Download - In-Sik Kang Climate and Environment System Research Center Seoul National University, Korea
In-Sik KangClimate and Environment System Research Center
Seoul National University, Korea
Chung-Kyu Park and Dong-Il LeeKorea Meteorological Administration
• Current Status of Global Climate Models• Multi-model ensemble prediction system• Computation and network environments • SNU-NASA multi-model prediction• Cyber Institute for Pacific-Asian Climate System
Multi-model Climate Prediction
Numerical Simulation of Earth Climate Atmospheric General Circulation Models (AGCMs)• Widely-used tools for Numerical Reproduction of Weather and Climate• Adapted to Seasonal Prediction Problem with the advance of High-performance Super Computing• Dynamic Equation Set
• Numerical Representation
• Super Computing
Run-type
Ensemble Members Integration Period Initial Conditions Boundary Conditions
SMIP 10May1979~Nov1999
7 months integrations for every year
00Z~12Z of 26Apr~30Apr for
every year
OISST(NCEP) and AMIP II climatological cycl
e Sea ice.
SNU AGCM Modeling and Climate Prediction
Model Resolution Dynamics Physics
SNUAGCM
T63L21 hybrid
vertical coordinate
Spectral model using semi-implicit
method
• 2-stream k-distribution radiation scheme (Nakajima and Tanaka 1986)• Simplified Arakawa-Schubert cumulus convection scheme based on RAS scheme (Moorthi and Suarez 1992)• Orographic gravity-wave drag (McFarlane 1987)• Dry adiabatic adjustment• Bona’s land surface model (Bonan 1996)• Mon-local PBL/vertical diffusion (Holtslag and Boville 1993)• Diffusion-type shallow convection• Modified CCM3 slab ocean/sea-ice.model
Experimental design for Seasonal Ensemble Prediction
SNU (Seoul National University ) AGCM description
Current Status of Global Climate ModelsSNUGCM Model Climatology (Summer)
(a) ObservationRainfall
(c) ObservationSea Level Pressure
(b) Model (d) Model
Climatology of Summer Rainfall (Various Models)
Super-Ensemble Prediction
- Superiority of a multi-model ensemble prediction compared to any of single prediction
- Applicability of superensemble technique to climate prediction
Training ForecastConventional Superensemble SVD
SVD Mean RMSE
Conventional Superensemble
Simple Ensemble
Superensemble Precipitation RMSE (Global)
Yun, Stefanovar and Krishnamurti (2002)
)(1
,
n
iitiit FFaOS
Asia-Pacific Climate Network (APCN)
To develop and maintain an infrastructure of a well-validated multi-model ensemble system (MMES) to produce the seasonal climate Prediction for Asian Pacific Economic Cooperation (APEC) member countries and to use it as an economic tool to effectively manage future weather and climate risks
The APCN-MMES will produce real-time seasonal forecasts and disseminated the forecast products to member countries.
Model Institute Resolution Experiment Type
NCEP NCEP T63L17 SMIP (10 member)
GDAPS KMA T106L21 SMIP (10 member)
GCPS SNU T63L21 SMIP (10 member)
NSIPP NASA 2ox2.5o L43 AMIP (9 member)
CWB Taiwan T42L18 AMIP (1 member)
Participated Model
Target of prediction
: Summer (JJA) mean precipitation
APCN Multi-Model Climate Prediction System
Dynamical
prediction
Dynamical
prediction
Dynamical
prediction
Dynamical
Prediction
Dynamical
Prediction
Correctedprediction
Corrected prediction
Corrected prediction
Corrected prediction
Corrected prediction
Statistical Downscaling (Post-processing)
Specio-Ensemble prediction
Multi Model Ensemble prediction
Multi Model Ensemble procedure
Statistical Prediction
Multi-Model Dynamical-Statistical Ensemble prediction
Conventional Multi-Model Ensemble prediction
Prediction skill – before downscaling / JJA Precipitation
Prediction skill – after downscaling
Summer Mean Precipitation (30S~60N)
Model Comp.
Superensemblewith MLRM
Superensemblewith SVD
(b) RMSE
(a) Pattern Correlation
Specio-ensemble prediction
Comparison of prediction skill for individual summer
Computational Resources (based on NEC/SX4)
Required CPU Time (Best guess, single node)
- Seasonal Hindcast experiments• AGCM 1 month integration (user time) = 1.6 hours • 7 months forecasts for 1 member = 1.6 hrs/month * 7 months = 11.2 hours• 10 member ensemble integrations = 11.2 hours/member * 10 member = 112 hours • 21 years hindcasts = 112 hours/1year * 21 years = 2352 hours• 4 seasons * 2352 hours = 9408 hours
Required Disks and Network Exchange - AGCM Integration 1 month = 0.7 GB) * 7 months * 10 members *21 years = ~ 1.03 TB * 4 seasons = ~ 4.12 TB
9,408 hours (~ 13 months ) CPU Time Needed
4.12 TB Disks Needed
Needed For One Prediction Center
Development of the SNU-NASA Multi-Model Ensemble Prediction System
SNU
NCEP
KISTI
KMAGPCPTRMM
NASA
Validation data
Forecast output
Model input
Model input
Forecast output
Forecast output
Forecast output
Model input
Tokyo. U
NCEP
COLA KMA
SNU NASA
Supported by National Computerization Agency
Network structure between SNU and NASA
CES
45Mbps
SeoulBackbone
node
USA
WebServer
DataServer
AnalysisServer
KMA
TaejonBackbone
nodeKISTI
Super Computer
Super Computer
NCEP NASA
AnalysisServer
DBServer
2.5Gbps
155Mbps
155Mbps
SNU
DBServer
FTPServer
Direct Conn.
KOREN
StarTap(APII-Test bed)
SNU Network
상용인터넷
국내의 KISTI, KMA, 국외의 NASA, NCEP ( 미국 ) 과
국제공동 기후 네트웍 확장
1Gbps
Network Traffic
• 2003. 01. 01 ~ 2003. 06. 30• Traffic amount : 112.13 TB
• Average Input : 5.28 Mbits• Average Output : 1.93 Mbits
2003 1 1 ~ 2003 6 30 5 년 월 일 년 월 일 분 평균네트웍 트래픽
0
10
20
30
40
50
60
70
80
2003- 01- 01 2003- 01- 21 2003-02-10 2003-03-02 2003-03-22 2003- 04- 11 2003-05-01 2003-05-21 2003-06-10 2003-06- 30날짜
(Mbi
t/sec
)트
래픽
Input output
초고속 선도망 및 APII-Testbed 활용도
0102030405060708090
100
데이터서버 모델서버 SNU-KISTI SNU-KMA SNU-NASA SNU-NCEP
네트
웍 속
도(M
bps) 개선전(2002) 개선후(2002말) 개선후(2003.6)
96 Mbps (학내 )60 Mbps (국내 )1.2 Mbps (미국 )
네트웍 Speed 개선
8 Mbps (학내 )5 Mbps (국내 )0.9 Mbps (미국 )
학내는 네트웍 대역폭 확대와 경로단축으로 큰 개선 효과 국외는 네트웍 경로문제로 속도개선이 크게 향상이 안됨
Network Speed after some attempt
SNU-NASA Joint Forecast for Washington D.C.Issued at Oct2002
• Site URL- http://147.46.56.215/cps/index.html• Provide real-time prediction for global and regional domains
Main PageGlobal Prediction
Regional Prediction
Web-based Operational Display System
⊙ Cyber Institute for Pacific-Asian Climate System Network
⊙CIPACS Main Page
About CIPACS Members Online Journal Forum Data News Links
Member`s Institute
The End