A Reanalysis of the 1916, 1918, 1927, 1928, and 1935 Tropical Cyclones of the North Atlantic Basin

1
A Reanalysis of the 1916, 1918, 1927, 1928, and
1935 Tropical Cyclones of the North Atlantic Basin
A Masters Defense By David A. Glenn
May 12th 2005
2
Introduction

• Thesis Objectives
• What is a tropical cyclone or hurricane?
• What is the hurricane database (HURDAT), how is
  it used, and why does it need revision?
• Data used in the reanalysis
• Methods for determining track and intensity
• Changes for 1916, 1918, 1927, 1928, and 1935
• Results of the reanalysis
• Results of the Accumulated Cyclone Energy (ACE)
  index

3
Objectives

• The purpose of this project is to reanalyze
  HURDAT for the years of 1916, 1918, 1927, 1928,
  and 1935. This involves revising the original
  HURDAT record by changing the track and intensity
  of tropical cyclones in the original hurricane
  database.
• II. The second objective is to assess the effects
  of revising HURDAT on the accumulated cyclone
  energy (ACE) index. The computed ACE index values
  indicate the amount of tropical activity
  therefore, if major alterations are suggested for
  HURDAT then the changes will be apparent in the
  ACE index computations.

4
The Hurricane

• A hurricane is defined as a warm core non-frontal
  synoptic-scale cyclone possessing winds greater
  than 64 knots.
• Low-level closed circulation about a well defined
  center.
• Maintained through latent heat of evaporation of
  warm ocean water.

Source NOAA COMET
5
Saffir-Simpson Scale
6
HURDAT

• The National Hurricane Center houses the North
  Atlantic Basins HURricane DATabase (HURDAT)
• Track and intensity estimates provided every 6
  hours
• Originally designed for statistical tropical
  cyclone guidance for the Apollo Space program
  (1960).
• Free online at www.nhc.noaa.gov/pastall.shtml

7
HURDAT

• Uses
• - Aid in forecast development and verification.
• - Climate trend assessment long term trends,
  seasonal forecasts, etc.
• - Building code standards and insurance rates
  for coastal communities
• - Risk assessment for emergency managers
  (recurrence intervals)
• HURDAT was not originally designed for all of
  these uses

8
Why Revise HURDAT?

• Why revise?
• - HURDAT contains many systematic and random
  errors (Neumann 1994)
• - Limitations in the original hurricane database
• Once daily estimates before 1930.
• - Advances in the understanding of hurricanes
  and analysis techniques
• - Lack of exact hurricane landfall parameters
  (i.e. location, time, intensity, central
  pressure, storm surge, and radius of maximum
  winds)

9
Beaufort Scale
Source Fitzpatrick, 1999
10
Random Error

• Original classification errors -

28320 10/30/1935 M10 6 SNBR 624 NOT NAMED
XING1 SSS2 L 28325 10/30 0 0 0
0326 609 35 0328 618 35 0330 628 40
0 28330 10/31331 637 40 0333 645 45
0334 653 50 0335 666 55 0 28335
11/01337 681 60 0337 697 60 0336 712
65 0335 726 70 0 28340 11/02332 738
70 0328 750 70 980322 757 70 0316 759
70 0 28345 11/03309 760 70 0302 760
70 0295 761 70 0286 764 65 0
28350 11/04277 771 65 0270 779 65
0265 787 65 0258 803 65 973 28355
11/05253 811 65 0249 825 65 0249 835
65 0251 843 65 0 28360 11/06252 850
60 0255 857 60 0260 863 55 0264
867 50 0 28365 11/07268 870 45 0275
871 40 0280 869 35 0281 858 30 0
28370 11/08278 847 25 0277 842 20
0276 837 15 0275 834 15 0 28375
HRCFL2
This is an example of a random error that is
easily corrected through the reanalysis process.
Category 2 83-95 kt.
11
Errors in HURDAT
12
Anemometer Bias

• Prior to the 1920s, Robinson 4-cup anemometers
• were used that generally overestimated winds
  speeds
• greater than 30 knots (Fergusson and Covert,
  1924).
• For example, a minimal hurricane force wind of 64
• kt actually reduces to 50 kt after bias
  correction.
• Most anemometers were compromised or
• destroyed before sampling the highest winds.

13
Original HURDAT Limitations

• From the period 1886 to 1930, the original HURDAT
  was created from a once daily 12 GMT track and
  intensity estimate (Jarvinen et al, 1984). The
  00, 06, and 18 GMT track and intensity estimates
  were interpolated from two days 12 GMT estimates.
• Problematic for storms undergoing rapid
  intensification or rapid decay.

