![Scatterplots of ATMS TB observations (limb corrected) on 28 September 2016 within [10˚N-20˚N] between (a) channels 5 and 7, (b) channels 6 and 8, and (c) channels 7 and 8. The colors show LWP values. The magenta lines are linear regression best-fit lines for clear-sky data (black dots, LWP < 0.03 g kg-1 ).](https://www.researchgate.net/profile/Xiaoxu-Tian-3/publication/327340702/figure/fig6/AS:666954532679681@1536025760620/Scatterplots-of-ATMS-TB-observations-limb-corrected-on-28-September-2016-within_Q320.jpg)
Content uploaded by Xiaoxu Tian
Author content
All content in this area was uploaded by Xiaoxu Tian on Oct 03, 2018
Content may be subject to copyright.
Hurricane Warm-Core Retrievals from AMSU-A
and Remapped ATMS Measurements
with Rain Contamination Eliminated
Xiaolei Zou
1
and Xiaoxu Tian
1
1
Earth Science System Interdisciplinary Center, University of Maryland, College Park, MD, USA
Abstract Due to a shorter effective integration time for each field of view of the Advanced Microwave
Temperature Sounder (ATMS) onboard the Suomi National Polar-orbiting Partnership (S-NPP) satellite
than that for the Advanced Microwave Sounding Unit-A (AMSU-A) onboard previous National Oceanic
and Atmospheric Administration (NOAA) polar-orbiting satellites NOAA-15 to NOAA-19, ATMS
temperature-sounding channels have higher observational resolutions and larger noise equivalent
differential temperatures than the corresponding AMSU-A channels. The high resolution of the ATMS allows
hurricane rainband features that are not resolvable by AMSU-A to be captured. But the larger noise
equivalent differential temperature of ATMS weakens this capability through the significant impact of
observational noise on warm-core retrievals. In this study, a remapping algorithm is applied to obtain
AMSU-A-like ATMS fields of view to suppress this noise. A modified warm-core retrieval algorithm, which
consists of two sets of training coefficients for clear-sky and cloudy conditions, is applied to limb-corrected
ATMS and AMSU-A measurements using collocated Global Positioning System radio occultation observations
in the previous month of the targeted hurricanes as training data sets. ATMS channels 5, 6, and 7 (AMSU-A
channels 4, 5, and 6) are excluded when training the coefficients for cloudy conditions to avoid cloud/rain
contamination. As a result, the abnormal cold core in the low and middle troposphere and the banded warm
structures in phase with rainbands are both successfully removed. The warm-core evolution of Hurricane
Matthew (2016) during its entire life span is temporally consistent on intensity as obtained from NOAA-15,
NOAA-18, and MetOp-B AMSU-A observations and S-NPP ATMS observations.
1. Introduction
Hurricane centers are warmer than their environments by more than several degrees and are often called
warm-core cyclones. An adiabatic warming of the atmosphere with a downward motion at the eye and the
associated latent heat release, and precipitation near and outside of the eye wall all contribute to the forma-
tion of hurricane warm-cores. A subsidence warming induced by convective bursts during the onset of rapid
intensification also contributes to the formation and/or intensification of warm-cores in the upper troposphere
(Chen & Zhang, 2013). La Seur and Hawkins (1963), Hawking and Rubsam (1968), and Hawkins and Imbembo
(1976) examined reconnaissance aircraft data and found that warm-cores were located around 250 hPa in the
temperature field, which was reconstructed using aircraft reconnaissance data at the 500- and 180-hPa pres-
sure levels. Halverson et al. (2006) used dropsonde observations to find that the primary warm-core of
Hurricane Erin (2001) was near 500 hPa. Durden (2013) generated composites of in situ observations (buoys,
measuring instruments mounted on ships and aircraft, dropsondes, and weather stations on islands) and con-
firmed that warm-core heights vary between 760 and 250 hPa. Through idealized model simulations, Stern
and Nolan (2012) concluded that primary warm-cores having maximum warm anomalies are located within
the 4–8 km (~700–150 hPa) vertical layer of the atmosphere. Although the exact altitudes of the maximum
centers of warm-cores vary from case to case, their intensities (the warm-core anomalies) are positively corre-
lated to hurricane intensities (Komaromi & Doyle, 2017). An early detection of warm-core formation could also
provide insights on the formation of hurricanes from tropical depressions (Dolling & Barnes, 2012).
In comparison with conventional in situ observations, satellite remote sensing observations are advanta-
geous in their horizontal, vertical, and temporal coverage over vast oceans. Of special interest to hurricanes
are microwave instruments onboard polar-orbiting operational environmental satellites that monitor the glo-
bal atmosphere twice daily. Advanced Microwave Sounding Unit-A (AMSU-A) temperature sounders have
been onboard U.S. National Oceanic and Atmospheric Administration (NOAA) polar-orbiting operational
ZOU AND TIAN 1
Journal of Geophysical Research: Atmospheres
RESEARCH ARTICLE
10.1029/2018JD028934
Key Points:
•A remapping algorithm is applied to
obtain AMSU-A-like ATMS fields of
view
•A rainband-like artificial warm-core
structure is caused by rain
contamination in ATMS channels 5–7
•Impacts of rain contamination on the
warm-core retrieval results in the low
troposphere and rainy areas are
removed
Correspondence to:
X. Zou,
xzou1@umd.edu
Citation:
Zou, X., & Tian, X. (2018). Hurricane
warm-core retrievals from AMSU-A and
remapped ATMS measurements with
rain contamination eliminated. Journal
of Geophysical Research: Atmospheres,
123. https://doi.org/10.1029/
2018JD028934
Received 2 MAY 2018
Accepted 17 AUG 2018
Accepted article online 30 AUG 2018
©2018. The Authors.
