close

Вход

Забыли?

вход по аккаунту

?

Satellite -based tropical cyclone intensity estimation using NOAA -KLM series advanced microwave sounding unit (AMSU) data

код для вставкиСкачать
INFORMATION TO USERS
This manuscript has been reproduced from the microfilm m aster. UMI films
the text directly from the original or copy submitted. Thus, som e thesis and
dissertation copies are in typewriter face, while others m ay be from any type of
com puter printer.
The quality of this reproduction is dependent upon the quality of the
copy submitted. Broken o r indistinct print, colored or poor quality illustrations
and photographs, print bleedthrough, substandard m argins, and improper
alignment can adversely affect reproduction.
In the unlikely event that th e author did not send UMI a com plete manuscript
and there are missing p ag es, th e se will be noted.
Also, if unauthorized
copyright material had to b e removed, a note will indicate the deletion.
Oversize materials
(e.g.,
m aps, drawings, charts)
are
reproduced
by
sectioning the original, beginning at the upper left-hand com er and continuing
from left to right in equal sections with small overlaps.
Photographs included in the original manuscript have been reproduced
xerographically in this copy.
Higher quality 6” x 9” black and white
photographic prints are available for any photographs or illustrations appearing
in this copy for an additional charge. Contact UMI directly to order.
P roQ uest Information and Learning
300 North Z eeb R oad, Ann Arbor, Ml 48106-1346 USA
800-521-0600
R e p r o d u c e d with p e r m i s s io n of t h e co p y rig h t o w n er. F u r th e r r e p r o d u c tio n prohibited w ith o u t p e r m is s io n .
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
SATELLITE-BASED TROPICAL CYCLONE INTENSITY ESTIMATION
USING NOAA-KLM SERIES ADVANCED MICROWAVE SOUNDING
UNIT (AMSU) DATA
By
KURT FREDERICK BRUESKE
A dissertation submitted in partial fulfillment o f
the requirements for the degree of
Doctor o f Philosophy
(Atmospheric and Oceanic Sciences)
at the
UNIVERSITY OF WISCONSIN-MADISON
2001
R e p r o d u c e d with p e r m i s s io n of t h e co p y rig h t o w n e r. F u r th e r r e p r o d u c tio n prohibited w ith o u t p e r m is s io n .
UMI Number: 3012541
___
®
UMI
UMI Microform 3012541
Copyright 2001 by Bell & Howell Information and Learning Company.
All rights reserved. This microform edition is protected against
unauthorized copying under Title 17, United States Code.
Bell & Howell Information and Learning Company
300 North Zeeb Road
P.O. Box 1346
Ann Arbor, Ml 48106-1346
R e p r o d u c e d with p e r m i s s io n of t h e co p y rig h t o w n er. F u r th e r r e p r o d u c tio n prohibited w ith o u t p e r m is s io n .
R eaders’ Page. This page is not to be hand-written except for
A dissertation entitled
Satellite—Based Tropical Cyclone Intensity Estimation
Using NOAA-KLM Series Advanced Microwave Sounding
Unit (AMSU) Data
submitted to the Graduate School of the
University of Wisconsin-Madison
in partial fulfillment of the requirements for the
degree of Doctor of Philosophy
by
R eaders’ Page. This page is not to be hand-written except for the signatures
Kurt Frederick Brueske
Date of Final Oral Examination:
Month & Year Degree to be awarded:
April 18, 2001
December
Approval Signatures of Dissertation Readers:
May 2001
August
Signature, Dean of Graduate School
R e p r o d u c e d with p e r m i s s io n of t h e co p y rig h t o w n e r. F u r th e r r e p r o d u c tio n prohibited w ith o u t p e r m is s io n .
i
TABLE OF CONTENTS
Table of Contents
i
Abstract
iii
List of Tables
vi
List of Figures
vii
Acknowledgements
ix
Section 1: Introduction
1
1.1 Background
3
1.2 Contemporary Issues
7
1.2.1 Scattering
8
1.2.2 Scan Geometry
10
1.2.3 Diffraction
12
Section 2: Data
21
Section 3: Methodology
25
3.1 Forward M odel
25
3.1.1 Maximum Likelihood Regression
29
3.1.2 TC UTWA Horizontal Scale
30
3.2 Dependent Data Set
Section 4: Independent Test
32
41
4.1 Implementation
42
4.2 Validation
44
R e p r o d u c e d with p e r m i s s io n of t h e co p y rig h t o w n er. F u r th e r r e p r o d u c tio n prohibited w ith o u t p e r m is s io n .
a
4.3 Results
Section 5: Case Studies
45
53
5.1 Inter-basin Applicability
53
5.2 Rapid Intensity Change
57
5.3 Tropical Cyclone Classification
59
5.4 Intensity Estimates Over Land
59
5.5 Comparison with Dvorak Estimates
61
Section 6: Analysis O f Errors
73
6.1 Analysis O f Estimate Error And Position
73
6.2 AMSU Synchronization
74
6.3 Instrument Limitations
75
6.4 Environmental Variability
77
6.5 Estimating TC Position
79
6.6 Algorithm Numerics
80
Section 7: Summary And Conclusions
84
7.1 General Comments
84
7.2 Application Considerations
86
7.3 Future Work And Recommendations
87
References
R e p r o d u c e d with p e r m i s s io n of t h e c o p y rig h t o w n e r . F u r th e r re p ro d u c tio n p rohib ited w ith o u t p e r m is s io n .
88
iii
Satellite-based Tropical Cyclone Intensity Estimation Using
NOAA-KLM Advanced Microwave Sounding Unit (AMSU) Data
Kurt Frederick Brueske
Under the supervision o f Associate Professor Steven A. Ackerman and Christopher S. Velden
ABSTRACT
Satellite-bome passive microwave radiometers are well suited to monitor tropical
cyclones (TCs) by virtue of their ability to assess changes in tropospheric warm core structure in
the presence of clouds. The temporal variability in TC warm core size, structure, and magnitude
provide vital information on changes in kinematic structure and minimum sea level pressure
(MSLP) through implicit thermodynamic and dynamic constraints. In this study, the efficacy of
a hybrid-statistical algorithm capable of estimating MSLP using Advanced Microwave Sounding
Unit (AMSU) temperature (AMSU-A) and moisture sounder (AMSU-B) data is demonstrated.
The proposed AMSU TC intensity estimation algorithm addresses tropospheric warm anomaly
(UTWA) sub-sampling through explicit convolution of an analytic function approximating the
horizontal distribution of the TC UTWA and the AMSU-A antenna gain pattern. Differences
between observed AMSU-A 54.94 GHz upper tropospheric limb-corrected brightness
temperatures (Tb) and those of a forward model based on the convolution function are
R e p r o d u c e d with p e r m i s s io n of t h e co p y rig h t o w n er. F u r th e r r e p r o d u c tio n prohibited w ith o u t p e r m is s io n .
minimized through maximum likelihood regression using known an-d estimated radiometric
noise, TC position and structure function errors.
Dependent data set results using 1999 Atlantic (ATL) and eastern Pacific (EPAC) basin
aircraft reconnaissance and AMSU-A observation pairs (n=22) indicate significantly increased
correlation between scan geometry/diffraction-corrected AMSU-A 5-4.94 GHz TC UTWA and
MSLP (R2 = 0.9) vs. using no correction (R2 = 0.7). Linear least squares regression coefficients
derived from the 1999 dependent sample were used to predict TC M S L P using a fully automated,
objective processing scheme in multiple ocean basins during 2000 anad a limited number of cases
in early 2001. ATL/EPAC independent test results (n=31) indicate tlhat substantial improve­
ments in correlation between AMSU-A TC UTWA and MSLP (R2 =00.94 vs. 0.80), predicted
MSLP mean error of 6.2 hPa vs. 7.5 hPa, and reduced standard devia*tion of 8.0 hPa vs. 9.9 hPa)
are possible using the proposed hybrid AMSU intensity estimation schem e.
Comparison o f AMSU ATL/EPAC independent test results w^ith subjective Dvorak
MSLP estimate mean error and standard deviation (7.8 hPa +/- 7.6 hEPa, n=31) analyses indicates
slightly superior AMSU MSLP estimate mean error with comparable: variance values. Limited
western North Pacific Ocean (NWPAC), Southern Indian Ocean (S IO ) and Southern Pacific
Ocean (SPAC) validation candidates (i.e., mean error of 5.3 hPa vs. 7?.4 hPa and standard
deviations 7.2 hPa vs. 9.7 hPa, n=l 1) suggest inter-basin operability although a larger sample
size is required to establish statistical significance.
R e p r o d u c e d with p e r m i s s io n of th e co p y rig h t o w n er. F u r th e r r e p r o d u c tio n p rohibited w ith o u t p e r m is s io n .
This study is the first successful attempt to explicitly correct for the interaction o f the TC
UTWA and AMSU-A antenna gain pattern using an objective, automated technique using both
dependent and independent near-real time data. Based on the results of this study, a fully
automated, objective, physically-based TC intensity estimation technique using AMSU passive
microwave radiance data is ready for future evaluation by operational forecasting centers.
Furthermore, when used in conjunction with other existing and/or planned satellite-based sensing
capabilities, the unique virtues o f the AMSU will assist overall TC preparedness and help
provide new insights into diverse areas of study from TC genesis to diagnosis of transition from
subtropical-tropical-extratropical state.
R e p r o d u c e d with p e r m i s s io n of t h e co p y rig h t o w n e r. F u r th e r r e p r o d u c tio n prohibited w ith o u t p e rm is s io n .
vi
LIST OF TABLES
Table 1. Horizontal Structure Function Values
38
Table 2. AMSU Sensor Information
52
Table 3. Independent Test Results
72
Table 4. Retrieval Performance vs. TC Position
81
R e p r o d u c e d with p e r m i s s io n of th e co p y rig h t o w n er. F u r th e r r e p r o d u c tio n prohibited w ith o u t p e r m is s io n .
vil
LIST OF FIGURES
Figure I. Tropical cyclone schematic diagram
13
Figure 2. Hurricane Inez warm core
14
Figure 3. Tropical cyclone warm anomaly/antenna gain interaction (weak and strong)
15
Figure 4. AMSU-A and AMSU-B weighting functions
16
Figure 5. MSU vs. AMSU horizontal resolution
17
Figure 6. Super Typhoon Zeb AMSU vs. MSU Tb(C) comparison 13-14 October 1998
18
Figure 7. Hurricane Floyd 14 September 1999 1238UTC AMSU-A-derived UTWA (C)
19
Figure 8. Bessel function (first kind, order one) representing AMSU-A intensity as a
function of off-axis scan angle (0), frequency (54.94 GHz) and instrument bore
sight aperture
20
Figure 9. Sample vortex data message
24
Figure 10. Color-enhanced AMSU-B 89.0 GHz T b (°C) image o f a mature TC
35
Figure 11. Schematic illustrating the determination of tropical cyclone eye size using
AMSU-B 89.0 GHz Tb (°C) data
36
Figure 12. Color-enhanced AMSU-B 89.0 GHz Tb (°C) image o f a weak TC
37
Figure 13. Sample pattern of 9 (nominal) AMSU-A 54.94 GHz FOV selected for retrieval
use, Hurricane Floyd, 14 September 1999 1238UTC
39
Figure 14. AMSU-A 54.94 GHz TC UTWA (AB) vs. MSLP scatter plot (1999)
40
Figure 15. Schematic diagram depicting the automated AMSU TC MSLP estimation routine
during 2000
50
Figure 16. Scatter plot of raw, retrieved and hybrid AMSU-A 54.94 GHz TC UTWA vs. in
situ MSLP estimates (1999/2000/2001)
R e p r o d u c e d with p e r m i s s io n of t h e co p y rig h t o w n er. F u r th e r r e p r o d u c tio n prohibited w ith o u t p e r m is s io n .
51
Figure 17. NCAR/NCEP reanalysis tropopause pressure and geopotential height (2.5°
resolution) fields: NWPAC vs. western ATL, 1948-1999
Figure 18. A comparison of AM SU vs. Dvorak TC intensity estimates, Super Typhoon Bilis
(18W), 20-24 August 2000
Figure 19. Garden Island, Taiwan, Doppler radar reflectivity (PPI) image, Super Typhoon
Bilis (18W), 1130 UTC 22 August 2000
Figure 20. Time series of JTWC intensity estimates, Super Typhoon Bilis (18W), 19-24
August 2000
Figure 21. AMSU-B 89.0 GHz Tb (°C) image, Tropical Cyclone Abigail (12P), 1046 UTC
26 February 2001
Figure 22. Hurricane Isaac composite image, 23 September 2000
Figure 23. WTNT KNHC forecast discussion bulletin
Figure 24. Hurricane Michael radar reflectivity composite image, 2145 UTC 19 October
2000 and 2304 UTC 19 October 2000 AMSU-B 89.0 GHz Tb (°C) image
Figure 25. Tropical Cyclone W insome (08P) AMSU-A Tb (°C) warm core observations,
1656 UTC 12 February 2001
Figure 26. Color-enhanced AMSU-A and AMSU-B images of Hurricane Keith, 1337 UTC
1 October 2000
Figure 27. Color-enhanced AMSU-B 89.0 GHz Tb (°C) image of Tropical Cyclone Ando
(04S), 0347 UTC 6 January 2001
R e p r o d u c e d with p e r m i s s io n of th e co p y rig h t o w n er. F u r th e r r e p r o d u c tio n prohibited w ith o u t p e r m is s io n .
ACKNOWLEDGEMENTS
Many individuals and organizations deserve credit for their support during my time as a
graduate student at the University of Wisconsin-Madison. First and foremost is my advisor
Steven A. Ackerman, Associate Professor, Department o f Atmospheric and Oceanic Sciences
(AOS) and Director of the University of Wisconsin-Madison Cooperative Institute for
Meteorological Satellite Studies (CIMSS). I am also grateful for the support o f my committee
members —Profs. John Young (AOS), Greg Tripoli (AOS), Zhengyu Liu (AOS), and John
Strikwerda (Computer Science) for helping to guide me through the Ph.D. process under
demanding time constraints. Chris Velden, Director, CIMSS Tropical Cyclone Research Group,
deserves special recognition for his outstanding leadership, oversight, and willingness to share
time and intellectual insight with me throughout my research program. The relevance and
potential future utilization o f the AMSU by the tropical meteorology community (operational
and academic) are largely due to Chris’ guidance, wisdom and support over the past three years.
Dr. Robert Merrill (formerly o f CIMSS) deserves credit for his pioneering work on the
overall methodology and development of the original ‘experim ental Tropical Cyclone Retrieval’
or ‘XTCR’ algorithm upon which this research initiative is based. The people and resources of
CIMSS also deserve special recognition. I rarely, if ever, lacked resources or willingness on the
part of CIMSS members to assist me perform vital functions of my research program. Dr. Hal
Woolf (NOAA/NESDIS retired) was instrumental in making available raw AMSU data,
computer code, and resources necessary to acquire, navigate, process, and limb correct raw
AMSU radiance data. Dr. Fred Nagle (NOAA/NESDIS retired) also helped provide source code
R e p r o d u c e d with p e r m i s s io n of th e co p y rig h t o w n er. F u r th e r r e p r o d u c tio n prohibited w ith o u t p e r m is s io n .
X
necessary to project individual AMSU sensor field-of-view patterns on the earth’s surface, a key
component of the AMSU intensity retrieval algorithm’s forward model numerics.
Exploitation o f internet-based tools and technologies proved an important part of my
research program, allowing the automated processing and visualization of AMSU tropical
cyclone-related data in near-real time as well as the ability to share these unique resources with a
global audience. Tim Olander, Steve Wanzong and Steve Barnet helped make this happen.
Liam Gumley kindly provided Interactive Data Language (IDL) programs, computer resources
and expertise required to interrogate and display AMSU limb corrected, geolocated radiance
data. External web-based resources also proved vital for both near-real time and historical
analysis of tropical cyclones in general. In this regard, I wish to personally thank Jeff Hawkins,
Naval Research Laboratory/Monterey (NRL/MRY) for his pioneering work and development of
the now-famous NRL/MRY tropical cyclone web page. The skilled assembly and fusion of
diverse satellite resources available on the NRL/MRY web page proved an incredibly powerful
tool in the course of analyzing the structure and development of tropical cyclones. Thanks are
also in order for Drs. Joanne Simpson and Jeff Halverson, NASA/GSFC, for their stimulating
discussions on tropical cyclone energetics and overwhelming support o f my work with the
AMSU. I’m indebted to Profs. James Head and Delores Knipp, USAF Academy Dept of Physics
(USAFA/DFP) for the opportunity to serve the USAF as an Air Force Institute of Technology
(AFIT) graduate student at the University of Wisconsin-Madison.
Last but not least, I would like to thank my parents, friends, and family for their
unwavering love and support over the years. AMSU research is sponsored by USAFA/DFP,
R e p r o d u c e d with p e r m i s s io n of t h e cop y rig h t o w n e r. F u r th e r r e p r o d u c tio n prohibited w ith o u t p e rm is s io n .
AFIT, the US Naval Research Laboratory M onterey (Mr. Jeff Hawkins) and US Navy Space
and Naval W arfare Systems Command (PMW-155) under PE 0603207N.
R e p r o d u c e d with p e r m is s io n of t h e cop y rig h t o w n e r. F u r th e r r e p r o d u c tio n prohibited w ith o u t p e r m is s io n .
1
1 Introduction
Tropical cyclones (hereafter referred to as TC) are warm core circulation systems whose
maintenance and intensification are the result of wind-driven evaporation and sensible heat flux
from the ocean surface (Frank, 1977a). As environmental air spirals in towards the TC surface
circulation center, its expands isothermally acquiring significant energy from the ocean surface.
The atmospheric response to air-sea energy flux, while complex, is simplified considering
principles of energy conservation (internal, mechanical and potential). The energy acquired
(latent ~ Lq) is redistributed vertically (potential energy ~ gz), transformed through condensation
processes (internal ~ CPT), and made available to the storm-scale kinematic structure
(mechanical ~ I©2) in the form o f the warm core. Riehl and Simpson (1979) emphasized the role
o f deep convection in the generation and maintenance o f the observed tropical tropospheric
vertical total energy distribution. Emanuel (1988a) further corroborated the role of cumulus-scale
convection by effectively reproducing the observed two-dimensional thermodynamic and
kinematic structure of a mature steady-state TC using a coupled ocean/atmosphere numerical
model based on planetary boundary layer 0e transfer parameterization.
