18
Current Status and Plans of Ensemble Prediction System at KMA Seung-Woo Lee Numerical Model Development Division Korea Meteorological Administration GIFS-TIGGE WG 11 th meeting, Exeter, UK

Current Status and Plans of Ensemble Prediction System at KMA Seung-Woo Lee Numerical Model Development Division Korea Meteorological Administration GIFS-TIGGE

Embed Size (px)

Citation preview

Page 1: Current Status and Plans of Ensemble Prediction System at KMA Seung-Woo Lee Numerical Model Development Division Korea Meteorological Administration GIFS-TIGGE

Current Status and Plans of Ensemble Prediction System at KMA

Current Status and Plans of Ensemble Prediction System at KMA

Seung-Woo Lee

Numerical Model Development Division

Korea Meteorological Administration

GIFS-TIGGE WG 11th meeting, Exeter, UK

Page 2: Current Status and Plans of Ensemble Prediction System at KMA Seung-Woo Lee Numerical Model Development Division Korea Meteorological Administration GIFS-TIGGE

2

ContentsContents

• Outline of KMA operational EPS (KMA EPSG)

• Sensitivity test of KMA Hybrid Ensemble-4dVAR

• Future plans of KMA EPSs

• Summary

Page 3: Current Status and Plans of Ensemble Prediction System at KMA Seung-Woo Lee Numerical Model Development Division Korea Meteorological Administration GIFS-TIGGE

3

Brief history of KMA EPSG for TIGGEBrief history of KMA EPSG for TIGGE

2006.07.~2010.12.2006.07.~2010.12. 2010.12~2011.052010.12~2011.05 2011.5~2012.6.2011.5~2012.6. 2012.6~2013.62012.6~2013.6 2013.7.~2013.7.~

Model Base GDAPS (JMA) UM (UKMO, ver7.5) UM ver7.7 UM ver7.9 UM ver7.9

AssimilationMethod

3D‐Var 4D‐Var 4D‐Var 4D-Var Hybrid Ensemble Hybrid Ensemble 4D‐Var 4D‐Var

HorizontalResolution

T213 (Gausian grid) 0.5625 degree in lat/lon

N320 (~40km)0.5625 in lon/ 0.375 in lat.

N320 (~40km)0.5625 in lon/ 0.375 in lat.

N320 (~40km)0.5625 in lon/ 0.375 in lat.

N320 (~40km)0.5625 in lon/ 0.375 in lat.

Vertical levels / top of model

40 / ~0.4 hPa 50 / ~63 km 70 / ~80 km 70 / ~80 km 70 / ~80 km

InitialTimes

00,12 00, 12 00,12 00,1200, 12 (06, 18 for cycled hybrid)

LeadTime

10 days 10 days 10 days 10 days 12 days12 days

OutputFrequency

6h 6h 6h 6h 6h to 240h,12h to 288

No. ofMembers (+control)

15+1 23+1 23+1 23+1 23+1

CoupledOcean

No No No No No

InitialPerturbations

Breeding + factor rotation

ETKF ETKF ETKF ETKF

ModelPerturbations

No RP, SKEB2 RP, SKEB2 RP, SKEB2 RP, SKEB2

SurfacePerturbations

No No No SST Perturbation SST Perturbation

Page 4: Current Status and Plans of Ensemble Prediction System at KMA Seung-Woo Lee Numerical Model Development Division Korea Meteorological Administration GIFS-TIGGE

4

Major change in EPSG in 2012~13Major change in EPSG in 2012~13

Trim obstore

OPS

ETKFETKF

UMUMN320L70N320L70

Varobsobstore

-6 hour-6 hourEPSG cycleEPSG cycle

-6 hour-6 hourEPSG cycleEPSG cycle

OPS background

OPS backgroundETKF background

SST SST statisticsstatistics

SST SST statisticsstatistics

GDPS(N512L70)GDPS(N512L70)OPS, VAR, UMOPS, VAR, UM

GDPS(N512L70)GDPS(N512L70)OPS, VAR, UMOPS, VAR, UM

Initial DumpReconfiguration

+6 hour+6 hourEPSG cycleEPSG cycle

+6 hour+6 hourEPSG cycleEPSG cycle

Trimmed obstore

Varobs,modelobs

Perts(u,v,p,q,t)

N320L70 T+0

Perts(SST)

Perts(u,v,p,q,t,SST)

ETKF backgroundFieldCalc

VarSCR_UMFileUnit

N512L70 T+0

VAR backgroundVAR background

GDPS(N512L70)GDPS(N512L70)4DVAR4DVAR

GDPS(N512L70)GDPS(N512L70)4DVAR4DVAR

2012. 6.2012. 6.