This is an open access article under the
terms of the Creative Commons
Attribution-NonCommercial-NoDerivs
License, which permits use and distri-
bution in any medium, provided the
original work is properly cited, the use is
non-commercial and no modifications
or adaptations are made.
environmental satellites NOAA-15 through NOAA-19 and the European Organization for the Exploitation of
Meteorological Satellites MetOp-A/-B since 1998. Merrill (1995) and Brueske and Velden (2003) developed
and tested an algorithm of retrieving tropical cyclone (TC) intensities from observations at a single AMSU
channel. A regression algorithm to retrieve the warm-core of Hurricane Bonnie (1998) from AMSU-A
temperature-sounding channels was developed and demonstrated by Zhu et al. (2002). Demuth et al.
(2004), Knaff et al. (2004), and Demuth et al. (2006) used AMSU-A measurements to estimate the maximum
sustained wind, minimum sea level pressure, and radii of winds of tropical cyclones. As a successor and com-
bination of the AMSU-A and the microwave humidity sounder, the Advanced Technology Microwave
Sounder (ATMS) onboard the Suomi National Polar-orbiting Partnership (S-NPP) satellite was successfully
launched into a Sun-synchronous orbit on 28 October 2011 (Weng et al., 2012). The algorithm developed
by Zhu et al. (2002) was applied to ATMS observations of Hurricane Sandy (2012) and a positive correlation
between the maximum upper level warm anomaly and the maximum sustained winds was found (Zhu
and Weng (2013). Due to scan biases found in the temperature retrievals obtained by the algorithm of Zhu
et al. (2002) and Zhu and Weng (2013), Tian and Zou (2016) proposed four modifications to the temperature
retrieval algorithm: (i) replace the simple scan correction term with regression coefficients that are explicit func-
tions of scan position; (ii) use only a subset of ATMS temperature-sounding channels for retrieving the tempera-
tures at certain pressure levels that are highly correlated (correlation coefficients greater than 0.5) to these
channels; (iii) obtain regression coefficients using limb-corrected ATMS and collocated Global Positioning
System (GPS) radio occultation (RO) observations in the month prior to the lifetime of the hurricane as training
data sets; and (iv) obtain two sets of regression coefficients, one for clear-sky conditions and another for
cloudy-sky conditions. The refined algorithm was applied to both AMSU-A and ATMS observations to demon-
strate that the warm-core structures for HurricanesMichael (2012) and Sandy (2012)during their lifetimes were
more realistic than those obtained by both the unmodified algorithm and the even more advanced one-
dimensional variational approach of the Microwave Integrated Retrieval System. Tian and Zou (2018) applied
the temperature retrieval algorithm in Tian and Zou (2016) to microwave temperature sounders onboard four
NOAA operational satellites to analyze the intensity variations in Hurricanes Harvey, Irma, and Maria (2017).
Although less contaminated by clouds and precipitation than infrared measurements (Chen et al., 2013;
Hilton et al., 2012; Janssen, 1994; Le Marshall et al., 2006), microwave measurements for low- and middle-level
temperature-sounding channels are still affected by optically thick and heavily precipitating clouds in hurri-
canes (Weng et al., 2003). As a result, in most TC cases populated with convection, only warm-cores in the
upper troposphere can be derived from satellite microwave measurements. Cold temperature anomalies
with extremely unrealistic magnitudes have appeared in all past warm-core retrieval results. Nonetheless,
when the TC eye is sufficiently large such that significant portions of the eyewall are not in the FOV or the
convection absent near the center, the warm-core can be resolved well into the middle troposphere chan-
nels. In this study, a remapping algorithm is introduced to reduce the random noise in ATMS observations
and to make the fields of view (FOVs) the same size as AMSU-A FOVs (section 2). The regression coefficients
for both clear-sky and cloudy conditions are obtained from GPS RO data and used as training data during the
two-week period prior to the occurrence of a hurricane. After eliminating temperature-sounding channels
with peak weighting functions (WFs) below 400 hPa in cloudy conditions, retrievals of warm-core structures
are then made (section 3). The warm-core structures and their temporal evolutions during the lifetime of
Hurricane Matthew (2016), the first Category 5 Atlantic hurricane since Hurricane Felix (2007), are retrieved
from all available AMSU-A and ATMS observations and compared (section 4). Finally, a summary and conclu-
sions are given in section 5.
2. Descriptions of ATMS Data and the Remapping Algorithm
2.1. ATMS and AMSU-A Data Characteristics
The ATMS and the AMSU-A provide measurements of atmospheric thermal emission in the microwave spec-
tral region between 23 and 90 GHz. The ATMS is onboard the S-NPP satellite and the AMSU-A is onboard the
NOAA-15, -18, -19, MetOp-A, and MetOp-B satellites. The ascending node of the S-NPP ATMS crosses the
equator at 1:30 pm local time, that is, the local equator crossing time. For atmospheric temperature profile
retrievals, ATMS channels 5–15 or AMSU-A channels 4–14 are used for hurricane warm-core retrievals. The
peak WFs and center frequencies of these channels from the lowest to the highest channel number change
from 850 hPa and 52.8 GHz to 2 hPa and 57.29 GHz, respectively. The antenna for ATMS channels 5–15 has a
10.1029/2018JD028934
Journal of Geophysical Research: Atmospheres
ZOU AND TIAN 2
beam width of 2.2°, while that for AMSU-A channels 4–14 has a beam
width of 3.3°. For the ATMS, the temperature-sounding channels (1–16)
are kept in the same scan mode as the humidity-sounding channels
(17–22) and with the same sampling time, a continuous-scan fashion is
adopted. Therefore, 96 scene resolution cells are sampled at an interval
of 8/3 s for each scan to cover 52.725° scan angles on both sides of the
subsatellite path. The integration time for each FOV is 18 ms. By contrast,
the AMSU-A scans the Earth scene within ±49.5° with respect to the
nadir direction and has a total of 30 FOVs. There are significant overlaps
among neighboring ATMS FOVs in both along-track and across-track
directions but not for the AMSU-A. The swath width of the ATMS scan is
2,600 km, which is wider than its predecessor AMSU-A (2,343 km), and
leaves almost no gaps near the equator, providing nearly full coverage
of the low latitudes where hurricanes develop and intensify over the
Atlantic and Pacific Oceans.
