Hurricane Frederic, Aug. 29-Sept. 13, 1979

[From the U.S. Government Printing Office, www.gpo.gov]

Natural Disaster Survey Report
IUOASTAL ZONE ~~~, ~o
INFORMATION CENTER
Hurricane Frederic

� Aug 29 - Sept 13, 1979 qESo,

U.S. DEPARTMENT OF COMMERCE
National Oceanic and Atmospheric Administration
National Weather Service, Silver Spring, Maryland

QC
945
. H8
1979

COASTAL ZONE
INFORMATION CENTER

Xft FOREWARD

Hurricane Frederic, one of the most severe hurricanes
to strike the U.S. in the past decade, caused enormous damage
s ~to parts of Alabama, Florida, and Mississippi as it came ashore
on Wednesday, September 12, 1979. An estimated quarter of a
million people heeded the warnings from NOAA's National Weather
Service offices and the advice of state officials. The extremely
~2 low death toll is due to the diligence of many officials,
agencies, volunteers, and governments working in a coordinated
and unified way.

NOAA sends a survey team into an area after a major
-4- natural disaster occurs. Their objective is to determine how
d4 the warning system performed. This report has several findings
of
and recommendations. It is gratifying that most are relatively
~0 minor and it is heartwarming to find that so many of the actions
< ~taken undoubtedly saved hundreds of lives.

C,) Richard E. Hallgren
Director, National Weather Service

Property of csc Library
-r
0

,~~Cl C2ll XECTIO

,.-

C)J ~Cover: This NOAA satellite view shows Hurricane Frederic at 11:00a.m. CDT on September 12, 1979, when the
O8~ ~ ) center was about 180 miles south of Mobile, Alabama.

co~ OU.S. DEPARTMENT OF COMMERCE NOAA
_ ~~ ~ ~'z~~ , CDCOASTAL SERVICES CENTER
2234 SOUTH HOBSON AVENUE
z , i'- CHARLESTON, SC 29405-2413

PREFACE

The NOAA survey team included:

Michael Glazer, Office of Policy and Planning, NOAA
Earl Estelle, Public Services Branch, NWS
Stanley Pearse, Atmospheric Services, NOAA
Stanley Spivey, Meteorological Services Division, NWS Southern Region
Doyle Cook, Meteorologist in Charge, WSFO, Charleston, W. Va.
Joseph Bird, Research and Development, NOAA
Dane Clark, Satellite Operations, NESS
Herbert Lieb, Consultant, Disaster Preparedness and Public Affairs

From September 18-21, 1979, the team visited the five
counties where the worse damage occurred -- Mobile and Baldwin
Counties, AL; Escambia County, FL; Jackson and Harrison Counties, MS.

In addition to interviewing NWS personnel at the WSO's
at Mobile, AL, and Pensacola, FL, the NHC and WSFO, New Orleans,
the team visited mayors, civil defense directors, county commission-
ers, television and radio executives, and newspaper editors and
reporters in the five counties.

The team is grateful for their time, patience, and inval-
uable insights.

GLOSSARY

NOAA: National Oceanic and Atmospheric Administration
NWS: National Weather Service, NOAA
NHC: National Hurricane Center, NWS
NMC: National Meteorological Center, NWS
WSFO: Weather Service Forecast Office, NWS
WSO: Weather Service Office, NWS
MIC: Meteorologist in Charge, NWS
NESS: National Environmental Satellite Center, NOAA

CONTENTS

Page

Forward ................................................. i

Preface ................................................. ii

Executive Summary ....................................... iv

Chapter 1. Hurricane Frederic - A Brief History ........ 1

Chapter 2. The Warning Chain:
Preparedness-Dissemination-Coordination-
Public Response ........................... 3

Chapter 3. Evaluation of Official Forecasts and
Guidance .................................. 5

Chapter 4. Monitoring .................................. 20

Aerial Reconnaissance
Satellite Support

EXECUTIVE SUMMARY

Hurricane Frederic, with winds reaching 145 mph, moved
over Dauphin Island (near the mouth of Mobile Bay) and inlane
just west of Mobile, Alabama, between 10 a.m. and 11 p.m. CDT
on Wednesday, September 12, 1979. Storm tides of 8 to 12 feet
above normal were reported from Pascagoula, Mississippi to
western Santa Rosa Island, Alabama. The damage estimate of
$2.3 billion makes Frederic the costiliest hurricane ever to hit
the United States, exceeding the $2.1 billion attributed to the
widespread floods from Hurricane Agnes in 1972. Based on inform-
ation from preparedness officials, 250,000 persons were safely
evacuated in advance of Frederic. It was the first hurricane to
strike Mobile directly since 1926. It was the most intense
hurricane of this century for the area from Mobile, Alabama, to
Pascagoula, Mississippi.*

That only five deaths have been directly attributed to
Frederic in the U.S. is testimony to the high level of preparedness
in the state, county, and city governments along the coast.
The survey team received the highest praise for the NWS's fore-
casts and warnings during Frederic.

The team's findings and recommendations follow:

*Finding 1

The combination of extremely good forecasts and warnings and
excellent hurricane preparedness resulted in a very low death
toll. Major credit must be given to several key Southern Region
employees and local emergency preparedness officials.

Recommendation 1

Awards and/or other recognition should be given to appropriate
individuals. (This recommendation was carried out at a special
ceremony in Mobile, Alabama, on December 13, 1979.)

'Finding 2

NOAA Weather Radio transmissions were lost at Gulfport, Mississippi
(after 6:30 p.m.); Mobile, Alabama (around 6:30 p.m.); Pensacola,
Florida (6-7 p.m.); and Panama City (7 p.m.). All outages due
to loss of electrical power at the transmitter site and through-
out the respective cities were due to storm damage. NWR at
Panama City was also struck by lightning. NWR is classified as
the sole Government-operated radio system to provide direct
warnings into private homes for both natural disasters and nuclear
attacks.

*Paul J. Hebert, National Hurricane Center, NWS, Atlantic
Hurricane Season of 1979, Monthly Weather Review, July 1980.

Recommendation 2

Emergency power must be provided for NWR transmitters, particularly
in vulnerable areas. Measures to provide lightning protection
should also be pursued.

*Finding 3

Tide gage data was lost early at both Mobile and Pensacola. In
Mobile the data was lost due to phone line outages. In Pensacola
the high tide destroyed the well and made the system inoperable.
These are critical losses which inevitably seem to occur during
every hurricane when the data is needed most.

