TPCNHC Tropical Cyclone Forecasting Process - PowerPoint PPT Presentation

Loading...

PPT – TPCNHC Tropical Cyclone Forecasting Process PowerPoint presentation | free to download - id: 68af6-M2Q4Z



Loading


The Adobe Flash plugin is needed to view this content

Get the plugin now

View by Category
About This Presentation
Title:

TPCNHC Tropical Cyclone Forecasting Process

Description:

TPCNHC Tropical Cyclone Forecasting Process – PowerPoint PPT presentation

Number of Views:173
Avg rating:3.0/5.0
Slides: 162
Provided by: MaxMay9
Learn more at: http://www.srh.noaa.gov
Category:

less

Write a Comment
User Comments (0)
Transcript and Presenter's Notes

Title: TPCNHC Tropical Cyclone Forecasting Process


1
TPC/NHC Tropical Cyclone Forecasting Process
  • Dr. Richard J. Pasch
  • Hurricane Specialist
  • Southern Region Marine Workshop
  • TPC/NHC
  • March 17, 2004

2
TPC/NHC Six-hour forecast cycle
3
TRACK FORECASTING
  • BASIC DYNAMICS
  • GUIDANCE MODELS
  • USE OF ENSEMBLES
  • USE OF INITIAL MOTION, CONTINUITY
    FROM PREVIOUS FORECAST, AND OTHER GUIDELINES
  • SYNOPTIC SURVEILLANCE
  • NHC TRACK FORECAST ACCURACY

4
CENTRAL PRESSURE VS. TIME FOR HURRICANE
ALLEN, 1980
TRACK OF HURRICANE ALLEN, 1980
INNER CORE VARIATIONS HAVE LITTLE INFLUENCE
ON TRACK
5
(No Transcript)
6
Factors Affecting TC Motion
  • Zero Order - Cons. of relative vorticity
  • Vortex Moves with Steering Flow
  • First Order - Cons. of absolute vorticity
  • Vortex induces beta-gyres and affects motion
  • General Dynamics and Physics
  • Vertical structure is important
  • Interaction with orography, friction, convection

7
Track Guidance Models
  • Zero Order - CLIPER, NHC90/91/98
  • First Order - BAM, LBAR, VICBAR
  • General Models - GFDL, GFS, NOGAPS, UKMET, ECMWF,
    ETA, NGM

8
Another way to classify models
  • Statistical CLIPER
  • Statistical/Dynamical - NHC/90/91/98
  • Purely Dynamical GFDL, GFS, NOGAPS, UKMET,
    ECMWF, ETA, MM5, et al.

9
CLIPER (CLImatology and PERsistence)
  • Statistical track model developed in 1972
  • Required Input
  • Current/12 h old speed/direction of motion
  • Current latitude/longitude
  • Julian Day, Storm maximum wind
  • Avg. 24, 48, 72 h errors 110, 230, 350 nautical
    miles
  • Used as benchmark for other models forecasts
    having errors greater than CLIPER are considered
    to have no skill.

10
72-HR FORECAST
48-HR FORECAST
36-HR FORECAST
24-HR FORECAST
12-HR FORECAST
INITIAL
CLIPER EXAMPLE HURRICANE DANIELLE, 1998
11
NHC90 (Atlantic), NHC91 (East Pacific)
  • Statistical/Dynamical track model
  • Required Input
  • CLIPER forecast tracks
  • Analyzed and forecast deep layer mean heights
    (1000-100 mb) from NCEP or UKMET global model
  • Input used as predictors for TC motion (along-
    cross-track components) in a multiple regression
  • New version for Atlantic with vortex removal
    scheme (NHC98)

12
Typical Predictor Locations for NHC90
13
THE SIMPLIFIED STEERING (OR TRAJECTORY) MODEL,
BAM (BETA AND ADVECTION MODEL)
-USES STEERING (TRAJECTORY) GIVEN BY
LAYER-AVERAGED WINDS FROM GLOBAL MODEL
(HORIZONTALLY SMOOTHED TO T25 RESOLUTION),
PLUS A CORRECTION TERM TO SIMULATE THE
SO-CALLED BETA EFFECT -THREE DIFFERENT
LAYER AVERAGES SHALLOW (850-700 MB) -
BAMS MEDIUM (850-400 MB) - BAMM DEEP
(850-200 MB) - BAMD
14
THE BETA DRIFT THE CIRCULATION OF A TROPICAL
CYCLONE, COMBINED WITH THE NORTH-SOUTH VARIATION
OF THE CORIOLIS PARAMETER, INDUCES
ASYMMETRIES KNOWN AS BETA GYRES. THESE
GYRES PRODUCE A NET STEERING CURRENT ACROSS
THE TC, GENERALLY TOWARD THE NW AT A FEW
KNOTS. THIS MOTION HAS COME TO BE KNOWN AS
THE BETA DRIFT.
15
HIGHER VALUES OF EARTHS VORTICITY
INDUCED STEERING
H
?vgt0
?vlt0
L
N
LOWER VALUES OF EARTHS VORTICITY
16
(No Transcript)
17
LBAR (Limited-area BARotropic)
  • Barotropic dynamics, i.e. 2-d motions
  • Shallow water equations on Mercator projection
    solved using sine transforms
  • Initialized with 850-200 mb average winds/heights
    from NCEP global model
  • Sum of idealized vortex and current motion vector
    added to large-scale analysis
  • Boundary conditions from global model