In the context of exploring TC energetics, Simpson et al. (1998) argued that deep,
undiluted ‘hot tower’ cumulonimbus clouds are a capable and necessary component in the
vertical transfer and deposition o f high equivalent potential temperature (0e a CpT + gz + Lq)
PBL air into the upper troposphere (Fig. 1). Furthermore, in order for nascent TCs to maximize
the potential contribution of condensational latent heat energy towards reduction of mean sea
R e p r o d u c e d with p e r m i s s io n of th e co p y rig h t o w n er. F u r th e r r e p r o d u c tio n prohibited w ith o u t p e r m is s io n .
2
level pressure (MSLP), energy must be retained locally above the surface vortex at an altitude
that maximizes the reduction o f MSLP through hydrostatic vertical integration. Such retention is
aided by the presence of pre-existing background vorticity made available by monsoon gyres,
monsoon troughs, easterly waves, or convective vortices known to form in regions o f stratoform
precipitation. The efficiency in which energy released on the cumulus-scale is made available on
the storm scale is determined by the local Rossby radius of deformation (LR) which itself is
influenced by atmospheric stability, vigor of the vortex circulation and background vorticity.
Ritchie et al. (1997) effectively demonstrated the efficacy of merging mesoscale convective
vortices as a means of reducing LRand assisting TC genesis. Rogers et al. (2000) successfully
demonstrated the role of convective redevelopment within MCVs and their role in maintaining
and strengthening the vigor of the mesoscale convective vortex itself.
The absence of conventional in-situ measurements of TCs over broad tropical oceanic
regions makes remote sensing appealing and necessary for several reasons. First, the stochastic
and geographically remote nature of TC genesis and intensification regions makes it virtually
impossible to observe important thermodynamic and kinematic features using conventional data
during critical stages of development. Furthermore, limited budgets and logistical constraints
dictate that aircraft resources be reserved for TCs threatening —or projected to threaten —urban
centers or densely populated coastal regions. In contrast, satellite-based observing methods are
cost-effective, offer global coverage, and are capable o f detecting and monitoring multiple
aspects influencing TC energetics and dynamics (i.e., sea surface temperatures, wind shear, etc.)
R e p r o d u c e d with p e r m i s s io n of t h e co p y rig h t o w n er. F u r th e r r e p r o d u c tio n prohibited w ith o u t p e r m is s io n .
3
over the entire duration o f their lifecycle. As a result, the diverse and unique capabilities of
satellite-borne sensors will continue to play a key role in the development of our understanding
of tropical cyclones in the broadest possible terms.
1.1 Background
Assuming that TCs are in hydrostatic balance at the storm-scale (Frank, 1977a), the
magnitude and distribution of MSLP should mirror the tropospheric temperature distribution —
namely, the size and shape of the upper tropospheric warm anomaly (hereafter referred to as
UTWA) formed and maintained through aforementioned processes as well as through localized
warming realized through adiabatic compression and subsidence within the eye. Several authors
(Kotewaram, 1967, Hawkins and Imbembo, 1976, Nunez and Gray, 1977) documented the
characteristic size and shape of a mature TC UTWA using composite analyses o f in-situ aircraft
reconnaissance and upper air sounding measurements. Peak wanning for mature TCs was
horizontally bounded within the TC eye wall region at an altitude corresponding with the 250
hPa pressure level (Fig. 2). Based on the analysis o f 15 years of western North Pacific
(NWPAC) tropical cyclone aircraft reconnaissance data, Jordan (1961) found that flight level
(500-700 hPa) temperature data (using hydrostatic constraints) could account for only a small
percentage (25%) of the observed MSLP variability emphasizing the dominant role of upper
tropospheric thermal forcing.
With the advent of satellite-borne passive microwave radiometers, several authors have
demonstrated the feasibility o f detecting TC UTWA and their relationship with storm intensity.
R e p r o d u c e d with p e r m i s s io n of t h e co p y rig h t o w n er. F u r th e r r e p r o d u c tio n p rohibited w ith o u t p e r m is s io n .
4
Kidder et al. (1978) first examined this relationship using coarse-resolution (145km at nadir)
Nimbus-6 Scanning Microwave Spectrometer (SCAMS) 55.45 GHz radiance anomalies (ATb)
and found a moderate, inverse correlation with MSLP anomalies (APsfc). Velden (1982) and
Velden et al. (1984) found an improved correlation between upper tropospheric 250 hPa constant
pressure level temperature anomalies (A T ^ ) and APsfc derived using NOAA/TIROS Microwave
Sounding Unit (MSU) 54.96 GHz radiance data and the TIROS Operational Vertical Sounder
(TOVS) retrieval package on the Man-computer Interactive Data Access System (McIDAS,
Suomi et al., 1983). Several factors were attributed to these improvements in correlation
including increased horizontal resolution (110km at nadir), radiometric accuracy (MSU NEAT of
0.2°C vs. 0.5°C with SCAMS), and utilization of all available MSU temperature sounding
channels (Smith et al., 1979).
Two primary factors make satellite-borne microwave-based TC UTWA intensity
techniques attractive including the lack o f sensitivity to non-precipitating clouds and the direct
physical relationship between the observed upper tropospheric warming and MSLP. The former
issue is very appealing as TCs are often accompanied by extensive cirroform cloud cover that
typically masks underlying tropospheric thermodynamic structure from satellite-based infrared
vertical sounding techniques requiring cloud-free fields of view (FOV). The latter issue enables
development of a statistical regression-based approach provided a statistically-significant sample
o f passive microwave observations/in-situ observations of TC MSLP are available. Once
developed, the regression coefficients derived from the dependent sample or training set can be
used independently to predict TC MSLP using passive microwave UTWA observations in
subsequent cases.
R e p r o d u c e d with p e r m is s io n of t h e cop y rig h t o w n e r. F u r th e r r e p r o d u c tio n prohibited w ith o u t p e r m is s io n .
5
As interest in passive microwave-based TC intensity estimates matured, Velden (1989)
and Velden et al. (1991) attempted to develop a passive microwave, regression-based approach
to TC intensity estimation using an expanded population o f MSU /in-situ observation pairs from
the both Atlantic (ATL) and western North Pacific (NWPAC) ocean basins (1980-1984). In the
ATL basin, standard errors of estimate (SEE) of 8 hPa/13 kts were found (n=108) in comparison
with an SEE of 13.7 hPa/16.7 kts in the NWPAC (n=82). It is important to note, for the purposes
o f this study, that in each case the authors restricted the TC position to fall within 6° in order to
mitigate the negative effects of reduced horizontal resoltuion near the scan limb. In doing so, the
adverse effects o f TC UTWA sub-sampling (and corresponding increases in standard estimate of
error) were minimized; however, such a strategy came with costs the most grievous being a
significant reduction in the total number o f future opportunities in which the dependent
regression equations could be applied in an independent predictive mode. As a result, the overall
operational utility of future MSU-based TC intensity estimation activities, already temporally
disadvantaged in comparison to more frequent intensity estimates available from geostationary
satellite data using the Dvorak (1974, 1984) technique, was significantly reduced.
In an effort to ameliorate the aforementioned TC position constraints, Merrill (1995)
attempted to develop a scheme capable of estimating TC intensity independent o f storm position
relative to scan swath. According to Merrill, TC UTWA sub-sampling is uniquely characterized
by the following factors (illustrated in Fig. 3 and discussed in following sections):
1) TC position (latitude/longitude)
2) Maturity o f the warm core
R e p r o d u c e d with p e r m i s s io n of t h e co p y rig h t o w n er. F u r th e r r e p r o d u c tio n p rohibited w ith o u t p e r m is s io n .
6
3) FOV scan angle (tp) defined by the geocentric latitude and longitude (deg) of
satellite position at the time of TC observation, satellite altitude (km), and the FOV
containing the TC UTWA peak wanning
4) FOV off-axis angle (0) defined by (2) and the TC position displacement
(km) relative to the FOV center containing the TC UTWA peak warming
5) Horizontal scale of the UTWA
any combination o f which varies between successive polar orbiting satellite observations or
between different passive microwave radiometers (i.e., MSU vs. DMSP SSM/T) each with
different effective observation accuracy, horizontal resolution and frequency-dependent antenna
gain patterns. Factors 1-5 are modeled through the explicit convolution of an a-priori analytic
function approximating the horizontal distribution o f a ‘typical’ TC UTWA and the passive
microwave radiometer antenna gain pattern. The former uses an analytic TC MSLP distribution
function patterned after Holland (1980) while the latter uses a model based on standard
principles of diffraction for a uniformly illuminated circular aperture (Ulaby et al., 1981 and
Hecht, 1987). The resulting forward model of TC MSU 54.96 GHz UTWA observations is then
iteratively adjusted using maximum-likelihood regression (Rogers, 1976) until the difference
between forward model radiances and actual passive microwave 55 GHz-region observations are
minimized. At this point, the adjusted amplitude (one of eight free adjustable parameters
defining the a priori horizontal structure function) o f the scan geometry, diffraction-corrected TC
UTWA is ready for regression against in-situ observations of TC MSLP.
R e p r o d u c e d with p e r m i s s io n of t h e co p y rig h t o w n e r. F u r th e r re p r o d u c tio n p ro hibited w ith o u t p e r m is s io n .
The success of this technique requires significant a priori information including an
accurate assessment o f T C position at the time o f observation, precise earth FOV navigation, an
accurate instrument/frequency-dependent antenna gain model and a representative a-priori
estimate of the horizontal scale of the UTWA. In Merrill (1995), the latter issue was particularly
troublesome and eventually proved insurmountable as no reliable microwave-based estimate of
TC UTWA horizontal scale was available for the 1980-1984 MSU data set. As a result, Merrill
resorted to using a fixed climatological eye size (a proxy for TC UTWA horizontal scale) radius
of 24km (Weatherford and Gray, 1988 hereafter referred to as WG88) which led to inferior
results (SEE = 17.8 hPa) when compared to those o f Velden et al. (1991) using the same data set.
1.2 Contemporary Issues
The next-generation Advanced Microwave Sounding Unit (AMSU) was launched on 13
May 1998. The AMSU temperature sounder (hereafter referred to as AMSU-A), like the
previous-generation M SU, is a cross-track, stepped-line scanning total power radiometer with an
instantaneous FOV of 3.3° at the half-power points providing a nominal spatial resolution o f
48km at nadir. The AMSU-A is divided into two separate modules, module A -l consists o f 13
channels and module A -2 two channels. When combined, both modules provide temperature
information from the earth’s surface to the mid-stratosphere (Fig. 4a). A significant
enhancement over the previous-generation MSU is the addition of the AMSU moisture sounder
(hereafter referred to as AMSU-B), a five channel, cross-track, continuous line scanning total
R e p r o d u c e d with p e r m i s s io n of t h e co p y rig h t o w n er. F u r th e r r e p r o d u c tio n prohibited w ith o u t p e r m is s io n .
8
power radiometer with an instantaneous FOV o f 1.1° at the half-power points corresponding to a
horizontal resolution of 16km at nadir. The AMSU-B, when combined with AMSU-A, is
capable of providing thermodynamic profiles (temperature and moisture) from the earth’s
surface to the upper-troposphere (Fig. 4b).
Several features of the AMSU offer potential improvements to microwave-based TC
intensity estimation techniques including improved horizontal resolution (Fig. 5), increased
vertical resolution, and the addition of moisture sounding capabilities (Kidder et al., 2000). As
mentioned in Section 1.1, the relatively coarse horizontal resolution of the MSU leads to TC
UTWA sub-sampling. Fig. 6 illustrates the effect o f improved AMSU-A horizontal resolution
and radiometric accuracy vs. MSU at similar observation frequencies (54.94 GHz vs. 54.96
GHz) for Super Typhoon Zeb, 13 October 1998. The AMSU-A resolves nearly 2°C of additional
warming (Tb=39.1°C) compared to the MSU (Tb=40.1°C) under similar scan geometry conditions
(i.e., TC position between nadir and limb of scan) and observing times (2326UTC vs.
1836UTC). A comparison of the AMSU-A TC UTWA and the 14 October 1998 0032UTC
GMS infrared also effectively demonstrates the retention and concentration of the TC UTWA by
the eye wall.
1.2.1 Scattering
Ideally, one would like to retrieve TC MSLP through integration o f the hypsometric
equation using AMSU-derived vertical temperature and moisture profiles in the eye. Based on
the improved vertical and horizontal resolution o f the AMSU-A, including the availability of
R e p r o d u c e d with p e r m i s s io n of t h e cop y rig h t o w n e r. F u r th e r r e p r o d u c tio n prohibited w ith o u t p e rm is s io n .
9
AMSU-B moisture sounding capabilities, the possibility of a hydrostatic TC MSLP approach is
once again revisited. Vertical temperature cross sections of Hurricane Floyd, 14 September 1999
1238UTC, derived using the NOAA/NESDIS Advanced TIROS Vertical Sounding (ATOVS)
package (Goldberg et al., 1999) are strikingly similar to historical composite analyses of
Hurricane Inez (Hawkins and Imbembo, 1966) of similar intensity (930 hPa vs. 929 hPa) using
in-situ aircraft and ground-based upper air observations (Fig. 7). Unfortunately, the AMSU-A
lower tropospheric channels (Channel 5 (53.6 GHz) and Channel 6 (54.4 GHz)) from which
lower tropospheric temperature are derived are strongly attenuated by large, mixed-phased
hydrometeor scattering below the melting level. While aircraft/radiosonde composites do
indicate weak lower tropospheric cooling in the proximity of the eye wall region, the size and
magnitude of cooling observed in the AMSU-A-derived temperature profiles is anomalously
large (AT in excess of -10°C).
Left uncorrected (the current form of the NOAA/NESDIS ATOVS retrieval algorithm
does not correct for the effects o f scattering), the unrealistically large (AT in excess of -10°C)
lower tropospheric temperature anomalies may lead to significant MSLP estimate errors using a
vertically-integrated hypsometric approach. The 14 September 1999 1238UTC Hurricane Floyd
example represents a special case in which scattering effects are minimized owing to the large
eye size reported by reconnaissance aircraft (approx. 73km diameter, 14 September 1999
1113UTC reconnaissance observation). Under these circumstances, the horizontal scale of
Floyd’s UTWA was well within the limits o f AMSU-A resolving limits (48km at nadir) to fully
resolve peak warming. Furthermore, eye wall convection was sufficiently displaced from the
R e p r o d u c e d with p e r m i s s io n of t h e co p y rig h t o w n er. F u r th e r r e p r o d u c tio n p rohibited w ith o u t p e r m is s io n .
10
surface circulation center such that the effects of scattering on the lower column radiances were
minimized.
A subsequent independent hydrostatic MSLP estimate o f 936 hPa using the 1238 UTC 14
September 1999 Hurricane Floyd AMSU-A-derived UTWA (Simpson, 2000) was in excellent
agreement with reconnaissance observations (930 hPa); however, subsequent tests have shown a
vertical compensatory effect (i.e., enhanced lower-tropospheric cooling concomitant with
enhancing the UTWA magnitude for the same atmospheric column) suggesting that AMSU-Aderived TC UTWA are not independent, or easily extricable, from lower-tropospheric
hydrometeor scattering effects. As a result, a single channel, AMSU-A Channel 7 (54.94 GHz)
approach is pursued due to the dominant contribution of TC UTWA signal and the relative
atmospheric opacity to surface emission, reflection and lower tropospheric thermal structure (and
scattering) at frequencies near 55 GHz.
1.2.2 Scan Geometry
As discussed in Section 1.2 and illustrated in Fig. 5, issues o f AMSU-A scan angledependent horizontal resolution and TC UTWA remain despite improved horizontal resolution.
By virtue of its through-nadir scanning characteristics, AMSU-A horizontal resolution varies as a
function of off-nadir scan angle <{), sampling progressively larger atmospheric emission kernels
defined by the product of a frequency-dependent weighting function and individual AMSU-A
FOV area represented by Wk(z) FLj(x,y) (pg. 27). As a result, AMSU-A horizontal resolution at
({>=48.3° (maximum scan angle) approaches that of the previous-generation MSU (-1 10km).
R e p r o d u c e d with p e r m i s s io n of t h e cop y rig h t o w n e r. F u r th e r r e p r o d u c tio n prohibited w ith o u t p e rm is s io n .
11
Furthermore, since AMSU-A tropospheric weighting functions overlap, upper-tropospheric
radiances can be influenced by absorption, emission, reflection and scattering elsewhere within
the atmospheric column. As <{>increases, so too does the potential for environmental factors
outside the warm core (i.e., hydrometeor scattering, cooler outer-core atmospheric thermal
structure, etc.) to dilute warm core physical attributes responsible for TC MSLP. If left
unaccounted, the resulting ‘volumetric smoothing’ or ‘beam filling’ can reduce the magnitude o f
the UTWA resolved by AMSU-A and lead to increased variance in statistical AMSU-A-derived
TC UTWAs/MSLP techniques.
The nature and degree of scan-angle induced AMSU-A FOV ‘beam filling’ depends on
several factors most notably the maturity of the TC and the horizontal scale of the UTWA. For
nascent or weak TCs, the lack of a partial eye wall and low mid-to-upper tropospheric inertial
stability enables radial transport of energy. As a result, the horizontal size of the UTWA is broad
and therefore less susceptible to sub-sampling regardless of TC position with respect to either
individual AMSU-A FOV or the entire scan swath. Under these circumstances, the effective
accuracy of the AM SU-A in measuring the TC UTWA magnitude is high. Such is not the case
for moderate to strong TCs (i.e., Tropical Storms with sustained 10 minute average wind speeds
>=34 kts). Concomitant with the primary circulation intensification and the formation of an eye
wall, the horizontal scale of upper tropospheric warming contracts to sub-AMSU-A horizontal
resolutions increasing the likelihood for sub-sampling. Under these circumstances, the position
of the TC relative to the individual AMSU-A FOV center containing peak warming (9) and
location of the UTWA relative to the scan swath (<()) become increasingly relevant. As
mentioned earlier, the effects o f scan geometry (c|>,0) are mitigated for large TCs like Hurricane
R e p r o d u c e d with p e r m i s s io n of t h e co p y rig h t o w n e r. F u r th e r r e p r o d u c tio n prohibited w ith o u t p e r m is s io n .