Recommendation 3

Some form of backup or emergency system needs to be developed and
implemented to guarantee access to data.

*Finding 4

While shaky at times, communications between the Weather Service,
local officials, and the news media never failed. However,
reliable back-up communication systems to local officials and the
media are lacking. Amateur radio groups played a crucial role in
providing emergency communications. Emergency power kept WSO's
Mobile and Pensacola in operation.

Recommendation 4

Providing reliable back-up communications to insure continuing
contact with the many vital action agencies is crucial, yet
technically difficult and costly. Time after time, amateur radio
groups, both Ham and CB, have effectively filled the need by
operating out of the WSFO or WSO. We should adopt this approach
as a matter of official policy and encourage the establishment of
such arrangements in appropriate areas of the country where they
do not exist now. It is an effective, economical, and practical
solution.

*Finding 5

Information on Hurricane Frederic reached Pensacola-area public
officials from several sources including the Naval Air station,
Pensacola, the Governor of Florida, and from one or more radio
stations with drops on the Florida NAWAS circuit. Radio and
television were outstanding in broadcasting the latest forecasts,
warnings, and advice until they were knocked out as the storm
approached the coast. There was some evidence of "editorializing"
by some broadcast personnel that led to misinformation. However,
the survey team found that it was not a significant problem.

Recommentation 5

Information about the hurricane should reach the public through
a single coordinated source. Multiple sources of information lead
to confusion and can have dire effects. A review should be made
of the dissemination of hurricane information in Florida to see
whether any changes are needed. 'With the media on NAWAS, it is
no longer a "private line" used by the decision-making action
teams. If this represents a new FEMA policy, other means of
having "hot-line"1 communications between NWS offices, Civil
Defense, law enforcement and other local action agencies must be
-found.

*Finding 6

The MIC, WSO Mobile, believed it would have been more effective
if he had had control of the NWR transmitter in southeastern
Mississippi which covers part of his warning area.

Recommendation 6

Transferring control of the McI-enry NWR transmitter to Mobile
WSO as the MIC requested has much merit in a warning situation.
The staff of the Mobile WSO would, however, be hard-pressed to
handle another transmitter with its current manpower. The
possible transfer should be explored.

*Finding 7

A local university meteorologist volunteered his services and
stayed at the Mobile Emergency Operating Center (EOC) throughout
the storm as a staff consultant to the Director of Civil Defense
and the County Commissioners. He helped them to interpret and
reconcile information they were receiving from WSO Mobile and
other sources (e.g., reports from state police patrol cars) as
Frederic was moving onshore. The meteorologist's assistance was
highly touted as contributing significantly to the feeling that
local officials were on top of the situation.

Recommendation 7

The feasibility of detailing meteorologists to the EOC's at the
center of the action should be explored.

*Finding 8

The high state of readiness in Alabama and Florida as Frederic hit
reflected a lot of hard work over a long period. Both the Pensacola
and Mobile MIC's are spending their own time and money to get the
preparedness job done. Their travel budget is inadequate to
cover their actual expenses and the demands on their time at the
office are such that they often have to use their own time to go
out and make contacts. The need for a principal assistant was
also raised by both MIC's.

Recommendation 8

We must provide more support to our MIC's and their staffs to do
the preparedness job.

*Finding 9

Although the majority of the satellite support was extremely
effective, two deficiencies were revealed which should be remedied.
Both of these were ongoing problems and not solely related to
the operations during Hurricane Frederic. Nevertheless, these
deficiencies did affect the overall performance of the meteorolo-
gical warning system. The first involved a complaint by the
National Hurricane Center. This involved NESS's recent change
from producing photographic prints from negatives to an automatic
picture display using UNIFAX machinery. Since the switchover,
NHC feels dissatisfied with the quality of their satellite data.
The second deficiency is the communication system NESS and NWS
use to pass satellite-derived rainfall estimates from satellite
meteorologists to NWS meteorologists and hydrologists. The
present phone relay system is deficient because often these estimates
fail to reach all NWS offices that could use this data. The
River Forecast Centers at New Orleans and Atlanta did not receive
these estimates during Hurricane Frederic.

Recommendation 9

The possibility of improving the quality of satellite imagery
available at NHC should be explored.

*Finding 10

The three aircraft in NOAA are not enough to sustain operations
for high density/accuracy data. An additional five aircraft
with this capability are needed for the Western Atlantic/
Caribbean/Gulf of Mexico area. This is a mutual NOAA/DoD
requirement.

Recommendation 10

NOAA must support the U.S. Air Force plan for an Improved Weather
Reconnaissance System to the maximum extent possible.

*Finding 11

The single side-band NWS inter-office emergency communications
equipment at both Mobile and Pensacola went down during Frederic
while conventional systems, for the most part, stayed in operation.

Recommendation 11

The inter-office emergency communications program should be reviewed.
If the single side-band system is to be maintained at coastal offices,
review of ways in which its survivability can be increased should
be considered.

Vii

CHAPTER I

HURRICANE FREDERIC - A BRIEF HISTORY

Hurricane Frederic's history spanned more than 2
weeks beginning in the Eastern Atlantic.

The system that gave birth to Frederic was detected
by satellite south of the Cape Verde Islands. Ship and
satellite data showed the formation of a depression on August 29.
It reached tropical storm strength on August 30 -- the same day
that Hurricane David was passing close to Puerto Rico and a day
before David devastated the Dominican Republic. David killed
more than 1,200 in the Dominican Republic. The history of
David and the news attention it received in the U.S. during its
Caribbean passage was a vital factor in encouraging public
response along the Gulf Coast as Frederic moved inland. The
NOAA National Disaster Survey Report, Hurricanes David and
Frederic as They Concerned Puerto Rico and the Virgin Islands,
August 26 - September 5, 1979, prepared by the NWS's Southern
Region details and compares the two storms and their impact
in the Caribbean.

Paul Hebert's article, cited earlier, also shows the
meteorological connection between Frederic and David.

On September 1, satellite photographs showed that
Frederic reached hurricane strength about 400 miles east of the
Lesser Antilles but David's influence weakened Frederic to
a tropical storm on September 2. From then until September 9
it moved generally westward, decellerated, and weakened more.

Frederic turned toward the northwest during the next
48 hours. On Sunday, September 9, forecasters at the National
Hurricane Center indicated that Frederic was over Cuba, regaining
strength, and posed a threat to the U.S. On the 10th, Frederic
was upgraded to a hurricane and was located 150 miles southwest
of Key West over the western end of Cuba.

