:AMNSOH>2.0.CO;2
Rosenkranz, P. W., and C. D. Barnet, 2006: Microwave radiative transfer model validation. J. Geophys. Res. Atmos., 111, D09S07. DOI:10.1029/2005JD006008
Saunders, R., P. Rayer, P. Brunel, et al., 2007: A comparison of radiative transfer models for simulating Atmospheric Infrared Sounder (AIRS) radiances. J. Geophys. Res. Atmos., 112, D01S90. DOI:10.1029/2006JD007088
Stern, D. P., and F. Q. Zhang, 2016: The warm-core structure of Hurricane Earl (2010). J. Atmos. Sci., 73, 3305–3328. DOI:10.1175/JAS-D-15-0328.1
Tian, X. X., and X. L. Zou, 2016: ATMS- and AMSU-A-derived hurricane warm core structures using a modified retrieval algorithm. J. Geophys. Res. Atmos., 121, 12630–12646. DOI:10.1002/2016JD025042
Vigh, J. L., and W. H. Schubert, 2009: Rapid development of the tropical cyclone warm core. J. Atmos. Sci., 66, 3335–3350. DOI:10.1175/2009JAS3092.1
Wang, Z., M. T. Montgomery, and T. J. Dunkerton, 2010: Genesis of pre-Hurricane Felix (2007). Part II: Warm core formation, precipitation evolution, and predictability. J. Atmos. Sci., 67, 1730–1744. DOI:10.1175/2010JAS3435.1
Weng, F. Z., 2007: Advances in radiative transfer modeling in support of satellite data assimilation. J. Atmos. Sci., 64, 3799–3807. DOI:10.1175/2007JAS2112.1
Weng, F. Z., and H. Yang, 2016: Validation of ATMS calibration accuracy using Suomi NPP pitch maneuver observations. Remote Sens., 8, 332. DOI:10.3390/rs8040332
Weng, F. Z., and X. L. Zou, 2014: 30-year atmospheric temperature record derived by one-dimensional variational data assimilation of MSU/AMSU-A observations. Climate Dyn., 43, 1857–1870. DOI:10.1007/s00382-013-2012-5
Weng, F., X. Zou, X. Wang, et al., 2012: Introduction to Suomi national polar-orbiting partnership advanced technology microwave sounder for numerical weather prediction and tropi-cal cyclone applications. J. Geophys. Res. Atmos., 117, D19112. DOI:10.1029/2012JD018144
Weng, F. Z., X. L. Zou, N. H. Sun, et al., 2013: Calibration of Suomi national polar-orbiting partnership advanced technology microwave sounder. J. Geophys. Res. Atmos., 118, 11,187–11,200. DOI:10.1002/jgrd.50840
Yang, H., and F. Z. Weng, 2016: Corrections for on-orbit ATMS lunar contamination. IEEE Trans. Geosci. Remote Sens., 54, 1918–1924. DOI:10.1109/TGRS.2015.2490198
Zhu, T., and F. Z. Weng, 2013: Hurricane Sandy warm-core structure observed from Advanced Technology Microwave Sounder. Geophys. Res. Lett., 40, 3325–3330. DOI:10.1002/grl.50626
Zou, X. L., F. Z. Weng, and H. Yang, 2014a: Connecting the time series of microwave sounding observations from AMSU to ATMS for long-term monitoring of climate. J. Atmos. Oceanic Technol., 31, 2206–2222. DOI:10.1175/JTECH-D-13-00232.1
Zou, X. L., L. Lin, and F. Z. Weng, 2014b: Absolute calibration of ATMS upper level temperature sounding channels using GPS RO observations. IEEE Trans. Geosci. Remote Sens., 52, 1397–1406. DOI:10.1109/TGRS.2013.2250981