Random noise in measurements made by microwave temperature soun-
ders AMSU-A and ATMS is estimated by standard deviations (Mo, 1996)
and/or Allan deviations calculated from warm target measurements
(Tian et al., 2015). These are known as the noise equivalent differential
temperatures (NEDTs). The NEDTs of the ATMS channels are slightly higher
than those of the corresponding AMSU-A channels because of the higher
sampling rate of the ATMS, that is, a shorter effective integration time of
each ATMS FOV (Weng et al., 2012). More detailed information of ATMS
instrument characteristics is provided in Table 1.
2.2. A Remapping Algorithm and Results
A remapping algorithm developed by Atkinson (2011) is employed to con-
vert the beam width of ATMS channels 5–15 to that of AMSU-A channels
4–14. The oversampling characteristics of ATMS observations make such
Table 1
ATMS Channel Characteristics
Channel no. Frequency (GHz) Specification (K) NEΔT (K) Beam width (deg) Peak WF (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
Figure 1. (a) Sizes of FOVs 40–56 for ATMS channels 5–15 over five consecu-
tive scanlines (black circles) and the AMSU-A-like ATMS FOVs along the third
scanline (nadir position at 15.25°N, 60.2°W) of the above five scanlines after
remapping (red and green circles in turn for clarity) observed on 28
September 2016. (b) The 96 ATMS FOVs after remapping over the single
scanline in (a) and the AMSU-A FOVs along the three consecutive scanlines
from NOAA-18 (blue circles) that are near the ATMS scanline, overlapped
onto GOES-13 imager channel 4 (10.7 μm) TB observations (unit K) at 1715
UTC 28 September 2016.
10.1029/2018JD028934
Journal of Geophysical Research: Atmospheres
ZOU AND TIAN 3
a remap meaningful to produce AMSU-A-like ATMS observations. Having a
consistent beam width between ATMS and AMSU-A is highly desirable
because it allows the ATMS derived warm-cores be linked to AMSU-A
derived ones that is required for several studies, such as examining the
diurnal change of hurricane warm-cores and climate change (Yang &
Zou, 2014).
Assuming the ATMS has a Gaussian beam, a modulation transfer function
(MTF), defined as the spatial frequency response function of the 3-dB beam
width (w), can be expressed as (Atkinson, 2011)
MTFwfðÞ¼exp πfw=2ðÞ
2
ln2
!
;(1)
where fis the spatial frequency and the reciprocal of the ATMS sampling
distance (1.1°). The beam shape, called the point spread function, is the
Fourier transform of MTF and can be written as
PSFwxðÞ¼exp x
w=2
2
ln2
!
:(2)
The MTF of both the ATMS intrinsic beam width (w= 2.2°) and the tar-
geted AMSU-A beam width (w= 3.3°) are first computed. The two-
dimensional brightness temperatures (BTs) at any given ATMS
temperature-sounding channel in the spatial domain [TATMS
bx;yðÞ], where
xand yrepresent the along-track and cross-track directions, respectively,
are converted to the frequency domain [F(ω
x
,ω
x
)] (Sorensen et al.,
1987). The function F(ω
x
,ω
x
) is then multiplied by the MTF modification
factor ( MTFAMSUA
3:3o=MTFATMS
2:2o) to obtain the MTF for the ATMS at the
AMSU-A-like beam width:
MTFAMSUAlike ATMS
3:3o¼Fωx;ωy
MTFAMSUA
3:3o=MTFATMS
2:2o:
Finally, MTFAMSUAlike ATMS
3:3ois converted back to the spatial domain to
obtain AMSU-A-like ATMS BTs [TAMSUAlike ATMS
bx;yðÞ] through the Fourier
transform.
Figure 1a provides an illustration of the sizes of FOVs 40–56 for ATMS chan-
nels 5–15 over five consecutive scan lines and the AMSU-A-like ATMS FOVs
along the third scan line of the above five scan lines after remapping
observed on 28 September 2016. Note that the remapped FOVs are con-
centric to the original FOVs with larger sizes. Although the sizes of the
AMSU-A-like ATMS FOVs generated by remapping are the same as those
of the AMSU-A, the feature of the continuous-ATMS-scan fashion is kept,
with 96 scene resolution cells covering 52.725° on both sides of the subsa-
tellite path. In other words, ATMS observations still have much higher (~3
times) spatial samplings than AMSU-A observations after remapping. This
is more clearly demonstrated in Figure 1b, which shows the ATMS FOVs
after remapping over the single scan line seen in Figure 1a overlapped
onto the AMSU-A FOVs along the three consecutive scan lines from
NOAA-18 that are near the ATMS scan line. BT observations of channel 4
(10.7 μm) from the GOES-13 imager at 1715 coordinated universal time
(UTC) 28 September 2016 are provided in Figure 1b to show the typical
scales of clouds in a hurricane environment. The FOV sizes at large scan
Figure 2. (a) ATMS channels 7 (red), 8 (black), and 9 (blue) TB observations
along the same scanline as in Figure 1 on 28 September 2016. (b) Same as
a except for after remapping.
Figure 3. Limb-corrected values of ATMS channel 8 [TCh8
bi;jðÞTCh8
b;LC i;jðÞ] TBs
(black contour lines at 2-K intervals) as a function of scan position and lati-
tude at 2° intervals, as well as scan variations in the global mean TBs (TmðÞ
biðÞ)
of channels 5–12 before (solid colored curves) and after (dashed colored
curves) the limb correction. ATMS observations are from January and
September of 2016.
10.1029/2018JD028934
Journal of Geophysical Research: Atmospheres
ZOU AND TIAN 4
angles are much larger than those near nadir, especially in the cross-track direction. In other words, small-
scale features of hurricanes, such as the eye, rainbands, and clear streaks between rainbands, are better
resolved near the swath center than around swath edges. This point will be illustrated later (see section 4).