2013. 7.2013. 7.

Page 5: Current Status and Plans of Ensemble Prediction System at KMA Seung-Woo Lee Numerical Model Development Division Korea Meteorological Administration GIFS-TIGGE

5

Sensitivity to ensemble membersSensitivity to ensemble members

OPER M22 M44

Observations KMA ODB KMA ODB KMA ODB

Data assimilation 4dVar Hybrid Ens. 4dVar

Hybrid Ens. 4dVar

Ensemble members excluding control

23 22 44

Model version UM 7.9 UM 7.9 UM 7.9

Background error Statistical BE 0.8*Statistical_BE + 0.5*Ens_BE

0.8*Statistical_BE + 0.5*Ens_BE

Page 6: Current Status and Plans of Ensemble Prediction System at KMA Seung-Woo Lee Numerical Model Development Division Korea Meteorological Administration GIFS-TIGGE

6

0

1

2

3

4

5

6

7

8

2012

0802

00

2012

0803

00

2012

0804

00

2012

0805

00

2012

0806

00

2012

0807

00

2012

0808

00

2012

0809

00

2012

0810

00

2012

0811

00

2012

0812

00

2012

0813

00

2012

0814

00

2012

0815

00

2012

0816

00

2012

0817

00

2012

0818

00

2012

0819

00

2012

0820

00

2012

0821

00

2012

0822

00

2012

0823

00

2012

0824

00

2012

0825

00

2012

0826

00

2012

0827

00

2012

0828

00

2012

0829

00

2012

0830

00

thet

a

Avg RMS of Perturbations

t_oper

t_m22

t_m44

Stable after 36 hours

Sensitivity to ensemble membersSensitivity to ensemble members

• Test period : 2012. 8. 3. 12Z -2012. 8. 3. 29. 12Z

RMS averaged for all perturbation members and levels

Unstable in model dynamics due to gravity wave drag parameterization.

Page 7: Current Status and Plans of Ensemble Prediction System at KMA Seung-Woo Lee Numerical Model Development Division Korea Meteorological Administration GIFS-TIGGE

7

NH Z500 error against with observation

Sensitivity to ensemble membersSensitivity to ensemble members

• Spread increased significantly in NH and Tropics, while the CRPSS and BSS are not significantly changed.

Page 8: Current Status and Plans of Ensemble Prediction System at KMA Seung-Woo Lee Numerical Model Development Division Korea Meteorological Administration GIFS-TIGGE

8

SH Z500 error against observation

Sensitivity to ensemble membersSensitivity to ensemble members

• Spread decreased significantly only in SH.

• M44 is a little better than M22 until T+144

• Only Spread of both M22 and M44 is

significant at the critical level=0.05

Page 9: Current Status and Plans of Ensemble Prediction System at KMA Seung-Woo Lee Numerical Model Development Division Korea Meteorological Administration GIFS-TIGGE

9

Impact on typhoon 4-day forecast (GDPS)Impact on typhoon 4-day forecast (GDPS)OPER M22

Analysis M44

Page 10: Current Status and Plans of Ensemble Prediction System at KMA Seung-Woo Lee Numerical Model Development Division Korea Meteorological Administration GIFS-TIGGE

10

RUN TIME (minute)

Operation M22 M44

Trim 3 3 3

OPS 6 6 10

ETKF 5 5 10

Reconfiguration 2 2 2

SST 1 1 1

Forecast (10d/9h) 70 70/6 70/6

Operation M22 M44

Trim 200M 200M 200M

OPS 6G 6G 12G

ETKF+SST 20G 20G 36G

Reconfiguration 3.5G 3.5G 3.5G

UM Forecast(10d/9h)

124G 131G/46G 265G/90G

TOTAL(1day) 308G 776G 1,484G

Data size: operation(2 times/day), M22/44(4 times/day)x ERLY/LATE

Considerations for implementationConsiderations for implementation

Page 11: Current Status and Plans of Ensemble Prediction System at KMA Seung-Woo Lee Numerical Model Development Division Korea Meteorological Administration GIFS-TIGGE

11

Sensitivity to cycle strategySensitivity to cycle strategy

ERLY LATE ERLY LATE ERLY LATE ERLY LATE

ERLY ERLY(10d)

ERLY ERLY(10d)

GDAPS

EPSF

06 UTC 12 UTC 18 UTC 00 UTCType 4

ERLY LATE ERLY LATE ERLY LATE ERLY LATE

LATE ERLY(10d)

LATE LATE ERLY(10d)

LATE

GDAPS

EPSG

06 UTC 12 UTC 18 UTC 00 UTCType 3

ERLY LATE ERLY LATE ERLY LATE ERLY LATE

LATE ERLY(10d)

