Tropical Cyclone

1
ISRO/MOP/SM-2.1
Tropical Cyclone Studies by Microwave Sensors

Chandra Mohan Kishtawal ASD/MOG Space
Applications Centre
2
Objectives TC Geolocation, Intensity
Estimation and prediction Using Microwave
observations
Data TMI observations for TC Over global
oceans during past 5 years ( more than 400 TMI
scenes analyzed). TC Track and Intensity data
was collected from NHC/TPC archives for algorithm
development and validation
3
Cyclone Geolocation Using Microwave Observations
4
Warned region is 3 times larger than the
region where actual damage takes place.
This proves Very Expensive. Also this shows the
importance of Even A marginal improve ment in
track prediction accuracy.
Warned region
Damaged Region
5
Impact of Initial Position Error on Track Forecast
6
A Comparison of
Microwave
and
Infrared
Observations of Tropical Cyclones
7
Sensitivity of different TMI frequencies to
TC-Rain
BT
8 km
85 GHz
BT
4 km
19-37 GHz
2 km
10 GHz
BT
TRMM - 33151
0
Rain Rate
50 mm/h
8
Two Main Microwave Sensing Channels for TCs
37 GHz Warm Precipitation Against Colder Ocean
Background
85 GHz Cold Precipitation Against Warmer Ocean
Background
Warm
Cold
Warm
Cold
9
PARALAX PROBLEM IN CONICAL SCAN
Paralax Errors 85 GHz 15-20 km 37 Ghz 5 km
Paralax Error
10
Example of Paralax
08-Aug-2000, 1057 UTC TC-JALAWAT
37 GHz
85 GHz
11
Differences between TMI derived TC centers from
Best-track Positions (IMD) ( After Paralax
Compensation)
12
Cyclone Intensity Estimation
13
Operational Centers worldwide still depend on
Dvoraks technique for TC intensity estimates
that uses manual pattern-analysis of VIS/IR
images. In operational set-up it proves
slow. We developed an automatic technique for
TC intensity assessment, that is quick, and
reliable.
14
CONVCTIVE ORGANIZATION WITHIN STORMS
15582
33108
15
Sensitivity of different TMI frequencies to
TC-Rain
BT
8 km
85 GHz
BT
4 km
19-37 GHz
2 km
10 GHz
BT
TRMM - 33151
0
Rain Rate
50 mm/h
16
2.5O
1.0O
17
Quantifying Isotropy of Convection
ISO ?i Øi /((n-1) A), n12
(5)
Øi (Loge(Ni1) A) if Loge(Ni1) ? A,
otherwise Øi 0
NI No of TMI pixels with PCT lt 240 K
ISOIN 0.621 ISOOUT 0.234
ISOIN 0.832 ISOOUT 0.523
18
ALGORITHM DEVELOPMENT BY GENETIC ALGORITHMS

• Randomized search and optimization technique
  guided by the principle of natural genetic
  systems.

19
GENETIC EVOLUTION OF PATTERNS
PARENT-2
PARENT-1
CHROMOSOMES
20
PARENT-2
PARENT-1
CHILDREN
21
A Simplified Concept of Genetic Algorithm
2
1
Random Initialization of Equation Population
Select the best individuals as per cost
1
2
Best ones get chance to reproduce
Offspring again reproduces as per merit
Mutation of a fraction of low-order population
Fittest individual emerges after N generations
22
PARAMETER LIST FOR INTENSITY ESTIMATION
MEAN BT10(H) 10-MAX(BT10H) 10-MIN(BT10H)
10-GHZ BT WITHIN 2 DEG RADIUS
23
Maximum Sensitivity Region
Distance from Center
24
CONVECTIVE ISOTROPY (SYMETRY OF THE REGION
DEFINED BY PCT lt 240 K)
ISO ?i Øi /((n-1) A), n12
(5)
Øi (Loge(Ni1) A) if Loge(Ni1) ? A,
otherwise Øi 0
NI No of TMI pixels with PCT lt 240 K
25
MSW(kt) a-d/(i-7.09)(ef-d)/
((-52.15c/b-f/(h-75.75))
(-21.96))b-168.17

26
SENSITIVITY OF DIFFERENT TERMS
27
Automatic Intensity Estimation Skill for Global
TCs
TC-CASES NIO NATL NEP
(Mean 11 kt)
Paper to appear in GRLApril-2005.
28
Automatic Intensity Estimation Case Studies
Depression
Severe Cyclone
18-Oct-2000
22-May-2001
JTWC 25 Kt Estimated 27 Kt
JTWC 60 Kt Estimated 52 Kt
29
Automatic Intensity Estimation Case Studies
Very Severe Cyclone-1
Very Severe Cyclone-2
16-Oct-1999
18-May-1999
JTWC 94 Kt Estimated 88 Kt
JTWC 110 Kt Estimated 120 Kt
30
Automatic Intensity Estimation Skill Levels
TMI estimated v/s JTWC Intensity
Correlations and RMS Error Training Set ( 60
TMI Scenes) 91 Verification Set ( 20
TMI Scenes) 90 Mean RMS error 12.53 Kt
Compare with
Bankert Tag-2002 RMSE 19.7 Kt NEP ATL IO
31
Cyclone Intensity Prediction
32
OBSERVATION-1 Intensification Process Of Weak
Cyclones ( Msw lt 64 Kt) is very much different
from that of strong cyclones (MSW gt 64 kt)
Area of cyclonic influence
(Rou/(fr) 1, core boundary)
Environmental forcing begins To take over.
Eye wall
Principal Band

• The outward edge of bands respond earliest to
  environmental flow
• Convective bands transport large cloud mass
  upward, much larger than
• eye-wall

33
Mean of 5 low frequency channels over the
un-masked region
Convective Mass in high CLW region ( BT-37H gt 240
K) Convective Mass ?CM CM(240-PCT)1.1 if PCT
lt 240 K , Else CM0
Minimum PCT in high CLW region
High CLW region
BT ( 37-H)
34
With the use of Cloud Mask, the correlations of
low frequency channels with 24-hour intensity
change improve, implying that much of the
signals arrive from outside the storm ( due to
wind ? SST ? ) However these are unusable if
storm intensity increases beyond 60 kt.
PCTmin is computed from masked area in both the
graphs. It is shown only for comparison
35
Convective Mass in Inner Core ( r lt 1.3o)
Convective Mass ?CM CM(230-PCT)1.1 if PCT lt
230 K , Else CM0
Convective Isotropy in Inner Core ( r lt 1.3o)
Convective Isotropy in outer Core ( 1.3o lt r lt
2.5o)
PCT (K)
PCT (K)
Low Isotropy Case
High Isotropy Case
36
Minimum PCT in inner core Average PCT in
inner core Average 10V BT inner core Average
10V BT in outer core
BT (37-H)
Convective SHEAR ( angular shift b/w high density
region of high BT(37H) and that of low PCT in 85
GHz image.
PCT
37
Picking the SST Signatures
Mean of 10 GHz (V) BT in 45o angular section
surrounding the direction of cyclone
motion during past 12 hours. A Pixel is
Considered only if BT(37-H) lt 185 K. This
parameter may pick SST signatures ahead of a TC
Direction of TC Motion in last 12 hours
38
Mean Histograms Of Decaying And Intensifying
Storms
39
(No Transcript)
40
BAR-CODING FOR SIGNAL ENHANCEMENT
41
(Accuracy 8 kt)
42
Thanks