ming hu 1,2 , yujie pan 3 , kefeng zhu 3 , xuguang wang 3 , ming xue 3 ,

15
Development of an EnKF/Hybrid Data Assimilation System for Mesoscale Application with the Rapid Refresh Ming Hu 1,2 , Yujie Pan 3 , Kefeng Zhu 3 , Xuguang Wang 3 , Ming Xue 3 , David Dowell 1 , Steve Weygandt 1 , Stan Benjamin 1 , Jeff Whitaker 4 , Curtis Alexander 1,2 1. Global System Division, ESRL/NOAA, Boulder, CO 2. CIRES, University of Colorado, Boulder, CO 3. CAPS, University of Oklahoma, Norman, OK 4. Physical Sciences Division, ESRL/NOAA, Boulder, CO 17 th conference on IOAS-AOLS Austin, TX 8 January 2013 1

Upload: cisco

Post on 22-Feb-2016

55 views

Category:

Documents


0 download

DESCRIPTION

17 th conference on IOAS- AOLS Austin , TX 8 January 2013. Development of an EnKF /Hybrid D ata A ssimilation S ystem for M esoscale A pplication with the Rapid Refresh. Ming Hu 1,2 , Yujie Pan 3 , Kefeng Zhu 3 , Xuguang Wang 3 , Ming Xue 3 , - PowerPoint PPT Presentation

TRANSCRIPT

Page 1: Ming  Hu 1,2 ,  Yujie  Pan 3 ,  Kefeng  Zhu 3 ,  Xuguang  Wang 3 , Ming Xue 3 ,

1

Development of an EnKF/Hybrid Data Assimilation System for Mesoscale Application with the Rapid Refresh

Ming Hu1,2, Yujie Pan3, Kefeng Zhu3, Xuguang Wang3, Ming Xue3,David Dowell1, Steve Weygandt1,

Stan Benjamin1, Jeff Whitaker4, Curtis Alexander1,2

1. Global System Division, ESRL/NOAA, Boulder, CO2. CIRES, University of Colorado, Boulder, CO

3. CAPS, University of Oklahoma, Norman, OK4. Physical Sciences Division, ESRL/NOAA, Boulder, CO

17th conference on IOAS-AOLSAustin, TX 8 January 2013

Page 2: Ming  Hu 1,2 ,  Yujie  Pan 3 ,  Kefeng  Zhu 3 ,  Xuguang  Wang 3 , Ming Xue 3 ,

2

Introduction o Rapid Refresh (RAP) is an operational hourly updated

regional numerical weather prediction system for aviation and severe weather forecasting

o GSI-3DVar is used for RAP data assimilationStephen S. Weygandt : Recent Rapid Refresh Enhancements to Improve Forecast Guidance for Aviation Weather Hazards and Improve Initial Fields for High Resolution Rapid Refresh Forecasts. 9.1 in 16th Conference on Aviation, Range, and Aerospace MeteorologyThursday, 10 January 2013: 8:30 AM.

o RAP evolves to a 6-member North American Rapid Refresh Ensemble in the future (2016)

o Testing of an hourly updating EnKF-3DVAR hybrid or EnKF capability for the RAP is underway• OU/CAPS, ESRL, and NCEP/EMC collaboration

Page 3: Ming  Hu 1,2 ,  Yujie  Pan 3 ,  Kefeng  Zhu 3 ,  Xuguang  Wang 3 , Ming Xue 3 ,

3

RAP hybrid/EnKF: Benefits

o High-resolution hourly update cyclesSituational Awareness NWP = flow dependent

o For surface and low level weather system• highly localized system• Vertical flow dependence, much needed for good surface

data analysiso For cloud analysis and severe weather

• Anisotropic distribution• Build better situation-dependent balance among T, Q and

cloud variables in analysis increment

Page 4: Ming  Hu 1,2 ,  Yujie  Pan 3 ,  Kefeng  Zhu 3 ,  Xuguang  Wang 3 , Ming Xue 3 ,

4

RAP hybrid/EnKF: Challenges

o High-resolution hourly update cycles• Huge computation cost• Short cut-off time: ensemble forecast needs to be

done within a short time• Ensemble convergence fast in hourly analysis

o For surface and low-level weather systems• Ensemble spread is usually poor in low levels

o For cloud analysis and severe weather• Ensemble requires special physical configuration

suitable for cloud and severe weather analysis

Page 5: Ming  Hu 1,2 ,  Yujie  Pan 3 ,  Kefeng  Zhu 3 ,  Xuguang  Wang 3 , Ming Xue 3 ,

5

Experiment system 1: RAP Hybrid System using RAP Ensemble

• Same 13km resolution and domain as operation RAP

• Hourly updated cycling with GSI Hybrid (2way) and EnKF

• Cold starts at 03Z May 30, 2012 and continue cycling 3 days

• 40 ensemble members

Using RAP configuration to build an hourly cycling 2-way hybrid system for testing the future implement of the Rapid

Refresh Ensemble

Page 6: Ming  Hu 1,2 ,  Yujie  Pan 3 ,  Kefeng  Zhu 3 ,  Xuguang  Wang 3 , Ming Xue 3 ,