12
Floyd; however, TC eye sizes and their associated UTWAs are typically much more compact as
was the case for Hurricane Opal (1995) in the Gulf of Mexico. Opal’s eye was reported as small
as 15km in diameter, thus the corresponding UTWA would have been significantly sub sampled
had AMSU-A observations been available.
1.2.3 Diffraction
Another equally important factor affecting the AMSU-A ability to quantify a TC UTWA
is the issue of diffraction. The width and shape of individual AMSU-A FOVs are uniquely
prescribed by the wavelength of radiation being sensed and the diameter o f the instrument
boresight aperture through which upwelling terrestrial microwave radiation must pass.
According to Fig. 8, the size of individual AMSU-A FOVs (for a given frequency, not
considering the effects of off-nadir scan angle illustrated in Fig. 5) is prescribed by the angle (<|))
at which the initial incident radiation intensity I (<j),<|) = 0) is reduced by a factor of one half (or 3
dBZ). For example, if we consider a TC UTWA intensity value o f unit magnitude displaced
from the center of an AMSU-A FOV by some distance ‘D,’ the effects of diffraction will reduce
the initial value I (<f>,(j>= 0) to 0.8. Furthermore, the magnitude of the diffractive contribution to
AMSU-A TC UTWA sub-sampling, given a displacement ‘D’ from the FOV center, will vary
depending on both AMSU-A FOV scan angle from nadir and off-axis scan angle (0). Therefore,
knowledge of TC position relative to both the AMSU-A scan swath and the FOV containing the
maximum upper-tropospheric warming are vital components to any scheme designed to
explicitly account for UTWA sub-sampling.
R e p r o d u c e d with p e r m i s s io n of t h e co p y rig h t o w n e r. F u r th e r r e p r o d u c tio n prohibited w ith o u t p e rm is s io n .
13
Warm core
Latent heat energy released
Rotation helps retain warm core aloft
Latent/sensible heat flax
Figure 1. A schematic diagram depicting a tropical cyclone warm core. Wind-driven latent
and sensible heat flux (1) is acquired at the atmosphere/ocean interface where it is transported,
released and deposited (2) in the upper troposphere by undiluted deep cumulus updrafts within
the eye wall region. Subsidence and high inertial stability within the inner eye wall region (3)
retain peak mid-upper tropospheric heating above the low-level circulation center helping to
further reduce minimum sea level pressure.
R e p r o d u c e d with p e r m i s s io n of t h e co p y rig h t o w n er. F u r th e r r e p r o d u c tio n prohibited w ith o u t p e r m is s io n .
14
eci
70
80
100
126
-2
2—
150
PRESSbR=, mb
175
?oo
250
300
12,
400
600
600
700
1000
400
300
200
IN ORTH}
<2>.
100
100
200
300
400
HAOfOS. <m
Figure 2. Vertical warm core cross section, Hurricane Inez, 28 September 1966, derived
from composite aircraft reconnaissance and upper air observations (Hawkins and
Imbembo, 1967)
R e p r o d u c e d with p e r m i s s io n of t h e co p y rig h t o w n e r. F u r th e r r e p r o d u c tio n prohibited w ith o u t p e r m is s io n .
15
Anom aly
Displacem ert
Figure 3. A graphical depiction of weak (8a) and strong (8b) TC UTWAs interacting with
the AMSU-A antenna gain pattern as a function o f TC position relative to scan swath (0
=0 at nadir). For a weak (strong) TC, the UTWA is diffuse (concentrated) and the ability
of AMSU-A to resolve peak upper tropospheric warming is unaffected (strongly affected)
by both storm position relative to individual FOV and variable nadir-limb FOV size.
R e p r o d u c e d with p e r m i s s io n of t h e co p y rig h t o w n e r. F u r th e r r e p r o d u c tio n prohibited w ith o u t p e r m is s io n .
16
20
80
Height (km )
60
C14
C13
C12
Nadir
Gear Sky
CIO
Cll
40
AMSU-B
Nadir
Clear Sky
Water Cloud
AMSU-A1
70
£ 12
C3
C8
C7
30
C4
10
0.0
C5
C15
0.1
0.15
0.2
0.05
W eighting Function (k m 1)
C2
Cl-^
0.0
0.05
oj
0.15
Weighting Function (k m 1)
Figure 4. AMSU-A and AMSU-B weighting functions. Source: Kidder et al., 2000
R e p r o d u c e d with p e r m i s s io n of t h e co p y rig h t o w n er. F u r th e r r e p r o d u c tio n prohibited w ith o u t p e r m is s io n .
03
17
•500
CROSS-TRRCK DISTRNCE (km)
500
Figure 5. Comparison of AMSU vs. MSU FOV size and coverage. The filled gray ellipses
illustrate the 110 km resolution (nadir) o f the MSU. The black outlined ellipses illustrate the
improved horizontal resolution of the AMSU-A instrument (48 k m at nadir). The black dots
represent sampling locations of the AMSU-B instrument Source: Kidder et. al (2000)
R e p r o d u c e d with p e r m i s s io n of t h e co p y rig h t o w n er. F u r th e r r e p r o d u c tio n prohibited w ith o u t p e r m is s io n .
18
•/ _
4V^*:Jftfl
^ ,I
v*
y a fc fc r
;r
It#
SATELLITE: NOAA-14
SENSOR: MSU C h*nd 3 (55GHe)
DATE/TIME: 130CT981836UTC
MAX TEMP: -4Q9C
z j i r j :i z h
r
: h i _ l czb
1n
120E
STORM TRACK
It#
129
*Jz
>ii
'.J%
IjJ
liu
SATELLITE: NOAA-15
SENSOR: AMSU C hand 7 (55GHz:)
DATE/TIME: 130CT98232SUTC
MAX TEMP: -39.IC
Super Typhoon Zeb (1998)
V l>n
CATS
Figure 6. A comparison o f MSU 54.96 GHz vs. AMSU-A 54.94 GHz upper
tropospheric warm anomaly resolving capabilities, Super Typhoon Zeb, 13 October
1998: (A) GMS-5 Infrared Image 14 Oct 98 00:30UTC, (B) NOAA-14 MSU Ch.3
Tb (°C), (C) NOAA-15 AM SU Ch. 7 Tb (°C), and (D) Zeb storm and intensity track
(courtesy of UW-CIMSS)
R e p r o d u c e d with p e r m i s s io n of t h e co p y rig h t o w n er. F u r th e r r e p r o d u c tio n prohibited w ith o u t p e r m is s io n .
19
tfcrfTfcan* Ftoyrf 14 S*ptc<nbar 190S 123SUTC
AMSU-A Derived T eraperattnt Anomcfy (Storm Center—Erviroftmeflt)
Contour fcttervol • 2R
H u r r ir n n * F tn x A ,t>nr U
-660000
—
82-5000
-79.0000
-75^000
Lort9%ude
-72-0000
-6a5000
Figure 7. NOAA-15 AMSU-A vertical temperature anomaly (°C), Hurricane Floyd,
1238 UTC 14 September 1999 (right). Floyd’s warm core is clearly evident with
temperature anomalies in excess of 18°C near 250hPa. Strong cooling (>-16°C) in
the convective eyewall region is fictitious and due to the lack of explicit correction
for scattering in the sounding retrieval package. AMSU-A 55.5 GHz, 54.94 GHz,
54.4 GHz and 53.6 GHz limb corrected brightness temperatures (Tb, top to bottom)
with corresponding peak radiance sampling levels (horizontal lines) are shown with
AVHRR inset (left).
R e p r o d u c e d with p e r m i s s io n of t h e co p y rig h t o w n e r. F u r th e r r e p r o d u c tio n prohibited w ith o u t p e r m is s io n .
20
1.1
0.9
or
as
at
o
^ =0
AMSU-A off-axis antenna scan angle (<f>)
Figure 8. Bessel function (first kind, order one) model of the diffraction pattern for
the uniformly illuminated AMSU-A aperture a t frequency 54.94 GHz. I(<|>)
describes intensity as a function of off-axis anCenna scan angle (<j>). D represents
the displacement of the storm center (and maximum warming) from AMSU-A
FOV center.
R e p r o d u c e d with p e r m i s s io n of t h e cop y rig h t o w n e r. F u r th e r r e p r o d u c tio n prohibited w ith o u t p e rm is s io n .
21
2 D ata
Global AMSU-A/B multi-spectral earth observation radiance data are stored on board the
satellite and downlinked via high-bandwidth NOAA/NESDIS Command and Data Acquisition
Stations (CDAS) in Fairbanks, Alaska and Wallops Island, Virginia, U.S.A. After a series of
internal quality control checks, AMSU-A/B data is forwarded by the CDAS to the Cooperative
Institute for Meteorological Satellite Studies (CIMSS) located within the University of
Wisconsin-Madison Space Science and Engineering Center (SSEC) via the General
Electric/RCA geostationary domestic telecommunications satellite (DOMSAT) stationed at 0°N
69°W. The time delay between earth observation and CIMSS receipt is typically within 0.5-2.5
hours from time of satellite earth observation. In extreme cases, the time delay may exceed 3-4
hours depending on observation time and DOMSAT bandwidth availability and NOAA/NESDIS
processing constraints.
Assembly of the 1999 dependent data set required manual identification and retrieval of
NOAA-15 AMSU-A/B orbits providing TC coverage. Orbital coverage (near-real time and post­
event) and local overpass time was determined using core McIDAS navigation and display
routines. Near-real tim e satellite orbital ephemeris data used to predict NOAA-15 TC coverage
was provided by NOAA/NESDIS. Relevant AMSU-A/B orbit files were then acquired from
either a CIMSS limited (one week), continuously refreshed on-line archive or restored from 8mm
magnetic media. For the automated independent 2000 AMSU TC intensity test, this process
was automated and is discussed in detail in Section 4. While AMSU-A/B data was available for
August-December 1998, on-board radio frequency interference (later characterized and
R e p r o d u c e d with p e r m i s s io n of t h e co p y rig h t o w n e r. F u r th e r r e p r o d u c tio n prohibited w ith o u t p e r m is s io n .
22
corrected) from a search and rescue (S AR) transponder contaminated AMSU-B radiance data
prohibiting its use.
ATL basin aircraft TC reconnaissance observations used to calibrate the 1999 dependent
data set AMSU-A 54.94 GHz TC UTWA vs. MSLP regression coefficients, and later to validate
the independent AMSU TC MSLP estimates in 2000, were provided by scheduled United States
Air Force Reserve (USAFR) WC-130 and NOAA P-3 reconnaissance flights. A sample aircraft
vortex data message (WMO bulletin header URNT KNHC) is provided in Fig. 9. Vortex data
messages were obtained through several sources including McIDAS, the National Hurricane
Center (NHC), and the Central Atlantic Storm Investigators (CASI) world wide web (WWW)
text weather archive. A limited number o f surface synoptic, buoy and ship observations o f TC
passage, particularly in the data-sparse NWPAC and Southern Pacific Ocean (SPAC, year 2001),
were provided by the Australian Bureau o f Meteorology, collected in near-real time from the US
AF Weather Information Network (AFWTN) WWW site as well as by NHC/JTWC forecast
discussion bulletins (when available).
All potential in situ observation candidates were scrutinized based on the time of
observation relative to AMSU observation time and proximity to the best estimate of TC
position. In general, only those aircraft/AMSU ‘match pairs’ that occurred within +/- 6 hrs and
for which the TC was not undergoing significant intensity change were considered. For cases
where TCs encountered in situ surface synoptic observation locations (independent test only),
special care was taken to consider only those ‘match pairs’ in which TC eyes came within 30km
of the reporting site and also during periods in which the TC was not undergoing significant
intensity change due to interaction with land, wind shear, etc.. Under these circumstances,
R e p r o d u c e d with p e r m i s s io n of t h e co p y rig h t o w n e r. F u r th e r r e p r o d u c tio n prohibited w ith o u t p e r m is s io n .
23
time constraints were increased to +/- 3 hrs.
R e p r o d u c e d with p e r m i s s io n of t h e co p y rig h t o w n e r. F u r th e r r e p r o d u c tio n prohibited w ith o u t p e r m is s io n .
24
- ~ /- .i': < ty < t >
“ rv.--vi.—■^ v ^ w ^ ^ j W v F v y i l f i e
E^Q50:DEG?80 N M ^
^
Figure 9. Sample vortex data message. A) Time o f observation, B) TC position, C)
flight level (hPa/m), D) maximum sustained winds (kts), E) bearing and range from
the center of the maximum surface wind, F) maximum flight level wind near storm
center, G) bearing and range from the maximum flight level wind, H) minimum sea
level pressure extrapolated from flight level, I) maximum temperature/pressure
altitude outside eye, J) maximum temperature/pressure altitude inside eye, K)
dewpoint temperature/sea surface temperature inside the eye, L) eye character, M)
eye shape (C=concentric, 25=25nm), N ) position fix code (l=penetration, 2=radar,
3=wind, 4=pressure, 5=temperature; 12345/7=all methods used at 700hPa flight
level), 0 ) navigation fix/meteorological accuracy, P) remarks section.
Source: National Hurricane Center
R e p r o d u c e d with p e r m i s s io n of t h e co p y rig h t o w n er. F u r th e r r e p r o d u c tio n prohibited w ith o u t p e r m is s io n .
25
3 Methodology
3.1 Forward Model
As discussed in previous sections, the AMSU-A ability to quantify TC UTWAs is limited
by improved, albeit relatively coarse, vertical and horizontal resolution, variable horizontal
resolution by virtue of its scan geometry, diffraction, and hydrometeor scattering. Complicating
matters further is the fact that the degree in which these factors contribute to TC UTWA sub­
sampling is largely a function o f TC horizontal size. The cumulative effects, when combined
with the ephemeral nature of TC position within the AMSU-A FOV scan swath between
successive satellite passes are non-thermodynamic, instrument scan geometry-based temporal
variations in TC UTWA unrelated to real changes in TC warm core structure and intensity.
Furthermore, differences in satellite (i.e., NOAA MSU/AMSU, DMSP SSM/T-1, etc.) passive
microwave radiometer horizontal resolution and radiometric accuracy will also yield unique TC
UTWA observations whose differences are also unrelated to real changes in TC warm core
structure and storm intensity. In summary, the number and complexity of the aforementioned
issues limit the effective accuracy o f any single passive microwave-based intensity estimation
technique using multiple satellite sensor observations of TC UTWA unless explicitly treated.
Merrill (1995) first formalized the conceptual framework surrounding this issue and
proposed a technique through which the effects of TC UTWA sub-sampling, diffraction and
satellite-sensor dependent differences in horizontal resolution and radiometric accuracy could be
minimized. According to M errill’s theory, the AMSU-A-observed TC upper tropospheric
radiance Bij kobs for the i-th row j-th column FOV of frequency k can be expressed as
R e p r o d u c e d with p e r m i s s io n of t h e co p y rig h t o w n er. F u r th e r r e p r o d u c tio n prohibited w ith o u t p e r m is s io n .
26
B y / " = I / J Wk(z)T(x,y,z)Fj.j(x,y)dV
(1)
V
provided large water or mixed-phase hydrometeors are absent and that atmosphere is sufficiently
opaque at frequency k so that surface reflection and emission can be considered negligible. Wk(z)
is the weighting function for AMSU-A frequency k, T(x,y,z) is the TC three-dimensional
temperature structure, and Fj j(x,y) is the projection of the AMSU-A antenna gain pattern for
FOV(i,j) on the earth’s surface. In order to specify Fy(x,y), accurate navigation o f the NOAA-15
geocentric satellite nadir at the time ( t ) of observation (a function of TC position and satellite
coverage on any given day) is required. Such a task, while complicated, is possible with a fairly
high degree of precision using observed or predicted specified satellite orbital ephemeris data
(i.e., inclination, eccentricity, mean anomaly, argument of the perigee, right ascension of the
ascending node, and semi-major axis), a time differential (At = t - 10 where t0is the epoch time),
and a Keplerian elliptical orbit prediction model (Kidder et al., 1995) such as the BrouwerLyddane (BROLYD) orbit prediction algorithm routinely used by NOAA/NESDIS.
Wk(z) and Fj j(x,y) are assumed to be normalized over the depth of the atmosphere and a
sufficiently large area so that their integrals are approximately equal to one. For the purpose of
this study, a single frequency form o f (1) can be written as
B,/*5 = JJ B*(x,y)FiJ-(x,y)dA
A
R e p r o d u c e d with p e r m i s s io n of t h e co p y rig h t o w n er. F u r th e r r e p r o d u c tio n prohibited w ith o u t p e r m is s io n .
(2 )
27
where Bjjobs is the observed brightness temperature in A M S U -A channel k a nd T(x,y,z) is
replaced by B*(x,y) representing the Planck function emission to space based o n the horizontal
thermal structure of the tropical cyclone of interest.
The form of the TC UTWA adopted for use in this study is based on the work of
Schloemer (1954) and Holland (1980) and is assumed to mirror (i.e., negatively correlated) the
sea level pressure distribution through storm-scale hydrostatic assumptions previously discussed.
In general, the ratio of differences between surface pressure p at radius r and T C central pressure
pc versus ambient pressure (i.e., the value of the first anticyclonically-curved isobar) pn and pc
(P-P c)/(P n-P c)
(3)
describes a family of hyperbolas which can be approximated by
r V f (pn - pc) / (p —pc) ] = A
(4)
where A and B specify the location relative to storm position and profile shape which can be
determined by either tropical cyclone observations or climatology. Taking the antilogarithm of
(4) and rearranging terms yields
P
=
Pn
+ pc(l - [exp(-A /r8 )])
R e p r o d u c e d with p e r m i s s io n of t h e co p y rig h t o w n er. F u r th e r r e p r o d u c tio n prohibited w ith o u t p e r m is s io n .
