Revised Pressure-Wind and Radial Profile Models: Implications for Tropical Cyclone Archives Mesoscale and Microscale Meteorology Division, ESSL National.

Revised Pressure-Wind and Radial Profile Models: Implications for Tropical

Cyclone Archives

Mesoscale and Microscale Meteorology Division, ESSLNational Center for Atmospheric Research

NCAR is Sponsored by NSF

Greg Holland

1

• What Use Tropical Cyclone Archives?– The single most important action we can take;– Most Important Parameters.

• Pressure-Wind Relationships– Role and Archive Requirements;– Pressure-Wind or Vice Versa.

• Reconstructing Historical Surface Winds– Approach and Archive Requirements;

Summary

2Holland, IBTraCS Workshop Asheville 0509

• Forecast Statistics;• Statistical Forecasting;• Assessing Climate Variability and Change;• Insurance Premiums;• Building Code Development and

Engineering of Major Structures.

What Use Tropical Cyclone Archives?


What is the single most important action we can take?

• Intensity;• Overall Wind Structure;• Translational Speed.

What are the Most Important Damage and Engineering Factors?


R2 : Vm3=27%, Vt=49%, Rh-2 =70%

Willis Hurricane Index α Vm3, Vt, Rh-2

Pressure-Wind Relationships


Plotted are:• All known P-W Relationships (after Harper 2002);• Black dashed is AH;• Red triangles are Atlantic 1995-2005 observations

Generally take the form: x

mv a p e.g. Atkinson and Holliday: 0.6443.4(1010 )m cv p

Alternative is Dvorak Look-Up Table, which can be translated to:

0.6443.92(1015 )m cv p

What are the Implications for Historical Archives?

• The Good: – They introduced an important level of objectivity;– They can be reproduced, provided we know what was used.

• The Bad:– Widespread use of different approaches without adequate

documentation;– Mixing objective and subjective approaches.

• Further:– Historical relationships were 1-1 and could not possibly capture the true

relationship; however, they were easily reproduceable;– New relationships are multi-variate, which enable capture of some of the

observed scatter; however, this is at the cost of increased difficulty in reproducing what was done, especially if there is inadequate documentation.

Implications of P-W Relationships


• AH tends to over-estimate intense hurricane winds (Knaff and Zehr 2007), and so does Dvorak (Holland 2008);

Some Known Issues


• Ad-hoc application, including:– Lack of proper peer review of

many relationships;– Variations in PW relationships;– Sometimes wind as the

primary parameter, other times pressure.

Fortunately, many of these can be reversed later, given adequate documentation

Issues: Western Australia


0 20 40 60 80

0

20

40

60

80

100

120

Vm (m/s)

DP

(h

Pa

)

Period 1970-2005 Lat 0-30 Long 120-135 Dir All

ActualAH1minAH10min

Atkinson and Holliday relationship using 10 min winds (blue) and 1 min winds (black), showing the incorrect use of AH with 1 min winds.

Improved documentation needed

• Knaff and Zehr (2007):

• Holland (2008):

Recent Revisions for P-W Relations


0.6443.92(1015 ) ,m cv p

mxtv a bs c d p e p fv

5 2 xct

0.5

pb= -4.4*10 p + 0.01 p + 0.03 - 0.014 +0.15v + 1.0,

tp

x= 0.6 (1- ) , and215

( )m

bv p

e

Holland Model: Dependent NAT Data


• The variable b reproduces about half the maximum scatter in HURDAT.

• But the scatter experienced by most storms is captured:

• RMSD=3.5 m/s

• 85%<5 m/s

Note Dvorak overestimate

The Importance of RMW


Black: Holland (2008) model, left original, right including RMWRed: Hurricane model data courtesy of Yuqing Wang (f-plane, 5 km grid, no environmental flow)

20 30 40 50 60 70 80 90

20

40

60

80

100

120

Vm (m/s)

DP

(h

Pa

)

Wang Data 1 Type= 5

ActualDerived

20 30 40 50 60 70 80 90

20

40

60

80

100

120

Vm (m/s)

DP

(h

Pa

)

Wang Data 1 Type= 5

ActualDerived

Should we start with pressure and derive winds, or vice versa?

• Must archive both and document how they were derived;

• If either or both are directly observed, no issue, archive both and document this;

• However, >90% of archives use a combination of satellite inference and subjective judgment;

• Should we estimate winds and infer central pressure, or vice versa?

Maximum Winds/Central Pressure?


• For Hurricanes, wind responds to changes in the mass field (i.e. heating);

• Pressures are much more conservative:– Easier to observe;– Not subject to wind definition issues;– Not subject to as much high-frequency, “random”

variation;– Less impact from cyclone translation speed variations;– Not affected by supergradient wind issues.