Based on the NHC advisories, the Pensacola and Mobile
MIC's alerted and began formal contacts with the Civil Defense
directors and others early on Monday, September 10, more than
2 full days before hurricane conditions occurred in coastal
areas. A hurricane warning was issued by the National Hurricane
Center in a special advisory at 9:30 pm. Tuesday, September 11,
approximately 22 hours ahead of observed hurricane conditions.
It read in part, "Hurricane warnings are being issued tonight
in order to alert local governments of the need for quick
action early Wednesday." This warning was quickly distributed
by the WS01s. Several interviewees stated that it was fortunate
that the National Weather Service did not wait until Wednesday

morning to issue the warning. It would have made successful
evacuation much more difficult -- or perhaps impossible. Sub-
sequent advices from both the National Hurricane Center and the
WSO's kept all concerned current on the progress of the storm.
Local action statements by the WSO's; particularly with respect
to specific evacuation recommendations, were generally very well
done and closely coordinated with local officials.

A hurricane watch was issued for Panama City, Florida,
to Vermillion Bay, Louisiana, at 6 p.m. Tuesday, September 11.
The storm was moving slowly northwest at this point and was 450
miles southeast of New Orleans.. At 9:30 p.m. a hurricane warning
was issued for Panama City, Florida, to Grand Isle, Louisiana,
in a special advisory.

Wednesday, September 12, an estimated 250,000 people
evacuated coastal areas during the day. The center of Frederic
moved over Dauphin Island, Alabama, and inland just west of
Mobile, Alabama, between 10 and 11 p.m. Landfall was at the
midpoint of the area placed under a Hurricane Warning earlier.
The maximum storm surge, usually the big killer, was observed
at 12 feet right near the middle of the forecast range of 10
to 15 feet. Highest observed winds were approximately 145 miles
per hour at Dauphin Island.

Thursday, September 13, the storm-weakened quickly as
it moved inland.

CHAPTER II

THE WARNING CHAIN - Preparedness - Dissemination - Coordination -
Public Response

There was a high level of hurricane preparedness in
the county/city governments along the coast. This undoubtedly
was a key factor in-the low death toll for a strong hurricane --
one which almost completely destroyed, or very heavily damaged,
several communities along the coast and caused over $2 billion in
property damage. The counties had good evacuation plans and used
them effectively in close coordination with the National Weather
Service and the news media. In fact, Mobile County issued an
updated and strengthened plan less than one month prior to
being hit by Frederic.

Certainly, the fact that Hurricane Camille (1969) and
Eloise (1975) had affected the same general section of the Gulf
within the last decade and that Hurricane David had killed so
many in the Caribbean less than 2 weeks earlier helped motivate
people to evacuate. However, the major credit for the preparedness
effort must be given to the NWS's preparedness program and to
the efforts of Southern Region officials, particularly
Neil Frank (Director, National Hurricane Center, Miami), David
Barnes (Meteorologist in Charge, WSFO New Orleans), Raymond Barnes
(MIC, WSO Mobile), and Phyllis Polland (MIC, WSO Pensacola),
in cooperation with local Civil Defense directors, the news
media and others.

Almost unanimously the Civil Defense directors attributed
the success achieved in Frederic to the year after year preseason
education and awareness programs and drills; to the close
alliance between Civil Defense and the NWS at the state-county-
city levels; and to the total cooperation and support of the
news media.

The NWS considers its preparedness and coordination
activities indispensible to the total warning program. Ever
since Camille in 1969 struck viciously at the Mississippi coast-
line a "tteam" of dedicated people from the states from Texas
to Florida have poured all their energies into a heightened public
awareness campaign as well as the exercising of disaster and
evacuation plans. The "team" consisted of NWS employees, civil
defense directors, law enforcement officials, the news media,
Red Cross workers, volunteer amateur radio operators, educators
and others.

Hurricane Eloise's impact along Florida's coastline
near where Frederic came in proved the value of such preparedness
efforts and even intensified those coordinated efforts.

The NWS, together with all county/city officials
responsible for public safety, hold conferences before the
hurricane season together with news directors of radio and
television and newspaper editors and reporters. The purpose is
to update disaster plans, coordinate communications systems and
enlist the help of the news media in a continuing education
venture.

Those type meetings were held in every area where
Frederic hit and have taken place annually since Camille hit a
decade before.

The role of radio and television before and during
Frederic's onslaught -- and afterwards, was superb. Their dedi-
cation to hurricane awareness and education before the season,
along with newspapers carrying safety information and advice,
were largely responsible for the wise and orderly public actions.

The NWS's dissemination channels, NOAA Weather Radio
(NWR), the NOAA Weather Wire Service (NWWS), and the use of
FEMA's National Warning System (NAWAS), worked well throughout
the storm until the NWR's at Gulfport, Pensacola, Mobile, and
Panama City lost commercial power. Communications between the
WSO's Pensacola and Mobile, the public officials, and the media
never completely failed.

The Mobile County Civil Defense had also installed a
Plectron radio system throughout Mobile County. This system is
used to distribute emergency information without dependence on
land lines. It links the Mobile County Emergency Operations
Center, the National Weather Service, and radio and TV stations
and certain other offices. The system produces a recorded message
at the receiving end and was used quite effectively during
Frederic. However, it was also partially out during the height
of the storm. The recipients included the local Emergency
Broadcast System (CPCS-1) station, WABB. The WABB chief engineer,
Mr. Rene Stiegler, felt that there should be a backup system to
link his station with WSO Mobile as a precautionary measure.

In accordance with a prearranged plan, an amateur radio
group set up an operation at WSO Pensacola early Wednesday
afternoon and maintained a communications link with the Emergency
Operations Center and the shelters until Thursday afternoon.
This was very helpful to all concerned. A'similar arrangement was
placed in operation at WSO Mobile. Mr. Prentiss Baughman,
Civil Defense Director of Baldwin County, said that the amateur
group was his only link with WSO Mobile after his phones went
out late Wednesday afternoon. He considered their contributions
invaluable.