(5)
28
relating surface pressure p to tropical cyclone central pc based on radial distance r and ambient
pressure pn (r —
representing the undisturbed environment. Recalling our earlier assertion
that the TC UTWA (or AMSU-A 54.94 GHz Tb distribution) near 250 hPa should mirror the
surface pressure distribution, we can write
B‘(x,y) = Benv + AB {l-exp[(-R/r) *] }+8B/3x(x - xc)+8B/9y(y - yc)
(a)
(b)
(c)
(6)
(d)
which is functionally equivalent to (5) describing the amplitude of the AMSU-A 54.94 GHz
emission based on position (x,y) relative to the tropical cyclone center (xc, yc). Term (6a)
represents the mean environmental contribution, (6b) a distance-weighted anomaly contribution,
and (6c,d) contributions based on latitudinal and meridional radiance derivatives. R is
considered to be the horizontal scale (km) o f the warm anomaly, distance from center r = [ (x —
xc)2 + (y - yc)2 ]l/2 (km), J a dimensionless storm shape parameter (aspect ratio). Substitution o f
(6) into (2) yields
B ,/Wd= I J [ Benv + AB {l-exp[(-R/r)T }+3B/3x(x - xc)+9B/ay(y - yc)] Fij(x,y)dA
(7)
A
representing a forward model of AMSU-A 54.94 GHz Tb observations for the FOV
corresponding to geographic location (x,y) based on frequency, storm horizontal thermal
structure, and antenna gain characteristics. The RHS of (6) constitutes the horizontal structure
function X = f[BenV, AB, R, y, 9B/8x, dB/dy, xc, yc ] uniquely characterizing the TC UTWA.
R e p r o d u c e d with p e r m i s s io n of t h e co p y rig h t o w n er. F u r th e r r e p r o d u c tio n prohibited w ith o u t p e r m is s io n .
29
3.1.1 Maximum Likelihood Regression
The experimental Tropical Cyclone Retrieval (XTCR) retrieval package (Merrill, 1994),
modified for use in this study, is patterned after that of Rodgers (1976). The technique seeks to
iteratively minimize the residual differences between a forward model and observations leading
to a maximum likelihood solution. Based on the discussion o f the previous section, if an priori
TC UTWA horizontal structure function X is accurately prescribed, it is possible to forward
calculate observed radiances Bi jfwd (and Tb) for each AMSU-A FOV(i,j) through convolution of
X and the antenna gain pattern F{j-(x,y). An iterative method that seeks to minimize the residual
between Bj/Wdand B ^
X ^ , = XCON+
+ Syy)-‘(Bobs -
- Km(XCON- X J )
(8)
will yield a maximum likelihood estimate o f X given uncertainties in reported storm position,
size, shape, instrument noise and satellite navigation. Xmis the m-th iterated storm structure,
XCON the constraint (a priori) storm structure, SxxCONthe constraint structure error covariance,
the forward model sensitivity to changes in the structure vector (i.e., d Bmfwd/9Xm), and S yy the
AMSU-A observation error covariance. Based on (8), the maximum likelihood solution for the
storm structure XRET (Xm, m—><») is achieved when the differences between the observed AMSUA radiances Bobs and forward model radiances Bmfwd and the forward model sensitivity-weighted
difference between the constraint storm structure XC0N and m-th iteration structure Xmare the
same.
R e p r o d u c e d with p e r m i s s io n of t h e co p y rig h t o w n e r. F u r th e r r e p r o d u c tio n prohibited w ith o u t p e r m is s io n .
30
The ability of the retrieval algorithm (8) to resolve X depends highly on the accuracy of
the constraint storm structure XCON. The solution storm structure XRETwill always fall between
XC0Nand the value of X that best produces the AMSU-A observed radiance distribution, Bobs.
Whether the AMSU-A observation or constraint is favored depends upon the expected constraint
structure error covariance SxxCON, the expected observation variance (noise) Syy, and the
sensitivity of the forward model to changes in the storm structure K,„. S„CON —>0 implies that
XC°Nis a good estimate of the storm structure and the solution will favor the constraint with zero
variance. On the other hand, SxxCON —
implies X001^ is a poor estimate of the storm structure
and the solution tends toward the inversion of the forward model. Syyand K,,, are coupled in the
sense that accurate AMSU-A observations (i.e., Syy —>0) or large forward model sensitivities
(i.e., favorable viewing geometry—^satellite nadir, FOV centered) cause the solution to favor the
solution of the forward model, whereas Syy —>oo or small
(i.e., unfavorable view geometry —>
satellite limb, non-centered) favor the constraint XCON.
3.1.2 TC UTWA Horizontal Scale
The XTCR technique is improved through the use of AMSU-B 89.0 GHz moisture
sounder radiance to define the TC eye size and the constraint structure function size parameter R.
AMSU-B 89.0 GHz channel radiance data provide useful information on the formation and size
of the TC eye due to its sensitivity to scattering by large liquid and mixed-phased hydrometeors
in the eye wall (Kidder et al., 2000). An example of a typical AMSU-B 89.0 GHz Tb distribution
R e p r o d u c e d with p e r m i s s io n of t h e co p y rig h t o w n e r. F u r th e r r e p r o d u c tio n prohibited w ith o u t p e r m is s io n .
31
for a mature TC is provided in Fig. 10. The initial failure of XTCR to improve upon the results
o f Velden (MSLP RMS error of 17.8 hPa vs. 13.7 hPa, n=79) was largely attributed to using a
fixed climatological eye size value (R=24km) due to a lack of independent satellite-based eye
size information. When the sample size was reduced to include only those storms with reported
eyes (n=53), the XTCR-retrieved MSLP estimates were superior to raw estimates in which no
corrections for scan geometry or diffraction were made.
Accurate specification of R is critical to retrieval performance as it defines the horizontal
distribution of maximum upper tropospheric warming and accuracy of the constraint storm
structure XCON. In this study, R is estimated by fitting an interpolating polynomial, F, based on
the radial distribution of AMSU-B 89.0 GHz radiance data as depicted in Fig. 11. Assuming that
AMSU-B 89.0 GHz radiances are significantly reduced due to scattering within the convective
eye wall region, the radial distance r that satisfies
32F(r) / ar2 = 0
(9)
will roughly coincide with the eye wall. Individual R values corresponding to each radial
direction (N, S, E and W in AMSU-A line/element space) are calculated from which a mean
value Rroean is estimated. The final R value is then chosen using only those individual R values
falling within one standard deviation (lor) o f RmeanDespite the relatively coarse resolution o f the AMSU-B (16 km at nadir) and the
simplicity of the technique, mean AMSU-B 89.0 GHz-derived R values for 1999 ATL TCs
(n=22) were found to be within 7km of reconnaissance reports. The overall mean R value of
R e p r o d u c e d with p e r m i s s io n of t h e co p y rig h t o w n e r. F u r th e r r e p r o d u c tio n prohibited w ith o u t p e r m is s io n .
32
27.2km is also in good agreement with the WG88 24km mean climatological value. The current
form of (9) is limited to cases in which at least a partial eye wall is present. A minimum of two
radial R values is required; if this condition is not satisfied, a maximum value o f R=65.2 km is
assigned (based on the algorithm numerics) and retrieval performance is degraded. The form of
(9) works for both strong and weak TCs and their accompanying differences in AMSU-B 89.0
GHz radiance signature. For strong TCs with either partial or fully-developed eye walls, the
AMSU-B 89.0 GHz FOV corresponding to TC center will be warm relative to neighboring cold
FOVs as a result of the combined effects of microwave emission and reflection from surrounding
cloud features. For weak TCs, or for systems with exposed low level circulation centers, the
spatial Tb distribution reverses and the FOV corresponding to TC center will appear cold relative
to neighboring FOV (Fig. 12). Cloud-free AMSU-B 89.0 GHz FOV (eye) will appear cold due
to the relatively low emissivity (0.4-0.5) o f the ocean’s surface; however, the combined effects
of elevated marine boundary layer equivalent potential temperature (0C) and small liquid-phase
hydrometeors characteristic of low-level stratoform clouds will serve to increase lower
tropospheric optical depth and the atmospheric contribution to the total emission observed by the
AMSU-B 89.0 GHz channel.
3.2 Dependent Data Set
Once the 1999 AMSU-A/B and in-situ observation match pairs were identified (based on
criteria specified in Section 2), the maximum likelihood regression algorithm was ready for use.
In each case, individual mean a priori structure function constraint values XCON and standard
R e p r o d u c e d with p e r m i s s io n of t h e cop y rig h t o w n e r. F u r th e r r e p ro d u c tio n prohibited w ith o u t p e rm is s io n .
33
deviations SxCONwere defined using values proposed by Merrill (1995) with the exception o f the
size parameter R (Table 1). Two separate tests for each case (n=22) were performed, one in
which R was fixed (Rfix) using the climatological value of 24km, the other using a variable form
o f R (R ^) defined using the two-dimensional AMSU-B 89.0 GHz Tb analysis algorithm (Eqn. 9)
discussed in the previous section. TC location was initially specified using published
NHC/JTWC position estimates. Each position estimate was subjectively analyzed and adjusted
to compensate for storm translation during the time interval between initial position estimates
and the time of AMSU observation. The final choice o f TC position was determined by a
combination o f consistency with estimated translation direction and speed, evaluation o f AMSUB 89.0 GHz radiance signature, and inclusion of the warmest AMSU-A 54.94 GHz Tb pixel
within the center 9 FOV cross-hair structure pattern (Fig. 13) assuming a near-axisymmetric
warm core.
Retrieval application results for 1999 (n=22) are provided in Fig. 14. The first test (red)
compares raw AMSU-A 54.94 GHz TC UTWA estimates with in situ MSLP estimates meeting
the aforementioned geographic and temporal constraints. The second test (green) compares
retrieved AMSU-A 54.94 GHz TC UTWA estimates with in situ MSLP estimates using Rfix,
while the third test (blue) compares retrieved AMSU-A 54.94 GHz TC UTWA estimates with insitu MSLP estimates using Rvar. The solid, color-coded lines represent linear least squares fits to
each sample with corresponding R2 values indicated in the legend. The horizontal structure
retrieval algorithm significandy improves the correlation between the AMSU-A-derived TC
UTWA vs. in situ MSLP estimates. Prior to applying the horizontal structure retrieval, AMSUA 54.94 GHz TC UTWA are found to be moderately correlated (R2 = 0.70)
R e p r o d u c e d with p e r m i s s io n of t h e co p y rig h t o w n e r. F u r th e r r e p r o d u c tio n prohibited w ith o u t p e r m is s io n .
34
with in situ MSLP estimates. However, after the retrieval is applied, the degree of correlation
improves (R2=0.82 for Rfix and R2=0.90 for R ^). The relatively small difference in retrieval
performance using R ^ vs. Rflx likely reflects the overall ssimilarity in average Rvar values
compared to the WG88 24km mean climatological value .
The results of the dependent study, namely the im proved correlation between AMSU-A
54.94GHz-derived TC UTWA estimates and MSLP afterr retrieval application, strongly supports
Merrill’s earlier hypothesis that the peak TC upper tropospheric warming will be sub sampled by
coarse resolution passive microwave radiometers. Furtherm ore, the retrieval effectively
increases the magnitude o f the TC UTWA concomitant vwith improvements in correlation which
is consistent with theory and supports the overall soundness of the methodology. Based on the
apparent success of the dependent test, the regression eqmations were then applied to an
independent sample, discussed in Section 4.
R e p r o d u c e d with p e r m i s s io n of t h e cop y rig h t o w n e r. F u r th e r r e p r o d u c tio n prohibited w ith o u t p e rm is s io n .
35
TYPHOON SAOMAI
AMSU—B Channel 16
Monday 11aep002S5
Figure 10. Color-enhanced AMSU-B 89.0 GHz Tb (°C) image of a mature tropical cyclone
R e p r o d u c e d with p e r m i s s io n of t h e cop y rig h t o w n e r. F u r th e r r e p ro d u c tio n prohibited w ith o u t p e rm is s io n .
36
-82
-5 0
-7 8
—
78
74
-72
-70
R
Figure 11. Schematic diagram illustrating
tropical cyclone eye size determination using
radial analysis of AMSU-B 89.0 GHz Tb data
for Hurricane Floyd, 14 September 1999 1238
UTC.
R e p r o d u c e d with p e r m i s s io n of t h e co p y rig h t o w n er. F u r th e r r e p r o d u c tio n prohibited w ith o u t p e r m is s io n .
37
TROPICAL STORM 33W
AMSU—B Channel i 6 (89 GHz) B rightness Tem perature (C)
Wednesday 29nov0033+T im e: 1039 UTC
12+
126
128
130
132
13 +
136
6 .9
-
3 .1
—13.1
-
23.1
-3 3 .1
—+3.1
-
5 3.1
-6 3 .1
-
73 .1
-8 3 .1
-9 3 .1
Figure 12. Color-enhanced AMSU-B 89.0 GHz Tb (°C) image with corresponding GMS
infrared image (inset).
R e p r o d u c e d with p e r m i s s io n of t h e co p y rig h t o w n er. F u r th e r r e p r o d u c tio n p rohibited w ith o u t p e r m is s io n .
38
Index
[i]
1
2
.5
4
5
6
7
8
Structure Function
X[i] (units)
AB [ In (K)]
B env (K)
R (km)
y (none)
dB/dx [ K (1000 km )1]
dB/dy [ K (1000 km )1]
yr ( latitude in deg)
xc (longitude in deg)
Constraint mean
XCON[i]
2.37
228.24
AMSU-B 89.0 GHz
0.874
-0.25
0.48
Variable
Variable
Constraint standard
Deviation SxCON[i]
0.53
1.73
9.0
0.122
0.54
0.48
0.20
0.20
Table 1. AMSU retrieval constraint values and standard deviations. Adapted from
Merrill (1995)
R e p r o d u c e d with p e r m i s s io n of t h e co p y rig h t o w n e r. F u r th e r r e p r o d u c tio n prohibited w ith o u t p e rm is s io n .
39
Figure 13. An example o f 9 (nominal) AMSU-A 54.94 GHz Tb (°C) FOV selected for
retrieval use, Hurricane Floyd, 1238 UTC 14 Septempber 1999: (A) NOAA-15 AVHRR
color-enhanced multi-spectral composite image, (B) AMSU-A 54.94 GHz Tb (°C)
observations, (C) FOV selected for use in the retrieval.
R e p r o d u c e d with p e r m i s s io n of th e co p y rig h t o w n er. F u r th e r r e p r o d u c tio n p rohibited w ith o u t p e r m is s io n .
40
1 9 9 9 A tla n tic S eason R esu lts (w /o Lenny)
NOAA—IS AMSU-A Ct>7 (54.94G H z) Tb v s. R econnaissance E stim ated MSLP
R2 = 0.69
No re trie v a l
■
R2 = o 82
R etriev al w / Raye Fixed (24km )
•
R2 = 0 .8 9
R etrieval w / Raya V a riab le (AMSU-B)
Tb Anomaly
(C)
4
1010
1000
990
980
97 0
960
950
9<10
930
920
NHC M SLP ( h P a )
Figure 14. AMSU-A 54.94 GHz TC UTWA (°C) vs. in situ MSLP (hPa) (n=22, 1999)
R e p r o d u c e d with p e r m i s s io n of t h e co p y rig h t o w n er. F u r th e r r e p r o d u c tio n prohibited w ith o u t p e r m is s io n .
41
4 Independent Test
Having validated the performance and suiitability of the AMSU-A horizontal structure
retrieval algorithm in Section 3, the resulting regression equations were used to predict MSLP in
2000 using AMSU-A-derived 54.94 GHz TC UTTWA data:
RAW
M S L P ^ = -(ATb R
AW- 68.88) / 0.07
( 10)
MSLP rex = -(ATb RET- 157.87) / '0.16
( 11)
where ATb RAWis the raw AMSU-A 54.94 GHz T 'C UTWA (no correction for scan geometry or
diffraction) and ATb RET is the retrieved A M SU -A 54.94 GHz brightness temperature anomaly
(corrected for potential TC UTWA sub-sampling) using the method described in Section 3. Due
to the near exclusivity o f in-situ observations in Che ATL basin, the majority o f comments in this
section apply to ATL (n=29). The performance o f the AMSU TC intensity estimation algorithm
in eastern Pacific (EPAC) ocean basin (n=2), N W PA C (n=5), SPAC (n=6, year 2001), and
southern Indian Ocean (SIO) (n=l, year 2001) is also noteworthy and is therefore included in the
final population statistics and discussions. S everal issues and concerns arise when using nonaircraft in situ observations to validate satellite-based intensity estimates, owing largely to the
large horizontal pressure gradients (e.g., 1 hPa k m '1) that are known to occur near the inner eye
wall region and the displacement between the T C path and nearest observing location. This will
be discussed and further addressed later in this se=ction.
R e p r o d u c e d with p e r m i s s io n of t h e co p y rig h t o w n e r. F u r th e r r e p r o d u c tio n prohibited w ith o u t p e r m is s io n .
42
4.1 Implementation
The independent test ran in an event-driven, automated manner commencing with the
formal declaration of a Tropical Depression (TD) by either the NHC/JTWC and continued for
the duration of the storm (Fig. 15). Several factors motivated this approach including 1) the
desire to maximize TC MSLP estimate objectivity, 2) investigating the overall feasibility of an
automated method, and 3) determining w hat factors, if any, might limit the potential accuracy of
estimate limits generated using a simulated future operational processing environment. The
frequency of subsequent AMSU TC intensity estimates were dictated by a number of factors
including position updates (every 6hrs in ATL/EPAC/NWPAC, 12hrs in SIO/SPAC),
corresponding AMSU TC observation coverage determined using the NOAA/NESDIS
BROLYD orbit prediction package, and data availability as described in Section 2. A
combination of limited swath width (~2200km), satellite orbit precession (~ 1° per day), TC
position and translation velocity provided as many as two AMSU intensity estimates per day and
occasional days in which no AMSU intensity estimates were available.
