Hurricane WRF(HWRF) 與Typhoon WRF(TWRF) 於西北太平洋颱風路徑及降雨預報表現之評估

蔡金成、蕭玲鳳、陳得松、包劍文與李清勝

Introduction of TAPEX TAiwan cooperative Precipitation Ensemble forecast eXperiment

2

Model configuration for ensemble members in TAPEX, 2015

編號-群組 model ICs LBCs Cumulus scheme

Microphysics scheme Boundary Layer note

11-C1 WRF partial cycle 3DVAR (CV5+OL) NCEPGFS KF Goddard YSU with blending, 2way

01-C2 WRF partial cycle 3DVAR (CV5+OL) NCEPGFS GD Goddard YSU

02-C3 WRF partial cycle 3DVAR (CV5+OL) NCEPGFS G3 Goddard YSU

03-C4 WRF partial cycle 3DVAR (CV5+OL) NCEPGFS BMJ Goddard YSU

06-C5 WRF cold start 3DVAR (CV5+OL) NCEPGFS KF(1) Goddard YSU

07-C6 WRF cold start 3DVAR (CV5+OL) NCEPGFS GD(3) Goddard YSU

08-C7 WRF cold start 3DVAR (CV5+OL) NCEPGFS G3(5) Goddard YSU

09-C8 WRF cold start 3DVAR (CV5+OL) NCEPGFS BMJ Goddard YSU

10-I1 WRF partial cycle 3DVAR (CV3) NCEPGFS KF Goddard YSU with blending

12-I2 WRF partial cycle 3DVAR (CV3) CWBGFS KF Goddard YSU with blending

04-I3 WRF partial cycle 3DVAR (CV5) NCEPGFS KF Goddard YSU

05-I4 WRF partial cycle 3DVAR (CV5+OL) NCEPGFS KF Goddard YSU no RO data; VS 11 with blending, 2way

13-I5 WRF cold start 3DVAR (CV3) NCEPGFS KF Goddard YSU

14-I6 WRF cold start 3DVAR (CV5) NCEPGFS KF Goddard YSU

15-I7 WRF cold start 3DVAR (CV5+OL) NCEPGFS KF Goddard YSU no RO data; VS 06

16-O1 CreSS cold start NODA NCEPGFS No CPS Cold rain Mellor and Yamada grid size 5km

17-O2 CreSS cold start NODA NCEPGFS No CPS Cold rain Mellor and Yamada grid size 2.5 km

18-O3 WRF cold start NODA NCEPGFS KF WSM5 YSU e_vert=28,two way,d03cu

19-O4 MM5 cold start NODA NCEPGFS Grell Goddard MRF

20-O5 MM5 cold start 4DVAR bogus NCEPGFS Grell Goddard MRF

21-H1 HWRF cold start NODA NCEPGFS SAS Ferrier NCEP GFS 27/9/3

22-H2 HWRF cold start NODA HWRF bogus NCEPGFS SAS Ferrier NCEP GFS 27/9/3, moving

nested

23-C9 WRF partial cycle 3DVAR (CV5+OL) NCEPGFS Tiedtke Goddard YSU cu=Tiedtke

24-C10 WRF cold start 3DVAR (CV5+OL) NCEPGFS Tiedtke Goddard YSU cu=Tiedtke

25-R1 WRF cold start 3DVAR (CV5) NCEPGFS KF Goddard YSU Radar, 30 hrs, d03NOcu

26-O6 WRF cold start NODA NCEPGFS KF Goddard YSU Large d03:301*241

27-R2 WRF cold start 3DVAR (CV7) 15/3km NCEPGFS KF Goddard YSU Radar, 30 hrs, d02NOcu,CV7

28-R3 WRF cold start 3DVAR (CV5) 15/3km NCEPGFS KF Goddard YSU No Radar, 30 hrs, d02NOcu

Hurricane WRF in Taiwan

• TTFRI and CWB collaborated with ESRL/NOAA and had successfully implemented the Hurricane WRF(HWRF) model in Taiwan since 2012.

• In Apr. 2013, TTFRI and CWB had the operational version of HWRF from EMC.

• TTFRI had added 2 HWRF members into Taiwan Cooperative Precipitation Ensemble Forecast Experiment (TAPEX) from 2012 to 2014.

Configuration of HWRF

Static domain

Moving domain

Experiment cold/warm analysis bogus Cu. M.P. PBL Note

HWRF (M22) cold start NODA HWRF

bogus SAS Ferrier NCEP GFS

43levels; P_top 50hPa; 2 way; 27/9/3km; moving; no ocean couple & no GSI; IC&BC from NCEP T574 atmospheric spectral data and include NCEP 0.5deg data.

HWRF_T (M21) cold start NODA no

bogus SAS Ferrier NCEP GFS

43levels; P_top 50hPa; 1 way; 45/15/5 27/9/3km; static; no ocean couple & no GSI; IC&BC from NCEP 0.5deg data.