14
ACE Index

• Accumulated Cyclone Energy (ACE) index is a tool
  used to quantify the amount of tropical activity.
• Uses wind velocities in HURDAT
• S (55 kt)2 (60 kt)2 (65 kt)2 (75 kt)2
  16475 kt2.
• The ACE index is considered a physically and
  statistically reasonable measure of the overall
  tropical activity for a given hurricane season
  (Bell et al, 2000).

15
Reanalysis Steps

• 1.) Gather all available raw data into a single
  database
• 2.) Determine track
• 3.) Determine intensity
• 4.) Document revisions (metadata file)
• Each day of each storm is reanalyzed. The
• process is usually fairly tedious and takes 2-3
• hours to discuss the changes for a single storm.

16
Data Sources

• Comprehensive Ocean-Atmosphere DataSet (COADS)
  (Woodruff et al, 1987)
• Monthly Weather Review (MWR) American
  Meteorological Society
• Original Monthly Records (OMR) Station Data
  from U.S. Weather Bureau or partner observation
  stations
• Historical Weather Maps (HWM) Created by the
  U.S. Navy and U.S. Weather Bureau
• National Climate Data Center individual ship
  and station observations
• Historical hurricane reports Barnes (1998a,b),
  Connor (1956), Dunn and Miller (1960), Ho (1989),
  Ho et al (1987), Ludlum (1963), Perez (2000),
  Tannehill (1938), Tucker (1982)

17
COADS
Source Woodruff et al, 1987
18
MWR
19
MWR
20
MWR
21
OMR
22
HWM
The example to the right is an annotated
Historical Weather Map. Some raw data are
gathered from the HWM but it is mainly used as
the synoptic background for plotting all raw data
from the sources of COADS, MWR, OMR, and other
historical reports. Observations are plotted
for the time interval 10-14Z to maximize data
near the synoptic time.
23
Data Aggregation and Conversion
Wind units are converted to knots and barometric
pressures are converted to millibars to be
consistent with current analyses.
24
Track Determination
The process for defining the center of a
tropical cyclone from observed winds. Two ship
observations (indicated by the red wind barbs)
roughly indicate the tropical cyclone center
(where the two black lines cross) assuming
cyclonic flow with a 20o inflow angle (Landsea
et al, 2004).
25
Time-Series Low-Pressures
Time-series low-pressures maps are useful for
landfalling tropical cyclones. The central path
follows the lowest pressures.
26
Intensity Determination

• It is important to locate the nearest lowest
  barometric pressure or observed wind speed (to
  the perceived center). When the analyst does not
  have observed wind speed, atmospheric pressure
  measurements (central or peripheral) can be used
  to provide estimates of the maximum sustained
  wind speeds in a tropical cyclone.
• Pressure-wind relationship
• Radius of maximum winds (RMW)
• Inland Wind Decay Model
• Collaborating pressures

27
Atlantic Pressure-Wind Relationship

• GMEX Wind (kt)10.627(1013-Po)0.5640 Sample size
  664 r0.991
• lt 25N Wind (kt)12.016(1013-Po)0.5337 Sample
  size 1033 r0.994
• 25-35N Wind (kt)14.172(1013-Po)0.4778 Sample
  size 922 r0.996
• 35-45N Wind (kt)16.086(1013-Po)0.4333 Sample
  size 492 r0.974
• Kraft Wind (kt)14.000(1013-Po)0.5000 Sample
  size 13

28
RWM

• Vickery et al. (2001) derived an equation for the
  radius of maximum winds (RMW) of north Atlantic
  tropical cyclones expressed in terms of central
  pressure (Po), environmental pressure (Pn), and
  latitude (L)
• ln(RMW) 2.363 0.00005086( Po - Pn)2
  0.0394899(L)
• RWM of 25 to 50 percent less than climatology had
  an adjustment of five knots while systems that
  had extremely smaller RMW (greater than 50
  percent) were adjusted by 10 knots (Landsea et
  al, 2004).

29
Kaplan and DeMaria (1995) Inland Wind Decay Model

• The Inland Wind Decay Model was used to determine
  the best track after landfalling tropical
  cyclones.
• Only used in the absence of observed winds.
• Designed for landfalling tropical cyclones over
  the southeastern U.S. where nearly all the
  regions within 150 nmi of the coast have
  elevations less than 650 ft.
• Not applicable for higher terrains such as Cuba,
  Hispaniola, and Mexico.

