26
Toshiro Inoue 1) Tomoo Ushio 2) , Daisuke Ka tagami 3) 1) Meteorological Research Institute 2) Osaka University Applications of Split Window Data for Rainfall Estimation

Toshiro Inoue 1) 、 Tomoo Ushio 2) , Daisuke Katagami 3) 1) Meteorological Research Institute

  • Upload
    august

  • View
    49

  • Download
    0

Embed Size (px)

DESCRIPTION

Applications of Split Window Data for Rainfall Estimation. Toshiro Inoue 1) 、 Tomoo Ushio 2) , Daisuke Katagami 3) 1) Meteorological Research Institute 2) Osaka University 3) Osaka Prefecture University. Cloud motion vector To construct high temporal rainfall map, - PowerPoint PPT Presentation

Citation preview

Page 1: Toshiro Inoue  1) 、 Tomoo Ushio  2) , Daisuke Katagami  3) 1) Meteorological Research Institute

Toshiro Inoue 1) 、 Tomoo Ushio 2), Daisuke Katagami 3)

1) Meteorological Research Institute

2) Osaka University

3) Osaka Prefecture University

Applications of Split Window Data for Rainfall Estimation

Page 2: Toshiro Inoue  1) 、 Tomoo Ushio  2) , Daisuke Katagami  3) 1) Meteorological Research Institute

1. Cloud motion vector   To construct high temporal rainfall map, we have to fill the observation gap from MW onboard LEO. 2. Convective/Stratiform rainfall type Dominant rainfall type depends on life stage of deep convection.  

Page 3: Toshiro Inoue  1) 、 Tomoo Ushio  2) , Daisuke Katagami  3) 1) Meteorological Research Institute

Cloud Motion Vector

• Filling the observation gap from the passive microwave onboard LEO becomes important issue to construct a higher temporal rainfall map.

• A method to estimate half-hourly rainfall map using cloud motion vector( Joyce et al., 2004) .

• Use of cloud motion vector is better than simple interpolation (Ushio et al., 2005) to fill the gap.

Page 4: Toshiro Inoue  1) 、 Tomoo Ushio  2) , Daisuke Katagami  3) 1) Meteorological Research Institute

Cloud motion vector by Split Window

Better rainfall area delineation by split window

Optically thicker cloud defined by the split window corresponds well to the rainfall area compared to the cloud area defined by single IR (Inoue and Aonashi, 2000; Inoue,1987)

Computing the cloud motion vector from split window is expected to be better than single IR.

Page 5: Toshiro Inoue  1) 、 Tomoo Ushio  2) , Daisuke Katagami  3) 1) Meteorological Research Institute

Data and MethodGOES-9 (over Japan)September, 2003Split Window (m,12m)Vector: 2D cross-correlation Template 6° lat/lon Every 3° lat/lon (Interpolated to 0.1 lat/lon)

Radar AMeDAS

Page 6: Toshiro Inoue  1) 、 Tomoo Ushio  2) , Daisuke Katagami  3) 1) Meteorological Research Institute

GSMaP (Global Satellite Mapping of Precipitation) is constructing hourly rainfall map over the globe with 0.1° lat/lon.

Considering current MW observation,we studied the time interpolation up to 5 hours.

Page 7: Toshiro Inoue  1) 、 Tomoo Ushio  2) , Daisuke Katagami  3) 1) Meteorological Research Institute

Radar AMeDAS image at 7 JST on 8th September, 2003 overlaid the cloud area defined as 273K or colder.

Radar AMeDAS image at 7 JST on 8th September, 2003 superimposed on the cloud area defined as 273K or colder with brightness temperature difference between the split window less than 2K.

Page 8: Toshiro Inoue  1) 、 Tomoo Ushio  2) , Daisuke Katagami  3) 1) Meteorological Research Institute

Same as former slides except for the 17 JST on 5th September, 2003.

Page 9: Toshiro Inoue  1) 、 Tomoo Ushio  2) , Daisuke Katagami  3) 1) Meteorological Research Institute

Cloud Motion Vector and Rainfall Motion Vector 18UTC 5th September, 2003

u

v

IR Split Window Radar AMeDAS

Page 10: Toshiro Inoue  1) 、 Tomoo Ushio  2) , Daisuke Katagami  3) 1) Meteorological Research Institute

1chBTD 2.5

BTD 2.0

BTD 1.5

BTD 1.0

BTD 0.5

1Hour0.594

0.600

0.710

0.678

0.617

0.544

2Hour0.430

0.460

0.521

0.477

0.515

0.351

3Hour0.378

0.418

0.438

0.341

0.417

0.225

4Hour0.399

0.355

0.373

0.263

0.289

0.151

5Hour0.207

0.187

0.236

0.164

0.117

0.108

Correlation coefficients between cloud tracked rainfall rate and Radar AMeDAS

on 5th September, 2003

Page 11: Toshiro Inoue  1) 、 Tomoo Ushio  2) , Daisuke Katagami  3) 1) Meteorological Research Institute

BTD 2.5

BTD 2.0

BTD 1.5

BTD 1.0

BTD 0.5

1Hour 0.98 1.08 1.29 1.20 1.16

2Hour 1.90 1.66 2.12 2.30 1.85

3Hour 1.83 1.65 1.72 2.05 1.56

4Hour 1.24 1.62 1.69 1.46 1.04

5Hour 1.42 1.08 1.34 1.24 1.62

Relative correlation coefficients (Split Window/ Single IR) We studied 30 cases during September, 2003.11 cases indicate better correlation coefficients.