The remapping done to obtain AMSU-like ATMS observations involves a Gaussian weighted average of origi-
nal ATMS observations in approximately nine ATMS FOVs. Therefore, the remapping will reduce the random
noise of the observations. Figure 2 shows the cross-track variation in ATMS BT observations of channels 7, 8,
and 9 along the same scan line shown in Figure 1 with its nadir position located at [15.25°N, 60.2°W] on 28
September 2016 before (Figure 2a) and after (Figure 2b) the remapping. As expected, the BT variations
become smoother after the remapping for all three channels shown. Similar results are obtained for other
channels and elsewhere (figures omitted). In fact, the noise reduction due to the remapping has significant
impacts on warm-core retrieval results which will be shown later (see Figure 10 in section 4).
2.3. A Brief Description of Limb Correction and Rain Contamination
As a cross-track scanning radiometer, the ATMS receives atmospheric emissions from a larger scan angle to
go over a longer optical path. Weather features could often be concealed in BTs by the prevailing scan
Figure 4. ATMS channels 5–7 TB observations (unit K) (left panels) before and (right panels) after the limb correction at
1716 UTC 28 September 2016. The black dotted lines show the nadir positions. (g) VIIRS infrared TB observations (10.7
um) at 1716 UTC 28 September 2016.
10.1029/2018JD028934
Journal of Geophysical Research: Atmospheres
ZOU AND TIAN 5
pattern. Goldberg et al. (2001) and Zhang et al. (2017) proposed a limb
effect correction algorithm to remove the scan-dependent patterns in
AMSU-A and ATMS BTs. The ATMS BTs after the limb correction (TkðÞ
b;LC )
are obtained as follows:
TkðÞ
b;LC i;j
ðÞ
¼bkðÞ
iþ∑
kþ1
m¼k1
amðÞ
iTmðÞ
bi;j
ðÞ
TmðÞ
bi
ðÞ
;(3)
where “i”,“j”, and “k”represent the scan position, scan line, and channel
number, respectively; TkðÞ
bi;jðÞis the BT of the k
th
channel at the i
th
scan
position and the j
th
scan line; TmðÞ
biðÞis the global mean of BTs as a function
of scan position i; and (ak1ðÞ
i,akðÞ
i, and akþ1ðÞ
i) are the regression coefficients.
The regression coefficients are obtained by minimizing the following cost
function:
Ja
kðÞ
i;bi
¼∑
φ
TkðÞ
b;reg i;φðÞTkðÞ
b;nadir φðÞ
2
;(4)
where φis the latitude and Tk
ðÞ
b;reg i;φðÞis the regression function defined as
Tk
ðÞ
b;reg i;φðÞ¼biþ∑
kþ1
m¼k1
am
ðÞ
iTm
ðÞ
bi;φðÞTm
ðÞ
biðÞ
hi
:(5)
The quantitiesTkðÞ
b;nadir φðÞand TmðÞ
bi;φðÞare the mean BTs in 2° latitudinal
bands at the nadir (48th and 49th FOVs) and for different FOVs, respec-
tively (Wark, 1993). For channels 5, 6, and 7, only observations over oceans
are included in the calculations.
Global ATMS BT observations made in January and September of 2016 are
used. The BTs at all scan angles are corrected with respect to the nadir
position where the limb correction is zero. An example of the limb correc-
tion is provided in Figure 3 for ATMS channel 8, in which the differences in
BT before [TCh8
bi;jðÞ] and after [TCh8
b;LC i;jðÞ] the correction are shown. The lar-
gest limb correction is at the largest scan angle near the equator (14 K)
and reduces to zero near the nadir (less than 0.2 K in magnitude for about
15 FOVs centered at the nadir). In Figure 3, we also show the scan varia-
tions in the global mean BTs of channels 5–12 before and after the limb correction. The scan variations of
ATMS channels 5–9 (channels 10–12) are of a downward curving (“upward curving”) shape as a result of a
decrease (increase) in atmospheric temperature with increasing altitude in the troposphere (stratosphere).
After the limb correction, the scan dependence of global mean BT observations for all temperature-sounding
channels (5–12) employed in ATMS warm-core retrievals is successfully removed.
Hurricane Matthew around 1716 UTC 28 September 2016 is characterized by colder BTs (Figures 4b, 4d,
and 4f) over cloudy regions based on Visible Infrared Imaging Radiometer Suite (VIIRS) BT observations
at 10.7 μm (Figure 4g). A closer match between ATMS and VIIRS observations at small scales is not
expected due to the much higher observational resolution of the VIIRS. The nadir FOV sizes of the
ATMS and AMSU-A temperature-sounding channels (31.6 and
48.6 km, respectively) are much coarser than that of the VIIRS infrared
channel at 10.7 μm (375 m).
3. ATMS Temperature Retrieval Algorithm for
Hurricane Matthew
3.1. Case Description
Hurricane Matthew (2016) was the first Atlantic hurricane that reached
category 5 intensity since Hurricane Felix in 2007. Figure 5 provides the
Figure 5. (a) The best track of Hurricane Matthew (2016) with the intensities
marked in different colored shapes from 0000 UTC 28 September to 1800
UTC 9 October 2016 at 6-h intervals. (b) The maximum surface wind (black)
and radii of 34- (blue), 50- (green), and 64-kt (red) winds from 28 September
to 9 October 2016. The background colors show the intensity categories
as defined in (a). TD, TS, and H1–H5 stand for tropical depression, tropical
storm, and hurricane categories 1–5.
Table 2
Data Count of RO Profiles Collocated With ATMS Observations From 1–27
September 2016
Satellite name All sky Clear sky
COSMIC 5,077 2,912
Metop-A 3,438 1,943
Metop-B 4,021 2,363
Total 12,536 7,218 (58%)
10.1029/2018JD028934
Journal of Geophysical Research: Atmospheres
ZOU AND TIAN 6
6-hourly best track (Figure 5a), the maximum surface wind as well as the radii of the 34-, 50-, and 64-kt winds
(Figure 5b), and the intensity category of Hurricane Matthew from 0000 UTC 28 September 2016 to 1800 UTC
9 October 2016. Hurricane Matthew was a tropical storm on 28 September, intensified to a category-1
hurricane on 29 September, and quickly reached category-5 intensity the next day. It soon turned from
moving westward to northward, weakened gradually, passed Cuba, and moved along and off the east coasts
of Florida and Georgia. Matthew made its landfall in South Carolina on 7 October 2016 as a category-1 hurri-
cane, immediately drifted away from land, and moved eastward into the ocean again. Hurricane Matthew
had catastrophic consequences, including hundreds of deaths and billions of dollars of economic losses in
five affected countries.