CHAPTER III

EVALUATION OF OFFICIAL FORECASTS, OBJECTIVE FORECASTS, AND NMC
GUIDANCE

A. National Hurricane Center (NHC) Evaluation

1. Introduction

Seven objective techniques are available to the hurri-
cane forecaster preparing an advisory. Forecasts through 72
hours are based on the four synoptic times of 0000, 0600, 1200, and
1800 GMT. IThe statistical techniques, NHC67, NHC72, HURRAN,
and CLIPER' are available every 6 hours (the HURRAN occasionally
does not have enough cases), and are received in time for the
forecaster to use them for the advisory based on the synoptic
time for which they were run. This means that the forecaster
can use the forecast positions from these techniques for the
same valid times as his 12, 24, 48, and 72 hour forecasts.

The three dynamic techniques - NHC73, SANBAR, and MFM
(NMC Movable Fine Mesh model 1) - require upper air data and are
not received until 6 to 9 hours after the 0000 or 1200 GMT
synoptic times on which they are based. This does not allow
them to be used as guidance for the advisory based on the synoptic
time for which these techniques are run, and requires the fore-
caster to interpolate or extrapolate to obtain forecast positions
compatible with the advisory being prepared when they become
available. In essence, this requires 36 and 48 hour forecasts
based on dynamic techniques to be as accurate as 24 and inter-
polated 36 hour official forecasts if one considers such guidance
in the most critical application - the issuance of hurricane
watches and warnings.

These considerations together with the impact of
directional or speed biases should be remembered in evaluating
the following tables and comments.

2. Verification

The following tables give various comparisons of the
* ~official and objective forecast technique errors based on the
official best track given in the included preliminary report
on Hurricane Frederic. Table I gives the verification for all
forecasts for the official and objective techniques for the
periods when Frederic was of tropical storm or hurricane
intensity. Table 2 gives the verification of the six cases
available for all forecasts when Frederic was in the Gulf of
Mexico.

1Attachment 1 gives a brief description of the various objective
techniques.

A significant consideration in evaluating the utility of
the various forecasts is the forecast errors during critical
forecast periods. Table 3 gives the average forecast errors for
two periods. The first is from 9/2/0000 GMT to 9/4/0000 GMT
when Frederic was approaching Puerto Rico and the Virgin Islands.
The second is the period from 9/10/0000 GMT to 9/12/1800 GMT
when Frederic was in the Gulf of Mexico approaching the U.S.
mainland. It is not known in advance which forecasts will be
exactly the critical ones for issuing hurricane watches and
warnings, because the forecasts have such large differences in
projected landfall times. However, a post-analysis can be made
from the best track based on actual time of landfall backwards
to the forecasts actually available to the forecaster 24 and 36
hours before landfall. These, of course, are the truly critical
individual forecasts upon which hurricane warnings and watches
are nominally issued. Table 4 gives these forecasts with those
for a watch circled and a warning boxed.

3. Discussion

In Table 1, results for the MFM are unrealistic for
comparison with the other techniques, the seven cases comprising
less than 20% of the official forecasts, and all occurring in
the Gulf of Mexico where data is plentiful. Table 2 gives a
homogeneous sample for the 6 cases 'in the Gulf of Mexico for which
the MFM and all other techniques were available. In this compari-
son the MFM has the poorest verification at 12 hours and second
poorest at 24 hours. However, it becomes the best at 48 hours.
The statistical techniques show the reverse with the official
and other dynamic techniques somewhere in between. Similar results
have been noted previously in larger samples.

Excluding the MFM, Table 1 shows the official forecasts
with the minimum forecast errors for all time periods except
the 72 hour NHC73 which was slightly better.

While detailed directional and speed bias analysis is
not available for this evaluation, the following comments can be
made with particular emphasis on forecasts in the Gulf of Mexico:

a) Official forecasts showed a slight bias to the right
of the observed track during early forecasts, especially for the
extended periods, but were generally on track as the center
approached the coast. Forecasters tend to forecast the center a
few tenths of a degree of latitude beyond that expected (faster)
to take into account the onset of hurricane conditions several
hours before the center makes landfall.

b) The NHC67 forecasts showed biases similar to the
official forecasts early in the period, but were very good as
the center approached the coast. The NHC72 is so affected by
the changes in analysis at NMC in recent years that forecasters
place little credence in it beyond 24 hours at present. Relatively

good verification for time periods of 12 and 24 hours is a result
of the large contribution of persistence-available from several
other techniques. After the NHC73, the NHC72 used to be one of
the forecasters' best guidance tools! Guidance frmteHRA
technique was good to excellent out to 72 hours as Frederic
approached the northern Leeward Islands and very good out to 24
hours from the time Frederic left Cuba to its U.S. mainland
approach. The CLIPER forecasts were not so good approaching the
Leeward Islands and Puerto Rico but verified well approaching
the mainland. However, they suffered from a strong bias to the
right (east), while the HURRAN had a slight bias to the left (west).

c) The NHC73 forecast verification appears good to
excellent throughout 72 hours both approaching the Leeward Islands
and the U.S. mainland. However, forecasts in the Gulf of Mexico
were all to the right (east) of the observed track and too slow.
This illustrates the importance of verifying the speed and direction
bias and landfall accuracy implications in terms of watch and
warning areas, based on the availability of the techniques and
their critical time frame forecasts. As mentioned earlier, in
the case of the dynamic techniques - MFM, SANBAR, NHC73 - the 36
hour and 48 hour forecasts are those available for 24 hour
warning and 36 hour watch decisions! SANBAR forecasts were
erratic approaching the Leeward Islands and Puerto Rico., In
the Gulf 6f Mexico they were slow but pretty much along the
observed track. The MFM model appears to have developed a
quirk whereby the forecast (vortex) goes west of the observed
initial motion for 12 to 24 hours and then recurves and acceler-
ates to cross the observed track by 48 hours. This contributed
to a slow bias for all forecasts.

4. Summary of Forecast Evaluation

A study by Neumann of NHC has shown the tendency of
the statistical techniques to do better in the shorter time
frames and the dynamic techniques in the longer time frames.
However, it is very difficult for the operational forecaster to
"hook-up"' the positive aspects of the different objective guidance
when the dynamic techniques have strong initial directional and
speed biases. For westward and northward moving storms/hurricanes
threatening land, official forecasts try to stay ahead of the
center and slightly to the left as a "least regret" philosophy
because of 'the obvious disastrous consequences of predicting a
* ~landfall too late or out of the hurricane warning area. The
tendency for most objective techniques to have a right bias
and be slow in critical Situations can have a compromising effect
on the forecaster's thinking and consequently the official
forecast.

While Tables I through 4 are for a single storm with a
relatively small number of cases, they are quite representative
in almost all cases of typical distributions of forecast errors
for the various types of forecasts for various samples and time

periods. As always, at the time of each individual forecast
the forecaster does not know which technique (even his own)
will turn out to be the best choice.

