Center Report from CMA Short, Medium-range NWP Xueshun Shen Center for Numerical Weather Prediction China Meteorological Administration WGNE, Tokyo, Japan,

Center Report from CMA

Short, Medium-range NWP

Xueshun ShenCenter for Numerical Weather Prediction

China Meteorological Administration

WGNE, Tokyo, Japan, 18-22 Oct. 2010

Outline

• Short & Medium-range NWP systems– Current status of operation system– New implementation– GRAPES_GFS: improvement toward

operation– Research activities

• New organization for NWP development and operation

CMA headquarter decided:

Freeze the current operational global NWP model: TL639L60Put most of the resources to improve and develop the GRAPES system

Big transition of NWP-related policy

Last WGNE presentation

Global SpectralModel

(TL639L60)

Meso Scale Model

(GRAPES_Meso)

10dayEnsemble(TL213L31)

Typhoon deterministic &

Ensemble forecast

Forecast range

Short- and Medium-range

forecast

Rainfall forecastShort-range

forecast10day forecast Typhoon forecast

Forecast domain Global East Asia

(8340km x 5480km) Global

Horizontal resolution TL639(0.28125 deg) 15km T213(0.5625 deg)

Verticallevels / Top

600.1 hPa

3110hPa

3110 hPa

ForecastHours

(Initial time)

240 hours(00 、 12 UTC)

72 hours(00, 12UTC)

240 hours(00 、 12 UTC)15 members

120 hours(00, 06, 12, 18

UTC)120 hours

(00 、 12 UTC)15 members

Initial Condition

Global Analysis(NCEP GSI)

GRAPES_3DVAR

NCEP SSI + Vortex relocation and Intensity adjustment

with ensemble perturbationsPerturbations are produced by

Breeding-method

Current NWP Operational models in CMA

1990-2010 年 NMC全球数值预报模式北半球500hPa高度场距平相关系数

T 500hPa系列模式北半球 高度预报月平均距平相关系数演变图

0

0. 2

0. 4

0. 6

0. 8

1

1. 2T42

9001

9007

9101

9107

9201

9207

9301

9307

9401

9407

9501

T63 9

507960

1960

7970

1T10

6 9707

9801

9807

9901

9907

00'01

00'07

01'01

01'07

02'01

02'07

T213

301307 401 407 501 507 601 607 701 707 801

T639

807901 907 1001

时间

距平相关

系数

48h

96h

144h

Evolution of ACC of 500 hPa Z

T42 T63 T106 T213 T639

20021990 2008

forecast verification 200909-201008 12UTCgeopotential 500hPa

Correlation coefficent of forecast anomalyNH Extratropics Lat 20.0 to 90.0 Lon -180.0 to 180.0

0.0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1.0

1 2 3 4 5 6 7 8 9 10

Forecast days

AC

C

T639JAPANECMWF

Performance of T213 medium-range ensemble forecastZ500 ACC (average for 20090901-20100831 :NH)

0.0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1.0

1 2 3 4 5 6 7 8 9 10Forecast days

AC

C

T213 Ensemble meanT213 Cntl. (0.5625°)

T639 (0.28125°)

0

50

100

150

200

250

300

350

400

450

JMA CMA JMA CMA JMA CMA EC

2007 2008 2009

Mea

n T

rack

err

ors(

km)

24h48h72h

TC Mean Track Errors from JMA, CMA and EC global models

GRAPES

WRF

Performance of operational GRAPES_Meso

New Implementationfor pre-operational test

GRAPES_MESO V3.0 with 4DVARGRAPES_MESO V3.0 with 4DVAR

Model:Model: GRAPES_MESO V3.0GRAPES_MESO V3.0Resolution:Resolution: 15 km (502x330), 31 levels 15 km (502x330), 31 levelsTime Step:Time Step: 300 seconds 300 seconds

Analysis System: GRAPES-4DVARAnalysis System: GRAPES-4DVAROuter loop resolution:Outer loop resolution: The same The same

resolution as the modelresolution as the modelInner loop resolution: Inner loop resolution: 45 km (167x111), 31 45 km (167x111), 31

levelslevelsPhysics process:Physics process: LSP; MRF PBL; CUDU LSP; MRF PBL; CUDU

convectionconvectionOuter loop:Outer loop: 1 iteration 1 iterationObs: TEMP, SYNOP, AIREP, SHIPSObs: TEMP, SYNOP, AIREP, SHIPSAssimilation Window:Assimilation Window: [-3, 0] [-3, 0]Analysis Time:Analysis Time: 00UTC and 12UTC 00UTC and 12UTCBackground Fields:Background Fields: T TLL639L60 12-hours 639L60 12-hours

forecastforecastForecast Range:Forecast Range: 48 hours 48 hours

ObservatioObservation Datan Data

Data Data ProcessingProcessing

GRAPES-4DVARGRAPES-4DVAR

GRAPES-ModelGRAPES-Model

BackgroundBackground

Analysis Analysis FieldsFields

ForecasForecastt

TTLL639L60 639L60

ForecastsForecasts

SiSi

Since Aug.2010

1-month averaged Ts score of 24 hour precipitation 1-month averaged Ts score of 24 hour precipitation forecast over whole Chinaforecast over whole China