18
(No Transcript)
19
THREE-DIMENSIONAL (MULTI-LEVEL) DYNAMICAL MODELS
(PRIMITIVE EQUATION MODELS) 1) THE NATIONAL
CENTERS FOR ENVIRONMENTAL PREDICTION (NCEP)
GLOBAL FORECAST SYSTEM, GFS (GLOBAL
SPECTRAL) 2) UKMET (GLOBAL GRID-POINT) 3) NOGAPS
(GLOBAL SPECTRAL) 4) GFDL (NESTED, LIMITED AREA
SPECIFICALLY DESIGNED FOR TC PREDICTION) 5) ECMWF
(GLOBAL SPECTRAL) 6) ETA, NGM, MM5, COAMPS
(LIMITED AREA GRID-POINT) THESE MODELS INCLUDE
LONG- AND SHORT-WAVE RADIATION, CUMULUS
CONVECTION, LARGE-SCALE PRECIPITATION, SURFACE
FLUXES, AND TURBULENT TRANSPORTS
20
EXPLICIT USE OF GLOBAL MODELS FOR TC
PREDICTION
  • MAY EMPLOY A BOGUSSING SCHEME TO INITIALIZE
    THE TC VORTEX (SYNTHETIC DATA SYSTEM)
  • MODEL RESOLUTION IS INADEQUATE TO DEFINE THE
    INNER CORE (EYE, EYEWALL, RADIUS OF MAXIMUM
    WINDS)
  • BECOMING USEFUL FOR FORECASTING TROPICAL CYCLONE
    SIZE / TROPICAL STORM FORCE WIND RADII
  • GLOBAL MODEL HAS NO LATERAL B.C. GOOD FOR
    LONGER-RANGE GUIDANCE

21
RECENT UPGRADES TO THE NCEP GLOBAL MODEL (GFS)
FOR IMPROVED TROPICAL PREDICTION - LATE IN 2002
THE RESOLUTION WAS INCREASED TO T254L64
(TRIANGULAR 254 WAVES, 64 VERTICAL LEVELS),
ROUGHLY EQUIVALENT TO A 55 KM, ½ DEGREE
HORIZONTAL GRID SIZE. - IN 2001, INCORPORATION OF
AMSU-B DATA FROM POLAR-ORBITING SATELLITES HAS
LED TO AN IMPROVED ANALYSIS OF THE TROPOSPHERIC
MOISTURE FIELD. -IN 2001, MOMENTUM MIXING BY
CUMULUS CONVECTION, AND A PROGNOSTIC EQUATION
FOR CLOUD WATER WAS ADDED.
22
NCEP global model had problems initializing
TCs 1000 mb Initial Wind/Vorticity 9/21/98 00
UTC (X indicates observed center of Hurricane
Georges)
X
X
23
  • - SINCE THE 2000 SEASON, THE GFS (NCEPS
    GLOBAL FORECAST SYSTEM) HAS NOT ROUTINELY
    USED BOGUSSING OF A TC. A NEW TECHNIQUE,
    VORTEX RELOCATION IS NOW EMPLOYED. THIS
    GIVES MUCH IMPROVED TC TRACK FORECASTS.
  • THE RELOCATION INVOLVES A RE-POSITIONING OF
    THE TC (TO THE POSITION REPORTED BY NHC)
    IN THE FIRST GUESS FIELD OF THE GLOBAL
    ANALYSIS. DETAILS MAY BE FOUND IN TECHNICAL
    PROCEDURES BULLETIN TPB472 AT THE FOLLOWING WEB
    SITE
  • http//www.nws.noaa.gov/om/tpb/472.htm
  • THE GFS IS NOW RUN EVERY 6 HOURS OUT TO 384 HOURS
    (16 DAYS).