5. NMC Numerical Prog Guidance

The numerical prog charts 'Limited-area Fine Mesh. (LFM),
Primitive Equation (PE), Barotropic, were examined for-tbe period
9/10/1200 GMT to 9/12/1200 GMT which coincides with the second
critical period considered previously for the official and
objective techniques. The guidance from these prog charts was
evaluated from two points of view. The first is to use the
low centers or vorticity maxima to track the hurricane itself,
and the second to use the prevailing synoptic scale featuers
other than the hurricane itself to make subjective estimates
of their effect on the hurricane.

In the case of Frederic, the numerical guidance
for the first aforementioned purpose contributed little directly
to the forecast. By 24 hours, the P.E. and Barotropic models
showed no evidence of the hurricane low which was present in
the initial analysis for any of the five forecast packages!
The LFM surface low always went west and southwest during the
48 hour period, the 700 mb low too far west and too slow, and
the 500 mb low disappeared with the vorticity maximum making
landfall only twice, and well east of the observed track. The
700 mb and 500 mb hurricane vortex appeared to merge with a cold
low near the Texas coast during latter forecast times as the
cold low moved east and northeast. Earlier prog packages showed the
low remaining stationary near the central Texas coast but it
came eastward as a stronger than forecast low during latter
forecast periods. Not surprisingly, the MFM and LFM appear to
have similar problems in the early portions of the forecast
period.

The forecasters were able to make the more general use
of the NMC prog guidance by noting the presence of the long wave
troughs and ridges, and, to a lesser extent, the influence of
the short waves in the westerlies as they interacted with the
hurricane. In this case, the guidance precluded to a high degree
of confidence a turn of Frederic to the west, and to a lesser
degree, a critical landfall in the New Orleans area.

6. Storm Surge Guidance

The SPLASH objective storm surge models were run at
0800 GMT on September 12 to obtain guidance for the storm surge
values to be used in the 1000 GMT advisory on Frederic. The
SPLASH I was run for a landfall 45 miles east of Gulfport,
Mississippi, or just west of Mobile Bay with an expected central
pressure of 940 millibars moving northnorthwest at 11 knots.

The accuracy of these forecast parameters determines the accuracy
of the storm surge profile along the coast. In this case the
landfall forecast was very good with the central pressure 946
millibars and the other parameters about as used. The model
gave a peak surge of 11.5 feet above MSL on the open coast
south of Mobile. The advisory forecast tides of 10 to 15
feet above normal near and to 100 miles east of where the center
makes landfall. The wording-of this statement is to allow for
higher tides at the heads of bays and estuaries and to allow for
* ~~forecast error and also not specify the landfall point too far
in advance. The highest tides reported thus far in the Mobile area
range from 8 to 12 feet above MSL.

An interesting note is that this hurricane with the same
parameters would result in a model forecast of 19 feet above MSL
near Bay St. Louis if the center made landfall just east of
New Orleans.

The guidance furnished by SPLASH I in this instance was
excellent.

TABLE 1. Initial position errors are displacement errors in nautical
miles for all forecasts of Frederic while it was of tropical
storm or hurricane intensity. Number of cases.are in parentheses.

INITIAL
POSITION 12-HR 24-HR 36-HR 48-HR 72-HR

OFFICIAL 20 37 71 147 217
(48) (43) (43) (0) (33) (28)

NHC67 19 47 97 141 172 283
(44) (42) (39) (35) (31) (26)

NHC72 20 44 87 152 261 347
(49) (47) (44) (40) (36) (31)

HURRAN 20 39 89 127 185 223
(43) (42) (41) (38) (35) (28)

CLIPER 20 37 81 119 157 225
(49) (47) (44) (40) (36) (31)

NHC 73 19 37 88 135 151 201
(21) (20) (19) (17) (15) (14)

SANBAR 19 42 80 117 184 344
(23) (22) (21) (19) (17) (15)

MFM ll 35 66 91 127
(7) (7) (7) (7) (7)

TABLE 2. Initial position errors and displacement errors in nautical
miles for the 6 cases on Frederic in the Gulf of Mexico
available for all forecasts.

INITIAL
POSITION 12-HR 24-HR 36-HR 48-HR 72-HR

OFFICIAL 12 29 55 160

NHC67 12 20 29 142

NHC72 12 18 57 252

HURRAN 12 23 66 225

CLIPER 12 20 59 202

NHC73 12 20 51 164

SANBAR 12 29 58 217

MFM 12 36 65 142

TABLE 3 Individual initial position errors and displacement errors in
nautical miles for critical forecasts as Frederic approached
1) Puerto Rico and 2) the United States mainland, Forecasts
which would be used for a hurricane watch are circled and for
a gale or hurricane warning are boxed.

1. PUERTO RICO
DATE/TIME INITIAL
(GMT) POSITION 24-HR 36-HR 48-HR

9/3/00 66 INTERPOLATED 1
OFFICIAL 9/3/06 12 414 t9
9/3/12 17 1 72 259

9/3/00 66 12 1 238
NHC67 9/3/06 12 93 229
9/3/12 17 121 252 497

9/3/00 66 90 186
NHC72 9/3/06 12 33 157
9/3/12 17 4 152 187

9/3/00 66 97 252
HURRAN 9/3/06 12 82 8 226
9/3/12 17 85 138 244

9/3/00 66 150 369
CLIPER 9/3/06 12 76 258
9/3/12 17 72 140 243

NHC73 9/2/12 6 33 41
9/3/00 66 67 1Oi

SANBAR 9/2/12 6 48 108
9/3/00 66 141 25 29
2. UNITED STATES
DATE/TIME INITIAL
(GMT) POSITION 24-HR 36-HR 48-HR

9/11/12 5 INTERPOLATED
OFFICIAL 9/11/18 13 9
9/12/00 5 1 47 143
9/12/06 5 24 204

9/11/12 5 6 58
NHC67 9/11/18 13 41 99 140
9/12/00 5 241 79 114
9/12/06 5 32 49 123

9/11/12 5 38 264
NHC72 9/11/18 13 61 258
9/12/00 5 7 179 334
9/12/06 5 73 192 403

TABLE 4 Average initial position errors and displacement errors in
nautical miles ;5r critical forecasts as Frederic approaches
Puerto Rico and the United States mainland. Number of
cases are in parentheses.

