Status of operational NWP system and satellite data utilization at JMA

APSDEU-8 Montreal, Canada, October 10-12, 2007

Masahiro KAZUMORINumerical Prediction Division, Japan Meteorological Agency

1-3-4, Otemachi, Chiyoda-ku, Tokyo 100-8122, JAPANkazumori@met.kishou.go.jp

Courtesy of H. Murata (MSC) and NPD colleagues

Contents

• Current NWP system at JMA• Recent developments• Plans for JFY2007 and after

Contents

• Current NWP system at JMA• Recent developments• Plans for JFY2007 and after

NWP system at JMA

2006.3

72.5

NH Balance Barotropic (381km)

70.4

75.1

NHM (381km L8)

83.3 NH Spectral (T42L12)

70.4 88.3 Global Spectral; GSM (T63L16)

Asia Quasi Geostrophic (304.8km L4) 89.1 GSM (T106L21)

73.10 GSM (T213L30)

Asia Fine Mesh; FLM (152.4km L6) GSM (T213L40)

FLM (127km L10) GSM (TL319L40)

83.3 FLM (127km L12)

88.3 Regional Spectral ASM (75km L16)

RSM (20km L36)

83.3 Very Fine Mesh; VFM (63.5km L11) RSM (20km L40)

84.1 VFM (63.5km L13)

88.3 J SM (40km L19) Hydro. MSM (10km L40)

92.3 J SM (30km L23) 04.9 Non- Hydro. MSM(10kmL40)

as of September 2006

1982.3

Non- Hydro.MSM (5kmL50)

2001.3

1959.3

IBM704

HITAC5020F

HITAC8800

1967.4 1973.8

HITACHISR11000K1

1996.3

05.2

HITACM200H

HITACS810K

HITACHIS3800

HITACHISR8000E1

NH Balance (381km L3)

NH Primitive; NHM (381km L4)

1987.12

8th NWP system

called NAPS-8

Numerical Analysis and Prediction System

Kiyose 2nd BuildingKiyose 1st Building

JMA Headquarters

Supercomputers of JMA

• HITACH SR11000(since Mar. 2006)– 2.1GFlops×16processors×80

nodes×2=21.5TFlops– 1.9GFlops×16processors×50

nodes×1=6.1TFlops– 13.1TB Main memory– 67TB Disk– 2PB Tape

HITACHI SR11000 -210nodes

50nodes 80nodes

80nodes

27.6Tflops

Main usage:50 nodes(sub1 system) : Re-analysis and climate model 80 nodes(sub2 system) : Pre-operational model (20km Global model)80 nodes(sub3 system) : Operational global and meso-scale model

Current Operational NWP Models

• Regional model (RSM)• Horizontal Resolution: 20 km• Updates: 2 times a day

• Global model (GSM)• Horizontal Resolution: 60 km• Updates: 4 times a day

• Mesoscale model(MSM)• Horizontal Resolution: 5 km• Updates: 8 times a day

Current NWP models at NPD/JMA

　 Global Model(GSM)

Regional Model(RSM)

Typhoon Model (TYM)

Mesoscale Model (MSM)

One-week Ensemble

Objectives Medium-range forecast Short-range forecast

Typhoon forecast Disaster reduction One- week

forecast

Forecast domain Global East Asia Typhoon and

its surroundingJapan and its surrounding Global

Grid size /Number of

grids

0.5625 deg640 x 320(TL319)

20 km325 x 257

24 km 271 x 271

5 km 721 x 577

1.125 deg320 x 160(TL159)

Vertical levels / Top

400.4 hPa

4010 hPa

2517.5 hPa

5021,800m

400.4 hPa

Forecast hours(Initial time)

90 hours (00 UTC)216 hours (12 UTC)

36 hours (06, 18 UTC)

51 hours (00, 12UTC)

84 hours (00, 06, 12, 18

UTC)

15 hours(00, 06, 12, 18 UTC)

33 hours(03, 09, 15, 21 UTC)

9 days (12 UTC)

51 members

Analysis 4D-VarInner model: T106

4D-VarInner model: 40km

Interpolated from Global

Analysis

4D-VarHydrostatic 10kmInner model: 20km

Global Analysis with

ensemble perturbations

Contents

• Current NWP system at JMA• Recent developments• Plans for JFY2007 and after

Recent developments

Oct. 2006 BUFR winds of GOES-11/12 and MTSAT-1R started to be used instead of SATOB winds.

Feb. 2007 Assimilation of AP-RARS (Asia-Pacific Regional ATOVS Retransmission Service) data started.

Mar. 2007 Assimilation of GPS-RO (Radio Occultation) data from CHAMP started.

Apr. 2007 Assimilation of NOAA18/AMSU-A, MHS started.

Jun. 2007 Direct assimilation of clear-sky radiances of water vapour channel from MTSAT-1R/IMAGER started.

Aug. 2007 Assimilation of EARS (EUMETSAT Advanced Retransmission Service) ATOVS data started.