Pressure vs Winds


e.g. Schubert et al (1980):

Winds Respond to Mass Changes


Hurricane Core

Wind responds to massMass responds to windE

ffic

ienc

y

Inverse of Deformation Radius

Deep TropicsLarge, weak systems

Mass adjusts to wind

Wind adjusts to mass

Wind vs Pressure Variation


Blue: cold SST= 26.3C

Red: warm SST=30.3C

Solid: dx = 1.67 km

Dashed: dx = 0.556 km

Wind “noise” at 556 m resolution is ~10 m/s

At 62 m resolution it is >20 m/s (gusts are explicitly simulated)

Model results from Rotunno et al (2008)

Roll Vortices at Landfall


• Hurricane Dean simulation using ARW• Vertical motion advects low-momentum air upward and

high momentum air downward.• Leads to low level wind variations of ~20 ms-1

Recommend pressure be the primary variable

Reconstructing the Surface Winds


CY CP L OT : I s ov e l s : V3 ( peak ) m/ s ; Vec t o r s : V3, T he t am Sc a l e : 1mm = 0 . 5 k m

Bur eau o f Me t eo r o l o gy : T C T r ac y _ 19 74Gr i d : C6 Run: T RACY _ 1974 T o t a l V3_ p eak ( max ) Env e l op e

25

30

30

35

35

40

40

45

45

50

55

60

65

70

75

80

85

8 5

85

85

90

90 90

90

9090

90 90

90

90

909

0

V 3 ( ma x ) =9 4 . 8 m/ s @ ( 9 3 , 1 1 6 ) ; v b =3 0 6 . 0 d e g , S t e p = 6 7 3 V e c t o r S c a l e : 2 0 . 0 m/ s

Da r wi n A MO Di ne s

S t a r t : 2 0 - D e c - 1 9 7 4 0 0 : 0 0 GMT E n d : 2 5 - De c - 1 9 7 4 1 1 : 5 0 GMTV 3 ( ma x ) : 2 4 - De c - 1 9 7 4 1 6 : 0 0 GMT

Hurricane Tracy Maximum Surface Wind Reconstruction Courtesy Bruce Harper

Needs: Track, Vm, Pc, Penv, RMWDesirable: Venv

Reconstructing the Surface Winds


Willoughby et al 1982

New approaches enable a good reconstruction of the surface wind field for historical storms, provided the archive includes information on: RMW, Pc, Vm, and Vne.

Vne is winds in the near environment and preferably 4 obs at orthogonal directions.

Also having an estimate of major secondary wind maxima would be useful, but not essential.

Examples of Profile Error Sensitivity


0 50 100 150 200 250 300-10

-5

0

5

10

Radius (km)

Azi

mu

tha

l W

ind

Dif

fere

nc

e (

m/s

)pn= 1005 pc= 950 bs= 18 vm= 56 vob= 56 RMW= 20 rob= 300

Error Range +5 to -5 m/s

a

0 50 100 150 200 250 300-15

-10

-5

0

5

10

15

Radius (km)

Azi

mu

tha

l W

ind

Dif

fere

nc

e (

m/s

)

pn= 1005 pc= 950 bs= 18 vm= 56 vob= 20000 RMW= 21 rob= 300

Error Range +5 to -5 km

b

0 50 100 150 200 250 300-10

-5

0

5

10

Radius (km)

Azi

mu

tha

l W

ind

Dif

fere

nc

e (

m/s

)


Error Range +5 to -5 m/s

c

0 50 100 150 200 250 300-10

-5

0

5

10

Radius (km)

Azi

mu

tha

l W

ind

Dif

fere

nc

e (

m/s

)


Error Range +100 to -100 km

d

Maximum Wind Error

External Radius ErrorExternal Wind Error

RMW Error

Needs: Track, Vm, Pc, Penv, RMW and Venv

Summary

Holland, IBTraCS Workshop Asheville 0509

• Tropical Cyclone Archives need to be customer driven, Climate/CAT-Modeling are very important customers:• Track, Pc, RMW, Vm, Penv, Venv,

Quality Flags;• Unless there are good quality direct

observations, Pc is recommended as the primary archive for intensity, with wind derived from a documented P-W relation and other knowledge of wind variations;

• Perhaps we should include an extra section in Annual Tropical Cyclone Reports, which discusses quality issues and uncertainty assessments for the season.

20

Revised Pressure-Wind and Radial Profile Models: Implications for Tropical Cyclone Archives Mesoscale and Microscale Meteorology Division, ESSL National.

Documents

ibtracs workshop asheville

dvorak holland

holland model

adequate documentation

historical relationships

nsf greg holland

relationships variations

pw relationships