A Study of Several Aspects of the
Dvorak Technique
A Study of Several Aspects of the
Dvorak Technique
Jack Beven, John Cangialosi, and Todd Kimberlain
National Hurricane Center
February 2016
The Dvorak TechniqueThe Dvorak Technique• A method for estimating the
intensity of tropical cyclones
(TCs) from VIS/IR imagery
based on “measurements” of
the TC cloud pattern and a set
of rules.
• Based on correlations between
the TC cloud patterns and the
intensity
• Comes in manual and
automated versions.
• Technique works best on
strong TCs but less so with
weaker ones; skill of the
analyst is extremely
important.
• Related technique for
subtropical cyclones.
• A method for estimating the
intensity of tropical cyclones
(TCs) from VIS/IR imagery
based on “measurements” of
the TC cloud pattern and a set
of rules.
• Based on correlations between
the TC cloud patterns and the
intensity
• Comes in manual and
automated versions.
• Technique works best on
strong TCs but less so with
weaker ones; skill of the
analyst is extremely
important.
• Related technique for
subtropical cyclones.
EYE TYPES
Dvorak Technique Cloud PatternsDvorak Technique Cloud Patterns
Advanced Dvorak TechniqueAdvanced Dvorak Technique
Dean (2007) Ike (2008)
Controversial Issues of the
Dvorak Technique
Controversial Issues of the
Dvorak Technique• Calibration of the winds/pressures
• The Dvorak Technique conceptual model
for development/weakening
• Potential regional variations in the cloud
patterns
• Possible shortcomings of the infrared eye
cloud pattern measurements
• Adjustment for cold-topped infrared
curved band cloud patterns
• Use of outer banding for infrared cloud
patterns
• Adjustment for eye clarity for visible eye
cloud patterns
• What to do when a system is over land?
• The constraints on development/
weakening
• Calibration of the winds/pressures
• The Dvorak Technique conceptual model
for development/weakening
• Potential regional variations in the cloud
patterns
• Possible shortcomings of the infrared eye
cloud pattern measurements
• Adjustment for cold-topped infrared
curved band cloud patterns
• Use of outer banding for infrared cloud
patterns
• Adjustment for eye clarity for visible eye
cloud patterns
• What to do when a system is over land?
• The constraints on development/
weakening
For all the issues, Dvorak Technique
intensity estimates compare pretty
well to aircraft data
For all the issues, Dvorak Technique
intensity estimates compare pretty
well to aircraft data
• Intensity Constraints – Reality from aircraft-based
best track data
• Intensity Constraints – Data-T (DT) Number
• Technique Validation – CI Number, Final-T (FT)
Number, and DT Number
• Intensity Constraints – Reality from aircraft-based
best track data
• Intensity Constraints – Data-T (DT) Number
• Technique Validation – CI Number, Final-T (FT)
Number, and DT Number
What We’ve Looked AtWhat We’ve Looked At
Why are there constraints in the
Dvorak Technique?
Why are there constraints in the
Dvorak Technique?
• Weak systems sometimes
lose all convection during the
diurnal minimum
• Cloud patterns for weak
systems sometimes look
unrealistically strong
• Strong systems sometimes
don’t intensify as quickly as
the cloud pattern suggests
• In weakening systems, the
decay of winds and pressures
usually somewhat lags
behind that of the cloud
pattern
• Weak systems sometimes
lose all convection during the
diurnal minimum
• Cloud patterns for weak
systems sometimes look
unrealistically strong
• Strong systems sometimes
don’t intensify as quickly as
the cloud pattern suggests
• In weakening systems, the
decay of winds and pressures
usually somewhat lags
behind that of the cloud
pattern Air recon intensity lagged behind Dvorak
estimates during Mitch’s intensification
Air recon intensity lagged behind Dvorak
estimates during Mitch’s intensification
0
20
40
60
80
100
120
140
160
180
10/
22/1
998
10/
23/1
998
10/
24/1
998
10/
25/1
998
10/
26/1
998
10/
27/1
998
10/
28/1
998
10/
29/1
998
10/
30/1
998
10/
31/1
998
11/
1/19
98
11/
2/19
98
11/
3/19
98
11/
4/19
98
11/
5/19
98
11/
6/19
98
Win
d (k
t)
Date
1998 Mitch BT vs. Fix Intensity
Best Track
TAFB
SAB
Hurricane Wilma (2005)Hurricane Wilma (2005)
75 kt/975 mb75 kt/975 mb (Images ~12 hr apart)(Images ~12 hr apart) 150 kt/892 mb150 kt/892 mb
Some tropical cyclones clearly violate the DT development constraints.