24
-THE NCEP GLOBAL MODEL PROVIDES INITIAL AND
BOUNDARY CONDITIONS FOR THE GFDL MODEL
(HOWEVER THE GLOBAL MODELS TC VORTEX IS
REMOVED IN THE GFDL INITIALIZATION PROCEDURE).
IT ALSO PROVIDES INITIAL CONDITIONS FOR THE LBAR
MODEL AND STEERING WINDS FOR BAMS, BAMM,
BAMD. -EXPLICIT TC TRACK FORECAST POINTS
(LAT/LON) FROM THE GLOBAL MODEL ARE DERIVED BY
RUNNING A TC TRACKER PROGRAM ON THE MODELS
OUTPUT. ADDITIONALLY, OF COURSE, THE FORECASTER
CAN INSPECT THE FORECAST FIELDS TO ASSESS THE
EVOLUTION OF THE TC AND ITS ENVIRONMENT.
25
GLOBAL MODELS SUCH AS THE NOGAPS (NAVY
OPERATIONAL GLOBAL ATMOSPHERIC PREDICTION
SYSTEM) AND THE UKMET (UNITED KINGDOM
METEOROLOGICAL ) OFFICE MODELS, CONTINUE TO
USE A BOGUSSING TECHNIQUE TO SPECIFY THE
TC CIRCULATION. CURRENTLY, THE ECMWF
(EUROPEAN CENTRE FOR MEDIUM-RANGE WEATHER
FORECASTS) GLOBAL MODEL DOES NOT HAVE A TC
BOGUS, ALTHOUGH THIS MODEL IS CAPABLE OF
SIMULATING MOST TCs (OCCASIONALLY IT FAILS
TO DETECT WEAK ONES).
26
(No Transcript)
27
IN 2002, THE NOGAPS MODEL WAS UPGRADED. THE
RESOLUTION WAS INCREASED FROM T159L24 TO T239L30
(239 WAVES, WHICH IS ABOUT A 1/2 HORIZONTAL
RESOLUTION, WITH 30 LEVELS) ALSO IN 2002, THE
U.K. MET. OFFICE MODEL WAS UPGRADED. THERE WERE
MANY IMPROVEMENTS TO THE MODELS DYNAMICS AND
PHYSICS.
28
GFS FORECASTS HURRICANE CLAUDETTE JULY 2003,
WITHIN 96 H OF LANDFALL
29
IMPACT OF POOR INITIALIZATION IN THE GFS
850 mb Vort / Wind
12Z 10 Sep 000 h GFS
PMSL / GOES-12 IR
30
850 mb Vort / Wind
18Z 10 Sep 006 h GFS
PMSL / GOES-12 IR
31
850 mb Vort / Wind
00Z 11 Sep 012 h GFS
PMSL / GOES-12 IR
32
850 mb Vort / Wind
06Z 11 Sep 018 h GFS
PMSL / GOES-12 IR
33
850 mb Vort / Wind
12Z 11 Sep 024 h GFS
PMSL / GOES-12 IR
34
850 mb Vort / Wind
18Z 11 Sep 030 h GFS
PMSL / GOES-12 IR
35
850 mb Vort / Wind
00Z 12 Sep 036 h GFS
PMSL / GOES-12 IR
36
850 mb Vort / Wind
06Z 12 Sep 042 h GFS
PMSL / GOES-12 IR
37
850 mb Vort / Wind
12Z 12 Sep 048 h GFS
PMSL / GOES-12 IR
38
850 mb Vort / Wind
18Z 12 Sep 054 h GFS
PMSL / GOES-12 IR
39
850 mb Vort / Wind
00Z 13 Sep 060 h GFS
PMSL / GOES-12 IR
40
850 mb Vort / Wind
06Z 13 Sep 066 h GFS
PMSL / GOES-12 IR
41
850 mb Vort / Wind
12Z 13 Sep 072 h GFS
PMSL / GOES-12 IR
42
850 mb Vort / Wind
18Z 13 Sep 078 h GFS
PMSL / GOES-12 IR
43
850 mb Vort / Wind
00Z 14 Sep 084 h GFS
PMSL / GOES-12 IR
44
850 mb Vort / Wind
06Z 14 Sep 090 h GFS
PMSL / GOES-12 IR
45
850 mb Vort / Wind
12Z 14 Sep 096 h GFS
PMSL / GOES-12 IR
46
850 mb Vort / Wind
18Z 14 Sep 102 h GFS
PMSL / GOES-12 IR
47
850 mb Vort / Wind
00Z 15 Sep 108 h GFS
PMSL / GOES-12 IR
48
850 mb Vort / Wind
06Z 15 Sep 114 h GFS
PMSL / GOES-12 IR
49
850 mb Vort / Wind
12Z 15 Sep 120 h GFS
PMSL / GOES-12 IR
50
850 mb Vort / Wind
18Z 15 Sep 126 h GFS
PMSL / GOES-12 IR
51
Hurricane Isabel Track Guidance 1800 UTC 10 Sep
2003
Note southward bias in AVNI track (12Z GFS
interpolated to 18Z initial position)
52
Hurricane Isabel
GFS Model Guidance for Sept. 10-21, 2003 18z
53
Hurricane Isabel
GFS Model Guidance for Sept. 13-21, 2003 00z
54
THE GEOPHYSICAL FLUID DYNAMICS LABORATORY (GFDL)
HURRICANE MODEL ALSO KNOWN AS THE MMM
(MULTIPLY-NESTED MOVABLE MESH MODEL), IT USES TWO
NESTED GRIDS. THE OUTER GRID COVERS A DOMAIN OF
75 LONGITUDE x 75 LATITUDE WITH A RESOLUTION OF
1/2 THE INNER (NESTED) GRID HAS A DOMAIN OF
11 x 11 WITH A RESOLUTION OF 1/ 6 . SINCE THE
RESOLUTION IS HIGH ENOUGH TO SIMULATE SOME OF THE
INNER TC STRUCTURE, THIS MODEL HAS SOME SKILL IN
INTENSITY PREDICTION.
55
GFDL MODEL (CONTINUED) THE INITIALIZATION
SCHEME ATTEMPTS TO MORE ACCURATELY SPECIFY THE
TC CIRCULATION, AS WELL AS OTHER DISTURBANCES IN
THE TC ENVIRONMENT THAT COULD INFLUENCE THE TRACK
AND INTENSITY. FIRST, THE GLOBAL MODELS VORTEX
IS REMOVED FROM THE GFS (OR NOGAPS) MODELS
INITIAL CONDITIONS BY USE OF A FILTER. THE
SPECIFIED TC VORTEX IS OBTAINED FROM A SEPARATE
MODEL WHICH IS NUDGEDTOWARD THE OBSERVED
INITIAL CONDITIONS (INTENSITY, WIND RADII) AS
SPECIFIED BY THE HURRICANE FORECASTERS.
56
GFDL MODEL (CONTINUED) INITIALIZATION
initial field global analysis - globally
analyzed NCEP (or NOGAPS) vortex specified
GFDL vortex global analysis basic field
disturbance field disturbance field
globally analyzed NCEP (or NOGAPS) vortex
nonhurricane component environmental
field basic field nonhurricane
component
57
(No Transcript)
58
(No Transcript)
59
Vortex from global Analysis
Vortex from spin-up procedure
60
M
NOTE MESH C IS NOW MESH M, AT A 1/2º
RESOLUTION, MESH M IS NOW MESH F, AT A 1/6º
RESOLUTION, AND THE ORIGINAL MESH F IS
DISCONTINUED.
F
61
  • (p/ps) LEVELS OF THE GFDL MODEL
  • THE MODEL HAS 42 s LEVELS, AT UNEQUALLY SPACED
    INTERVALS, FROM THE SURFACE TO ABOUT 2 MB. FOR
    A SURFACE PRESSURE OF 1000 MB, THE LOWEST
    ATMOSPHERIC LEVEL IS AT A PRESSURE OF ABOUT 996
    MB, AND 12 LEVELS ARE BELOW 800 MB WHILE 10
    LEVELS ARE ABOVE 100 MB.