I Leeward Islands, Virgin Islands, Puerto Rico

INITIAL
POSITION 12-HR 24-HR 36-HR 48-HR 72-HR

OFFICIAL 24 (9) 36 (9) 72 (9) - (O) 188 (9) 289 (7)
NHC67 24 (9) 45 (9) 108(9) 166(9) 204 (9) 353 (7)

NHC72 24 (9) 44 (9) 71' (9) 139'('9) 165 (9) 190 (7)

HURRAN 24 (9) 38 (9) 76 (9) 120(9) 165 (9) 224 (7)

CLIPER 24 (9) 42 (9) 95 (9) 166(9) 227 (9) 320 (7)

NHC73 23 (5) 31 (5) 71 (5) 111(5) 109 (5) 64 (4)

SANBAR 23 (5) 30 (5) 75 (5) 143(5) 211 (5) 312 (4)

2 United States

INITIAL
POSITION 12-HR 24-HR 36-HR 48-HR 72-HR

OFFICIAL 13 (12) 23 (12) 49 (12) (0) 152 (11) 233 (7)

NHC67 13 (12) 33 (12) 77 (12) 77 (12) 125 (11) 274 (7)

NHC72 13 (12) 24 (12) 59 (12) 119(12) 230 (11) 323 (7)

HURRAN 13 (11) 28 (11) 72 (11) 137(11) 137 (11) 306 (6)

CLIPER 13 (12) 28 (12) 71 (12) 135(12) 209 (11) 336 (7)

NHC73 15 ( 6) 20 ( 6) 48 ( 6) 105( 6) 159 ( 6) 230 (4)

SANBAR 15 (6) 32 (6) 61 (6) 123( 6) 234 ( 6) 471 (4)

MFM 12 (7) 38 (6) 70 (6) 88( 6) 120 (6) (0)

TABLE 4 CONTINUED.
UNITED STATES

DATE/TIME INITIAL
(GMT) POSITION 24-HR 36-HR 48-HR

9/11/12 5 61 217
HURRAN 9/11/18 13 69 297
9/12/00 5 7 156 267
9/12/06 5 W 195 336

9/11/12 5 53 198
CLIPER 9/11/18 13 85 Q 236
9/12/00 5 150 196
9/12/06 5 152 249

NHC73 9/11/00 18 28 105
9/11/12 5 54 129

SANBAR 9/11/00 18 68 130 2
9/11/12 5 68 13

MFM 9/11/00 18 121 129
9/11/12 5 39 72

Attachment 1

Objective Forecast Techniques

The following tropical cyclone prediction models were
used at the National Hurricane Center for forecasting motion on
an operational basis:

1. NHC-67 (Miller, Hill, Chase, 1968). A stepwise screening
regressing model using predictors derived from the current and
24-hour old 1000, 700, and 500 mb data, and includes persistence
during the early forecast periods.

2. SANBAR (Sanders and Burpee, 1968). A filtered barotropic
model using input data derived from the 1000 to 100 mb pressure
weighted winds. The model requires the use of "bogus" data in
data-void areas. The system was modified by Pike (1972) so that
the initial wind field near the storm would conform to the
current storm motion.

3. HURRAN (Hope and Neumann, 1970). An analog system using as
a data base the tracks of all Atlantic tropical storms and
hurricanes dating back to 1886.

4. CLIPER (Neumann, 1972). Stepwise multiple screening regression
using the predictors derived from climatology and persistence.

5. NHC-72 (Neumann, Hope, Miller, 1972). A modified stepwise
multiple screening regression system which combines the NHC-67
concept and the CLIPER system into a single model.

6. NHC-73 (Neumann and Lawrence, 1973). Similar in concept to
the NHC-72 except it also used the "perfect prog" and MOS
(model output statistics) methods to introduce NMC (National
Meteorological Center) numerical prognostic data into the
prediction equations.

7. NMC MFM MODEL (Hovermale, 1975). A ten-level baroclinic
model which uses a moving fine mesh (MFM) grid nested within
the coarser NMC fixed grid primitive equation (PE) model.
It is capable of predicting both track and intensity changes.

---
; '� 9/13

HURRICANE FREDERIC
9/12 ) Aug. 29 - Sep. 14,1989

........ Tropical Depression
-- --Tropical Storm
9/10 d ! Hurricane
>.9/9 �, x x x Extra Tropical
9/8
9/7
9/..... 9/6

9/5
9/4 /3

_,~~~~ ".o..9/2

8/31. - - .8/30
8/31

B. NMC Quantitative Precipitation Forecasts

The swath of heaviest rainfall was to the left of the
observed track of Frederic' s circulation center, in contrast to
David where the heaviest rainfall occurred to the right of its
track.

Frederic' s rainfall maximum of 8" - 10" over southeast
Mississippi tapered off to 5"1 - 6"1 as the decaying storm center
moved northward west of the Appalachians.

All numerical precipitation guidance was available
throughout the storm. The areal extent of 1"1 rainfall was fairly
well forecast by the Limited Fine Mesh (LFM) but was unable
to indicate either the magnitude of the steep gradient of the
rainfall. The MOS probability of precipitation amount (POPA)
guidance largely resembled the explicit solutions, with the
exception that it added substantially sized areas of spurious
and erroneous I" - 2"1 rainfall. The MFM precipitation forecasts
showed larger amounts than the LFM (in fact, at times too large)
but overall were not very good at predicting either speed of
the maximum or the axis of heaviest rainfall.

NMC's subjective QPF's were by far the best guidance
issued for rainfall amounts. These forecasts correctly indicatedl
8" or more associated with landfall and consistently showed a
5"1 - 6" forecast along and slightly to the left of the predicted
track. Early forecasts of the northward extension of the rains
were slow, corresponding to the slowness of the forecast movement
of Frederic. Frederic was forecast slightly too far to the
east in the Appalachians, which helped cause a corresponding
small error in the placement of the relatively narrow maximum
rainfall.

Excessive rainfall potential forecasts from Forecast
Division were timely and accurate, adjusting quickly to expected
deviations from the forecast track. For example, the 0730Z
scheduled outlook indicated possible rainfall in excess of 5"1
from Meridian, Mississippi, to central West Virginia in an
area about 90 miles wide, completely consistent with the
predicted track of Frederic. At 1430Z the expected axis of
maximum rainfall was correctly shifted west about 70 miles.
Maximum indicated rainfall of 5"1 - 8" was correct. Later that
evening an adjustment for speed was indicated which correctly
brought excessive rainfall into western New York by 12Z the
following morning.