3.2. A Brief Description of the Warm-Core Retrieval Algorithm
In order to examine the warm-core structures of Hurricane Matthew, the three-dimensional atmospheric
temperature fields during the life cycle of Matthew are first retrieved from the remapped and limb-
corrected ATMS BT observations. Specifically, the atmospheric temperatures at different pressure levels
Figure 6. Scatterplots of ATMS TB observations (limb corrected) on 28 September 2016 within 10°N–20°N between
(a) channels 5 and 7, (b) channels 6 and 8, and (c) channels 7 and 8. The colors show LWP values. The magenta lines are
linear regression best fit lines for clear-sky data (black dots; LWP <0.03 g/kg).
10.1029/2018JD028934
Journal of Geophysical Research: Atmospheres
ZOU AND TIAN 7
are expressed as functions of ATMS BT observations at each ATMS/AMSU-A FOV location by the following
equation:
TpðÞ¼C0pðÞþ ∑
i2;p
i¼i1;p
CipðÞTb;ipðÞ;(6)
where pis pressure and (i
1, p
,i
1, p
+1,i
1, p
+2,⋯,i
2, p
) represents a subset of AMSU-A channels 4–14 (ATMS
channels 5–15) that are correlated with the atmospheric temperatures at the pressure level p(Tian & Zou,
2016). Equation (6) is similar to the refined temperature retrieval algorithm proposed by Tian and Zou
(2016) except for removing the scan-angle dependence of the regression coefficients. In order to remove rain
contamination from the retrieved atmospheric temperatures in the lower troposphere, ATMS channel 5–7
and AMSU-A channels 4–6 are not used in cloudy conditions. The regression coefficients (C
0
and C
i
) are
obtained from a training data set, namely, ATMS observations over the ocean and collocated GPS RO tem-
perature profiles from 1 to 27 September 2016. Data are collocated if they fall within a 3-h time window
and a 100-km spatial distance. The total number of RO profiles collocated with ATMS observations from
the Constellation Observing System for Meteorology, Ionosphere, and Climate (COSMIC), MetOp-A, and
MetOp-B satellites between 60°S and 60°N are provided in Table 2. Out of the 12,536 RO profiles that are col-
located with ATMS data, 58% represent clear-sky conditions (liquid water path (LWP) <0.03 g/kg).
Figure 6 shows scatterplots of ATMS BT observations after limb correction on 28 September 2016 within
[10°N–20°N], that is, channel 5 versus channel 8 (Figure 6a), channel 6 versus channel 8 (Figure 6b), and chan-
nel 7 versus channel 8 (Figure 6c). The values of LWP for cloudy (LWP ≥0.03 g/kg) and clear-sky
(LWP <0.03 g/kg) observations are shown as colored and black dots, respectively. Three corresponding linear
regression functions are derived from clear-sky data. Cloud contamination is seen in the three low-level ATMS
channels (5–7). BTs are either elevated or lowered in the presence of clouds more significantly for the lower-
level channels 5 and 6, and less for channel 7. In most TC cases, cloud contamination in ATMS channel 8 is
negligible. But for some small-cored, intense convection TCs, cloud contamination in ATMS channel 8 might
be more significant.
Figure 7. Regression coefficients for ATMS temperature retrievals at 400 (black), 250 (blue), and 200 hPa (red) trained with
collocated RO data during the period 1–27 September 2016 in (a) clear-sky and (b) cloudy conditions. The intercepts for
temperature retrievals at 400, 250, and 200 hPa are 67.78, 156.70, and 112.72 for clear-sky conditions and 85.95,
83.30, and 39.11 for cloudy conditions.
10.1029/2018JD028934
Journal of Geophysical Research: Atmospheres
ZOU AND TIAN 8
The regression coefficients for retrieving atmospheric temperatures at the 400-, 250-, and 200-hPa pressure
levels are provided in Figure 7. The 400-, 250-, and 200-hPa pressure levels are where the WF peaks of chan-
nels 7, 8, and 9, respectively, are located. The regression coefficients are the largest for the channel with its WF
peak located at the same 400-, 250-, and 200-hPa pressure levels, respectively, except for channel 8 in cloudy
conditions. In clear-sky conditions (Figure 7a), channels 6, 8, and 9 also contribute to the temperature retrieval
at 400 hPa (channel 7’s WF peak); channels 6, 7, 9, and 10 contribute to the temperature retrieval at 250 hPa
(channel 8’s WF peak); and channels 6, 7, 9, 10, and 11 contribute to the temperature retrieval at 200 hPa
(channel 9’s WF peak). Channels whose WF peaks are further away from the pressure level on which the tem-
perature is to be retrieved have little impact on the retrieval. The variations in the magnitudes of the regres-
sion coefficients in cloudy conditions are similar for temperature retrievals at the 250- and 200-hPa pressure
levels (Figure 7b). Since channel 7 is not used in cloudy conditions, the regression coefficients for channel 8
are the largest for temperature retrievals at 400 hPa.
4. Numerical Results of Matthew’s Warm-Core Structures and Evolutions
Temperature anomalies at 250 hPa on 0600 UTC 2 October 2016 calculated from ATMS BT observations with-
out and with remapping are presented in Figures 8a and 8b, respectively. Here the temperature anomalies
are calculated by subtracting the mean temperature averaged within a 15° latitude/longitude box centered
at the storm center but outside of the 34-knot wind radius circle. The temperature retrieval without applying
the remapping algorithm (Figure 8a) contains significant noise (~1 K), which is successfully mitigated by the
Figure 8. Temperature anomalies at 250 hPa at 0643 UTC 2 October 2016 calculated from ATMS observations (a) without
and (b) with remapping. (c) Temperature anomalies calculated with ECMWF interim reanalysis data. The black cross shows
the location of the observed hurricane center.