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

1 2 3 4

3DVAR

4DVAR

LightLight ModerateModerate HeavyHeavy TorrentialTorrential

48-hour accumulated precipitation48-hour accumulated precipitation00Z03JUL2009-00Z04JUL200900Z03JUL2009-00Z04JUL2009

Initial time: 00Z02JUL2009Initial time: 00Z02JUL2009

Obs CTRL

3DVAR 4DVAR

dQv and w difference between 3DVAR and 4DVARdQv and w difference between 3DVAR and 4DVAR

Setup of GRAPES global forecast system since Jul.

2007• GRAPES_GFS1.0 : medium-range global forecast

– GRAPES_Global 50km L36 with model top at 10 hPa

– GRAPES_3DVAR at 1.125 degree (global version)

– 6-hourly cycle

– 240 hour forecast (12UTC)

– Assimilated Obs.

GTS conventional data

NOAA15 、 16 、 17

METEOSAT-9 & MTSAT AMV

MODIS polar AMV

Presented in last WGNE

GRAPES_GFS 1.0global model

• SISL dynamical core with mass fixer• Physics

– Radiation: RRTMG LW(V4.71)/SW(V3.61)– Cumulus: Simplified Arakawa Schubert with modified

entrainment and detrainment rates– Grid-scale precipitation: WSM-6– Cloud: Xu & Randall diagnostic cloud– Land surface: CoLM – PBL: Modified Hong & Pan nonlocal PBL

– Gravity wave drag: McFarlane 1987

GRAPES_GFS 1.0global 3DVAR

• Incremental analysis• Digital filter• Recalculated background error covariance –

NMC method • 1.125x1.125 resolution, 17 standard pressure

levels• Bias correction scheme of satellite radiances

based on simple linear regression (Harris and Kelly,2001): (1) 1000-300 hPa thickness, (2)200-50 hPa thickness.

Efforts in improving the forecast skill of GRAPES_GFS

-toward operation-• Improve accuracy of initial values: data assimilation

– ATOVS(NOAA-18,19,METOP,FY3)– GPS Reflectivity (COSMIC)– AIRS– IASI

• Improve model performance– Improve the accuracy of finite difference scheme, especially, for PGF

calculation– Hybrid vertical coordinate: from terrain-following to terrain-following & Z– Tuning of physical processes

• Radiation: RRTMG• Land surface: SLAB to CoLM• GWD• SSO replaces the effective roughness length• Cumulus scheme tuning• Radiative energy budget (cloud-radiation)

– Improve the forecast of synoptic evolution and accuracy of local weather elements, particularly those which have large impact on East Asian weather

CAMS/CMA

Time series of Innvoationand Residual: Height(Sonde)N.H.

500hPa

100hPa

850hPa

CAMS/CMA

Using EC analysis as Reference(NCEP,GRAPES)500hPa

old

new

Zonal mean temp. biasJJA (3d fcst.)

By introducing the new radiationScheme: RRTMG

New cloud cover parameterization

New cloud water path parameterization

Radiative effect of fractional cloud

Radiative effect of cloud inhomogeneity

Cloud-radiation interaction Improvement

Original scheme: binary cloudcloud=1 , when QC+QI >1.0e-6cloud=0 , when QC+QI <1.0e-6effective cloud drop radius: liquid 10µm 、 ice 80µm

New scheme Liang and Wang （ 1995）Cloud cover: Combine Slingo and Slingo （ 1991 ） and Kiehl et al. （ 1994 ） ; 4 cloud genus : convective cloud, anvil cirrus, inversion stratus and stratiform cloud. Fractional cloud cover and vertical cloud overlapping are consideredeffective cloud drop radius: liquid cloud: Savijarvi 1997 ice cloud: Kiehl et al. 1996

Cloud cover parameterization

Cloud cover compared with ISCCP satellite data

Zonal mean total (TCC), high (HCC), middle (MCC) and low (LCC) cloud cover (%) of the ISCCP data (dashed) and the 5th day forecast by GRAPES using the ORG (thin solid) and NEW (thick solid) cloud scheme.

