Highlight of the CWB WRF development

51
Highlight of the CWB WRF development

description

Highlight of the CWB WRF development . OUTLINE. Re-center the EAKF using the blending scheme Hybrid variational /ensemble data assimilation . Blending Scheme. Guess Blending. Analysis Blending. Fields to be Blended ( 13) - U, V, T, QVAPOR, PH, P, MU, - U10, V10, T2, Q2, PSFC, TH2. - PowerPoint PPT Presentation

Transcript of Highlight of the CWB WRF development

Page 1: Highlight of the CWB WRF development

Highlight of the CWB WRFdevelopment

Page 2: Highlight of the CWB WRF development

OUTLINE Re-center the EAKF using the blending scheme Hybrid variational/ensemble data assimilation

Page 3: Highlight of the CWB WRF development

Blending Scheme

RF

anal

RF

analanal MeanGFSMean ...

RF

fcsthr

RF

analfcsthr MeanGFSMean 6.6Guess

Blending

Analysis Blending

Fields to be Blended (13) - U, V, T, QVAPOR, PH, P, MU,- U10, V10, T2, Q2, PSFC, TH2

CLS=1200 km

Page 4: Highlight of the CWB WRF development

EAKF

Xa1

Xa2

Xa32

Xb1

Xb2

Xb32

guess

analy_mean

analysis

guess_meanBlend

analysis

Ensemble Forecast

6hr

Guess Blending

guess_meanBlend

RECENTER

Xb1

Xb2

Xb32

WRF 72hr

EAKF

Page 5: Highlight of the CWB WRF development

Xa1

Xa2

Xa32

Xb1

Xb2

Xb32

guess

Blend

Ensemble Forecast

6hr

Analysis Blending

Blend

analysis

Xa1

Xa2

Xa32

RECENTER

analy_meanBlend

EAKF

WRF 72hr

analy_mean

Page 6: Highlight of the CWB WRF development

Experimental Design

Model Version: CWB OP24 Run: Full Cycling Domain: CWB WRF Domain 1 (45KM) Period: 2008/09/04 00UTC ~ 2008/09/23 00 UTC T-PARC Experiments :

1) Without Blending (woBL, CTL)2) Guess Blending (guessBL)3) Analysis Blending (analyBL)4) analyBL + Ensemble Mean Analysis with Bleinding (analyBL_mBL)

Page 7: Highlight of the CWB WRF development

Ensemble Performance

T U V

guessBL

analyBL

woBL

over

small

Page 8: Highlight of the CWB WRF development

SCORE of Ensemble Mean Analysis Forecast 72hr (against NCEP)

H T U V24hr

woBL :代表無 Blending guessBL : Guess BlnedinganalyBL : Analysis Blending analyBL_m : blending again on analyBL ensemble mean

Page 9: Highlight of the CWB WRF development

H T U V48hr

SCORE of Ensemble Mean Analysis Forecast 72hr (against NCEP)

woBL :代表無 Blending guessBL : Guess BlnedinganalyBL : Analysis Blending analyBL_m : blending again on analyBL ensemble mean

Page 10: Highlight of the CWB WRF development

H T U V

72hr

SCORE of Ensemble Mean Analysis Forecast 72hr (against NCEP)

woBL :代表無 Blending guessBL : Guess BlnedinganalyBL : Analysis Blending analyBL_m : blending again on analyBL ensemble mean

Page 11: Highlight of the CWB WRF development

TC Track Forecasts - SINLAKU (1)

woBL analyBLguessBL analyBL_m

from 2008/09/09 00UTC to 2008/09/13 00UTC

woBL :代表無 Blending guessBL : Guess BlnedinganalyBL : Analysis Blending analyBL_m : blending again on analyBL ensemble mean

Page 12: Highlight of the CWB WRF development

TC Track Forecasts - SINLAKU (2)

from 2008/09/09 00UTC to 2008/09/13 00UTC

woBL :代表無 Blending guessBL : Guess BlnedinganalyBL : Analysis Blending analyBL_m : blending again on analyBL ensemble mean