62
42
1 medium
1/2
(150 X 150)
75
60
45
2 fine
11
20
? Turbulence GFS boundary layer parameterization
scheme ? Surface flux Monin-Obukhov framework,
interfacial layer included ? Cumulus convection
simplified Arakawa-Schubert scheme, very similar
to the GFS version includes parameterized
vertical momentum transports by convection


63
(No Transcript)
64
1
4
7
9
2
3
5
6
8
65
(No Transcript)
66
OVERALL DOMAIN OF THE GFDL MODEL - 75 x 75
67
SFC (10 M) WIND SWATH FROM GFDL HURRICANE MODEL
68
THE GFDL MODEL CAN PRODUCE USEFUL OBJECTIVE
RAINFALL GUIDANCE
69
INFORMATION ON THE GFDL HURRICANE MODEL IS ALSO
CONTAINED IN A TECH. PROCEDURES BULLETIN AT THE
FOLLOWING WEB SITE
http//www.nws.noaa.gov/om/tpb/492.htm
70
GFDL CHANGES IMPLEMENTED IN 2003
  • Run from T254, 64 level GFS
  • Ocean Coupling over entire Ocean Domain, Gulf
    Stream Assimilation
  • Increase of Vertical Levels from 18 to 42
  • Simplified Arakawa-Schubert Cumulus
    Parameterization (SAS)
  • GFS Boundary Layer Scheme (Replaces MY)
  • More Accurate Pressure Gradient Computation

71
CHANGES IMPLEMENTED IN 2003 (cont.)
  • Improved Mass Initialization
  • (Reduces Noise over Mountains )
  • Improved Interpolation of Winds over mountains
  • Improved Vortex removal algorithm in
    Initialization
  • More consistent target wind in initialization
  • Bug fix in computation of environmental field

72
Track Forecasts for Hurricane Michelle from 2001,
2002 and 2003 GFDL Forecast Systems
73
NOTE TWO VERSIONS OF THE GFDL MODEL ARE
CURRENTLY OPERATIONAL ONE WHICH USES THE NCEP
GLOBAL MODEL (GFS) FOR INITIAL AND BOUNDARY
CONDITIONS AND ONE WHICH USES THE NAVYS GLOBAL
MODEL (NOGAPS) FOR INITIAL AND BOUNDARY
CONDITIONS. THE LATTER IS REFERRED TO AS THE
GFDN MODEL. THE GFDN WILL LIKELY BE RUN 4
TIMES PER DAY OUT TO 126 HOURS FOR TC FORECASTS
THIS COMING HURRICANE SEASON.
74
AVAILABILITY OF MODELS FOR OPERATIONAL TC
FORECASTING LATE VS. EARLY DYNAMICAL MODELS
SUCH AS THE AVN, UKMET, NOGAPS AND THE GFDL ARE
NOT AVAILABLE UNTIL ABOUT 4 TO 6 HRS AFTER THE
INITIAL SYNOPTIC TIME (I.E. THE 12Z RUN IS NOT
AVAILABLE UNTIL 18Z IN REAL TIME). SIMPLER
MODELS SUCH AS LBAR, NHC90, BAM, AND CLIPER ARE
AVAILABLE SHORTLY AFTER THE SYNOPTIC TIME (USING
6- THROUGH 78-HR GLOBAL MODEL FORECAST FIELDS).
75
UP UNTIL JUST A FEW YEARS AGO, STATISTICAL/DYNAMIC
AL MODELS, SUCH AS THE NHC90, WERE TYPICALLY
THE MOST RELIABLE, AND HAD THE LOWEST FORECAST
ERRORS OF ALL OBJECTIVE TRACK PREDICTION
TECHNIQUES. HOWEVER IN RECENT YEARS (IN THE
ATLANTIC BASIN), DYNAMICAL MODELS HAVE BECOME
THE MAINSTAY FOR TC TRACK PREDICTION, AND
OFFER THE PROMISE FOR CONTINUED REFINEMENTS
THROUGH IMPROVED PHYSICS, HIGHER RESOLUTION
AND A MORE REALISTIC INITIALIZATION.
76
48-Hour Track Model Errors 1965-2003 which is
the best model? Depends on the year!
77
Hurricane Isabel
Model Guidance for Sept. 08-19, 2003 00z 12z
78
ENSEMBLE FORECASTS IN THE CLASSIC ENSEMBLE
FORECAST APPROACH, A NUMBER OF FORECASTS
ARE MADE WITH A MODEL USING PERTURBED
INITIAL CONDITIONS THAT (HOPEFULLY) REPRESENT
THE LIKELY INITIAL ANALYSIS ERROR
DISTRIBUTION. THE ENSEMBLE METHOD HOLDS
PROMISE FOR PROVIDING A MEASURE OF
CONFIDENCE IN A FORECAST. IF THERE IS A
RELATIVELY SMALL SPREAD AMONGST MEMBERS OF
THE ENSEMBLE, THEN ONE MIGHT EXPECT TO
BE ABLE TO MAKE A RELATIVELY HIGH
CONFIDENCE TC TRACK FORECAST.
79
NOTE A GROUP OF FORECAST TRACKS FROM
DIFFERENT PREDICTION MODELS (GFDL, UKMET,
NOGAPS, ETC.) FROM THE SAME INITIAL TIME
CAN ALSO BE CONSIDERED AS AN ENSEMBLE A
MULTI-MODEL ENSEMBLE. IN FACT, A MULTI-MODEL
ENSEMBLE IS USUALLY SUPERIOR TO AN ENSEMBLE FROM
A SINGLE MODEL, SINCE THE DIFFERENT MODELS
TYPICALLY HAVE DIFFERENT BIASES WHICH TEND TO
CANCEL OUT WHEN THEY ARE COMBINED. THE
MULTI-MODEL ENSEMBLE IS OFTEN CALLED A CONSENSUS
FORECAST. AT THE NHC, THE MOST COMMONLY-USED
CONSENSUS FORECAST IS CALLED GUNA, WHICH IS
SIMPLY THE AVERAGE OF THE GFDL, UKMET, NOGAPS,
AND AVN (GFS) TRACKS.
80
OVER THE LAST FEW YEARS, IT HAS BECOME APPARENT
THAT A CONSENSUS TC TRACK FORECAST DERIVED FROM
GOOD DYNAMICAL MODELS IS SIGNIFICANTLY BETTER
THAN THE FORECAST FROM ANY OF THE INDIVIDUAL
MODELS.
81
2001-2003 Atlantic GUNA Ensemble TC Forecast
Error (nm)
619
358
176
Number of Forecasts
467
229
82
VERIFYING POSITION
EXCELLENT EXAMPLE OF GUNA CONSENSUS HURRICANE
ISABEL, 1200 UTC 11 SEP 2003
83
VERIFYING POSITION
NOT-SO EXCELLENT EXAMPLE OF GUNA CONSENSUS
HURRICANE KATE, 1800 UTC 29 SEP 2003 (OFFICIAL
5-DAY FORECAST WAS NEAR 34N 42W)
84
CAN WE USE THE SPREAD (DEFINED TO BE THE AVERAGE
DISTANCE OF THE INDIVIDUAL MODEL FORECASTS FROM
THE CONSENSUS) TO HELP US DETERMINE THE EXPECTED
ERROR (I.E., CONFIDENCE) OF A GIVEN TRACK
FORECAST?
85
(No Transcript)
86
Hurricane Isabel 00Z September 13,
2003 Predicted CONU 76 Confidence Radius
87
Hurricane Kate 00Z September 30, 2003 Predicted
CONU 76 Confidence Radius
88
INITIAL (CURRENT) MOTION
  • VERY IMPORTANT FOR THE SHORT-TERM FORECAST
  • 12-HOUR FORECAST TRACK IS WEIGHTED HEAVILY
    BY THE INITIAL MOTION
  • NOT ALWAYS EASY TO DETERMINE…PARTICULARLY
    FOR SYSTEMS WITH POORLY-DEFINED CENTERS