Using ISCCP Simulator

CERES ISCCP ORG NEW

Surface radiation balance (W m-2)

Upwelling LW 400 405 404 408

Downwelling LW 353 357 334 365

Net LW -47 -48 -71 -44

Upwelling LW CRF -- 2 0 1

Downwelling LW CRF -- 30 16 32

Net LW CRF 31 28 15 31

Upwelling SW 22 17 25 21

Downwelling SW 184 177 231 182

Net SW 162 160 206 161

Upwelling SW CRF -- -5 -3 -7

Downwelling SW CRF -- -57 -27 -71

Net SW CRF -42 -52 -24 -64

Comparison of surface radiation budget

Errors reduce 14-47 1-8

Improvement on radiation budget

TOA radiation balance (W m-2)

OLR 244 240 278 233

Net LW CRF 27 27 3 35

Upwelling SW 96 100 66 106

Net SW 235 231 265 225

Net SW CRF -48 -50 -21 -62

CERES ISCCP ORG NEW

Comparison of TOA radiation budget

Errors reduce from 29-34 3-8

Comparison of net flux

Statistical verification

Low-level southerly bias

GRAPES 3-day forecast

To mitigate southerly bias:• Introduce mountain blocking effect in GWD parameterization• Introduce small scale orography-induced form drag

24/48hr V-wind difference (20090701-0720)

By introducing the mountain blocking effect in GWD parameterization

Partly alleviated the low-levelsoutherly bias.

0 40 80 120 160 200

t(h)

-1

0

1

2

3

bia

s/r

mse(m

/s)

Impact of SSO on low-level wind prediction

Partly alleviated the low-levelsoutherly bias.

New version of GRAPES_GFS

• GRAPES_GFS1.2.0 : medium-range global forecast– GRAPES_Global 50km L36 with model top at 10 hPa

– GRAPES_3DVAR at 1.125 degree

– 6-hourly cycle

– 240 hour forecast (00,12UTC)

– Assimilated Obs.• GTS conventional data• NOAA15 、 16 、 17 、 18 、 19• METOP-2• AIRS• FY-3 radiance• METEOSAT-9 & MTSAT AMV• MODIS polar AMV• COSMIC Refraction

More satellite data assimilated

Improved model performance

forecast verification 200906-200908 12UTCgeopotential 500hPa

Correlation coefficent of forecast anomalySH Extratropics Lat -20.0 to -90.0 Lon -180.0 to 180.0

0.0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1.0

1 2 3 4 5 6 7 8

Forecast days

AC

C

T639

GRAPES

forecast verification 200906-200908 12UTCgeopotential 500hPa

Correlation coefficent of forecast anomalyNH Extratropics Lat 20.0 to 90.0 Lon -180.0 to 180.0

0.0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1.0

1 2 3 4 5 6 7 8

Forecast days

AC

C

T639

GRAPES

grapes

operation

N. Hemsiphere

S. Hemisphere

ACC 500hPa ZJJA 2009

Near future upgrade activities

• Global GRAPES_3DVAR– Arakawa-A & pressure level to Model grid space analysis– RTTOV: RTTOV7->RTTOV93– VarBC– More satellite data: IASI, GRAS etc.

• GRAPES global model– Hybrid vertical coordinate– Increase the vertical resolution– Conservative scalar SL advection: CSLR– Improve SL numerics

• SETTLS: Stable extrapolating two-time-level semi-Lagrangian scheme

– Continuous improvement of model physics

Research Activities

• More satellite data• Cloud microphysics parameterization• Global GRAPES-4DVAR• Yin-Yang GRAPES

– To avoid polar singularity problem– More homogeneous grid size

Progress of GRAPES Yin-Yang gridThe Helmholtz equation of GRAPES in the Yin-Yang overset grid are solved.The transplant of the whole GRAPES dynamical core is finished. However,some bugs exist and it need to be debuged in the next step.

2 2 H Helmholtz equation:

• Finish the coding of tangent linear & adjoint model

• Finish the accuracy check

Development of Global GRAPES_4DVAR

a F(a)10-1 0.999333329810-2 1.002537421210-3 1.134534128310-4 1.430711269510-5 1.000000206510-6 1.000000095910-7 0.999999987810-8 1.000001025810-9 1.000015335910-10 1.000092107110-11 1.000736213010-12 1.0499615795

)(M

)()()(

Xa

XMXaXMaF

1)(lim0

aFa

wleft = 362468.822258871398

wright = 362468.822258874832

XXXX T )),(M(M)(M)(M ，Adjoint code check

Office

CMA Numerical Prediction Center

Development Branch

动力过程组

物理过程组

区域模式组

诊断与图形处理组

并行计算组

系统集成与测试组

观测资料预处理、质量控制

资料同化组 集合预报组

Operation Branch

台风预报组

系统中试与运行 版本管理

与信息技术

模式及应用检验

后处理与产品开发

Staff ： 73