Several factors made automated implementation of the AMSU independent test
challenging, particularly determining TC position at the time of AMSU observation. The ability
to explicitly correct for AMSU-A TC UTW A sub-sampling requires precise knowledge of TC
position combined with accurate individual AMSU-A FOV earth navigation capabilities (i.e.,
Fjj(x,y) in Eqn. 1). In order to characterize the effects of diffraction, the scheme must be able to
determine the displacement of TC position from the center of the FOV believed to contain the
R e p r o d u c e d with p e r m i s s io n of t h e cop y rig h t o w n e r. F u r th e r r e p ro d u c tio n prohibited w ith o u t p e rm is s io n .
43
maximum upper tropospheric temperature. Furthermore, optimization of the maximum
likelihood regression constraint structure function XCON, upon which the solution structure vector
XRETdepends, requires accurate specification o f the size parameter R which is highly sensitive to
errors in TC position. Further complicating matters is the non-zero time differential between the
latest NHC/JTWC position estimate and the nearest AMSU observation. The algorithm
addressed this issue by estimating TC position at the time of AMSU observation by linear
extrapolation o f NHC/JTWC initial and 12 hour forecast positions given the time differential
between the AMSU observation and initial NHC/JTWC position estimate. A final adjustment of
up to 1° from the extrapolated TC position is tolerated based on a laplacian analysis o f AMSU-B
89.0 GHz Tb data
V2Tb= ( Tb
+ Tb,^[ + Tb,v.l4- Tbl>l- 4 T b, ) / d 2
(12)
where i and j are the two-dimensional array line/element indices and d is the latitude-dependent
distance (km) between the outer elements under consideration. The AMSU-B 89.0 GHz Tb
signature corresponding to the location o f either a weak/strong TC, as discussed in Section 3.1.2,
should coincide to the 1° position-constrained location of the maximum absolute value of (12).
As one might expect, the accuracy o f the extrapolated TC position (and thus the retrieved
TC UTWA) depends on the accuracy of the initial position as well as the estimated translation
velocity. Optimal results typically occurred for well-defined TCs with small-to-moderate time
differentials (i.e., 0-3hrs). Under these circumstances, well defined multi-spectral geostationary
and polar orbiting satellite imagery features aid NHC/JTWC initial position estimates.
R e p r o d u c e d with p e r m is s io n of t h e cop y rig h t o w n e r. F u r th e r r e p r o d u c tio n prohibited w ith o u t p e r m is s io n .
44
Furthermore, consensus amongst numerical models and statistical forecasting aids likely results
in accurate 3-6hr forecast positions from which reliable inter-hour position estimates can be
derived. In contrast, inferior results typically accompanied weak TCs characterized by poor
satellite signatures or multiple low level circulation centers. Under these circumstances
numerical models and statistical aids often produce divergent 6hr forecast positions. As a result,
even small time differentials between NHC/JTWC position estimates and AMSU observations
can lead to significant errors in extrapolated TC position. In either case, linear extrapolation will
fail to account for either short-term, mesoscale changes in TC track, translation speed or both
leading to storm position estimate errors.
4.2 Validation
Aircraft reconnaissance frequently offer the best (or only) estimate of true TC position
and MSLP (Fig. 16). This is obviously the case over data sparse oceanic regions where few if
any in situ surface observations are available; however, it is also true even for circumstances
when surface in situ observations are available due to the potentially large pressure gradients
(lhPa km '1) found in the TC inner eyewall region (Gray and Shea, 1973). Based on the TC
intensity, the size of the surface vortex and the displacement of the observing location from the
true TC center, displacements as small as 5 -10km can introduce potential errors of 5-10 hPa or
larger for intense systems with generally more relaxed displacement constraints for weaker TCs.
As a result, it is entirely possible for a TC to pass through a matrix of surface observing locations
providing what appears to be an excellent validation opportunity when, in reality, it may
R e p r o d u c e d with p e r m i s s io n of t h e co p y rig h t o w n e r. F u r th e r r e p r o d u c tio n prohibited w ith o u t p e r m is s io n .
45
introduce considerable estimate error into the independent sample. As a result, these factors
must be considered and rigorously applied when screening surface in situ observation validation
candidates.
TC Dera (Fig. 17) is an excellent example of a SIO AMSU TC MSLP estimate validation
candidate. At approximately 2000UTC 9 March 2001, TC Dera passed within 10 nm (18.3km)
o f Europa Island (white dot near 21.3S 40.6E) in the Mozambique Channel. Similar to AMSU-B
89.0 GHz, Tropical Rainfall Measurement Mission (TRMM) 85.0 GHz channel radiances are
also significantly attenuated (i.e., reduced Tb) by heavy precipitation in the eyewall region
thereby serving as an effective proxy for TC eye size. According to Fig. 17, Europa Island is
virtually centered within the 65km diameter eye (inferred from TRMM 85GHz data) and
therefore the MSLP o f 973 hPa observed near 2000UTC can be considered representative,
possibly a few hPa higher, than conditions found at Dera’s center location. The corresponding
1611UTC 9 March 2001 NOAA-15 AMSU MSLP using a manually-specified R = 32.0 km
yielded 972 hPa in excellent agreement with Europa Island MSLP observations.
4.3 Results
The results of the independent test are summarized in Table 2. ‘A Time’ denotes the AMSU
TC observation time, ‘O Tim e’ the in situ observation time, ‘D Tim e’ the difference between
AMSU and in situ observation times, ‘O MSLP’ the observed in situ MSLP, ‘Raw MSLP’ the
MSLP derived using raw AM SU-A 54.94 GHz TC UTWA observations and dependent sample
regression coefficients (Eqn. 10), ‘Ret MSLP’ the retrieved MSLP using scan geometry and
R e p r o d u c e d with p e r m i s s io n of t h e co p y rig h t o w n e r. F u r th e r r e p r o d u c tio n prohibited w ith o u t p e r m is s io n .
46
diffraction-corrected AM SU-A 54.94 GHz TC UTWA observations and dependent sample
regression coefficients (Eqn. 11), ‘O-Raw’ and ‘O-Ret’ the difference between AMSU-Aderived MSLPs and in situ observed MSLP, while ‘Hybrid’ combines the results o f ‘O-Raw’ and
‘O-Ret’ based on an analysis o f variance using different raw AMSU-A 54.94 GHz TC UTWA
thresholds. In this case, the choice of ‘Hybrid’ candidates is based on a raw UTWA threshold of
0.5°C. For circumstances in which the raw UTWA is less than or equal to 0.5°C (<= 1000 hPa)
values below or equal to this value, raw MSLP estimates are considered whereas when raw
UTWA exceed 0.5°C (> 1000 hPa), retrieved UTWA are used. The rationale for considering a
hybrid approach is discussed later in this section.
The ‘ALL’ row indicates AMSU TC MSLP estimate RMS errors (hPa) using raw, retrieved
and hybrid UTWA candidates from all basins (i.e., ATL, EPAC, NWPAC, SIO, and SPAC),
n=42 whereas ‘ATL/EPAC’ reflect only ATL and EPAC cases (n=31). The ‘Raw M SLP’ results
(mean error of 7.4 hPa vs. 7.5 hPa with standard deviations o f 9.7 hPa vs. 9.9 hPa respectively)
likely reflect gains realized through increased horizontal resolution (48km at nadir) and
improved radiometric accuracy of the AMSU-A when compared to the previous results of
Velden et al. (1989) and Merrill (1994) using MSU (1 10km at nadir) radiance data. Retrieved or
‘Ret MSLP’ results (mean error of 5.7 hPa vs. 6.2 hPa with standard deviations of 7.9 hPa and
8.0 hPa respectively) improve upon ‘Raw’ MSLP estimates and demonstrate the efficacy o f the
retrieval’s ability to resolve additional TC upper tropospheric warm by explicit treatment o f scan
geometry and diffraction effects. The best results are achieved using a hybrid approach (mean
error o f 5.3 hPa vs. 5.5 hPa with identical standard deviation values of 7.2 hPa).
R e p r o d u c e d with p e r m i s s io n of t h e cop y rig h t o w n e r. F u r th e r r e p r o d u c tio n prohibited w ith o u t p e rm is s io n .
47
Consideration of a hybrid AMSU TC intensity approach is based on fundamental principles
o f TC warm core structural morphology, retrieval mechanics and recognizing when retrieval use
is warranted. During the formative stages of TC development, the UTWA horizontal scale is
large and relatively diffuse representing the small residual increase in upper tropospheric energy
based on the competing processes of adiabatic cooling and latent heat release within organized
deep cumulus updrafts. The absence of an eye wall limits the ability of organized convection to
retain and concentrate heat to a horizontal scale below the resolving limits of the AMSU-A
instrument. Under these circumstances, scan angle (<f>) and off-antenna-axis scan angle (0) as
depicted in Fig. 8 are largely irrelevant. Under these circumstances, any attempt to define R
using either fixed WG88 climatological values or those found through spatial analysis of
relatively ambiguous AMSU-B 89.0 GHz Tb distribution is physically ill-founded and will
ultimately lead to biased estimates of TC UTWA magnitude and unrealistically low MSLP
values. The role of the eye wall and associated maximum tangential winds is appreciation using
simple thermodynamic and dynamic constraints. For an axisymmetric vortex in gradient wind
balance, the relationship between the radial temperature gradient and the vertical shear of
tangential winds in cylindrical coordinates can be expressed as (Holton, 1979):
3v/3z (f + 2v/r) = (R/H) 3T/3r
(13)
where v is the tangential wind velocity (ms"1) , / the coriolis parameter (s'1), r the horizontal scale
(m), R the gas constant (Jkg'lBCl), H the scale height (m) and T the temperature (K). Using
representative values for a nascent TC (3r ~ 100 k m ,/ ~ 5 x 10"5s"‘, H ~ 8 km and v ~ 10ms"1),
R e p r o d u c e d with p e r m i s s io n of t h e co p y rig h t o w n e r. F u r th e r r e p ro d u c tio n prohibited w ith o u t p e rm is s io n .
48
the associated radial temperature gradient 9T ~ IK. By way of comparison, AMSU-A observed
TC UTWAs in excess o f 16K require tangential velocities in excess of 50m s'1.
The combined results of the dependent and independent test comparing raw, retrieved
and hybrid AMSU-A 54.94 GHz TC UTWA vs. in situ MSLP estimates are provided at Fig. 16.
Individual raw UTWA estimates vs. MSLP are denoted by the red squares and linear least
squares fit line (R2 = 0.79). The increased scatter, particularly for TCs characterized by MSLP
values 980-950 hPa, is apparent and is quite similar to the results o f previous studies using the
MSU (Velden et al., 1989, 1991) and reflect TC UTWA sub-sampling (i.e., degree of scatter
increases as TC MSLP decreases ) consistent with theory. The individual black diamonds
(including those with white centers for MSLP <=1000 hPa) and black linear least squares fit line
(R2 = 0.94) represent the results of applying the retrieval and correcting for TC UTWA sub­
sampling. Finally, the individual black diamonds with white centers (all MSLP values) and dark
green linear least squares fit line (R2 = 0.95) represent the hybrid solutions.
Finally, the results of the independent test provide compelling evidence that issues of TC
UTWA remain despite improvements in AMSU-A horizontal resolution and radiometric
accuracy. To more fully assess the value of the proposed retrieval methodology, it is important
to quantify the impact o f improvements in horizontal resolution and radiometric accuracy alone
when comparing AMSU vs. MSU MSLP estimation performance. This is done by evaluating
only those independent test raw AMSU-A 54.94 GHz TC UTWA-derived MSLP estimates for
TCs positioned with 6° o f NOAA-15 nadir similar to Velden (1989) using the MSU. In doing so,
the MSLP estimate RMS error decreases to 7.1 hPa (n=12) vs. 10.4 hPa for the entire sample
(n=42) which equates to a 1 hPa improvement over the MSU (MSLP RMS error ~ 8 hPa).
R e p r o d u c e d with p e r m i s s io n of t h e co p y rig h t o w n er. F u r th e r r e p r o d u c tio n prohibited w ith o u t p e r m is s io n .
49
AMSU TC MSLP estimate error decreases to 4.9 hPa using the same position constrained
sample (n=12) after applying the retrieval. Such results suggest that improvements in horizontal
resolution and compensation for scan geometry and diffraction both lead to improved
quantification of TC upper tropospheric warming; however, the retrieval serves a vital role not
only in further improving the accuracy o f AMSU TC UTWA-derived MSLP estimates but
significantly improves the potential exploitation of the entire AMSU-A scan swath with very
little penalty to the accuracy of the estimate. These results offer much promise for potential
future satellite applications and will be discussed in Section 7.3.
R e p r o d u c e d with p e r m i s s io n of t h e co p y rig h t o w n er. F u r th e r r e p r o d u c tio n prohibited w ith o u t p e r m is s io n .
50
-VMM
iH M a t r w
(Ret T;— )
VTXS31 PG T V 222100
1.TROPICAL CYCLONE 068
01 ACTIVE TROPICAL CYCLONE .
MAX SUSTAINED V IN D 8 BASED
(Raw T _ _ )
VAKNING POSITION:
221800Z3 — NEAR 19.484 72.1E0
M0 VEMENT PAST SIX HOURS -
<Ret)
MSLP
{R aw )
Tim
MSLP
Figure 15. Schematic diagram depicting automated AMSU TC MSLP estimation
algorithm usage: (A) declaration of numbered system by NHC/JTWC, (B) position
estimated for AMSU observation time —eye size estimated from AMSU-B radiance
data and forward model run, (C) TC UTWA raw and retrieved values estimated, (D)
UTWA values used in conjunction with regression coefficients derived from the 1999
dependent sample, and (E) MSLP estimated
R e p r o d u c e d with p e r m i s s io n of t h e co p y rig h t o w n er. F u r th e r r e p r o d u c tio n prohibited w ith o u t p e r m is s io n .
51
AMSU-A D erived TC UTWA v s . MSLP
■
Raw AMSU-A TC UTWA R2 = 0.79
+
R et AMSU-A TC UTWA R2 = 0.94
O
Hyb AMSU-A TC UTWA R2 = 0.95
- “ Hyb Tk T hreshold (0 .5 * 0
UTWA (T* in C)
N = 61
♦ ♦
•---------------------- 1---- w
1020
1010
------------ 1-----------------------1
1000
990
t
■i
i
i
980
970
960
950
1
i
940
r
.................
930
r
k
.
920
In s itu MSLP (hPa)
Figure 16. Scatter plot of raw, retrieved and hybrid AMSU-A 54.94 GHz TC UTWA vs.
in situ MSLP estimates (1999, 2001, and early 2001)
R e p r o d u c e d with p e r m i s s io n of t h e co p y rig h t o w n e r. F u r th e r r e p r o d u c tio n prohibited w ith o u t p e r m is s io n .
52
81
AMSU-A 5
AMSU-A 6
AMSU-A 7
AMSU-A 8
AMSU-B 16
P B andw idtnli
.^^E alilraiiioE ils||i
s Km sh I
53.60
54.40
54.94
55.50
89.00
170
400
400
330
6000
0.25
0.25
0.25
0.25
n/a
wsimmomk
1.5
1.5
1.5
1.5
n/a
Table 2. AMSU-A and AMSU-B channel characteristics. Source: NOAA/NASA
Polar Orbiting Environmental Satellite (POES) Guide, NP-1997-12-052-GSFC
R e p r o d u c e d with p e r m is s io n of t h e cop y rig h t o w n e r. F u r th e r r e p r o d u c tio n prohibited w ith o u t p e r m is s io n .
53
5 Case Studies
The previous section discussed results of the independent test and validated Merrill’s
(1995) earlier hypothesis that given the availability of information concerning TC UTWA
horizontal scale, as well as a ability to explicitly model the interaction of the AMSU-A antenna
gain pattern with a first-guess analytic function approximating the horizontal distribution o f TC
upper tropospheric warming, one could account for the amount o f UTWA sub-sampling and
produce a more accurate MSLP technique. This was successfully demonstrated and is the
primary focus o f this effort; however, in the process of running and evaluating the retrieval
performance both in real and post-event mode, several events occurred that are noteworthy. The
events include evidence of AMSU TC intensity estimation algorithm skill in multiple basins, the
virtue of monitoring the warm core during periods of rapid intensity change, the potential o f the
AMSU as a tropical cyclone classification tool, and the possibility of AMSU MSLP estimation
capabilities for TCs over land. In each case, there are several illustrative examples; however, for
the purposes o f brevity, a single representative example will be discussed.
5.1 Inter-basin Applicability
TC are a worldwide phenomena whose formation and energetics share increasingly well
known underlying physical principles necessary for their development and maintenance. Despite
broad inter-basin similarities in large scale conditions such as warm sea surface temperatures and
R e p r o d u c e d with p e r m i s s io n of t h e co p y rig h t o w n e r. F u r th e r r e p r o d u c tio n prohibited w ith o u t p e rm is s io n .
54
mixed-layer depth, weak vertical environmental wind shear, among others (Hastenrath, 1991),
differences in environmental background vorticity and tropopause height lead to differences in
average TC horizontal and vertical scale. The monsoon trough and monsoon gyres in the
NWPAC generally support spatially larger TCs, an average of 15% larger than in other basins
(Merrill, 1984). In addition, climatological analyses indicate a 1-2 km higher tropopause height
in the NWPAC compared to the western ATL during northern hemisphere TC season (Hoinka,
1999) as illustrated using NCAR/NCEP reanalysis data climatological (1948-1999) tropopause
height fields (Fig. 17). Both factors were believed responsible for the historically larger MSU
55GHz region passive microwave TC warm anomaly observations for NWPAC TCs compared
to ATL TCs o f similar intensity (Velden et al., 1991). As a result, regression relationships
derived from MSU 55GHz region passive microwave TC UTWA measurement/in situ MSLP
observation pairs in the NWPAC where not shown to be valid for use in the ATL and vice versa.