10.1029/2018JD028934
Journal of Geophysical Research: Atmospheres
ZOU AND TIAN 9
remapping algorithm (Figure 8b). The warm-core center of Hurricane Matthew has a temperature anomaly of
about 5 K and is centered at the observed hurricane location. The warm-core in the European Centre for
Medium-Range Weather Forecasts (ECMWF) ERA-interim temperature analysis (Figure 8c) is misplaced to
the east by more than 2 degrees of longitude, and is too broad and too weak, which is existent in global
large-scale analyses (Dee et al., 2011).
The impact of rain contamination on the warm-core retrieval obtained without eliminating channels 5, 6,
and 7 for the regression in cloudy conditions is shown in Figure 9. The temperature anomalies at
250 hPa (Figure 9a) contain a rainband structure that is in phase with the LWP distribution at 0600
UTC 2 October 2016 (Figure 9b). The maximum temperature anomaly at the warm-core center
(Figure 9a) is more than 3 K higher than that in Figure 8b. The vertical cross section of the temperature
anomalies (Figure 9c) shows a series of cold anomalies in the lower troposphere, with larger magnitudes
over areas with stronger LWPs. Such unrealistic cold anomalies below the upper-level warm-core is not
seen in the retrieval that neglects channels 5, 6, and 7 in cloudy conditions (Figure 9d) nor in the
ECMWF interim reanalysis (Figure 9e). It should also be pointed out that the warm-core intensity around
250 hPa becomes stronger due to rain contaminations on channels 5–7 (Figure 9c). The observed bright-
ness temperatures of channels 5 and 6 are lower in the presence of rain due to scattering effects
Figure 9. (a) Temperature anomalies at 250 hPa (unit K), (b) liquid water path (unit kg/m
2
), and (c) the vertical cross section
of temperature anomalies following the dashed line in a at 0634 UTC 2 October 2016 obtained from remapped ATMS TB
observations without eliminating channels 5, 6, and 7. The regression is made using data representing cloudy
conditions. (d) Same as (c) except for results obtained from remapped ATMS TB observations with channels 5, 6, and 7
eliminated in cloudy conditions. (e and f) Same as (c) except for results from (e) the ECMWF interim reanalysis at 0600
UTC and (f) the MIRS retrievals at 0634 UTC 2 October 2016.
10.1029/2018JD028934
Journal of Geophysical Research: Atmospheres
ZOU AND TIAN 10
(Figure 6), and their correlations with the upper tropospheric (200 and 250 hPa) atmospheric
temperatures are negative (Tian & Zou, 2016). When channel 6 is included, the retrieved atmospheric
temperatures in the upper troposphere were elevated due to a negative regression coefficient
(Figure 7a) and lower brightness temperatures. Impacts of channel 5 on the temperature retrieval are
small since the regression coefficient for channel 5 is near zero (see Figure 7a). The temperature
anomaly obtained from the Microwave Integrated Retrieval System (Figure 9f) also displays a warm-
core around 200 hPa. An extremely warm center (>7 K) peaking near the surface and small
horizontal-scale oscillating features are found in the Microwave Integrated Retrieval System results,
both of which seem not physical. The ATMS-derived warm-core is much more compact and located at
a higher altitude than that in the ECMWF reanalysis. The maximum LWP from ATMS is located to the
northeast of the warm-core and hurricane center.
The radial profiles of ATMS-derived temperature anomalies from 30 September to 9 October 2016 are shown
in Figure 10a (without averaging) and Figure 10b (with averaging, and with exclusion of channels 5, 6, and 7
in cloudy conditions). The modified hurricane warm-core retrieval algorithm for ATMS observations gives
more compact warm-core structures and more consistent temporal evolutions.
The temporal evolution of the warm-core at 250 hPa is provided during the time period from 0134 local
standard time 2 October to 0115 local standard time 3 October 2016 based on AMSU-A observations from
Figure 10. ATMS-derived temperature anomalies at 250 hPa as a function of distance from the center of Hurricane
Matthew in the direction from west to east during the period 30 September 30 to 9 October 2016 calculated (a) without
and (b) with remapping and the elimination of channels 5, 6, and 7 for the regression in cloudy conditions. The solid
and dashed colored lines are for different days.
10.1029/2018JD028934
Journal of Geophysical Research: Atmospheres
ZOU AND TIAN 11
Metop-B, NOAA-15, and NOAA-18, and S-NPP ATMS observations (Figure 11). NOAA-19 AMSU-A
observations are not used since the NEDTs for channels 7 and 8 are far above the specifications.
Warm-core structures and intensities vary with time. The warm-cores near the nadir (the swath middle)
Figure 11. Hurricane Matthew’s warm-core temperature anomaly structures at 250 hPa (unit K) on and between 0134 local
standard time (LST) 2 October to 1756 LST 3 October 2016 retrieved from Metop-B, NOAA-15, and NOAA-18 AMSU-A and
S-NPP ATMS observations. The LSTs and satellite names are indicated above each panel.
10.1029/2018JD028934
Journal of Geophysical Research: Atmospheres
ZOU AND TIAN 12
are generally smaller than those near the swath edges, which are a result of the scan dependence of the
observational resolution. AMSU-A-derived warm-cores are larger than those from the ATMS due to the
coarser observational sampling of the AMSU-A. Also, the maximum of the warm anomaly is located at
exactly the best-track hurricane center, suggesting that Hurricane Matthew maintained a straight
vertical structure during this time period. The strongest warm-core intensity is observed at 1647 UTC 2
Figure 12. Same as Figure 11 except for LWP.