89
CONTINUITY
  • PREVIOUS OFFICIAL FORECAST PROVIDES
    CONSTRAINTS ON THE CURRENT FORECAST
  • CREDIBILITY CAN BE HURT BY MAKING BIG
    CHANGES FROM ONE OFFICIAL FORECAST TO THE
    NEXT
  • SO, CHANGES TO THE PREVIOUS FORECAST ARE
    NORMALLY MADE IN SMALL INCREMENTS

90
ADDITIONAL GUIDLINES
  • DO NOT MAKE SUDDEN CHANGES IN DIRECTION
    FROM ONE FORECAST INTERVAL (12-24 HR, 24-36
    HR, etc.) TO THE NEXT.
  • DO NOT MAKE DRASTIC CHANGES IN FORWARD
    SPEED FROM ONE FORECAST INTERVAL TO THE
    NEXT. HOWEVER, WHEN DECELERATION OR
    ACCELERATION IS PREDICTED, MAKE THE CHANGES
    GRADUAL.

91
INITIAL INPUT DATA
AS SOON AFTER THE SYNOPTIC TIME (00Z, 06Z,
12Z, 18Z) AS POSSIBLE, THE FORECASTER
PROVIDES THE INITIAL CONDITIONS OF THE TC
FOR INPUT INTO THE PREDICTION MODELS. THESE
DATA ARE THE CURRENT T-12 HR POSITION /
MOTION, T-24 HR POSITION, CURRENT MAXIMUM
WINDS / MINIMUM PRESSURE, RADIUS OF MAXIMUM
WINDS, RADIUS AND VALUE OF THE OUTERMOST
CLOSED ISOBAR, AND RADII (BY QUADRANT) OF
34- AND 50-KNOT WINDS.
92
(No Transcript)
93
HOW TO USE TRACK PREDICTION MODELS TO MAKE A
FORECAST??
  • DO NOT SIMPLY GO WITH A GIVEN MODEL.
  • A CONSENSUS OF HIGH-QUALITY DYNAMICAL MODELS IS
    USUALLY A GOOD FIRST GUESS
  • CONTINUITY FROM PREVIOUS FORECAST IS A STRONG
    CONSTRAINT (AVOIDANCE OF THE WINDSHIELD WIPER
    EFFECT).
  • EVALUATE LARGE-SCALE SYNOPTIC ENVIRONMENT
    (CONVENTIONAL DATA AND SATELLITE, E.G. WATER
    VAPOR INFORMATION) AND TRY TO ASSESS THE
    STEERING INFLUENCES. IN OTHER WORDS, BE A
    GOOD METEOROLOGIST, NOT A MECHANICAL MODEL
    FOLLOWER. THIS SYNOPTIC EVALUATION MAY ALSO
    HELP TO DETERMINE OUTLIERS IN THE MODEL SUITE.

94
USE OF MODELS (CONT.)
  • IN SOME CASES, WHEN A GIVEN MODEL APPEARS TO BE
    DOING WELL , AND THAT MODELS TRACK LOOKS
    REASONABLE FOR THE SYNOPTIC SITUATION, THE
    OFFICIAL FORECAST MIGHT FOLLOW THIS MODEL FAIRLY
    CLOSELY.
  • THE OVERALL SPREAD OF THE OBJECTIVE TRACKS MAY
    GIVE SOME MEASURE OF FORECAST CONFIDENCE.
  • EVEN THE MOST RELIABLE MODELS CAN FAIL MISERABLY.

95
NOAA G-IV AIRCRAFT A SYNOPTIC SURVEILLANCE
PLATFORM
96
RADIOSONDE AND GPS DROPSONDE WINDS
97
OBSERVED 72 H LOCATION
EXAMPLE OF IMPACT OF GPS DROPSONDE DATA ON 72 H
MODEL FORECASTS
98
MEAN TRACK FORECAST ERRORS IN THE NCEP GLOBAL
MODEL (GFS) WITH WITHOUT GPS DROPSONDES,
1999-2003. NOTE THAT THE BIGGEST IMPROVEMENTS
ARE IN THE 36-48 H TIME FRAME.
99
(No Transcript)
100
BIG IMPROVEMENTS IN TC TRACK FORECASTS OVER
RECENT DECADE, ESPECIALLY BEYOND 2 DAYS
ATTRIBUTABLE TO NWP!
101
CONCLUDING REMARKS (TRACK FORECASTING)
  • ? Dynamical models have become much more skillful
    in predicting TC track
  • ? Big improvements in NHC official track
    forecasts over the last couple of decades are
    directly attributable to these improvements in
    dynamical models (i.e. to NWP)
  • ? The consensus track forecast has become a very
    powerful tool and need to be exploited further