30
Original 1916 HURDAT
31
(No Transcript)
32
1916 Revisions
33
1916 Revisions cont.
34
(No Transcript)
35
(No Transcript)
36
1918 Revisions
37
(No Transcript)
38
(No Transcript)
39
1927 Revisions
40
(No Transcript)
41
(No Transcript)
42
1928 Revisions
43
(No Transcript)
44
(No Transcript)
45
1935 Revisions
46
Metadata

• It is estimated that the storm became a hurricane
  on September 1st around 12 UTC just south of
  Andros Island in the
• Bahamas. Over the next 24 hours the system nearly
  doubled in intensity from 65 kt to 120 kt on the
  2nd. A central
• pressure of 924 mb at 21 UTC of the 2nd implies
  winds of 132 kt from the southern pressure-wind
  relationship. 140 kt is
• chosen for 18 UTC HURDAT as later evidence
  suggests a smaller than usual radius of maximum
  winds (RMW) for this
• system. Available observations estimate landfall
  to have occurred near 02 UTC on the 3rd at Craig
  Key. Intensity at
• landfall has been estimated from pressure
  observations at Craig Key 892 mb indicating 155
  kt winds from the
• southern pressure-wind relationship or 140 kt
  from the subtropical pressure-wind relationship.
  An average of these two
• pressure-wind relationships (since the system was
  near the border of the two) suggests winds of
  around 148 kt.
• Regarding the probable RMW at landfall in the
  Keys, "it is estimated that the calm center was
  perhaps 8 miles in
• diameter" (MWR). Available observations suggests
  the RWM to be 6 nmi (Ho et al). RMW were also
  noted as 6 nmi
• from Schwerdt et al. 6 nmi RMW is smaller than
  what might be expected from climatology of this
  central pressure and
• landfall latitude (9-10nmi - Vickery et al). Thus
  this tiny RMW suggests a significant boost from
  the standard pressure
• wind relationship and 150 kt is chosen for 00 UTC
  in HURDAT. Peak observations of lowest pressure
  (and implied
• highest winds) were observed at 02 UTC and
  therefore an addition to the normal HURDAT 6 hour
  time interval is
• suggested such as in the case with Hurricane
  Andrew (1992). 160 kt is chosen for 02 UTC in
  HURDAT. This maintains
• the hurricanes intensity as Category 5 but
  increases its previous peak intensity from 140 kt
  to 160 kt. After passing the
• Keys, the hurricane moved toward the northwest
  and eventually north on the 4thbefore making its
  second landfall in
• northwest Florida. An observed peripheral
  pressure of 941 mb early on the 4th suggests at
  least 119 kt winds from the

Metadata can take several hours per storm and is
the bulk of the work for this thesis.
47
Reanalysis Results Synopsis
In general, track estimates during the 1910s,
1920s, and 1930s needed minor revisions.
Intensity estimates for the period saw the
greatest change with the single largest change of
30 knots occurring in 1916 and 1927. The addition
of four tropical storms to the original HURDAT
record was offset by the deletion of three
tropical storms, increasing the total number of
storms for the period to 41. Only one storm had
no revisions in intensity from the original
HURDAT. No major changes were found for systems
that made landfall in the United States. A new
database for tropical cyclones that affected the
United States was also completed allowing future
HURDAT users to obtain more information about
such tropical cyclones.
48
Tropical Cyclone Activity

• Klotzbach and Gray (2004) - normal
• 9.6 named systems, 5.9 hurricanes (2.3 major
  hurricanes)
• Bell et al. (2000) - ACE Index
• 6.5 to 10105 kt2
• - Based on climatology from 1950-2000

49
ACE
50
Discussion

• Mexico data needs to be utilized.
• This process can be repeated if newly discovered
  data becomes available.
• The project can also be utilized in other basins
  (e.g. West Pacific Ocean)
• This research will be evaluated by the Best Track
  Change Committee for final acceptance of
  revisions to HURDAT.

51
Acknowledgments

• I would like to thank Dr. John Rodgers for his
  endless guidance and assistance during the
  masters program.
• I would like to also sincerely thank Dr. Chris
  Landsea for allowing me to be part of such a
  meaningful project. His assistance, direction,
  and insight were invaluable throughout the
  masters degree.

52
Acknowledgments

• Special recognition is also needed for several
  research assistants at the Hurricane Research
  Division, namely Mr. Steve Feuer (HRD), Mr.
  Lenworth Woolcock (FIU), and Mr. Lyle Hufstetler
  (UM).
• I would also like to thank my beautiful fiancé
  Melissa for her endless support!
• Also, thanks to Mountain Dew and Dave Matthews
  Band for their energy and support throughout the
  last two months.