6

GSI 3D-VarHMObs

Experiment system 2: RAP GSI hybrid using bkg error cov from GFS Ensemble

GSI 3D-Var

Obs

Cloud Anx

DigitalFilter

HMObs

ReflObs18 hr fcst

GSI 3D-Var

Obs

Cloud Anx

DigitalFilter

1 hr

fcst

HMObs

ReflObs18 hr fcst

Obs

Cloud Anx

DigitalFilter

ReflObs18 hr fcst

13z 14z 15z13 km RAP

1 hr

fcst

current real-time RAP configuration

Page 7: Ming  Hu 1,2 ,  Yujie  Pan 3 ,  Kefeng  Zhu 3 ,  Xuguang  Wang 3 , Ming Xue 3 ,

7

GSI HybridHMObs

Experiment system 2: RAP GSI hybrid using bkg error cov from GFS Ensemble

GSI Hybrid

Obs

Cloud Anx

DigitalFilter

HMObs

ReflObs18 hr fcst

GSI Hybrid

Obs

Cloud Anx

DigitalFilter

1 hr

fcst

HMObs

ReflObs18 hr fcst

Obs

Cloud Anx

DigitalFilter

ReflObs18 hr fcst

13z 14z 15z13 km RAP

1 hr

fcst

80 member GFS EnKF Ensemble forecast valid at

15Z (9-h fcst from 6Z)

Available 4 times a day valid at 03, 09, 15, 21Z

Page 8: Ming  Hu 1,2 ,  Yujie  Pan 3 ,  Kefeng  Zhu 3 ,  Xuguang  Wang 3 , Ming Xue 3 ,

8

Single observation test for GSI hybrid using bkg error cov from GFS Ensemble

GSI 3D-Var

GSI Hybrid(β=0)

Horizontal cross section of analysis increment from single T obs with 1.0 degree innovation

T

T V

VU

U

Page 9: Ming  Hu 1,2 ,  Yujie  Pan 3 ,  Kefeng  Zhu 3 ,  Xuguang  Wang 3 , Ming Xue 3 ,

9

Real-time Test for RAP hybrid using bkg error cov from GFS Ensemble

RMS profile for analysis – soundings from 1000-100mb

o Compare RAP development with GSI hybrid to RAP primary cycle with GSI-Var• Real-time test from Nov 22 to Dec 22, 2012 • GSI hybrid with half static BE and half BE from GFS

Ensemble forecasts

RAP hybrid RAP

TUV RH

Page 10: Ming  Hu 1,2 ,  Yujie  Pan 3 ,  Kefeng  Zhu 3 ,  Xuguang  Wang 3 , Ming Xue 3 ,

10

Forecast results: RMS profile

RMS profile for 3-h forecast – soundings from 1000-100mb

RMS profile for 12-h forecast – soundings from 1000-100mb

RAP hybrid RAP

TUV RH

TUV RH

Page 11: Ming  Hu 1,2 ,  Yujie  Pan 3 ,  Kefeng  Zhu 3 ,  Xuguang  Wang 3 , Ming Xue 3 ,

11

Forecast results: RMS time series

RAP hybrid RAP

UV RH

TUV

TRMS time series for 12-h forecast – soundings from 1000-100mb

RHRMS time series for 3-h forecast – soundings from 1000-100mb

Page 12: Ming  Hu 1,2 ,  Yujie  Pan 3 ,  Kefeng  Zhu 3 ,  Xuguang  Wang 3 , Ming Xue 3 ,

12

Conclusion

o GSI hybrid (using background error covariance from GFS ensemble) is very promising: the statistical results are clearly better than GSI Var• Wind is improved most, next is humidity• Temperature is improved mainly for 3-h but is neutral

for 12-h forecast• Middle to upper-air levels show clear improvement but

low levels are neutralo Successful ensemble forecasts used by GSI hybrid is

key of a successful GSI hybrid analysiso Need to improve RAP hybrid structure

Page 13: Ming  Hu 1,2 ,  Yujie  Pan 3 ,  Kefeng  Zhu 3 ,  Xuguang  Wang 3 , Ming Xue 3 ,

13

Future Work

o Tuning parameters• localization, ratio of ensemble BE and static BE, vertical

variance of this ratioo GFS ensemble forecast every 1 h rather than 3 h

• forecast valid at analysis timeo RAP ensemble forecast initialized from GFS EnKF

ensemble• Increase spread in low level• Create WRF special physical fields (such as cloud field)

o North American Rapid Refresh Ensemble by 2016, co-development between ESRL and NCEP/EMC

Page 14: Ming  Hu 1,2 ,  Yujie  Pan 3 ,  Kefeng  Zhu 3 ,  Xuguang  Wang 3 , Ming Xue 3 ,

14

Initial Test Results from RAP hybrid (EnKF/var) assimilation (1h, 13km)

RAP hybrid RAP

RMS profile for analysis – soundings from 1000-100mb

T

T

Q

Q

UV

UV

RMS profile for 3-h forecast – soundings from 1000-100mb

Page 15: Ming  Hu 1,2 ,  Yujie  Pan 3 ,  Kefeng  Zhu 3 ,  Xuguang  Wang 3 , Ming Xue 3 ,

15

Diagnosis of RAP hybrid using RAP ensemble

Horizontal distribution of Standard Deviation of surface pressure perturbation at 03z, 06z, 09z, 12z, 15z, 18z of May 30, 2012

Time series of prior observation-space ensemble standard deviation 13km

Rapid Refresh