Page 13: Highlight of the CWB WRF development

TC Track Forecasts - HAGUPIT (1)

woBL

analyBL

guessBL

analyBL_m

from 2008/09/20 00UTC to 2008/09/23 00UTC

woBL :代表無 Blending guessBL : Guess BlnedinganalyBL : Analysis Blending analyBL_m : blending again on analyBL ensemble mean

Page 14: Highlight of the CWB WRF development

TC Track Forecasts - HAGUPIT (2)

from 2008/09/20 00UTC to 2008/09/23 00UTC

woBL :代表無 Blending guessBL : Guess BlnedinganalyBL : Analysis Blending analyBL_m : blending again on analyBL ensemble mean

Page 15: Highlight of the CWB WRF development

Summary Guess-blending outperforms the other

experiments, especially for the typhoon track More detail analysis is undergoing to answer why

guess blending is the best The effect of the blending scheme is bounded

by the GFS performance, the blending strategy should be re-evaluated as the GFS was improved, e.g. 2012 NCEP GFS

Page 16: Highlight of the CWB WRF development

Performance of the hybrid system

Single Observation Tests in 3DVAR-Hybrid - Tests of Tuning Factors in Hybrid

Localization Scale Ensemble Covariance Weighting Vertical localization

The use of the ensemble perturbation

Page 17: Highlight of the CWB WRF development

Single Observation Tests in 3DVAR-Hybrid

Observation Setting

Tuning Factors in Hybrid Localization Scale Ensemble Covariance Weighting

• Temperature• innov = 1 K , obs_err = 1 K• Lon ~ 137.143 ( x = 150 )• Lat ~ 28.2 ( y = 69 )• Lev ~ 860 mb ( z = 11 )

Page 18: Highlight of the CWB WRF development

Localization Scale 200 km

375 km

750 km

Ensemble System : EAKFShade : IncrementGreen Line : Geopotential Height of Ensemble Forecast 6hr Mean.

Page 19: Highlight of the CWB WRF development

VAR

50%

75%

Full

Ensemble Covariance Weighting Factor

Ensemble System : EAKFShade : IncrementGreen Line : Geopotential Height of Ensemble Forecast 6hr Mean.

Page 20: Highlight of the CWB WRF development

EnSRF HGSI

Y-Z Plane

Lev=11 (~860mb

)

HVAREAKF

Page 21: Highlight of the CWB WRF development

Lev=21 (~520mb

)Y-Z

Plane EnSRF HGSI

HVAREAKF

Page 22: Highlight of the CWB WRF development

Lev=30 (~250mb

)Y-Z

Plane EnSRF HGSI

HVAREAKF

Page 23: Highlight of the CWB WRF development

𝑳𝒄=𝟏𝟎𝒏𝒛 ×𝒌𝒄

𝛒 (𝒌 ,𝒌𝒄 )=𝒆𝒙𝒑 [− (𝒌−𝒌𝒄 )𝟐 /𝑳𝒄𝟐 ]

Page 24: Highlight of the CWB WRF development

Summary Localization scale of 200 km and 75%

ensemble BE are used in the hybrid analysis. Vertical localization in WRF VAR-Hybrid

should be tuned. The moisture field has the most dramatic

change as including the ensemble BE.

Page 25: Highlight of the CWB WRF development

Evaluation of the 3DVAR-Hybrid

PART-1: Ensemble Members of EAKF without/with Blending PART-2: 3DVAR vs. Hybrid-EAKFPART-3: Hybrid-WEPS vs. Hybrid-EAKFPART-4:WEPS + EAKF

Page 26: Highlight of the CWB WRF development

Evaluation of the 3DVAR-Hybrid

3DVAR vs. Hybrid-EAKFExperimental Design: Model Version: CWB OP25 Run: Partial Cycling Domain: CWB WRF Domain 1 (45KM) Period: 2008/09/09 00UTC ~ 2008/09/30 12 UTC T-PARC Ensemble members of Hybrid-EAKF from “analyBL”

case. Verified against the NCEP GFS analysis and

RAOB/dropsound

Page 27: Highlight of the CWB WRF development

H T U V

3DVAR HEAKF

ana

Page 28: Highlight of the CWB WRF development

H T U V

3DVAR HEAKF

12hr

Page 29: Highlight of the CWB WRF development

H T U V

3DVAR HEAKF

24hr

Page 30: Highlight of the CWB WRF development

H T U V

3DVAR HEAKF

72hr

Page 31: Highlight of the CWB WRF development

Shade :Difference ( 3DVAR - HEAKF )