Differences in warm core structure, combined with the 1987 decommissioning o f Air
Force tropical cyclone aircraft reconnaissance in NWPAC, significantly limited use of MSUbased TC UTWA MSLP estimation techniques outside of the ATL basin. Furthermore, the lack
o f systematic in situ upper tropospheric warm core observations by reconnaissance aircraft do
not permit the investigation and validation of passive microwave-observed inter-basin
differences in TC UTWA strength. While increased NWPAC TC horizontal scale is certainly a
leading candidate for improved MSU (and AMSU) TC UTWA sampling when compared to the
ATL, the potential contribution to inner core sub-pixel warming due to enhanced subsidence as a
by product o f greater mean tropopause elevation and associated increased convective available
R e p r o d u c e d with p e r m i s s io n of th e co p y rig h t o w n er. F u r th e r r e p r o d u c tio n prohibited w ith o u t p e r m is s io n .
55
potential energy (Velden, 2001), while plausible, is tenuous with little or no historical data to test
this hypothesis.
As mentioned earlier, the lack of aircraft reconnaissance in most basins limits satellitebased intensity estimate validation opportunities to sporadic landfall surface synoptic
observations, radar observations (e.g. WSR-88D observations from Anderson AFB, Guam,
Kadena AB, Japan, etc.) or encounters with ocean buoys. In-situ synoptic reports for landfalling
NWPAC, SIO and SPAC TCs were acquired in near-real time from the Air Force Weather
Agency (AFWA), the Australian Bureau o f Meteorology (ABoM), and Meteo-France. For this
limited set o f cases (n=l 1), AMSU-derived MSLP RMS errors (again, using regression
coefficients derived from the 1999 ATL dependent data set) averaged 7.0,4.1 and 4.5 hPa for
raw, retrieved and hybrid MSLP estimates. For the NWPAC cases (n=4), AMSU-derived MSLP
errors averaged 8.0, 4.0 and 5.25 hPa. While the sample sizes for both o f these examples are
arguably small, the magnitude o f the MSLP estimate error and dramatic improvements offered
through application of the retrieval are promising.
A noteworthy NWPAC example illustrating the virtues o f the proposed AMSU TC
UTW A approach is Super Typhoon (STY) Bilis (18W). Several days prior to striking Taiwan,
STY Bilis rapidly intensified with Dvorak-estimated sustained winds of 140KTS corresponding
to a MSLP of 898 hPa (Shewchuck and Weir, 1980) at 1200 UTC 22 August 2000 (Fig. 18).
Automated AMSU intensities are denoted by circled MSLP values and corresponding SaffirSimpson scale color coding while Dvorak color-coded intensity estimates are indicated by color
changes along the storm best track. The AMSU intensity estimate at 1118 UTC 22 August 2000
R e p r o d u c e d with p e r m i s s io n of t h e co p y rig h t o w n er. F u r th e r r e p r o d u c tio n prohibited w ith o u t p e r m is s io n .
56
was 904 hPa with an AMSU-B 89.0 GHz estimated eye size (R) of 10km in excellent agreement
with both JTWC estimates (11km) and Garden Island, Taiwan Doppler radar reflectivity data
(Fig. 19). At the time of impact, Cheng Kon (WMO 46761) reported sustained winds between
120-147KTS (914-898 hPa).
A time series of JTWC intensity estimates for STY Bilis, as well as estimates from other
agencies is provided in Fig. 20. Several features deserve comment, most notably the timing and
magnitude of the AMSU-based intensity estimate during STY Bilis’ peak intensity. Extending
past findings of observed differences in MSU-based NWPAC vs. ATL TC UTWA thermal
structure to the present suggests that use of ATL regression coefficients would lead to a gross
underestimate o f Bilis’ intensity near 1200 22 August 2000. Instead, the use of ATL coefficients
with AMSU-A UTWA and AMSU-B eye size estimates leads to an intensity estimate strikingly
similar to those produced independently by several centers. Further analysis reveals general
agreement between AMSU and Dvorak-based intensity estimates prior to landfall; however,
intensity estimates diverge shortly after landfall. While AMSU intensity estimates and UTWA
depict rapid weakening, subjective Dvorak intensity estimates decrease only marginally in the
12hr period following landfall. Interestingly, of all Dvorak-based estimates, the UW-CIMSS
Objective Dvorak Technique (ODT, Velden et al., 1998) most closely resembles the post-landfall
AMSU 12hr MSLP tendency.
Differences in ODT vs. subjective Dvorak intensity estimates for STY Bilis are likely due
to differences in internal time-average constraints. In the case of ODT, cases suspected of
undergoing rapid intensity change (inferred by analysis satellite infrared cloud top temperature
measurements) use a curtailed 3hr intensity estimate averaging interval (vs. 12hr in the
R e p r o d u c e d with p e r m i s s io n of t h e co p y rig h t o w n e r. F u r th e r r e p r o d u c tio n prohibited w ith o u t p e r m is s io n .
57
subjective method used operationally). In circumstances in which rapidly weaken, traditional
subjective Dvorak intensity estimate constraint rules (i.e., 1 T number per 12hr period) may lead
to artificially high TC intensities and unrealistic short-term trends. In a separate related study,
Bender et al. (1987) investigated the effects of landfall over Taiwan using a high resolution (1/6°
inner mesh) version of the GFDL model and an idealized TC. In the control experiment (no
land/topography), MSLP is marginally affected (decreases slightly); however, when more
realistic land topography and reduced evaporative heat flux are introduced (several mountain
peaks approach elevations of 3km), the resulting impact is dramatic and qualitatively similar to
both STY Bilis’ AMSU-derived UTW A and MSLP trends (Fig. 18).
Another excellent example suggesting AMSU TC intensity estimate inter-basin skill is
TC Abigail (12P) 1046 UTC 26 February 2001 (Fig. 21). TC Abigail passed directly over the
Momington Island, Australia (16.65S 139.17E, BSN 942560) early on 26 February 2001. The
1046 UTC AMSU-derived MSLP o f 976 hPa was found to be in excellent agreement with 0920
UTC 974 hPa observation.
5.2 Rapid Intensity Change
As discussed in the previous section, Dvorak method constraints limit TC intensity
estimate temporal variability and may potentially lead to erroneous estimates in cases o f rapid
TC deepening and weakening. The AMSU algorithm is inherently sensitive to short-term trends
in UTWA magnitude (AMSU-A) and eye size (AMSU-B) both o f which are important
R e p r o d u c e d with p e r m i s s io n of t h e cop y rig h t o w n e r. F u r th e r r e p r o d u c tio n prohibited w ith o u t p e rm is s io n .
58
diagnostic indicators of TC intensity change. Hurricane Isaac (Fig. 22) effectively demonstrates
this unique capability of the AMSU intensity technique. On 23 September 2000 1000 UTC
(5AM EDT), Isaac’s rapid organization was masked by an infrared Central Dense Overcast
(CDO) with NHC-estimated sustained winds of 55 KTS. AMSU- B 89.0 GHz observations at
roughly the same time (0954 UTC) suggest the presence of an eye and hurricane strength winds
(i.e., minimum of 64 KTS sustained). Analysis of AMSU-B 89.0 GHz data yielded a small 10.2
km (radius) eye size leading to a MSLP estimate of 974 hPa, approximately 26 hPa less than
MSLP estimates derived using the raw UTWA values (1000 hPa). Six hours later at 1800UTC
(11AM EDT), NHC continued classifying Isaac as a Tropical Storm. By 0600UTC (11PM EDT)
on 24 September, Isaac was reclassified a major hurricane having sustained winds of 105KTs
and a 9.2km (radius) eye in excellent agreement with the previous AMSU-B 89.0 GHz estimate
earlier in the day.
Owing to a lack of in-situ observations, whether Isaac had indeed reached hurricane
strength (i.e., sustained 10 minute average winds in excess of 64 kts) early on 23 September
2000 is a matter of debate. Post-storm analysis, particularly the formation of a small eye and
Dvorak-based sustained wind estimates of 105 kts in a 12 hr period, as well as the excellent
agreement between AMSU-generated and visually observed eye size later in the day, corroborate
earlier AMSU-based intensity estimate indicators. Finally, the poor MSLP estimate (1000 hPa)
derived from the 0954 UTC raw AMSU 54.94 GHz TC UTWA estimate, despite relatively good
viewing geometry (i.e., well within the AMSU-A scan swath), once again draws attention to the
need to correct for TC UTWA sub-sampling - exacerbated in the case of Isaac due to a small eye
- and the potential ‘value added’ by the retrieval.
R e p r o d u c e d with p e r m i s s io n of t h e co p y rig h t o w n er. F u r th e r r e p r o d u c tio n prohibited w ith o u t p e r m is s io n .
59
53 Tropical Cyclone Classification
One last event that deserves special recognition was the timely, and quite possibly the
first, operational use o f AMSU-A warm core observations by NHC forecasters. As discussed by
Knaff et al. (2000), the unique vertical temperature profiling capabilities of the AMSU-A allows
forecasters the ability to discern warm-core vs. cold-core temperature features for systems whose
outward appearances are nearly identical at visible or infrared wavelengths. During the late
evening hours of 16 October 2000, AMSU-A warm core information made available by the UWCIMSS web page (created by the author) facilitated the NHC reclassification of Michael from
subtropical low to a Tropical Depression (Fig. 23). For the next 48hrs, AMSU-A warm core
observations played a vital role in NHC maintenance o f Michael’s hurricane status up through
extratropical transition early on 20 October 2000.
The last AMSU intensity estimate before Michael lost warm core characteristics occurred
at 2304 UTC 19 October 2000 shortly after landfall in New Foundland. An eroding and
elongated 50-60 km diameter eye wall is evident in a 2145 UTC 20 October 2000 radar
reflectivity image (Fig. 24); corresponding AMSU-B 89.0 GHz-derived eye size
estimates of 72 km yielded an improved MSLP measurement of 976 hPa (vs. 981 hPa raw) in
excellent agreement with several in situ observations of 974-976hPa reported within 0-30km o f
Michael’s path.
5.4 Intensity Estimation Over Land
R e p r o d u c e d with p e r m i s s io n of t h e co p y rig h t o w n e r. F u r th e r r e p r o d u c tio n prohibited w ith o u t p e r m is s io n .
60
Another capability of the AMSU related to TC analysis and forecasting is the ability to
monitor the morphology of the warm core after landfall. After moving over land, oceanatmosphere evaporative heat flux processes essential for TC maintenance and intensification
abruptly cease. The subsequent loss of convergent, high equivalent potential energy sub-cloud
layer air diminishes the ability of deep moist convection within the eye wall region, and
associated latent heat release, to maintain the TC UTVVA and accompanying MSLP deficit.
Provided that environmental factors that might otherwise influence the integrity o f the vertical
warm core structure (e.g., horizontal wind shear, ventilation, etc.) are absent, the warm core can
remain remarkably intact long after TC landfall and thus maintain reduced MSLP.
An analysis o f TC Winsome illustrates the longevity of the warm core after landfall and
the ability of the AMSU TC intensity estimation technique (Fig. 25). TC Winsome developed
along the Australian monsoon trough/shear line oriented E-W through the northern Gulf of
Carpentaria. At the time of landfall (-0000 UTC 11 February 2001), ABoM surface pressure
analyses indicate an MSLP of 984 hPa with gale force winds displaced approximately 275km
from the surface circulation center. By 1700UTC 12 February 2001, TC Winsome had been
over land nearly two days and traversed approximately 540 km over land characterized by
serozems, desert soils and semi-fixed sands (Times Atlas, 1988). The 1656UTC 12 February
2001 AMSU-A observations still depict indications of a warm core yielding AMSU-derived
raw/retrieved MSLP estimates of 993/987 hPa respectively. Wave Hill (17.37S 131.1E, BSN
942290), located approximately 4km from the position used by the AMSU retrieval, reported a
MSLP of 984 hPa at 1700UTC.
R e p r o d u c e d with p e r m i s s io n of t h e co p y rig h t o w n e r. F u r th e r r e p r o d u c tio n prohibited w ith o u t p e r m is s io n .
61
5.5 Comparison with Dvorak Estimates
The subjective Dvorak technique (Dvorak, 1975, 1984) is currently the worldwide
standard TC intensity estimation technique and uses primarily geostationary visible and infrared
imagery. Dvorak TC intensity estimates are based on imagery analysis using pattem-recognition
and rules defined by features such as infrared-based cloud top temperatures, areal coverage,
banding features, and trends. Several agencies including NHC, NOAA/SAB and AFWA
routinely generate and share Dvorak intensity estimates to exploit independent analysis in a
general search of TC intensity estimate consensus. Analysis of the most recent imagery yields a
current intensity or (Cl) number which is modified, using rules that limit temporal variations of
TC intensity, to produce a final intensity (Final T).
To further assess AMSU TC intensity estimation algorithm performance in the
ATL/EPAC, analysis was extended to include a direct comparison with Dvorak intensity
estimates for the same TC at approximately the same time. Individual agency intensity estimates
where usually performed at or near the same time —typically no more than 1 hour apart. As one
would expect, differences in independent Final T numbers occurred and, as a result, were
averaged here for comparison with the AMSU intensity estimates. For the ATL/EPAC
independent sample, Dvorak MSLP (derived from Final T numbers) mean error and standard
deviations were 7.8 hPa and 7.6 hPa respectively which are slightly inferior to AMSU intensity
estimates (Table 3). In general, the subjective Dvorak technique did very well on average with
the exception of systems undergoing rapid intensity change (e.g., Hurricane Keith), weak TCs
interacting with land or undergoing moderate-strong vertical wind shear (e.g., Tropical Storm
R e p r o d u c e d with p e r m i s s io n of t h e co p y rig h t o w n e r. F u r th e r r e p r o d u c tio n prohibited w ith o u t p e r m is s io n .
Debby), or for TCs undergoing extratropical transition (e.g., Hurricane Michael). Under these
circumstances, the AMSU TC intensity estimation technique dramatically outperformed
subjective Dvorak estimates. In the latter case, subjective Dvorak Final T-derived MSLPs were
almost 30 hPa weaker than observed as Michael passed over Newfoundland at Category 1
hurricane strength.
R e p r o d u c e d with p e r m is s io n of t h e cop y rig h t o w n e r. F u r th e r r e p r o d u c tio n prohibited w ith o u t p e r m is s io n .
63
40 N
SO w
10°N
120*ff
locAr
I
I
40 N -
ao*r
«oV
ao4*
40*¥
> fc i* *.» fc y .z m A ? .? A a
30 ft H
*w>w>w?v-
20 Ct H
io sr H
t>
H
io a
BO*B
MO B
1204B
140*E
ibo
Figure 17. NCAR/NCEP 2.5° grid resolution reanalysis (1948-1999)
tropopause geopotential height (km) for the ATL (top) and NWPAC (bottom)
basins. Contour interval 1 km below 15 km, 0.25 km above 15 km
R e p r o d u c e d with p e r m i s s io n of th e co p y rig h t o w n er. F u r th e r r e p r o d u c tio n prohibited w ith o u t p e r m is s io n .
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
65
22 Aug 00
11:30
C. Intensity
dBz
Product Type :PPI Corrected Intensity T ilt: 1 Elevation:
PRF :1180Hz
Gatewidth : 500a
Pulse Width : 0.8us
0.9
Hax Range :112iai
Sanples : 65
Clutter Filter : 3
Site Nane :GRN ISL, TAIWAN
Radar Type :DHSR-92C
Gates : 224
Unfolding :4:3
Range Normalization :0n
Antenna Height : 284m
Figure 19. Garden Island, Taiwan 5cm plan position indicator (PPI) radar reflectivity
image of STY Bilis (18W), 1130 UTC 22 August 2000
R e p r o d u c e d with p e r m i s s io n of t h e co p y rig h t o w n e r. F u r th e r r e p r o d u c tio n prohibited w ith o u t p e r m is s io n .
66
Super Typhoon (STY) Bills 17-26 August 2000
Satellite-based Intensity Estimate Comparison
S5
80
75
• JTWC
Current Intensity (Cl) Number
70
55
• AFW A
50
O SA B
55
0 KADENA
50
45
t JM A
40
■ ODT
35
B E ST
TR A C K
I
30
25
20
• T -N U M
15
★
a m su
10
05
17/00
18/00
19/00
20/00
21/00
22/00
23/00
24/00
25/00
25/00
Date/Time (UTC)
Figure 20. Time series of STY Bilis (18W) intensity estimates, 17-26 August 2000
R e p r o d u c e d with p e r m i s s io n of t h e co p y rig h t o w n e r. F u r th e r r e p r o d u c tio n prohibited w ith o u t p e r m is s io n .
67
TROPICAL CYCLONE ABIGAIL
AMSU—B Channel 1 6 (89 GHz) Brightness T em perature (C)
Monday 26feb01Q 57 if m e 1046 UfC
132
13+
136
136
140
1+2
1++
146
6.9
—
3.1
-1 3 .1
-
23.1
-3 3 .1
-4 3 .1
-
53.1
-6 3 .1
-
73.1
-8 3 .1
132
13+
136
136
140
1+2
1+ 4
146
Figure 21. Color-enhanced AMSU-B 89.0 GHz Tb (°C) image, Tropical
Cyclone Abigail, 1046 UTC 26 February 2001
R e p r o d u c e d with p e r m is s io n of t h e cop y rig h t o w n e r. F u r th e r r e p r o d u c tio n prohibited w ith o u t p e r m is s io n .
-9 3 .1
68
Figure 22. Hurricane Isaac, 23 September 2000. METEOSAT-7 IR image 1000 UTC
23 September 2000 (upper left), AMSU intensity retrieval diagnostic output 0954
UTC (upper right), NHC forecast discussion bulletin segments 23 September 2000
(lower left), and color-enhanced AMSU-B 89.0 GHz Tb 0954 UTC image (lower
right)
R e p r o d u c e d with p e r m i s s io n of t h e co p y rig h t o w n er. F u r th e r r e p r o d u c tio n prohibited w ith o u t p e r m is s io n .
Figure 23. Segments of NHC forecast discussion bulletin W TNT42 KNHC, 16/19
October 2000
R e p r o d u c e d with p e r m i s s io n of t h e co p y rig h t o w n e r. F u r th e r r e p r o d u c tio n prohibited w ith o u t p e r m is s io n .