102
TROPICAL CYCLONE INTENSITY FORECASTING
Photo M. Black
103
LITTLE OR NO IMPROVEMENT IN TC INTENSITY
FORECASTS BEYOND 2 DAYS. THUS FAR, NWP HAS NOT
PROVIDED MUCH HELP IN PREDICTING TC INTENSITY
CHANGES.
104
WHAT IS THE INTENSITY OF A TROPICAL CYCLONE?
  • Maximum sustained surface wind Maximum wind,
    averaged over a 1 minute interval at an altitude
    of 33 ft (10 m), associated with the circulation
    of the tropical cyclone at a given point in time.
  • With very, very few exceptions, direct
    observations of the maximum sustained surface
    wind in a tropical cyclone are not available.

105
HOW DO WE ESTIMATE INTENSITY?
  • Satellite imagery using the Dvorak technique.
  • Surface observations ships, buoys, land stations
    (infrequent)
  • Aircraft reconnaissance flight-level winds
  • GPS dropwindsondes

106
Dvorak Technique
Tropical cyclones have characteristic cloud
patterns that correspond to stages of development
and certain intensities.
107
Dvorak Technique Cloud Patterns
Typical accuracy 15 MPH (1 SSHS Category)
108
The Dvorak Technique has a standard (normal) rate
of development of a TC, which can provide a first
guess for an intensity forecast
Normal development rate is one Dvorak T number
per day If at t0, the TC is a 30-kt tropical
depression, corresponding to a T2.0, then
typically -At 24 h, the TC will be a T3.0, or 45
kt -At 36 h, the TC will be a T3.5, or 55 kt -At
48 h, the TC will be a T4.0, or 65 kt -At 72 h,
the TC will be a T5.0 or 90 kt
109
SURFACE OBSERVATIONS CAN OCCASIONALLY BE USED TO
ESTIMATE INTENSITY
110
AIRCRAFT RECONNAISSANCE (WHEN THERE IS A THREAT
TO LAND)
AFRC/53WRS C-130
NOAA/AOC P-3
Hurricane Georges Eyewall Photo M Black
HRD
111
90 kt
95 kt
90 kt
105 kt
RECON FLIGHT-LEVEL (10,000 FT) WINDS HURRICANE
GEORGES 9/20/98 20-23Z
112
BUT NOBODY LIVES AT 10,000 FT!
Because the hurricane is warm-core, winds
increase downward from flight-level (10000 ft).
Friction decreases wind in the lowest 1500 ft of
the eyewall.
113
GPS DROPWINDSONDE
  • Developed in conjunction with the NOAA
    Gulfstream-IV jet aircraft. First systematic
    use for intensity was in 1998s Hurricane Bonnie.
  • GPS dropsondes provide, for the first time, the
    ability to obtain direct measurements of the
    winds at low levels in the hurricane eyewall.
  • Dropsonde data reveal that the structure of the
    eyewall is very complex, and can vary
    tremendously from storm to storm.

114
ALTHOUGH THE NORMAL RATIO OF SURFACE TO FLIGHT
WIND SPEED IS 90, SOME HURRICANES HAVE TROUBLE
BRINGING THEIR HIGH WINDS DOWN TO THE SURFACE.
115
THE SURFACE WINDS CAN EXCEED THE FLIGHT-LEVEL
WIND AT TIMES.
116
NEW DROPSONDE WIND SPEED MEASUREMENT IN A
HURRICANE (ISABEL IN THE ATLANTIC, 12 SEPTEMBER
2003) 203 KNOTS NEAR 4500 FEET.
New dropsonde wind speed record in a hurricane
(Isabel in the Atlantic) 203 kt.
117
ALTHOUGH CENTRAL PRESSURE IS APPROXIMATELY
RELATED TO INTENSITY, THERE CAN BE BIG
VARIATIONS…AS MUCH AS TWO SAFFIR/SIMPSON
CATEGORIES FOR THE SAME CENTRAL PRESSURE!
118
FACTORS AFFECTING TROPICAL CYCLONE INTENSITY
  • Sea surface temperature / upper ocean heat
    content.
  • Interaction with land.
  • Vertical wind shear.
  • Trough interactions.
  • Temperature and moisture patterns in the storm
    environment.
  • Eyewall replacement cycles.

119
UPPER OCEANIC HEAT CONTENT
SEA SURFACE TEMPERATURE
120
INLAND WIND DECAY MODEL EXPONENTIAL
DECREASE OF INTENSITY WITH TIME
121
VERTICAL WIND SHEAR
45000 ft
DEEP CONVECTION
30000 ft
20000 ft
10000 ft
EXPOSED CENTER
5000 ft
1000 ft
122
TROUGH INTERACTION HURRICANE BERTHA, JULY 12 1996
0600 UTC
123
BERTHA STRENGTHENED JUST BEFORE LANDFALL
1200 UTC
124
(No Transcript)
125
DRY AIR IN THE MIDDLE LEVELS OF THE ATMOSPHERE
PROBABLY CAUSED THE STORM TO WEAKEN
50
10
30
126
TROPICAL CYCLONE INTENSITY STATISTICAL FORECAST
MODELS
  • SHIFOR (Statistical Hurricane Intensity
    FORecast) Based solely on climatology and
    persistence.
  • SHIPS (Statistical Hurricane Intensity Prediction
    Scheme) Based on climatology, persistence, and
    current/predicted environmental conditions.
  • DSHIPS (Decay SHIPS) same as SHIPS except when
    track forecast points are over land when a
    decrease in intensity following an inland decay
    model is included.
  • DSHIPS modified to include information about
    oceanic heat content and inner core convection
    (using infrared satellite imagery).