10.1029/2018JD028934
Journal of Geophysical Research: Atmospheres
ZOU AND TIAN 13
October by NOAA-15 AMSU-A. It agrees with what was reported by Stewart (2017) that Hurricane
Matthew underwent an intensification period on 2 October and reached a peak intensity on the same
day later. Similar to Figures 9a and 9b, it is further confirmed that the temperature anomalies in the upper
troposphere at all times shown in Figure 11 do not have rainband-like structures that are in phase with
the LWP distribution at the corresponding times (Figure 12).
5. Summary and Conclusions
ATMS BT observations have random noise with relatively large magnitudes that obscures the ATMS-retrieved
atmospheric temperature fields. A remapping algorithm is applied to ATMS observations to reduce their ran-
dom noise. Even though the impact of this noise on ATMS warm-core retrievals is eliminated after applying
the remapping algorithm to ATMS observations, an unrealistic rainband-like warm-anomaly structure is
found in hurricane rainband areas. This is caused by cloud contamination on ATMS channels 5, 6, and 7
(AMSU-A channels 4, 5, and 6). The elimination of these channels for ATMS/AMSU-A temperature retrievals
in cloudy conditions avoids not only the unrealistic rainband-like warm anomaly but also the unrealistic cold
anomalies in the lower troposphere. A limb correction is finally applied to ATMS and AMSU-A observations so
that there are enough GPS RO data collocated with AMSU-A/ATMS data to serve as a training data set.
The 3-D temperature anomaly fields during the life cycle of Hurricane Matthew (2016) are then retrieved
using the same temperature retrieval algorithm as that developed by Tian and Zou (2016) except for the
above-mentioned three changes. Compared with the temperature anomalies from the ECMWF interim rea-
nalysis, the warm-core structures retrieved from ATMS observations are much more compact and closer to
the best-track positions. The unrealistic rainband-like warm-core structures that are seen in the temperature
retrieval from the remapped ATMS BT observations, which are in phase with the rainbands revealed by the
VIIRS 10.8-μm radiance observations, are mostly eliminated. It is finally shown that AMSU-A observations
from Metop-B, NOAA-15, and NOAA-18, and S-NPP ATMS observations together illustrate the temporal evo-
lution of the warm-core at ~1.5-h intervals.
It may be possible to observe a better-than-hourly evolution of hurricane warm-cores during the 2018 hurri-
cane season since NOAA-20 ATMS observations are now available. The warm-core results, along with
microwave-imager-retrieved total precipitable water, surface wind speed, and sea surface temperature, as
well as S-NPP and NOAA-20 Ozone Mapping and Profile Suite total column ozone, will be incorporated into
a new satellite-observation-based vortex initialization as a follow-on study.
References
Atkinson, N. (2011). Annex to AAPP scientific documentation: pre-processing of ATMS and CrIS. Document (NWPSAF-MO-UD-027).
Brueske, K. F., & Velden, C. S. (2003). Satellite-based tropical cyclone intensity estimation using the NOAA-KLM series advanced microwave
sounding unit (AMSU). Monthly Weather Review,131(4), 687–697. https://doi.org/10.1175/1520-0493(2003)131<0687:SBTCIE>2.0.CO;2
Chen, H., & Zhang, D.-L. (2013). On the rapid intensification of Hurricane Wilma (2005). Part II: Convective bursts and the upper-level warm
core. Journal of the Atmospheric Sciences,70(1), 146–162.
Chen, Y., Han, Y., & Weng, F. (2013). Detection of Earth-rotation Doppler shift from Suomi national polar-orbiting partnership cross-track
infrared sounder. Applied Optics,52(25), 6250–6257. https://doi.org/10.1364/AO.52.006250
Dee, D. P., Uppala, S. M., Simmons, A. J., Berrisford, P., Poli, P., Kobayashi, S., et al. (2011). The ERA-interim reanalysis: Configuration and perfor-
mance of thedata assimilation system.Quarterly Journal of the RoyalMeteorological Society,137(656), 553–597. https://doi.org/10.1002/qj.828
Demuth, J. L., DeMaria, M., & Knaff, J. A. (2006). Improvement of advanced microwave sounding unit tropical cyclone intensity and size
estimation algorithms. Journal of Applied Meteorology and Climatology,45(11), 1573–1581. https://doi.org/10. 1175/JAM2429.1
Demuth, J. L., DeMaria, M., Knaff, J. A., & Vonder Haar, T. H. (2004). Evaluation of advanced microwave sounding unit tropical-cyclone intensity
and size estimation algorithms. Journal of Applied Meteorology and Climatology,43(2), 282–296. https://doi.org/10.1175/1520-
0450(2004)043<0282:EOAMSU>2.0.CO;2
Dolling, K., & Barnes, G. M. (2012). Warm-core formation in tropical storm Humberto (2001). Monthly Weather Review,140(4), 1177–1190.
Durden, S. L. (2013). Observed tropical cyclone eye thermal anomaly profiles extending above 300 hPa. Monthly Weather Review,141(12),
4256–4268. https://doi.org/10.1175/MWR-D-13-00021.1
Goldberg, M. D., Crosby, D. S., & Zhou, L. (2001). The limb adjustment of AMSU-A observations: Methodology and validation. Journal of
Applied Meteorology and Climatology,40(1), 70–83.