BUFR winds of GOES-11/12 and MTSAT-1R

•Data assimilation of BUFR coded AMV data with QI (2006.10)Improved thinningImproved use of QIImproved QCDiscontinue of Obs. Error adjustmentUse of MTSAT-1R hourly AMVs

QIQI

Wind speed （ Jan.2005, 300hPa, Right:Test ， Left:Cntl ）

RMSE of Obs. against Guess

Number of used data

Bias of Obs. against Guess

Kiyose (Tokyo)Seoul

Beijing

Guangzhou

Urumuqi

Crib Point 1Crib Point 2(Melbourne)

Perth

Syowa Station

Network Configuration(as of August 2007)

GTS circuits

Domestic links

HRPT receiving stations

Regional ATOVS Retransmission System (RARS)

Modified by H. Murata

Singapore

Darwin

AP-RARS and EARS data

Available data coverage

AP-RARS (red) and EARS (blue)

Available data coverage

AP-RARS (red) and EARS (blue)データの利用可能な時刻（積算割合）

0%10%20%30%40%50%60%70%80%90%

100%

0:00 1:00 2:00 3:00 4:00 5:00 6:00 7:00 8:00 9:00 10:00 11:00 12:00

データの利用可能な時刻（積算割合）

0%10%20%30%40%50%60%70%80%90%

100%

0:00 1:00 2:00 3:00 4:00 5:00 6:00 7:00 8:00 9:00 10:00 11:00 12:00

GTS (global data)GTS (global data)

EARS ATOVS (GTS route)EARS ATOVS (GTS route)

ATOVS data latency at JMAATOVS data latency at JMA

RARS data is expected to bring positive impacts on early analysis and forecast at JMA.

Cut off time of Meso-scale analysis Cut off time of global early analysis

NOAA18 AMSUA

NOAA15 AMSUA

without AP-RARSwithout AP-RARS

with AP-RARSwith AP-RARS

AP-RARS NWP Benefits

Analysis difference (early - cycle) at 500hPa geopotential height

NOAA18 AMSUA

NOAA16 AMSUA

without EARSwithout EARS

with EARSwith EARS

EARS NWP Benefits

NOAA15 AMSUA

Analysis difference (early - cycle) at 500hPa geopotential height

GPS Radio Occultation data

• Refractivity from CHAMP

N (n 1) 106 c1

P

T c2

e

T 2

106

C1=77.6ｘ 10-6 C2=0.373

n 1 c1

P

T c2

e

T 2

Forward

N : Refractivity

n : Refractive index

e

T

Tc

T

ec

cn

2232

1 12

TＰ－ ２

n

Tc

T

ec

c

e

T ~1

2T

~

~

22

321

Ｐ－ ２

Tangent 　Linear

Adjoint

Low computational costs and effective method.

--- Refractivity assimilation ---

●Distributed data

① 　 Refractivity 　 n1 　　　　　　　　　　　　　　

② 　 Position 　 vector 　 at 　 the 　tangent 　 point

●Procedure of data assimilation 　　　　①　 Calculate the refractivity ( n2 ) at the

tangent point using forward model from the first guess.

② Sum of n2-n1 and distance from the first guess was minimized from first guess using JMA global 4D-Var system.

GPS Radio Occultation data

Data distribution of CHAMP Analysis increment by CHAMP

Assimilation of CHAMP refractivity showed small positive impact on forecast skill.

Implemented in Operational DA system in Mar.2007

Hope to see large impacts from COSMIC data.

NOAA-18/AMSU-A, MHS radiance data

3 AMSU-A (NOAA15,16,Aqua)

4 AMSU-A (NOAA15,16,18,Aqua)

3 AMSU-B (NOAA15,16,17)

4 AMSU-B (NOAA15,16,17,18)

NOAA-18 data bring wide data coverage in the early analysis and better forecasts.

Expanded data coverage by NOAA-18 data in JMA global early analysis

Red ： W NOAA-18Blue: W/O NOAA-18

Red ： W NOAA-18Blue: W/O NOAA-18

N.H. Z500 RMSE (Aug. 2006) N.H. Z500 RMSE (Jan. 2007)

MTSAT-1R CSR• Clear Sky Radiances (CSR) from MTSAT-1R WV channel

– Meteorological Satellite Center (MSC) of JMA produces– ready for dissemination to NWP community. (Current user: ECMWF,CMC)

clear ratio

O-B

mean=-1.7std=2.18

O-B map

10

-10

-5

5

0 00 UTC Sep.14 2006

MTSAT-1R CSR

Verification period: August 2006

Red ： Without CSR

Blue: With CSR

• CSR provide moisture information in middle and upper troposphere.

• Using CSR data together with polar-orbiting satellite data (AMSU-B) bring much moisture information around Japan.

• Assimilation CSR data showed positive impact on typhoon forecast score.

• Operational use started in 7 June 2007.