Coordination calls were held with most of the affected
WSFO's concerning expected rainfall. The quality, credibility
and timeliness of the scheduled forecasts and specials undoubtedly
significantly reduced the possible number of calls.

In summary, the LFM provided perhaps the best numerical
guidance for the extent of I" - 2"1 rainfall. Scheduled 24-hour
period QPF's were consistent with the official forecast track
and correctly indicated the rainfall amounts, although typically
were a little to the east and a little slow. Excessive rainfall
potential forecasts -- both scheduled and unscheduled -- showed
the flexibility desired in adapting to errors in circulation
forecasts by being able to modify the important aspects of
earlier forecasts.

/ 14 Sep

-r~~~-

i ~ ~~~II

13 Sep I

Frederic
for
I U\ rz \ ~~~~~~~00 Z
Sep # II S Observed
* Forcasted

~ s12 Sep 9
I

S1110Sep ,

) 14 Sep__
ooz

1 3 Sep
I,7

Frederic
for
12 Z
Observed
0 Forcasted

12 Sep S

c1 Sep

CHAPTER IV

MONITORING

A. Aerial Reconnaissance

The National Hurricane Operations Plan states that
the "Department of Defense (DoD) will attempt to fulfill all
Department of Commerce (DoC) requirements; however, based on
stated DoC needs, DoD will normally be prepared to generate
up to five reconnaissance aircraft sorties per day. Requirements
exceeding this capability will be executed on a resource
permitting basis. Research aircraft of NOAA's Research
Facilities Center (RFC) may be diverted to fulfill urgent
operational requirements." The Director of the National
Hurricane Center is responsible for identifying the total
operational requirements for hurricane reconnaissance in the
Atlantic, Caribbean, Gulf of Mexico, Eastern Pacific, and
Central Pacific. He is responsible for advising the Chief,
Aerial Reconnaissance Coordination, All Hurricanes, (CARCAH)
of these requiements. The CARCAH is responsible for coordination
and the final preparation of the Tropical Cyclone Plan of the
Day (TCPOD) and for scheduling of aircraft required to meet the
provisions of the Plan. The CARCAH will coordinate the TCPOD
with NHC, Keesler Weather Reconnaissance Coordination Center
and the RFC before publication. The RFC may be requested to:

(1) Augment the U.S. Air Force (USAF) for operational
aircraft reconnaissance with high-density/accuracy data, when
storms are within 24 hours of landfall of the continental
United States.

(2) Augment capabilities for USAF aircraft reconnaissance
when DoC needs exceed the capabilities of DoD resources.

(3) Assume responsibility for hurricane reconnaissance
over foreign airspace that may be restricted for military
operations.

(4) RFC may also conduct research flights which assume
an operational responsibility to the hurricane forecast centers.
As is apparent all of this requires extensive effort on the part
of CARCAH to insure that operational requirements are fully
met, recognizing problems of maintenance, logistics, communi-
cations, and in the case of Frederic, evacuation of an operational
base (Keesler).

A review of the aerial reconnaissance activity, based
upon the records retained by CARCAH, reveals that 49 scheduled
fixes were requested by NHC for Frederic and all were obtained
on time. An additional 18 extra fixes were obtained. To
accomplish this 28 missions were flown with a total of 273 hours.
The 920th Weather Reconnaissance Group (WRG) flew 15 missions;
the 53rd Weather Reconnaissance Squadron (WRC) flew 8 missions;
and the RFC flew 5 missions.

This outstanding record resulted from hard work and
fine cooperation by all individuals and units involved. CARCAH
coordinated activities to meet the reconnaissance requirements
of NHC and on several occasions was able to maintain the
continuous support required by exchanging reconnaissance
commitments among the various participating units.

The first scheduled reconnaissance mission on Frederic
was flown by the 53rd WRS on September 2, 1979, with the 1200Z
position at 15:37N and 57:13W. Lowest sea level presure was
996 mb with a maximum surface wind of 70 knots. The 53rd flew
the next two missions.

The 920th WRG flew the next two missions with the storm
decreasing in intensity as it moved over Puerto Rico. The storm
continued to move in its weakened condition with aerial reconnais-
sance maintained at six hourly intervals. The track was over
Hispanola and the length of Cuba. Finally it freed itself
from Cuba later on September 10th. Mission 19 on Frederic was
flown by the 920th with the surface pressure 985 mb and surface
winds of 45 knots. By September 11 at 1210Z the storm had
quickly intensified to hurricane force with 980 mb surface
pressure and surface winds of 95 knots. This was the first
Frederic mission flown by the RFC.

The RFC flew five missions among the last nine that
were flown on Hurricane Frederic. The high density data from
the research aircraft are extremely useful, especially in the
24 hours prior to landfall. The high density and precision
information on location, wind speed, and direction at flight level
are needed for input to newly developed numerical prediction
models. These new models use grid spacings of about 30 Km and
produce significant improvements in predictions of both storm
path and intensity. An example of the data which are routinely
available to the NHC from NOAA's aircraft is shown in Figure 1.
These data are transmitted from the aircraft via satellite to
Wallops Island, Virginia, and then via landline through the
National Environmental Satellite Service (NESS) computer in
Washington, D.C., and on to NHC. At NHC the data are plotted
by an on-line computer. This system allows for a rapid visual
display of this information.

It is essential that some of the USAF aircraft that
reconnoiter storms be equipped to provide more precise and detailed
data similar to the data available on the RFC aircraft. At this
time the maximum winds reported -by USAF aircraft are estimated
since the observed winds are an average over about a ten minute
time interval. The winds in a storm can make some significant
changes. in ten minutes.

For example, in NOAA flight 2709 on Frederic 9/13/79
(Figure 1) from 02107 to 02207 the wind increased from 25 knots
to 95 knots then decreased to 65 knots. This would be observed

+11+0024
0 +09 1835 +09

+12 "+10 -00
+12, 1"
+12 S-+10 -01
+12 . +1080
+13 +1I1-03E 0

+15\-,n456+i os'
+1 828 ; a
-0446 &+IoSio0

-389
+1 824
tJ4410 1814
-0195
1816

NOAR82 2009 FREDERIC

9/I11/79 SCALE 1/1500000
24K TRUE ST 22.5K
NATIONAL HURRICANE CENTER. MIAMI. FLA.