127
THE SHIPS MODEL
Statistical multiple regression model relating
tropical cyclone intensity change to various
climatological, persistence, and environmental
predictors.
  • () SST POTENTIAL (VMAX-V) Difference between
    the maximum potential intensity (depends on SST)
    and the current intensity.
  • (-) VERTICAL (850-200 MB) WIND SHEAR Current
    and forecast.
  • () PERSISTENCE If its been strengthening, it
    will probably continue to strengthen, and vice
    versa.
  • (-) UPPER LEVEL (200 MB) TEMPERATURE Warm
    upper-level temperatures inhibit convection
  • () THETA-E EXCESS Related to buoyancy (CAPE)
    more buoyancy is conducive to strengthening
  • () 500-300 MB LAYER AVERAGE RELATIVE HUMIDITY
    Dry air at mid-levels inhibits strengthening

128
THE SHIPS MODEL (cont.)
Statistical multiple regression model relating
tropical cyclone intensity change to various
climatological, persistence, and environmental
predictors.
  • () 850 MB ENVIRONMENTAL RELATIVE VORTICITY
    Vorticity is averaged over a large area, about
    10 radius. Intensification is favored when the
    storm is in an environment of cyclonic low-level
    vorticity.
  • (-) ZONAL STORM MOTION Intensification is
    favored when TCs are moving west
  • (-) STEERING LEVEL PRESSURE intensification is
    favored for storms that are moving more with the
    upper level flow. This predictor usually only
    comes into play when storms get sheared off and
    move with the flow at very low levels (in which
    case they are likely to weaken).
  • () 200 MB DIVERGENCE Divergence aloft enhances
    outflow and promotes strengthening
  • (-) CLIMATOLOGY Number of days from the
    climatological peak of the hurricane season

129
Improvement of Parallel vs. Operational
SHIPS (Real-time 2002 Atlantic Forecasts)
Incorporation of Ocean Heat Content and Satellite
Data Predictors in the Statistical Hurricane
Intensity Prediction Scheme (SHIPS)
N 238 219 199 185 158
133 118
130
VERIFYING 145 KNOTS
131
(No Transcript)
132
(No Transcript)
133
IN ADDITION TO LARGE-SCALE INFLUENCES,
TROPICAL CYCLONE INTENSITY CHANGE CAN BE
CAUSED BY INNER-CORE PROCESSES, SUCH AS
EYEWALL REPLACEMENT CYCLES. AFTER A PERIOD
OF SIGNIFICANT INTENSIFICATION, AN INNER
EYEWALL IS OFTEN REPLACED BY AN OUTER EYEWALL.
FIRST WE SEE CONCENTRIC EYEWALLS, AND THEN THE
INNER EYEWALL DISSIPATES LEAVING AN OUTER
EYEWALL (LARGER DIAMETER). SOME WEAKENING
USUALLY OCCURS DURING THIS REPLACEMENT
PROCESS. IN SOME CASES, THE OUTER EYEWALL
CONTRACTS AND THE HURRICANE RE-STRENGTHENS TO
COMPLETE THE CYCLE. THIS CYCLE COULD REPEAT
SEVERAL TIMES IN THE LIFETIME OF THE HURRICANE.
134
CONCENTRIC EYEWALL CYCLE HURRICANE FLOYD
13 / 0116Z
13 / 1122Z
13 / 1347Z
13 / 2240Z
14 / 0104Z
14 / 1110Z
14 / 2228Z
135
CONCENTRIC EYEWALL CYCLE HURRICANE FLOYD
13 / 0116Z
13 / 1122Z
13 / 1347Z
13 / 2240Z
14 / 0104Z
14 / 1110Z
14 / 2228Z
136
EYEWALL REPLACEMENT
137
CENTRAL PRESSURE VS. TIME FOR HURRICANE
ALLEN, 1980 LARGE FLUCTUATIONS LARGELY DUE TO
EYEWALL REPLACEMENT CYCLES
138
TROPICAL CYCLONE INTENSITY DYNAMICAL FORECAST
MODELS
  • GFDL, NCEP Global Model (AVN/MRF), UKMET (U.K.
    Met Office), NOGAPS (U.S. Navy), ECMWF (European)
  • These models are of limited use, because of…
  • sparse observations
  • inadequate resolution (need to go down to a few
    km grid spacing the GFDL, our highest-resolution
    operational hurricane model, is currently about
    18 km)
  • incomplete understanding and simulation of basic
    physics of intensity change
  • biases in upper-level wind forecasts.

139
LACK OF CONSISTENCY IN GFDL INTENSITY FORECASTS
IN EARLY STAGES OF TCS EXAMPLE OF GFDL SFC WIND
FORECAST FOR HURRICANE KATE AS A TD INITIALIZED
ON 25 SEP 2003 18Z
140
GFDL SFC WIND FORECAST FOR HURRICANE KATE AS A TD
INITIALIZED ON 26 SEP 2003 00Z
141
GFDL SFC WIND FORECAST FOR HURRICANE KATE AS A TD
INITIALIZED ON 26 SEP 2003 06Z
142
RECENT IMPROVEMENTS TO THE PHYSICS OF THE GFDL
MODEL SHOULD REDUCE LOW BIASES IN INTENSITY
FORECASTS FOR THE EARLY STAGES OF TROPICAL
CYCLONES IN 2004.
143
NHC OFFICIAL INTENSITY FORECAST
  • Based on statistical guidance from SHIPS and
    SHIFOR, qualitative guidance from dynamical
    models.
  • Persistence is used quite a bit!
  • Obvious signs in the environment, i.e. cooler
    waters, increasing upper-level winds, are taken
    into account
  • Generally corresponds to what is normal for a
    storm in any particular situation (e.g. the
    standard Dvorak development rate).
  • Tends to be conservative extreme events are
    almost never forecast.
  • For forecasts 24 h and beyond, the average error
    is roughly 1 SSHS Category (about 20 mph).