70
Figure 24. Plan projection indicator (PPI) radar reflectivity (dBz) image of Hurricane
Michael, 2145 UTC 19 October 2000 Oeft) and color-enhanced AMSU-B 89.0 GHz Tb
image, 2304 UTC 19 October 2000 (right). The magenta circle center denotes the
approximate location o f Michael. PPI radar reflectivity image provided courtesy of
Mr. Chris Fogarty, Newfoundland Weather Centre, Environment Canada.
R e p r o d u c e d with p e r m i s s io n of t h e co p y rig h t o w n er. F u r th e r r e p r o d u c tio n prohibited w ith o u t p e r m is s io n .
a
in
a
t
w
ur
iu* u us u* m ui m
r
Uw
m
u
in
us
ui
i5i
iji
iu
Figure 25. Tropical Cyclone Winsome (08P) AMSU-A warm core (A: 55.5GHz, B:
54.94GHz, C: 54.46GHz, D: 53.6 GHz) and AMSU-B 89.0 GHz window channel (E)
R e p r o d u c e d with p e r m i s s io n of t h e co p y rig h t o w n er. F u r th e r r e p r o d u c tio n prohibited w ith o u t p e r m is s io n .
72
•..System ;:
Debbv
Debbv
Debbv
Debbv
Rorence
R orence
Gordon
Gordon
Helene
Helene
Helene
Jovce
Jovce
Keith
Keith
Keith
Keith
Keith
Keith
Keith
Keith
Keith
Keith
Michael
Michael
Michael
Michael
Michael
ST Low
Carlotta
Rosa
Tembin
Saomai
Bebinca
32W
Terri
Winsome
Abiaail
Abiqail
Paula
ST Low
Dera
D ate:
22-Auq
22-Auq
23-Auq
24-Auq
12-Sep
14-SeD
14-SeD
17-Sep
20-Sep
21-Sep
22-Sep
1-Oct
1-Oct
29-Sep
29-Sep
1-Oct
1-Oct
2-Oct
2-Oct
3-Oct
4-Oct
5-Oct
5-Oct
17-Oct
17-Oct
18-Oct
18-Oct
19-Oct
26-Oct
20-Jun
7-Nov
18-Jul
12-Sep
2-Nov
9-Nov
30-Jan
12-Feb
26-Feb
27-Feb
28-Feb
8-Mar
9-Mar
f f M
ATL
ATL
ATL
ATL
ATL
ATL
ATL
ATL
ATL
ATL
ATL
ATL
ATL
ATL
ATL
ATL
ATL
ATL
ATL
ATL
ATL
ATL
ATL
ATL
ATL
ATL
ATL
ATL
ATL
EPAC
EPAC
NWPAC
NWPAC
NWPAC
NWPAC
SO
SPAC
SPAC
SPAC
SPAC
SPAC
SO
i
m
m
1157
2306
1132
0024
1222
131 5
1320
1350
0017
1400
0114
1157
2302
0013
1242
0108
2302
0046
1315
0023
1408
0117
1345
1231
2346
1207
2324
2300
2342
1354
0956
2148
1004
1051
1442
2317
1656
1046
1706
0749
0946
161 1
w
m
1131
2322
1722
0425
1201
1736
1240
1328
2155
132 2
0301
0748
2037
2001
1814
2118
2037
2304
1104
2300
1730
2307
11 0 7
1717
1906
0757
1857
1700
2131
1901
1000
0000
0955
1200
1656
0100
1700
0920
1700
0500
0700
1955
SH W M Si
00 :2 6
0 0 :1 5
0 5 :5 0
04:01
00:21
04:21
0 0:4 0
0 0 :2 2
0 2 :2 2
00:38
0 1 :4 7
04:0 9
02:2 5
0 4 :1 2
0 5 :3 2
0 3 :5 0
02:2 5
0 1 :4 2
02:11
01:23
0 3 :2 2
02 :1 0
02:38
04 :4 6
04 :4 0
04:10
04 :2 7
06:0 0
0 2 :1 2
05:0 7
0 0 :0 4
0 2 :1 2
0 0:0 9
0 1:0 9
0 2:1 4
01:43
0 0 :0 4
01:2 6
00:0 6
02:49
02:4 6
03:4 4
Bill
999
998
1006
1012
991
995
1008
990
1010
1007
1000
1007
1008
1004
1000
966
942
958
979
990
996
988
984
990
988
986
979
966
997
977
1000
993
953
994
1007
984
984
974
997
980
992
973
3RMMSU& ■ m i n t r '-:Onwritif tff iS !
995
998
4
10 0 7
994
-9
10 1 0
997
-4
1005
999
7
990
992
1
1000
991
-5
10 0 7
1002
1
994
992
-4
10 0 5
995
5
994
10 0 6
1
1006
991
-6
1010
1005
-3
100 7
1002
1
1008
1000
-4
1005
996
-5
976
969
-10
975
961
-33
993
982
-35
990
968
-11
10 0 0
990
-10
10 0 4
990
-8
992
979
-4
993
980
-9
986
996
4
984
985
4
98 0
988
6
972
984
7
989
9 62
-23
99 8
995
-1
996
981
-19
10 0 6
996
-6
1001
990
-8
969
959
-16
987
988
7
10 0 6
1006
1
989
986
-5
993
987
-9
972
976
2
995
990
2
990
977
-10
994
987
-2
988
972
-15
ALL(MEAN)
7 .4
ALLfSTD DEV)
9 .7
ATL/EPAC(MEAN)
7 .5
ATL/EPAC(STD DEV)
9.9
O -flsc sHybiMi
"■•hRa-J
1
4
-9
9
-4
7
13
-1
-1
4
-5
1
6
-2
-2
15
5
1
13
9
-6
-3
2
6
1
4
-4
4
-5
-3
-3
m m
1
-11
-11
-2 4
11
0
6
9
4
-6
3
-2
-5
4
2
-4
4
3
-6
6
1
-2
-3
-2
7
3
5
7
5 .7
7 .9
6 .2
8 .0
-24
11
-10
6
9
4
-6
3
-2
-5
4
2
-4
-6
-8
-6
6
1
-2
-3
-2
7
3
5
7
5.3
7 .2
5.5
7 .2
Table 3. Independent test results (n=42). ‘A Time’ is AMSU observation time, ‘O Tim e’ is
in situ observation time, ‘D T im e’ the time differential, ‘O MSLP’ the in situ MSLP, ‘Raw
MSLP’ the raw MSLP estim ate, ‘Ret MSLP’ the retrieved MSLP, ‘O-Raw’ the difference
between observed and raw M SLP, ‘O-Ret’ the difference between observed and retrieved
MSLP, and ‘Hybrid’ the hybrid results using a raw MSLP threshold of 1000 hPa (0.5°C).
All estimates for 2000 excluding SIO/SPAC (2001)
R e p r o d u c e d with p e r m i s s io n of t h e co p y rig h t o w n er. F u r th e r r e p r o d u c tio n prohibited w ith o u t p e r m is s io n .
|
73
6 Analysis Of Errors
The proposed AMSU TC intensity estimation algorithm is not immune to diverse sources
error, many of which have already been discussed. It is important to document the
circumstances leading to optimal AMSU TC intensity estimation algorithm performance as well
as those that degrade estimate skill. First, assuming that the AMSU algorithm is considered
worthy for operational use by civilian or military forecasting agencies, it is vital that analysts
interpreting results recognize potential factors limiting estimate skill so that adequate confidence
limits can be established. Documentation of algorithm limitations will also help direct
subsequent algorithm modification efforts and/or the development of future, satellite-based
sensor technologies designed to meet noted shortfalls —for parochial purposes, TC intensity
estimation.
6.1 Analysis Of Estimate Error And Position
In Section 5, we validated the overall suitability and soundness of the AMSU TC
intensity estimation algorithm methodology through the analysis o f statistical quantities derived
from the entire independent test population (n=31 ATL/EPAC only, n=42 all basins). However,
this does not necessarily imply that algorithm skill is uniformly consistent for all possible
permutations of TC strength and/or position within the AMSU-A scan swath. Some insight is
gained by evaluating independent test performance as a function o f TC position relative to
satellite nadir (Table 4). Consistent with the hypothesis that TC UTWA strength is increasingly
R e p r o d u c e d with p e r m i s s io n of t h e c o p y rig h t o w n e r . F u r th e r re p ro d u c tio n p rohib ited w ith o u t p e r m is s io n .
74
sub-sampled with increased scan angle (|>, AMSU MSLP estimate skill is maximized for TC
position within 3 AMSU-A FOV from nadir. Furthermore, in virtually all cases, explicit
correction for scan geometry and diffraction reduces MSLP estimate error compared to estimates
derived from raw TC UTWA values and dependent regression coefficients. Theory portends that
raw AMSU MSLP estimate bias would be increasingly negative (i.e., UTWA sub sampled) with
increased <j>; however, the largest estimate standard deviations (15.3 hPa and 9.8 hPa) occur for
TCs falling between 3-9 FOV from AMSU-A scan swath nadir. Reasons that contribute to this
discrepancy include the larger individual sample average in situ MSLP observations near the
scan limb (994 hPa) vs. mid-swath averages (981.5 hPa and 990.3 hPa) and the smaller sample
size (n =4 vs. 11), the former enabling accurate specification of TC UTWA amplitude for
reasons previously discussed in Section 1.2.2 and illustrated in Fig. 3.
6.2 AMSU Synchronization
In early July 2000, the NOAA-15 Advanced Very High Resolution Radiometer (AVHRR)
experienced instrument problems leading to the loss of NOAA-15 High Resolution Infrared
Sounder (HIRS), AMSU-A and AMSU-B data. On-board processing combines data from all
sensors and uses the AVHRR to synchronize all spacecraft instrumentation. Accurate
synchronization enables quantitative data analysis by assigning accurate time information to the
beginning of each instrument scan line. Errors in time assigned to individual AMSU-A scan
lines can lead to inaccurate navigation. Assuming that the NOAA-15 satellite is in a circular
orbit, an altitude of approximately 810 km translates to an orbital velocity of 7450 m s'1. As a
R e p r o d u c e d with p e r m i s s io n of th e co p y rig h t o w n er. F u r th e r r e p r o d u c tio n prohibited w ith o u t p e r m is s io n .
75
result, a synchronization error o f 8s (time interval between successive AMSU-A scan lines)
translates to a geocentric displacement error o f approximately 60km, slightly larger than half of
one degree. Observed synchronization error impacts range from imperceptible to obvious was
the case for Hurricane Keith, 1337 UTC 1 October 2000 (Fig. 26).
Under the current methodology, an entire AMSU-A orbit file (approx. 100-110 minutes of
data) is sectorized based on T C position and estimates o f NOAA-15 position using satellite
orbital ephemeris data and the NOAA/NESDIS Brouwer-Lyddane prediction algorithm.
Beginning/end times used to sectorize the entire orbit file are assigned to individual AMSU-A
scan lines by the AVHRR. Synchronization error results in a time differential between true
AMSU-A scan line begin time and the anomalously-assigned AVHRR time. The end result is a
displacement in the along-track. direction of the sector file the magnitude of which is related to
the error in seconds. In this case, the automated AMSU TC intensity estimate was in gross error
due to the exclusion of the entire warm anomaly (A) from the AMSU-A sector used by the
retrieval.
In order to center the AM SU -A sector over Keith (B), a time adjustment (decrement) of 160
s was required leading to an approximate 11° adjustment in the geocentric latitude/longitude
corresponding to the center o f the AMSU-A TC sector file. While this is arguably an extreme
case, it does represent a source o f navigation error and potential exclusion from use during near
real-time processing.
6.3 Instrument Limitations
R e p r o d u c e d with p e r m i s s io n of t h e co p y rig h t o w n e r. F u r th e r r e p r o d u c tio n prohibited w ith o u t p e r m is s io n .
76
AMSU TC intensity estimates are also potentially affected by AMSU-A instrument noise
and the relative broad nature and overlap in AMSU-A weighting functions. AMSU-A instrument
noise (NEAT of 0.25°Kat 290K) contributes to uncertainties in AMSU-A 54.94 GHz Tb (and
therefore ATb) which, for weak TCs, approaches the magnitude of the observed UWTA. In
addition, the relatively broad nature and overlap of frequency-dependent AMSU-A weighting
functions (Fig. 4) makes it possible for lower tropospheric phenomena (e.g., scattering by
hydrometeors) to influence AMSU-A upper tropospheric channels including 54.94 GHz. On rare
occasions, TCs with extremely cold cirrus canopies (commonly referred to as ‘CCCs’ by TC
satellite analysts) are observed to produce cold anomalies in AMSU-A 54.94 GHz (-250 hPa)
and even in AMSU-A 55.3 GHz (-150 hPa) Tb observations largely considered ‘immune’ to
scattering effects. In such cases, the overall impact is a reduction o f the AMSU-A TC UWTA
which adversely affects Gowers) ATb and raises (i.e., weakens) AMSU MSLP estimates.
Another non-negligible potential source of error is related to the atmospheric effects
associated with variable AMSU-A scan angle. As the AMSU-A scans away from nadir, the
atmospheric path length (and optical depth) increases which effectively elevates the level o f the
frequency-dependent emission source. This is commonly referred to as the ‘limb effect.’ In the
troposphere, increased scan angle is usually accompanied by limb ‘darkening’ or reduced Tb due
to the negative lapse rate (i.e., temperature decreasing with increasing altitude). However, due to
the proximity of the 54.94 GHz weighting function peak to the tropopause, increased scan angle
is typically accompanied by limb ‘brightening’ or increased Tb due to the lapse rate reversal in
the lower stratosphere. Although techniques to correct for limb effects are well documented and
generally very effective, the proposed AMSU TC intensity technique is susceptible to inadequate
R e p r o d u c e d with p e r m is s io n of t h e cop y rig h t o w n e r. F u r th e r r e p r o d u c tio n prohibited w ith o u t p e r m is s io n .
77
treatment of the limb effect. The outermost AMSU-A FOV (Fig. 13) are averaged (Benv or term
A in Eqn. 6) and used to define AB in the retrieval. Quite frequently these FOV fall on o r near
the edge o f the AMSU-A scan swath and are therefore susceptible to errors in limb correction.
Furthermore, if the center o f a TC falls on or near the edge o f the scan swath, the combined
effects o f limb correction error and/or radiometric noise can exacerbate the issue of TC UTWA
sub-sampling.
6.4 Environmental Variability
The results of previous MSU-based TC intensity studies all relate TC UTWA magnitude
to the surface pressure anomaly or AMSLP between the TC center and the undisturbed
environment. In this study, the authors have regressed AMSU-A-derived TC UTWA to MSLP,
not taking into account the surrounding environmental pressure distribution. This potential error
contribution is non-trivial especially for TCs forming in regions characterized by lower than
average pressure (i.e., monsoon trough, monsoon gyres, etc), latitudinal variations in
environmental surface pressure, or for systems undergoing extratropical transition. This issue
warrants future study; however, it is desirable that a satellite-based estimate of environmental
surface pressure be used—possibly using temperature profiles derived from the AMSU itself—to
maintain the integrity o f a purely objective, satellite-based method.
Environmental variability may also contribute to AMSU TC intensity estimate error
under any circumstances in which the time differential between AMSU-estimated and in situobserved MSLP is non-zero, especially during periods in which a TC is undergoing rapid
R e p r o d u c e d with p e r m is s io n of t h e cop y rig h t o w n e r. F u r th e r r e p r o d u c tio n prohibited w ith o u t p e r m is s io n .
78
deepening/weakening. Once again, Hurricane Keith (Fig. 26) serves as an excellent example. In
the preceding 24hrs prior to the 1337 UTC 1 October 2000 AMSU MSLP estimate of 957 hPa,
Keith explosively deepened from a 0740 UTC 30 September 2000 aircraft reconnaissance
estimated MSLP of 982 hPa to 942 hPa observed at 0743UTC I October 2000. During the next
24hrs, Keith dramatically weakened (reconnaissance estimated MSLP of 979 hPa at 1104 UTC 2
October 2000) undergoing landfall over the Yucatan peninsula. The nearest aircraft
reconnaissance flight after Keith’s brief period of maximum intensity (2304 UTC 1 October
2000) reported a MSLP of 958 hPa and a 15 nm (27.4 km) diameter eye. While this MSLP
measurement is in excellent agreement with the earlier 1337 UTC AMSU MSLP estimate of 957
hPa (AMSU-B 89.0 GHz eye size of 22.2 km diameter), it is in greater violation of the allowable
6 hr time difference constraints (10 hrs vs. 5 hrs) and therefore is not considered ‘near­
coincident’.
Removing the aforementioned 1337 UTC October 2000 Keith case from the independent
AMSU TC intensity estimate population, which based on previously discussed problems with
synchronization and environmental variability appears justifiable, slightly improves overall
AMSU TC intensity estimate performance indicators (Table 3). However, removal also draws
attention to the dangers of drawing universal conclusions o f AMSU TC intensity estimation
algorithm skill when removal of a single case can moderately influence the statistics drawn from
our relatively small independent population. For lack of additional in situ data, and the fact that
the case met the +/6 hr time difference threshold, the author chose to retain the AMSU MSLP
estimate as part of the independent test results.
R e p r o d u c e d with p e r m i s s io n of t h e co p y rig h t o w n e r. F u r th e r r e p r o d u c tio n prohibited w ith o u t p e r m is s io n .
79
6.5 Estimating TC Position
A recurrent theme throughout much o f this thesis is the AMSU TC intensity estimation
algorithm dependence on precision TC positioning and AMSU-A FOV navigation at the time of
observation. As discussed in Section 1, inaccurate TC positioning at the time o f AMSU-A
observation may result in sub-optimal treatment o f UTW A diffraction, erroneous AMSU-B 89.0
GHz eye size estimates, and in the worst case, entirely exclude the TC UTWA peak warming
observed in AMSU-A 54.94 GHz. An excellent example of retrieval impact based on poor
position assignment is the SIO case of TC Ando (04S), 0347 UTC 6 January 2001. The position
at the time of observation, 18.8°S 54.3°E, was estimated using the method described in Section
4.1 using a 1800 UTC 5 January 2001 JTWC forecast discussion bulletin (WTXS31 PGTW).
Visual inspection o f the 0347 UTC AMSU-B 89.0 GHz color-enhanced Tb image (Fig.