Wilma deepened from 970 mb to 882 mb in ~12 hr.
How often does this happen? Does it happen every time it is observed
in the cloud pattern? Do the constraints need to be changed?
Some tropical cyclones clearly violate the DT development constraints.
Wilma deepened from 970 mb to 882 mb in ~12 hr.
How often does this happen? Does it happen every time it is observed
in the cloud pattern? Do the constraints need to be changed?
Testing the Dvorak ConstraintsTesting the Dvorak Constraints
• Used the Atlantic best track database
from 1998-2012 (the GPS dropsonde era).
• Used only those cases when the TC
stayed over water and reconnaissance
aircraft data were available within 3 h.
• Converted the observed intensity
changes at 6, 12, 18, and 24 h intervals to
the equivalent Dvorak T-number
changes.
• The results only use the best track
intensity changes. They do not examine
the corresponding Dvorak Technique
estimates.
• Changes of less than 0.25 T-numbers in
the specified time period were treated as
no change – a steady-state TC
• Used the Atlantic best track database
from 1998-2012 (the GPS dropsonde era).
• Used only those cases when the TC
stayed over water and reconnaissance
aircraft data were available within 3 h.
• Converted the observed intensity
changes at 6, 12, 18, and 24 h intervals to
the equivalent Dvorak T-number
changes.
• The results only use the best track
intensity changes. They do not examine
the corresponding Dvorak Technique
estimates.
• Changes of less than 0.25 T-numbers in
the specified time period were treated as
no change – a steady-state TC20
30
40
50
60
70
80
90
100
110
120
130
10/21 10/23 10/25 10/27 10/29 10/31
BEST TRACKSat (TAFB)Sat (SAB)ADTAC (sfc)AC (flt->sfc)AC (DVK P->W)ScatterometerSurfaceDrop (sfc)Drop (LLM xtrp)AMSUAnalysis
Win
d S
pee
d (
kt)
Date (Month/Day)
Hurricane SandyOctober 2012
2030405060708090
100110120130140150160170180190200
10/20 10/21 10/22 10/23 10/24 10/25
BEST TRACKSat (TAFB)Sat (SAB)ADTAC (sfc)AC (flt->sfc)AC (DVK P->W)ScatterometerDrop (sfc)Drop (LLM xtrp)AMSU
Win
d S
pee
d (k
t)
Date (Month/Day)
Hurricane Patricia20 - 24 October 2015
Results for the 24-h changesResults for the 24-h changes
• Out of 559 cases for intensifying
systems, 1.1% of the best track intensity
changes broke the 24-h constraints.
• Out of 273 cases for weakening systems,
none of the best track intensity changes
broke the 24-h constraints.
• Out of 559 cases for intensifying
systems, 1.1% of the best track intensity
changes broke the 24-h constraints.
• Out of 273 cases for weakening systems,
none of the best track intensity changes
broke the 24-h constraints.
Results for the 18-h changesResults for the 18-h changes
• Out of 557 cases for intensifying systems,
0.7% of the best track intensity changes
broke the 18-h constraints.
• Out of 271 cases for weakening systems,
none of the best track intensity changes
broke the 18-h constraints.
• Out of 557 cases for intensifying systems,
0.7% of the best track intensity changes
broke the 18-h constraints.
• Out of 271 cases for weakening systems,
none of the best track intensity changes
broke the 18-h constraints.
Results for the 12-h changesResults for the 12-h changes
• Out of 579 cases for intensifying systems,
1.7% of the best track intensity changes
broke the 12-h constraints.
• Out of 318 cases for weakening systems,
0.3% of the best track intensity changes
broke the 12-h constraints.
• Out of 579 cases for intensifying systems,
1.7% of the best track intensity changes
broke the 12-h constraints.
• Out of 318 cases for weakening systems,
0.3% of the best track intensity changes
broke the 12-h constraints.
Results for the 6-h changesResults for the 6-h changes
• Out of 492 cases for intensifying systems,
1.2% of the best track intensity changes
broke the 6-h constraints.
• Out of 265 cases for weakening systems,
0.8% of the best track intensity changes
broke the 6-h constraints.