5

2013-2014 TTFRI HWRF vs. EMC HWRF

26

53

75

101

128

156

188

28 51

72

94

125

155

188

0

50

100

150

200

0(704) 12(650) 24(593) 36(532) 48(475) 60(417) 72(362)

Track err

or(km)

Forecast hour(# of cases)

2013-14EMC_HWRF

2013-14TTFRI_HWRF

MSLP of HWRF_T(45/15/5km) has Bull’s eye

06/20 18Z +18fhr

06/24 00Z +12fhr

07/11 00Z +42fhr

07/12 00Z +24fhr

Reduce terrain resolution >> adjust ptsgm level from 200hPa to 420hPa(HWRF)

06/20 18Z +18fhr

06/24 00Z +12fhr

07/11 00Z +42fhr

07/12 00Z +24fhr

Bull’s eye of MSLP had significantly reduced when increasing the resolution(27/9/3km)

HWRF with static domain • In typhoon season 2014

– Increase the horizontal resolution of 3-nested domain from 45/15/5 km to 27/9/3 km.

– Reduce the vertical resolution from 43 to 28 levs for limited computing resource.

– But the track performance was poor after 24 forecast hour.

9

44 70

124

168

212

266

324

44 70

111 145

162 205

235

35 72

101 132

158 196 205

0

50

100

150

200

250

300

350

0(54) 12(51) 24(46) 36(43) 48(39) 60(35) 72(31)

Trac

k Er

ror

Forecast hour(# of cases)

typhoon MATMO and FUNG-WONG(2014)

M21_28levs_d1

M21_43levs_d1

M11_d1(TWRF)

HWRF with moving nested domain

10

44 70

124

168

212

266

324

44 70

111

145 162

205

235

35

72 101

132 158

196 205

63 78

91 117

158

191

229

0

50

100

150

200

250

300

350

0(54) 12(51) 24(46) 36(43) 48(39) 60(35) 72(31)

Trac

k Er

ror

Forecast hour(# of cases)

typhoon MATMO and FUNG-WONG(2014)

M21_28levs_d1

M21_43levs_d1

M11_d1(TWRF)

M22_d1(HWRF)

Track performance of HWRF and TWRF

11

37 54

109

165

229

294

367

37 51

97

144

189

239

283

38 56

88 117

143

194 207

50 70 80

114 153

187 223

0

50

100

150

200

250

300

350

400

0(18) 12(22) 24(26) 36(26) 48(26) 60(22) 72(18)

Tra

ck E

rror

(km

)

Forecast hour(# of cases)

MATMO and FUNG-WONG(2014), heavy rainfall period

M21_28levs_d1

M21_43levs_d1

M11_d1(TWRF)

M22_d1(HWRF)

Rainfall forecast skill score of typhoon FUNG-WONG&MATMO (18 cases)

12

0 0.1 0.2 0.3 0.4 0.5 0.6

50 80 130 200 350 500

Thre

at S

core

Threshold

MATMO&FUNG-WONG(2014) 00-24 fhr

M21_43Lev M21_28Lev M11 M22

0 0.5

1 1.5

2 2.5

3

50 80 130 200 350 500

BIAS

Threshold

MATMO&FUNG-WONG(2014) 00-24 fhr

M21_43Lev M21_28Lev M11 M22

0 0.1 0.2 0.3 0.4 0.5 0.6

50 80 130 200 350 500

Thre

at S

core

Threshold

MATMO&FUNG-WONG(2014) 24-48 fhr

M21_43Lev M21_28Lev M11 M22

0 0.5

1 1.5

2 2.5

3

50 80 130 200 350 500

BIAS

Threshold

MATMO&FUNG-WONG(2014) 24-48 fhr

M21_43Lev M21_28Lev M11 M22

0 0.1 0.2 0.3 0.4 0.5 0.6

50 80 130 200 350 500

Thre

at S

core

Threshold

MATMO&FUNG-WONG(2014) 48-72 fhr

M21_43Lev M21_28Lev M11 M22

0 0.5

1 1.5

2 2.5

3

50 80 130 200 350 500

BIAS

Threshold

MATMO&FUNG-WONG(2014) 48-72 fhr

M21_43Lev M21_28Lev M11 M22

Summary & Future work

• We successfully implemented HWRF system in Taiwan and evaluated the track and rainfall forecast performance of HWRF over WNP from 2013-2014.

• For static domain HWRF configuration with higher vertical(43levs) and horizontal(3 km) resolution can have better capabilities in forecasting typhoon track and rainfall.

• To update the 2015 operational HWRF version and continue evaluating the performance over the WNP.

13

2015/10/16 14

Data Application of TAPEX

TAPEX

Super computer

Integral universities and government

agencies Rainfall

Track

Hydrological modeling

Storm Surge

Wind

Temperature

MSLP

Wind Power

Solar

Inundation