Green Line :Geopotential Height of HEAKF

U VOne Month Mean

100 mbana

12hr

72hr

Page 32: Highlight of the CWB WRF development

U VOne Month Mean

300 mbana

12hr

72hr

Shade :Difference ( 3DVAR - HEAKF )

Green Line :Geopotential Height of HEAKF

Page 33: Highlight of the CWB WRF development

One Month Mean

850 mbana

12hr

72hr

Shade :Difference ( 3DVAR - HEAKF )

Green Line :Geopotential Height of HEAKF

T Q

Page 34: Highlight of the CWB WRF development

OBSLocation

Page 35: Highlight of the CWB WRF development

12hr

T WIND Qs

3DVAR HEAKF

Page 36: Highlight of the CWB WRF development

T WIND Qs

24hr 3DVAR

HEAKF

Page 37: Highlight of the CWB WRF development

T WIND Qs

72hr 3DVAR

HEAKF

Page 38: Highlight of the CWB WRF development

Summary

• Compare to the 3DVAR, the forecast performance is slightly better in the hybrid system.

• The major differences are high level wind, low level temperature and moisture.– The reason to cause the difference is not clear yet.

Page 39: Highlight of the CWB WRF development

Evaluation of the 3DVAR-Hybrid

Hybrid-WEPS vs. Hybrid-EAKF

Hybrid-EAKF

Hybrid-WEPSϐe : ϐb = 0.75 : 0.25

Spin-up

Page 40: Highlight of the CWB WRF development

HWPES HEAKF

SCORE of Ensemble Mean Analysis Forecast 72hr (against NCEP)

T ANA.

24hr

48hr

72hr

Page 41: Highlight of the CWB WRF development

T

Qv

3DVAR HEAKF HWEP

S HEAKF850 mb

Analysis Difference in SINLAKU Period

Page 42: Highlight of the CWB WRF development

3DVAR

SINLAKU Typhoon Track

HEAKF HWEPS

from 2008/08/09 12UTC to 2008/08/13 00UTC

Page 43: Highlight of the CWB WRF development

3DVAR

HAGUPIT Typhoon Track

HEAKF

from 2008/08/20 00UTC to 2008/08/23 00UTC

HWEPS

Page 44: Highlight of the CWB WRF development

3DVAR

from 2008/08/24 12UTC to 2008/08/28 00UTC

JANGMI Typhoon Track

HEAKF HWEPS

Page 45: Highlight of the CWB WRF development

Evaluation of the 3DVAR-Hybrid

Hybrid-EAKF+WEPS vs. Hybrid-EAKF

32 members EAKF

20 members WEPS

52 membersEAKF+WEPS 3DVAR-Hybrid

Page 46: Highlight of the CWB WRF development

Analysis Difference in SINLAKU Period 850 mb T, Qv

HEKWP - HEAKF

HWEPS - HEAKF

HEKWP

T Qv

Page 47: Highlight of the CWB WRF development

12hr

24hr

48hr

72hr

SCORE of Ensemble Mean Analysis Forecast 72hr (against DROPSONDES_SPECIFIC_HUMIDITY)

HEKWPHEAKF

Page 48: Highlight of the CWB WRF development

1.0𝛽e Lev=11 (~860mb)

T Q U V

HWEPS

HEAKF

HEKWP

Full ensemble mode

Page 49: Highlight of the CWB WRF development

1.0 Lev=30 (~250mb)𝛽e

T Q U V

HEAKF

HEKWP

HWEPS

Full ensemble mode

Page 50: Highlight of the CWB WRF development

Summary The difference between HEAKF,

HWEPS, and HEKW is limited.HWEPS has the moisture spin-up

issue.

Page 51: Highlight of the CWB WRF development

Flow dependent BE

Initial perturbation

Model pert.

BC pert.

Deterministic prediction

Ensemble prediction system

The advanced data assimilation system

Recentered-EAKF

Ensemble data assimilation system play a key role to support the ensemble and deterministic prediction

Dual-resolution3DVAR hybrid DA

4D VAR (hybrid)

Rapid updated radar DA

blending