Halverson, J. B., Simpson, J., Heymsfield, G., Pierce, H., Hock, T., & Ritchie, L. (2006). Warm core structure of hurricane Erin diagnosed from high
altitude dropsondes during CAMEX-4. Journal of the Atmospheric Sciences,63(1), 309–324. https://doi.org/10.1175/JAS3596.1
Hawking, H. F., & Rubsam, D. T. (1968). Hurricane Hilda, 1964. II, structure and budgets of the hurricane on October J 1964. Monthly Weather
Review,96(9), 617–636. https://doi.org/10.1175/1520-0493(1968)096<0617:HH>2.0.CO;2
Hawkins, H. F., & Imbembo, S. M. (1976). The structure of a small, intense hurricane—Inez 1966. Monthly Weather Review,104(4), 418–442.
https://doi.org/10.1175/1520-0493(1976)104<0418:TSOASI>2.0.CO;2
Hilton, F., Armante, R., August, T., Barnet, C., Bouchard, A., Camy-Peyret, C., et al. (2012). Hyperspectral Earth observation from IASI: Five years
of accomplishments. Bulletin of the American Meteorological Society,93(3), 347–370. https://doi.org/10.1175/BAMS-D-11-00027.1
10.1029/2018JD028934
Journal of Geophysical Research: Atmospheres
ZOU AND TIAN 14
Acknowledgments
This study was supported by NOAA
grant NA14NES4320003 (Cooperative
Institute for Climate and Satellites
(CICS)) at the University of
Maryland/ESSIC. The ATMS and VIIRS
observations employed in this study
were obtained from NOAA’s
Comprehensive Large Array-data
Stewardship System at the website
http://www.class.ncdc.noaa.gov/saa/
products/welcome, and ECMWF ERA-
Interim Reanalysis data were obtained
from the website at http://www.ecmwf.
int/en/research/climate-reanalysis/era-
interim. Results used to generate the
figures in this study are available at
http://jlrdata.umd.edu/Tianshare/
HurricaneMatthew.html.
Janssen, M. A. (1994). Atmospheric remote sensing by microwave radiometry. International Journal of Climatology,14(9), 1062–1063.
Knaff, J. A., Seseske, S. A., DeMaria, M., & Demuth, J. L. (2004). On the influences of vertical wind shear on symmetric tropical cyclone structure
derived from AMSU. Monthly Weather Review,132(10), 2503–2510. https://doi.org/10.1175/1520-0493(2004)132<2503:OTIOVW>2.0.CO;2
Komaromi, W. A., & Doyle, J. D. (2017). Tropical Cyclone Outflow and Warm Core Structure as Revealed by HS3 Dropsonde Data. Monthly
Weather Review,145(4), 1339–1359.
La Seur, N. E., & Hawkins, H. F. (1963). An analysis of hurricane Cleo (1958) based on data from research reconnaissance aircraft. Monthly
Weather Review,91(10), 694–709. https://doi.org/10.1175/1520-0493(1963)091<0694:AAOHCB>2.3.CO;2
Le Marshall, J., Jung, J., Derber, J., Chahine, M., Treadon, R., Lord, S. J., et al. (2006). Improving global analysis and forecasting with AIRS. Bulletin
of the American Meteorological Society,87(7), 891–895. https://doi.org/10.1175/BAMS-87-7-891
Merrill, R. T. (1995). Simulations of physical retrieval of tropical cyclone thermal structure using 55-GHz band passive microwave observations
from polar-orbiting satellites. Journal of Applied Meteorology,34(4), 773–787. https://doi.org/10.1175/1520-0450(1995)034<0773:
SOPROT>2.0.CO;2
Mo, T. (1996). Prelaunch calibration of the advanced microwave sounding unit-A for NOAA-K. IEEE Transactions on Microwave Theory and
Techniques,44(8), 1460–1469.
Sorensen, H. V., Jones, D., Heideman, M., & Burrus, C. (1987). Real-valued fast Fourier transform algorithms. IEEE Transactions on Acoustics,
Speech, and Signal Processing,35(6), 849–863.
Stern, D. P., & Nolan, D. S. (2012). On the height of the warm core in tropical cyclones. Journal of the Atmospheric Sciences,69(5), 165–1680.
Stewart, S. R., 2017: Tropical cyclone report: Hurricane Matthew (AL142016). 7 April 2016, 96 pp., http://www.nhc.noaa.gov/data/tcr/
AL142016_Matthew.pdf.
Tian, X., & Zou, X. (2018). Capturing Size and Intensity Changes of Hurricanes Irma and Maria (2017) from Polar-Orbiting Satellite Microwave
Radiometers. Journal of Atmospheric Sciences,75(8), 2509–2522.
Tian, X., & Zou, X. (2016). ATMS- and AMSU-A-derived hurricane warm core structures using a modified retrieval algorithm. Journal of
Geophysical Research,121.
Tian, M., Zou, X., & Weng, F. (2015). Use of Allan Deviation for Characterizing Satellite Microwave Sounder Noise Equivalent Differential
Temperature (NEDT). Ieee Geoscience and Remote Sensing Letters,12(12), 2477–2480.
Wark, D. Q. (1993). Adjustment of TIROS Operational Vertical Sounder data to a vertical view (NOAA Tech. Rep. NESDIS-64, 36 pp).
Weng, F., Zhao, L., Ferraro, R. R., Poe, G., Li, X., & Grody, N. C. (2003). Advanced microwave sounding unit cloud and precipitation algorithms.
Radio Science,38(4), 8068. https://doi.org/10.1029/2002RS002679
Weng, F., Zou, X., Wang, X., Yang, S., & Goldberg, M. (2012). Introduction to Suomi NPP ATMS for NWP and tropical cyclone applications.
Journal of Geophysical Research,117, D19112. https://doi.org/10.1029/2012JD018144
Yang, H., & Zou, X. (2014). Optimal ATMS remapping algorithm for climate research. IEEE Transactions on Geoscience and Remote Sensing,
52(11), 7290–7296.
Zhang, K., Zhou, L., Goldberg, M., Liu, X., Wolf, W., Tan, C., & Liu, Q. (2017). A Methodology to Adjust ATMS Observations for Limb Effect and Its
Applications. Journal of Geophysical Research,122(21), 11,347–11,356.
Zhu, T., & Weng, F. (2013). Hurricane Sandy warm-core structure observed from advanced technology microwave sounder. Geophysical
Research Letters,40, 3325–3330. https://doi.org/10.1002/grl.50626
Zhu, T., Zhang, D.-L., & Weng, F. (2002). Impact of the advanced microwave sounding unit measurements on hurricane prediction. Monthly
Weather Review,130(10), 2416–2432. https://doi.org/10.1175/1520-0493(2002)130<2416:IOTAMS>2.0.CO;2
10.1029/2018JD028934
Journal of Geophysical Research: Atmospheres
ZOU AND TIAN 15