27) indicates that TC Ando’s true movement was further west than JTWC 12hr position forecast
estimates. As a result, the poor eye size estimate (r = 62.4 km) led to an erroneously high
AMSU-derived MSLP value of 978 hPa (weak Safflr-Simpson Category 1 TC) at a time when
Dvorak-based estimates classified Ando a Category 4 (935 hPa/125 kts) TC. Subjective manual
adjustment of TC Ando position to 19.4°S 53.8°E, based on visual inspection o f the 0347 UTC
AMSU-B 89.0 GHz Tb image, yielded more realistic eye size (r =
10.0 km) and MSLP (947 hPa, borderline Category 3/4) estimates more closely resembling
independent Dvorak estimates from NHC/SAB/AFWA.
R e p r o d u c e d with p e r m i s s io n of t h e co p y rig h t o w n e r. F u r th e r r e p r o d u c tio n prohibited w ith o u t p e r m is s io n .
80
6.6 Algorithm Numerics
Finally, several sources of error are possible due to retrieval algorithm inadequacies
including inaccurate modeling of the antenna gain pattern, variability in true TC UTWA
structure not taken into account by the horizontal structure function X, and potential vertical
displacement of the UTWA to an altitude other than that coinciding with the 54.94 GHz
weighting function peak. In its current form, the AMSU TC intensity estimation algorithm is
based on the assumption that peak warming will always occur near 250 hPa. Three years of
AMSU-A TC observations, as well as limited aircraft and radiosonde composite observations
(Kotewaram, 1967, Hawkins and Imbembo, 1976, Nunez and Gray, 1977), generally support this
assumption; however, AMSU-observed TC UTWA vertical structure does depart from
climatology particularly during times of strong vertical wind shear and during periods of
extratropical transition as previously discussed. AMSU-A observations also suggest that strong
wind shear not only causes horizontal displacement and non-axisymmetric warm core but may
also lead to ventilation and/or subsidence-induced lower tropospheric warming unrelated to
either retention of eye wall latent heat release or adiabatic compression associated with the
secondary circulation.
R e p r o d u c e d with p e r m i s s io n of th e co p y rig h t o w n er. F u r th e r r e p r o d u c tio n p rohibited w ith o u t p e r m is s io n .
81
Bias
Ave
STD
N
P
Raw
-0.2
4.8
2.7
Ret
-1.5
2.8
1.7
6
983.5
Raw
-6.5
10.5
15.3
Ret
-4.0
8.0
4
981.5
15
Raw
-4.2
8.4
9.8
Ret
2.5
7.9
6.9
10
990.3
Raw
-4.6
63
43
Ret
4.5
6.4
4.4
11
990.5
Raw
-7.7
73
53
Ret
13
4.0
2.4
11
994.0
Raw
-2.7
7.1
9.5
Ret
-3.1
4.9
53
10
982.7
Table 4. A comparison of AMSU TC intensity estimate skill as a function of TC
position relative to AMSU-A scan swath. Raw and retrieved ‘Ret’ AMSU TC intensity
estimates are binned according to the AMSU-A scan line element ‘ELE’ (range 0-15)
relative to nadir (E=0). ‘Bias’ indicates the sample mean bias (hPa), ‘Ave’ indicates the
sample mean absolute deviation between AMSU estimate and in situ MSLP observation
(hPa), ‘STD’ the standard deviation (or root-mean-square (RMS) error), ‘N’ the sample
size, and ‘P’ the sample size mean MSLP value (hPa).
R e p r o d u c e d with p e r m i s s io n of t h e co p y rig h t o w n er. F u r th e r r e p r o d u c tio n prohibited w ith o u t p e r m is s io n .
82
Figure 26. An example of AMISU-A synchronization error and impact, Hurricane Keith,
1337UTC 1 October 2000. ‘A .’, ‘B \ and ‘D’ represent the AMSU-A 54.94 GHz Tb (°C)
distribution while ‘C ’ depicts th e AMSU-B 89.0 GHz Tb (°C) distribution. AMSU-B was
unaffected by synchronization in this case.
R e p r o d u c e d with p e r m i s s io n of t h e co p y rig h t o w n e r. F u r th e r r e p r o d u c tio n prohibited w ith o u t p e r m is s io n .
83
TROPICAL CYCLONE 04S
AMSU—0 Channel 1 6 (09 GHz) B rig h tn e ss T em perature (C)
S a tu rd a y 6jan01G 06 lim e : 0 3 4 7 lilC
6.9
- 3 .1
- 1 3 .1
- 2 3 .1
—33.1
—43.1
-S 3 .1
- 6 3 .1
- 7 3 .1
- 8 3 .1
- 9 3 .1
Figure 27. Tropical Cyclone Ando (04S) AMSU-B 89.0 GHz Tb (°C) image, 0347 UTC 6
January 2001
R e p r o d u c e d with p e r m i s s io n of t h e co p y rig h t o w n er. F u r th e r r e p r o d u c tio n prohibited w ith o u t p e r m is s io n .
84
7 Summary And Conclusions
7.1 General Comments
The primary goal of this research is the development o f a TC intensity estimation
technique using passive-microwave data from the AMSU. In order for the algorithm to be
operationally viable, it must at a minimum yield MSLP estimates equivalent or superior to those
commonly produced by existing satellite-based techniques. Despite improvements in AMSU-A
horizontal resolution and radiometric accuracy, the aforementioned dependent and independent
test results effectively demonstrate that issues of TC UTWA sub-sampling remain. This study is
the first successful attempt to explicitly correct for the interaction of the TC UTWA and AMSUA antenna gain pattern using an objective, automated technique using both dependent and
independent near-real time data.
The incorporation of AMSU-B 89.0 GHz radiance data as a proxy for TC eye size, and
the notable improvement in correlation between AMSU-A 54.94 GHz-derived TC UTWA and
MSLP, validates Merrill’s (1995) earlier hypothesis concerning the primary cause for the initial
failure of algorithm using the MSU and strongly supports a linear relationship between the two
quantities —a relationship that has been relatively tenuous until now. Furthermore, despite the
relatively naive nature of the AMSU-B eye size algorithm, it’s overall positive contribution to
MSLP retrieval using the proposed methodology in conjunction with AMSU-A data is
undeniable.
R e p r o d u c e d with p e r m i s s io n of t h e co p y rig h t o w n er. F u r th e r r e p r o d u c tio n prohibited w ith o u t p e r m is s io n .
85
The results of the independent test (Table 3) are very promising and are slightly superior
to the operational subjective Dvorak estimates (ATL/EPAC mean error and standard deviations
o f 7.8 hPa and 7.6 hPa respectively) making the proposed AMSU TC intensity estimation
algorithm a strong candidate for future operational use. Early indications of inter-basin
applicability are also encouraging (entire population MSLP mean error and standard deviation
5.7 of 7.9 hPa respectively, 5.3 and 7.2 hPa using a hybrid approach); however, based on the
small sample size (n=l 1), more cases are required to establish statistical significance before the
technique can be considered truly basin-independent.
Operational requirements for 6-hourly TC position and intensity estimates demand that
every AMSU orbit be exploited to maximize temporal coverage. Past microwave-based intensity
estimation techniques (Velden, 1989) constrained TC position to fall within +/- 6° of satellite
nadir in order to mitigate the effects of TC UTWA sampling and achieve MSLP RMS errors of
approximately 8.0 hPa. As a result, sample size was reduced approximately 30% and placed
limitations on operational use of the MSU. In contrast, the proposed AMSU TC intensity
technique effectively resolves this issue by enabling use of virtually all of the available AMSU
passes concomitant with superior results and no artificial position constraints.
Finally, as discussed in Section 5, the proposed AMSU technique offers several unique
capabilities for TC intensity estimation under circumstances in which Dvorak estimates are either
suspect or not available. Many o f these situations involve cases of rapid TC intensification,
weakening and/or interaction with land —circumstances in which the need for accurate
assessment of TC intensity and intensity change are most vital.
R e p r o d u c e d with p e r m is s io n of t h e cop y rig h t o w n e r. F u r th e r r e p r o d u c tio n prohibited w ith o u t p e r m is s io n .
86
7.2 Application Considerations
AMSU TC intensity estimation technique performance is strongly influenced by errors in
storm position and/or eye size (R). While linear extrapolation of the storm position often leads
to an accurate estimate of TC position at the time of AMSU observation, occasionally the results
are less than satisfactory. A method allowing for subjective manual adjustment o f erroneous
automated positioning through either post-analysis of AMSU-B 89.0 GHz Tb data used in the
retrieval or through the inclusion o f superior, satellite-based ancillary TC position information
(i.e., TRMM, SSM/I, QuikSCAT) may lead to improved results. This is especially true for
basins in which TC position estimates are updated only twice daily (i.e., SIO and SPAC) where
Unear extrapolation is particularly vulnerable to non-linearities in storm motion.
Despite the fact that the AMSU retrieval is capable o f handUng systems near the scan
limb, increased AMSU-B 89.0 GHz FOV size often leads to unrealistically large R values (and
corresponding high MSLP estimates) especially for systems with small eyes. Under these
circumstances, the AMSU TC intensity technique may again benefit from the timely availability
o f ancUlary eye size information from additional sateUite resources like TRMM or SSM/I 85GHz
data with superior storm/scan geometries, horizontal resolution and near constant cross-track
resolution or radar observations. The near-real time processing activities of several agencies
(i.e., Naval Research Laboratory (NRL)-Monterey or A ir Force Weather Agency (AFWA))
coupled with the abUity to share TC-specific digital data via high-speed communication
networks offer promise in the future.
R e p r o d u c e d with p e r m i s s io n of t h e co p y rig h t o w n e r. F u r th e r r e p r o d u c tio n prohibited w ith o u t p e r m is s io n .
87
7.3 Future Work And Recommendations
The incorporation o f NOAA-16 AMSU data into UW-CIMSS processing starting in 2001
offers unprecedented insight into short-term TC UTWA morphology including 6 hourly
(nominal) microwave-based TC intensity estimates. Plans are underway to adopt the algorithm
for use with the next-generation DMSP Special Sensor Microwave Imager Sounder (SSMIS).
AMSU processing will also be coordinated with NHC/JTWC to identify suspect genesis regions
through coupling with the Automated Tropical Cyclone Forecast (ATCF) package currently used
to initiate NRL-Monterey TC multi-spectral satellite products. Several additional studies are
planned including the impact of the adjusted AMSU 54.94 GHz radiances on TC eye soundings,
2-D radial wind distribution profiles and numerical weather prediction model vortex
initialization. Finally, the CAMEX-4 field campaign scheduled for 2001 offers an outstanding
opportunity to validate AMSU-derived products in near-real time.
Throughout the course of this study, several questions and potentially interesting AMSU
TC research topics emerged including 1) the development o f a hydrostatic approach to TC MSLP
estimation (including treatment o f hydrometeor scattering) using eye soundings, 2) a multiplechannel AMSU TC MSLP estimate scheme, 3) improvements to the objective TC positioning
scheme and 4) an exhaustive analysis of multi-year, multi-satellite AMSU TC data focused
towards the development o f a warm core climatology. The latter would likely help improve the
results o f this study by improving the a priori specification o f the UTWA structure and might
possibly lead to a AMSU-based TC short-term intensity prognostication capability.
R e p r o d u c e d with p e r m is s io n of t h e cop y rig h t o w n e r. F u r th e r r e p r o d u c tio n prohibited w ith o u t p e r m is s io n .
88
REFERENCES
Bender, M. A., R. E. Tuleya and Y. Kurihara, 1987: A Numerical Study of the Effect of Island
Terrain on Tropical Cyclones. Mon. Wea. Rev.. 115, 130-155.
Dvorak, V. F., 1975: Tropical cyclone intensity analysis and forecasting from satellite imagery.
Mon. Wea. Rev.. 103, 420-430.
Dvorak, V. F., 1984: Tropical Cyclone Intensity Analysis Using Satellite Data. NOAA Tech.
Rep. NESDIS 11, National Environmental Satellite Data Information Service, Washington, D.C.,
47pp.
Frank, W. M., 1977a: The structure and energetics of the tropical cyclone, I: Storm structure.
Mon. Wea. Rev.. 105, 1119-1135.
Goldberg, M. D., 1999: Generation of retrieval products from AMSU-A: methodology and
validation. 10th International TO VS Study Conference, Boulder, Colorado, 27 January—2
February.
Gray, W. M., 1979: Tropical Cyclone Intensity Determination Through Upper-Tropospheric
Aircraft Reconnaissance, Bull. Amer. Soc.. Vol. 60, 1069-1074.
R e p r o d u c e d with p e r m i s s io n of t h e co p y rig h t o w n er. F u r th e r r e p r o d u c tio n prohibited w ith o u t p e r m is s io n .
89
Grody, N. and W. C. Shen, 1982: Observations of Hurricane David (1979) using the microwave
sounding unit. NOAA TR, NESS 88, NOAA, Washington D. C., 52pp.
Hastenrath, S., 1991: Climate Dynamics o f the Tropics, Kluwer Academic Publishers, 488 pp.
Hoinka, K. P., 1999: Temperature, Humidity and Wind at the Global Tropopause. Mon. Wea.
Rev.. 127, 2248-2265.
Holland, G. J., 1980: An analytic model of the wind and pressure profiles in hurricanes. Mon.
Wea. Rev.. 108, 1212-1218.
Holton, J. R., 1992: An Introduction to Dynamical Meteorology, 3rd Edition, Academic Press,
511 pp.
Jordan, C. L., 1961: Marked changes in the characteristics o f the eye of intense typhoons
between deepening and filling stages, J. Meteor.. 15, 779-789.
Kidder, S. Q., W. M. Gray, and T. H. Vonder Haar, 1978: Estimating tropical cyclone central
pressure and outer winds from satellite microwave sounder data. Mon. Wea. Rev.. 108, 14581464.
R e p r o d u c e d with p e r m i s s io n of t h e co p y rig h t o w n e r. F u r th e r r e p r o d u c tio n prohibited w ith o u t p e r m is s io n .
90
Kidder, S.Q., M.D. Goldberg, R.M. Zehr, M. DeMaria, J.F.W. Purdom, C.S. Velden, N.C.
Grody, and S.J. Kusselson, 2000: Satellite Analysis o f Tropical Cyclones Using The Advanced
Microwave Sounding Unit (AMSU), Bull. Amer. Meteor. Soc.. 81, 1241-1259.
Knaff, J. A., R.M. Zehr, M.D. Goldberg, and S.Q. Kidder, 2000: An Example of Temperature
Structure Differences in Two Cyclone Systems Derived from the Advanced Microwave Sounder
Unit. Weather and Forecasting. 15, 476-483.
Koteswaram, P., 1967: On The Structure Of Hurricanes In The Upper Troposphere And Lower
Stratosphere. Mon. Wea. Rev. , 95, 541-564.
Merrill, R. T., 1984: A comparison of large and small tropical cyclones. Mon. Wea. Rev., 112,
1408-1418.
Merrill, R. T., 1994: Typhoon Monitoring Using Passive Microwave Observations.
USAF/AFGL (Phillips Laboratory) Tech. Rep. 95-2012, 28 pp.
Merrill, R. T., 1995: Simulations of Physical Retrieval of Tropical Cyclone Thermal Structure
Using 55-GHz Band Passive Microwave Observations from Polar-Orbiting Satellites. J. Appl.
Meteor.. 34, 774-787.
R e p r o d u c e d with p e r m i s s io n of t h e co p y rig h t o w n e r. F u r th e r r e p r o d u c tio n prohibited w ith o u t p e r m is s io n .
91
Rodgers, C. A., 1976: Retrieval of atmospheric temperature and composition from remote
measurements of thermal radiation. Rev. Geophvs. Space Phvs.. 14, 609-624.
Schloemer, R. W., 1954: Analysis and synthesis of hurricane wind patterns over Lake
Okechobee, FL. Hydromet Rep. 3 1 ,49pp. {Govt. Printing Office, No. C30.70:31].
Shewchuck, J. D. and R.C. Weir, 1980: An evaluation of the Dvorak technique for estimating
tropical cyclone intensity from satellite imagery. NOCC/JTWC 80-2, USNOCC, JTWC,
COMNAVMARIANAS, 25 pp.
Simpson, J., 2000: Personal communication.
Spencer, R. W. and W. D. Braswell, 2000: Tropical Cyclone Monitoring with AMSU-A:
Estimation of Maximum Sustained Wind Speeds. Submitted to Mon. Wea. Rev.
Times Atlas Of The World, 1988, 7th Edition, John Bartholomew & Son, Air Weather Service
Technical Library.
Velden, C. S. and W. L. Smith, 1983: Monitoring tropical cyclone evolution with NOAA
satellite microwave observations. J. Climate Appl. Meteor.. 22, 714-724
R e p r o d u c e d with p e r m i s s io n of th e co p y rig h t o w n er. F u r th e r r e p r o d u c tio n p rohibited w ith o u t p e r m is s io n .
92
Velden, C. S., 1989: Observational analysis of North Atlantic tropical cyclones from NOAA
polar-orbiting satellite microwave data. J. Appl. Meteor.. 28, 59-70.
Velden, C. S., B. M. Goodman, and R. T. Merrill, 1991: Western North Pacific tropical cyclone
intensity estimation from NOAA polar-orbiting satellite microwave data. Mon. Wea. Rev.. 119,
159-168.
Velden, C. S., T. L. Olander, and R.M. Zehr, 1998: Development of an Objective Scheme to
Estimate Tropical Cyclone Intensity from Digital Geostationary Satellite Infrared Imagery.
Weather and Forecasting. 13: (1), 172-186.
Velden, C. S., 2001: Personal communication.
Weatherford, C. L., and W. M. Gray, 1988: Typhoon structure as revealed by aircraft
reconnaissance. Part I: Data analysis and climatology. Mon. Wea. Rev.. 116, 1032-1043.
R e p r o d u c e d with p e r m i s s io n of t h e co p y rig h t o w n er. F u r th e r r e p r o d u c tio n prohibited w ith o u t p e r m is s io n .
Документ
Категория
Без категории
Просмотров
0
Размер файла
4 578 Кб
Теги
sdewsdweddes
1/--страниц
Пожаловаться на содержимое документа