LATE LATE ERLY(10d)

LATE

GDAPS

EPSG

06 UTC 12 UTC 18 UTC 00 UTCType 2

ERLY LATE ERLY LATE ERLY LATE ERLY LATE

ERLY LATE ERLY(10d)

LATE ERLY LATE ERLY(10d)

LATE

GDAPS

EPSG

06 UTC 12 UTC 18 UTC 00 UTCType 1

Page 12: Current Status and Plans of Ensemble Prediction System at KMA Seung-Woo Lee Numerical Model Development Division Korea Meteorological Administration GIFS-TIGGE

12

Number of ingested observationsNumber of ingested observations

• Period: 2012. 6. 26. 00 ~ 2012. 7. 11. 18 UTC

• About 85~90% of satellite data are ingested in the early cycle experiments.

Page 13: Current Status and Plans of Ensemble Prediction System at KMA Seung-Woo Lee Numerical Model Development Division Korea Meteorological Administration GIFS-TIGGE

13

Difference between each variant and 1st variant (Type 1)

RMSE and Spread RMSE and Spread

Page 14: Current Status and Plans of Ensemble Prediction System at KMA Seung-Woo Lee Numerical Model Development Division Korea Meteorological Administration GIFS-TIGGE

14

Relative performancesRelative performances

• Independent early cycle (Type 3 and 4) showed

improved ensemble spread.

• Type 1 for NH, Type 4 for SH, and Type 1 or 3 for

Tropics

• Type 2 reveals poorer performance than other types of

hybrid

1 4 3 3 3 3 2 4

1 4 3 1 4 4 3 3

2 4 1 3 3 4 3 4

Page 15: Current Status and Plans of Ensemble Prediction System at KMA Seung-Woo Lee Numerical Model Development Division Korea Meteorological Administration GIFS-TIGGE

15

0

1

2

3

4

5

6

7

8

RMSE RMSE RMSE RMSE SPREADSPREADSPREADSPREAD

NH SH TR AR NH SH TR AR

ratio

average of the difference ratio against the best for Z500

1안

2안

3안

4안

Type 1

Type 2

Type 3

Type 4

0

1

2

3

4

5

RMSE RMSE RMSE RMSE SPREADSPREADSPREADSPREAD

NH SH TR AR NH SH TR AR

ratio

average of the difference ratio against the best for t850

1안

2안

3안

4안

Type 1

Type 2

Type 3

Type 4

Verification against with observationVerification against with observation

• Hybrid implementation of type 3 showed improved ensemble spread for Northern and Southern Hemisphere.

• Over the tropics and Asian region, type 2 and 4 showed improved performances.

3/1 4 2 1 3 3 2 4 3 1/4 3 1 3 3 2 4

Page 16: Current Status and Plans of Ensemble Prediction System at KMA Seung-Woo Lee Numerical Model Development Division Korea Meteorological Administration GIFS-TIGGE

16

Seamless prediction from medium range to sub seasonal scale• Increased spatial resolution and ensemble members EPSG, which

covers forecast range of medium to sub-seasonal scale of 3~4-weeks.

Data Assimilation• Further optimization of Hybrid Ensemble 4DVAR system (in 2013)

• Introducing of 4D Ensemble-Var (next generation EPSG, in 5 years)

- Aiming at direct ensemble data assimilation with 4dVar

Coupling of ocean model• Implementation of extended EPSG with coupled ocean model (Operation

planned in 2014)

- Plans to evolve EPSG covers one-month period of forecast.

Future plans of KMA EPSsFuture plans of KMA EPSs

Convective scale ensemble prediction system• Developing a convective scale EPS to provide short-range probabilities of

high impact weather over local area (Operation planned in 2015)

Page 17: Current Status and Plans of Ensemble Prediction System at KMA Seung-Woo Lee Numerical Model Development Division Korea Meteorological Administration GIFS-TIGGE

17

SummarySummary

• KMA has been operating and developing a global EPS.

− introducing SST perturbation, hybrid ensemble 4dVar.

−sensitivity test shows a minor improvement in 44-members of hybrid ensemble 4dVar, and a similar effect for each configuration of operating strategies.

• KMA has plan to operate a global high-resolution EPSG, which has forecast lead times from medium-range up to 3-weeks in 2016.

−with the coupling of ocean model and aim at development of one month forecast EPSG.

• Research and development for the convective scale ensemble prediction system are conducted.

− targeting short-range probabilistic forecast of local high impact weather.

Page 18: Current Status and Plans of Ensemble Prediction System at KMA Seung-Woo Lee Numerical Model Development Division Korea Meteorological Administration GIFS-TIGGE