• Out of 492 cases for intensifying systems,
1.2% of the best track intensity changes
broke the 6-h constraints.
• Out of 265 cases for weakening systems,
0.8% of the best track intensity changes
broke the 6-h constraints.
How often do the changes in the
Data-T numbers break constraints?
How often do the changes in the
Data-T numbers break constraints?
DT numbers, while
subjective, are not
constrained in the
Dvorak Technique.
They violated the FT
constraints about 1%
of the cases for 24-h
changes up to about
8% of the cases for 6-h
changes – higher
percentages than for
the best track data.
Note: The DT data set
includes all available
DT regardless of the
presence of aircraft.
The two data sets are
not homogeneous!
DT DT
BT BT
• Dvorak calibration and validation studies were
made in the 1970’s using the aircraft data of that
era.
• Current aircraft-based best track allows for a
more rigorous assessment of the accuracy and
calibration of the Dvorak Technique than was
possible prior to 1998.
• Dvorak calibration and validation studies were
made in the 1970’s using the aircraft data of that
era.
• Current aircraft-based best track allows for a
more rigorous assessment of the accuracy and
calibration of the Dvorak Technique than was
possible prior to 1998.
Dvorak Technique Errors
A Closer Examination
Dvorak Technique Errors
A Closer Examination
R2=0.928
R2=0.899
R2=0.931
R2=0.935
R2=0.934
• In the Atlantic basin, the vast majority of the observed 6,
12, 18, and 24-h hour intensity changes (based on
aircraft-based best track data) fall within the constraints
of the Dvorak Technique.
• The technique constraints appear to need little or no
change in the Atlantic basin.
• The constraints for the CIMSS Automated Dvorak
Technique (which are looser than those of the subjective
Dvorak Technique) may need adjustment.
• However, there is a fundamental question: Should the
constraints be large enough to cover every case, or just a
statistically significant majority of the cases?
• In the Atlantic basin, the vast majority of the observed 6,
12, 18, and 24-h hour intensity changes (based on
aircraft-based best track data) fall within the constraints
of the Dvorak Technique.
• The technique constraints appear to need little or no
change in the Atlantic basin.
• The constraints for the CIMSS Automated Dvorak
Technique (which are looser than those of the subjective
Dvorak Technique) may need adjustment.
• However, there is a fundamental question: Should the
constraints be large enough to cover every case, or just a
statistically significant majority of the cases?
Summary and ConclusionsSummary and Conclusions
• Data-T number changes fall outside of the constraints of
the technique more often than the best track data.
• For the TAFB satellite fixes, the various Dvorak numbers
show a low intensity bias for weaker system and a high
intensity bias for stronger ones.
• The Data-T numbers by themselves produce slightly less
accurate results, suggesting that the other components
of the technique are adding value to the analyses.
• Data-T number changes fall outside of the constraints of
the technique more often than the best track data.
• For the TAFB satellite fixes, the various Dvorak numbers
show a low intensity bias for weaker system and a high
intensity bias for stronger ones.
• The Data-T numbers by themselves produce slightly less
accurate results, suggesting that the other components
of the technique are adding value to the analyses.
Summary and ConclusionsSummary and Conclusions
• Put the current results through more rigorous
statistical significance tests.
• Stratify the various analyses by intensity.
• Investigate the constraints issue in other basins,
especially the western North Pacific (need ground
truth).
• Experiment with modifications to some of the
rules in the Dvorak Technique to see if they can
be improved.
• Put the current results through more rigorous
statistical significance tests.
• Stratify the various analyses by intensity.
• Investigate the constraints issue in other basins,
especially the western North Pacific (need ground
truth).
• Experiment with modifications to some of the
rules in the Dvorak Technique to see if they can
be improved.
Future WorkFuture Work
D-24 D-18 D-12 D-6 W-6 W-12
Keith (2000) Wilma (2005) Bret (1999) Bret (1999) Henri (2003) Lili (2002)
Wilma (2005) Dean (2007) Keith (2000)Humberto
(2001)Paloma (2008)
Felix (2007) Felix (2007)Humberto
(2001)Iris (2001)
Iris (2001) Felix (2007)
Katrina (2005)
Wilma (2005)
Felix (2007)
Lorenzo (2007)
Gustav (2008)
The Rule Breaking TCsThe Rule Breaking TCs