+21402 +21 -1420 coSAEleooo

17 0 0 9 ~ ~~~~TRLIE AT 22.5K 1iy
'~~t~~;1391 ~~~j NATIONAL HURRICANE CENTER, MIAMI. FLA 3 ''- ''
+19 '0210I

+17 '0211

I~~4161

213 D1 +17 -0312
+1 10234
C7 890 17 -0286

+Is-as+18 0215 +7'%

19-0643 +703
+1r0217 ~~~~~~~~~~~~+1S o2nca
f~\2;560 +1 0-

L\~>44 +1? -0373

+i~1044 221

+ 222 9/13/79

by USAF reconnaissance aircraft as 68 knots and the maximum wind
estimated to be something higher than that. But the detailed wind
speed distribution within that 10 minutes (or about 55 miles) is
lost. This is important information to the forecasting model as
well as the analysts at NHC.

Some areas along the coast cannot evacuate in such a
timely manner as did Mobile citizens. Many places require more
than 24 hours for such an orderly evacuation. This requires high
density and accurate information while the storm is still 36 to
48 hours from landfall. This cannot be accomplished with only
the three NOAA aircraft. Only with dedication and good luck was
RFC able to generate the five missions in two days for Hurricane
Frederic. USAF must also be equipped to provide this data for
the longer time needed for specific and more advanced warnings
required for these areas.

B. Evaluation of Satellite Systems

The National Environmental Satellite Service (NESS)
provided continuous and comprehensive support to the National
Weather Service during Hurricane Frederic. Geostationary Operational
Environmental Satellite (GOES) data displays and communication
networks performed effectively at all NESS and NWS field offices.
Gridding accuracy was quite acceptable -with only minor difficulties.

I. NWS Support

(1) The Kansas City Satellite Field Service Station
(SFSS) and the Gulf Support Unit provided Weather Service Forecast
Offices with detailed information prior to and after landfall.
Satellite Interpretation Messages (SIMS) and numerous telephone
coordination calls provided guidance for marine forecasts, local
forecasts, and special weather statements issued by forecast'offices.

The Gulf of Mexico SIMS, issued every 6 hours included
information about position, movement, and intensity of Hurricane
Frederic. Satellite meteorologists carefully described atmospheric
conditions surrounding the hurricane to provide forecasters impor-
tant information over sparse data regions. SIMS also included
high-level and low-level cloud motion vectors prepared by the
Interactive Processing Group (IPG) of NESS. The Gulf Support Unit
provided telephone coordination calls to the New Orleans WSFO
which supplemented SIMS messages and kept forecasters aware of
the latest satellite indicated trends.

(2) The Miami SFSS provided the National Hurricane
Center (NHC) with timely and detailed analysis of Hurricane Frederic
from August 29 to September 13, 1979. GOES data were utilized to
determine storm location, intensity and movement. Hurricane
forecasters were briefed routinely, and constantly were informed
of the latest changes. SFSS meteorologists also provided input

parameters, calculations, and interpretation for the NHEML rainfall
estimate model. This model produced rainfall potential guidance
for NHC.

Examination of maximum wind information available to NWS
forecasters revealed several noteworthy facts. Figure I shows
maximum wind information from reconnaissance and satellite analysis
during the 15 hour period preceding landfall. Even though NHC
maintained 115 kts in the official advisories, reconnaissance
and satellite information depicted the extreme variability exper-
ienced as the storm approached the coastline.

The Miami SFSS used the current Dvorak Satellite Tropical
Cyclone Classification System during Frederic. The final product
of this classification system produced an intensity (CI) number or
maximum wind estimate (Figure 1). This maximum wind estimate,
passed to NHC, relates closely with the official advisories. But
since reconnaissance information is usually considered "ground
truth," the CI number did not accurately represent the short term
variability of the storm. (In fairness to the system, the CI
number concept was not designed to define the short term variability
of tropical cyclones).

Another part of the Dvorak classification system did
correlate positively with aircraft wind measurements (Figure 1).
This relatively new phase of the classification system utilizes
enhanced infrared (EIR) imagery and a quasi-objective decision
ladder. There is also some evidence that trends in these EIR
intensity measurements actually preceded similar trends in maximum
wind strength as measured by reconnaissance aircraft.

(3) The Synoptic Analysis Branch (SAB) together with the
Kansas City SFSS supplied timely and extremely accurate satellite-
derived rainfall estimates to the NWS (Figure 2.a). Rainfall
estimates were especially useful for NWS forecasters because
hurricane conditions substantially reduced the effectiveness of
the rainfall reporting network. WSFO's used these estimates,
together with radar data, to issue Flash Flood Warnings. The
maximum satellite estimate of 12 inches was used directly in a
Flash Flood Warning issued by the Jackson WSFO. This figure was
used to represent a maximum potential and to update the 5 to 10
inch rainfall forecast carried in the Hurricane Advisories.
Post-analysis of the maximum rainfall area indicated an 11 inch
gage near Pascagoula, Mississippi, had overflowed during the
hurricane (Figure 2.b).

II. Media Support

The Miami SFSS supplied special prints to the UPI and
the SAB periodically prepared special prints of Hurricane Frederic
for the AP. A short descriptive text accompanied these prints.I
Local and national television stations along with local newspapers

utilized this special product.

SAB and the Washington SFSS also cooperated with local
news media. They provided use of the image analyzer for the "tlive"?
television broadcasts as Frederic approached and made landfall.
Satellite meteorologists were also interviewed "live"' to help
interpret the iamgery for TV audiences.

Kansas City SFSS and Miami SFSS also cooperated with the
news media providing extra pictures and professional interpretation
of the data.

Fig. 1 Maximum winds of Hurricane Frederic:
A comparison of Reconnaissance data with
satellite derived wind and intensity estimates

130K -

120K - SATELLITE MAX WIND
| ESTIMATES

110K -

X IX- l--x - - ___-x
100K -

LANDFALL

90K

80K -
DVORAK

ENHANCED IR -RECON

CLASSIFICATION .

70K -

SEP12 12Z 15Z 18Z 21Z ooz 03Z
SEP 13

2 10 B 4 2 *as :
2

' ~ ~~~~..
~4

2
Fig. 2a Satellite derived rainfall estimates
(2130Z Sep 12 to 1100Z Sep 13)

,: - _

f. I:
I 4

27
' I

- ~ ~~~~~~~~ ,,

s*3h. 4' 2'-. ..

Fig. 2b Observed 24-hour rainfall
(1200Z Sep 12 to 1200Z sep 13)