Typhoon track forecast verification

Performance of GSMNH Z500 (Day 2) SH Z500 (Day 2)

NH Z500 (Day 5) SH Z500 (Day 5)

Performance of GSM

JMA Global Model Typhoon Center Position Error

0

100

200

300

400

500

600

700

800

900

1000

1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006Year

Posi

tion

Erro

rs（k

m）

120mean 12096mean 9672mean 7248mean 4824mean 24

Recent Operational Changes to MSM

• Extended forecast time from 15 hours to 33hours (03, 09, 15, 21 UTC Initial)

• Improvement of surface and boundary layer scheme, radiation scheme, cumulus parameterization scheme

• Use of Non-hydrostatic model in 4D-Var outer loop forecast model(2007.5)

Improvement of heavy rain forecasts.3-hour precipitation forecasts at 15hour forecasts from 21UTC 29 June, 2006 Initial. Old (left), New (center), Radar-AMeDAS observation (right)

New-MSMOld-MSM Observation

Data use in Global Analysis (1/2)

SYNOPBUOYSHIP

AIRCRAFTPAOB

RAOBPROFILER

AMV

Data use in Global Analysis (2/2)

SeaWinds

GPS-ROSSMITMIAMSR-E

AMSU-AAMSU-B

CSR

Data use in Mesoscale Analysis

ATOVS retrievalsATOVS retrievalsAMV, SeaWindsAMV, SeaWindsMicrowave radiometer (TPW, Rain rate)Microwave radiometer (TPW, Rain rate)

SYNOP,SHIP, BUOYSYNOP,SHIP, BUOY RAOB, Wind Profiler, Doppler radar

RAOB, Wind Profiler, Doppler radar

Air craftAir craft

Contents

• Current NWP system at JMA• Recent developments• Plans for JFY2007 and after

NWP models at JMA in the near future(Late 2007-)

　 Global Model(GSM)

Typhoon Ensemble

One-weekEnsemble

Mesoscale Model (MSM)

One-monthEnsemble

ObjectivesShort-range, medium-

range and typhoon forecast

Typhoon track forecast

One-weekforecast Disaster reduction One-month

forecast

Forecast domain Global Global Global Japan and its

surrounding Global

Grid size /Number of

grids

0.1875 deg1920 x 960

(TL959)

0.5625 deg640 x 320(TL319)

0.5625 deg640 x 320(TL319)

5 km721 x 577

1.125 deg320 x 160(TL159)

Vertical levels / Top

600.1 hPa

600.1 hPa

600.1hPa

5021,800m

600.1hpa

Forecast hours(Initial time)

84 hours (00, 06, 18 UTC)

216 hours (12 UTC)

84 hours (00, 06, 12, 18UTC)

9 days(12 UTC)

51 members

15 hours(00, 06, 12, 18 UTC)

33 hours(03, 09, 15, 21 UTC)

Once a week

Analysis 4D-VarInner model: T159

Global Analysis with SV

perturbations

Global Analysis with SV

perturbations

4D-Var based on non-hydrostatic meso-

modelJCDAS

TL959L60 Global Model (20km-GSM)

Psea RMSE(hPa)

0

0.5

1

1.5

2

2.5

3

0 12 24 36 48

GSM20km 00UTCGSM20km 06UTCGSM20km 12UTCGSM20km 18UTCRSM 00UTCRSM 12UTC

Z500 RMSE(m)

0

5

10

15

20

25

0 12 24 36 48

GSM20km 00UTCGSM20km 06UTCGSM20km 12UTCGSM20km 18UTCRSM 00UTCRSM 12UTC

Psea RMSE(hPa)

0

0.5

1

1.5

2

2.5

3

0 12 24 36 48

GSM20km 00UTCGSM20km 06UTCGSM20km 12UTCGSM20km 18UTCRSM 00UTCRSM 12UTC

Z500 RMSE(m)

0

2

4

6

8

10

12

14

16

18

0 12 24 36 48

GSM20km 00UTCGSM20km 06UTCGSM20km 12UTCGSM20km 18UTCRSM 00UTCRSM 12UTC

Comparison of New 20km-GSM and RSM (Regional model)Forecast score against initial

Dec. 2005 and Jan.2006Aug. and Sep. 2004

Performance of the new 20km-GSM is superior to JMA Regional model (RSM 20km resolution).

Parallel run of 20km-GSM started 26 Sep. 2007 and 20km-GSM will be implemented into operational system in Nov. 2007.

Plans for JFY2007 and after

Nov. 2007• Implementation 20km Global model• Typhoon Ensemble • 0.25 deg. SST analysis data for GA• Use of Metop ATOVS data for GA

Mar. 2008• MTSAT-1R hourly AMV for MA• GOES11/12 and Meteosat7,9 CSR data for GA• ASCAT: need level2 ocean surface wind data• AIRS radiance data for GA• SSMIS radiance data for GA• GRACE GPS-RO data for GA• Improved assimilation of scatterometer Ocean Vector Winds (ambiguity winds)• Non-hydrostatic based 4D-Var for MA

After• New satellite data (IASI) and cloud/rain-affected radiances data• Radiance assimilation in MA

Pre-operational run started in 26 Sep. 2007