Page 12: Toshiro Inoue  1) 、 Tomoo Ushio  2) , Daisuke Katagami  3) 1) Meteorological Research Institute

The use of Split Windowfor computing Cloud Motion Vector

• We studied 30 rainfall cases during September 2003, and split window data indicated better score in 11 cases out of 30.

• The correlation coefficients are improved by more than 50% for the 11 cases.

Page 13: Toshiro Inoue  1) 、 Tomoo Ushio  2) , Daisuke Katagami  3) 1) Meteorological Research Institute

Convective/Stratiform rainfall type

Split Window can classify Ci and Cb

The % of Ci within the deep convection is a good indicator to tell the life stage.

Depending on the life stage, dominance of rainfall type (conv/stra) is different..

Page 14: Toshiro Inoue  1) 、 Tomoo Ushio  2) , Daisuke Katagami  3) 1) Meteorological Research Institute

Data: Hourly data of GOES-W

Period: Jan., 2001-Dec., 2001 (except Sep., 2001)

Deep Convection: Cloud area colder than 253K

Life Cycle: Birth: First time of DC detected   End: Last time of DC detected

Case: No merge No split

% of Ci defined by Split Window within DC

Life Cycle of Deep Convection Defined by Split Window

Page 15: Toshiro Inoue  1) 、 Tomoo Ushio  2) , Daisuke Katagami  3) 1) Meteorological Research Institute

BTD Image by MSG(Ci is whiter, Cb/Cu black)

IR Image by MSG

Page 16: Toshiro Inoue  1) 、 Tomoo Ushio  2) , Daisuke Katagami  3) 1) Meteorological Research Institute

Cloud Type Map Classified by the Split Window

GOES-W 2 hourly from 12UTC May 09, 2001

Page 17: Toshiro Inoue  1) 、 Tomoo Ushio  2) , Daisuke Katagami  3) 1) Meteorological Research Institute

0

5

10

15

20

25

30

35

40

45

1 2 3 4 5 6 7 8 9 10 11

Time from DC Initiation

% of C

i with

in 253

K (D

C) 3HR

4HR 5HR 6HR 7HR 8HR 9HR10HR11HR

% of Ci within DC for each life time category

Page 18: Toshiro Inoue  1) 、 Tomoo Ushio  2) , Daisuke Katagami  3) 1) Meteorological Research Institute

0

5

10

15

20

25

30

35

40

1 2 3 4 5 6 7 8 9 10 11

Time from DC Initiation

Ci % w

ithin

253K Mean+s

MeanMean-s

Ci %

wit

hin

DC

developing

mature

decaying

Life Stage of Deep Convection

Page 19: Toshiro Inoue  1) 、 Tomoo Ushio  2) , Daisuke Katagami  3) 1) Meteorological Research Institute

Comparison between % of Ci within 253K cloud area and PR/TRMM rainfall data (3G68)

% of convective rain and % of Ci within DC

Rainfall Rate and % of Ci within DC

Page 20: Toshiro Inoue  1) 、 Tomoo Ushio  2) , Daisuke Katagami  3) 1) Meteorological Research Institute

% of Ci within DC

Good indicator the life stage of deep convection

Good indicator for dominance of Convective /Stratiform rainfall type classification as a system

Possible to apply the passive microwave estimation which has a tendency of overestimation at decaying stage comparing to PR estimation

Page 21: Toshiro Inoue  1) 、 Tomoo Ushio  2) , Daisuke Katagami  3) 1) Meteorological Research Institute

Thank you for your patience.

Page 22: Toshiro Inoue  1) 、 Tomoo Ushio  2) , Daisuke Katagami  3) 1) Meteorological Research Institute
Page 23: Toshiro Inoue  1) 、 Tomoo Ushio  2) , Daisuke Katagami  3) 1) Meteorological Research Institute
Page 24: Toshiro Inoue  1) 、 Tomoo Ushio  2) , Daisuke Katagami  3) 1) Meteorological Research Institute
Page 25: Toshiro Inoue  1) 、 Tomoo Ushio  2) , Daisuke Katagami  3) 1) Meteorological Research Institute
Page 26: Toshiro Inoue  1) 、 Tomoo Ushio  2) , Daisuke Katagami  3) 1) Meteorological Research Institute