144
Operational Intensity forecast example
145
UPPER LEVEL TROUGH
L
L
146
TROPICAL STORM MITCH DISCUSSION NUMBER
7 NATIONAL WEATHER SERVICE MIAMI FL 11 AM EDT FRI
OCT 23 1998 MITCH IS NOT STRENGTHENING AT THIS
TIME. ALTHOUGH THE LOW-LEVEL CIRCULATION IS WELL
DEFINED...THE CONVECTION IS WEAK. UNEXPECTED
UPPER-LEVEL WESTERLY WINDS HAVE BECOME
ESTABLISHED OVER THE SYSTEM DISRUPTING THE
EXCELLENT OUTFLOW WHICH WAS OBSERVED YESTERDAY.
INITIAL INTENSITY IS KEPT AT 50 KNOTS IN THIS
ADVISORY...BUT A RECONNAISSANCE PLANE WILL
DETERMINE IF MITCH HAS WEAKENED AS SUSPECTED.
ALL AVAILABLE MODELS SUGGEST THAT AN UPPER-LEVEL
ANTICYCLONE SHOULD FORM OVER THE WESTERN
CARIBBEAN WITHIN THE NEXT 24 HOURS OR SO. SUCH A
PATTERN WOULD PROVIDE EXCELLENT CONDITIONS FOR
STRENGTHENING AND IN FACT...BOTH SHIPS AND THE
GFDL MODELS BRING MITCH TO A 75- AND 100-KNOT
HURRICANE... RESPECTIVELY. HOWEVER...IT IS
DIFFICULT TO VISUALIZE SUCH INTENSIFICATION WITH
THE PRESENT POORLY DEFINED PATTERN OBSERVED ON
SATELLITE IMAGES AND KNOWING THAT THE GLOBAL
MODELS TEND TO GET RID OF WESTERLIES TOO FAST.
AT THIS TIME...THE OFFICIAL FORECAST CALLS FOR NO
CHANGE IN INTENSITY DURING THE NEXT 24 HOURS AND
IF MITCH SURVIVES...THEN..STRENGTHENING SHOULD
BEGIN. ltADDITIONAL DISCUSSION DELETED
HEREgt FORECAST POSITIONS AND MAX WINDS INITIAL
23/1500Z 12.7N 77.9W 50 KTS 12HR VT
24/0000Z 12.9N 78.0W 50 KTS 24HR VT
24/1200Z 14.0N 78.2W 50 KTS 36HR VT
25/0000Z 14.8N 78.6W 65 KTS 48HR VT
25/1200Z 16.0N 79.0W 75 KTS 72HR VT
26/1200Z 17.0N 81.0W 80 KTS
VERIFYING 145 KNOTS
147
MITCH NEAR PEAK INTENSITY
148
OFFICIAL FORECASTS AND BEST TRACK INTENSITY
149
HAVE WE BEEN WE GETTING MUCH BETTER?
(2000)
70 kt
150
A LITTLE BETTER?
151
LANDFALL
152
(No Transcript)
153
(No Transcript)
154
A STATISTICAL TECHNIQUE TO AID IN THE FORECAST OF
RAPID INTENSIFICATION
The 7 predictors and rapid intensification (RI)
thresholds used to estimate the probability of RI
(RI is defined as an increase in maximum wind
speed of at least 30 kt over 24 h)
155
(No Transcript)
156
(No Transcript)
157
(No Transcript)
158
(No Transcript)
159
HURRICANE MICHELLE DISCUSSION NUMBER 16 NATIONAL
WEATHER SERVICE MIAMI FL 10 AM EST FRI NOV 02
2001 EVEN THOUGH THE FLIGHT-LEVEL WINDS WERE
NOT QUITE THERE YET... MICHELLE WAS UPGRADED TO
A HURRICANE EARLIER THIS MORNING BASED ON TWO
DROPWINDSONDE SURFACE WIND REPORTS OF 69 KT AND A
THIRD OF 63 KT. MICHELLE IS BEGINNING TO TAKE
ON A MORE CLASSICAL CIRCULAR APPEARANCE ON
SATELLITE IMAGERY AND EVIDENCE OF SHEAR IS
DIMINISHING. FOR THE FIRST TIME...ALL 5 OF THE
STATISTICAL PREDICTORS FROM THE HURRICANE
RESEARCH DIVISIONS EXPERIMENTAL RAPID
INTENSIFICATION MODEL HAVE BEEN SATISFIED.
GIVEN THIS...THE IMPROVED SATELLITE
APPEARANCE... INCLUDING THE FIRST EVIDENCE OF A
DEVELOPING EYE...THE OFFICIAL INTENSITY FORECAST
IS ABOVE ALL THE EXPLICIT GUIDANCE IN CALLING
FOR MICHELLE TO REACH MAJOR HURRICANE STATUS IN
THE NORTHWESTERN CARIBBEAN. …ADDITIONAL TEXT
DELETED... FORECASTER FRANKLIN FORECAST
POSITIONS AND MAX WINDS INITIAL 02/1500Z
17.9N 83.9W 65 KTS 12HR VT 03/0000Z 18.5N
84.0W 75 KTS 24HR VT 03/1200Z 19.6N
84.1W 90 KTS 36HR VT 04/0000Z 20.7N 84.0W
100 KTS 48HR VT 04/1200Z 21.5N 83.5W 95
KTS 72HR VT 05/1200Z 23.5N 81.0W 85 KTS

VERIFYING 115 KNOTS
160
CONCLUDING REMARKS
  • TC intensity forecasting is not as advanced as
    track forecasting. Compared to climatology and
    persistence, there is only very limited skill
    essentially none by about 3 days. There is a
    message here about the quality of intensity
    forecasts at 4 and 5 days!
  • Current guidance is provided mainly by
    statistical models. Improvements in these
    models, using oceanic heat content, inner core
    convection, and rapid intensification criteria,
    are currently being made.
  • Dynamical models have shown little or no skill in
    intensity forecasting. However, in time,
    improvements to these models (higher resolution
    and better physics), and enhanced observations,
    particularly in the inner core of the TC, offer
    hope for the future.

161
(No Transcript)
About PowerShow.com