Projection of the Global Climate change with the high-resolution AGCM

-1-

Projection of the Global Climate change with the high-resolution AGCM

based on the RCP Scenarios

Nov. 18, 2013

Jai-Ho Oh 1 , Sumin Woo 1 and Kyoung-Min Lee 2

1

Dept. Env. & Atmos. Sci., Pukyong National Univ., Busan, S. Korea

2

CRAY Korea Inc., Seoul, S. Korea

(jhoh@.pknu.ac.kr)

-2-

• Climate change has caused the large casualties and economical losses in the world and it will continue for centuries. So understanding and predicting the Climate change are so important to respond the adaption policies in the field of national agriculture and water resources that many climate modeling groups are participating the climate research.

• But, users of climate results produced by global climate models with coarse grid-spacing have a lot of dissatisfactions with the insufficient mismatch of spatial scale in particularly area with complex terrain like East Asia region.

• The various downscaling techniques (Statistical & Dynamical downscaling) have been conventionally used to obtain the climate scenarios with local- to regional-scale (10 to 100 km) information from larger-scale models or data analyses.

• Among these methods, the regional downscaling are used in a lot of research activities (CORDEX, ENSEMBLES, NARCCAP, RMIP, etc), but it has a

limitation from lateral boundary conditions which regional climate models have inherently.

Introduction

-3-

• Experiment on dependence of domain size and location of RCMs

Limits of Regional Downscaling Methods

Domain 1 Domain 2 Domain 3

D1 D2 D3

Model WRF community model

Resolution 20 km

Ver. layer 27 layers (model top : 50hPa)

IC/BC GME 40km data

Centered Lat · Lon

lat : 39.5 lon : 125.0

lat : 39.5 lon : 125.0

lat : 45.46 lon : 113.1 No. grid 400X300 330X230 400X300

1. Set of 3 different domains including the East Asia and Korean peninsula 2. Seasonal prediction for 1981-1982 using WRF regional model

Common area

-4-

(a) Mean Precipitation of 3 domains (b) Difference (D1 – 3domains)

• When the domain size and location are changed, it shows the difference in spatial temperature (not shown) and especially precipitation pattern.

-50 -40 -30 -20 -10 10 20 30 40 50 5 20 40 60 80 120 160 200 240

(c) Difference (D2 – 3domains) (d) Difference (D3 – 3domains)

Limits of Regional Downscaling Methods

• Mean and Difference of Precipitation (July, 1982)

-5- 0.5% 0.5%

2.5%

5%

2.5%

5%

-2.54 -1.96 -1.10 0.0

Area of

no significant difference

1.10 1.96 2.54

∙∙ ∙∙∙

∙∙ ∙ ∙ ∙∙

∙∙ ∙ ∙ ∙ ∙∙ ∙ ∙ ∙ ∙ ∙ ∙ ∙∙∙∙∙∙

 : 1%

 : 5%

 : 10%

 : no difference

Area of

significant difference

Area of

significant difference

Limitation of Regional Downscaling

• Compared with the regional downscaling, global downscaling method can avoid the lateral boundary problems.

• So In this study, we suggest the climate simulation using atmospheric global climate model with horizontally high-resolution grid.

• Student’s T-test of Precipitation (1981)

-6-

◦ GME model : Atmospheric GCM used the Operation NWP of German weather Service by DWD

◦ Grid Structure: Icosahedral-Hexagonal grid

◦ Resolution : 40 km mesh size

 368,642 gridpoints/layer

◦ Layer : hybrid (sigma/pressure) layer

◦ Prognostic variables : ps, u, v, t, qv, qc, qi, o3

◦ Time integration: semi-Lagrangian scheme

◦ Convection Scheme : Tiedtke, 1989

• Advantages of GME

- Avoid pole problem, so CFL for advection is not an issue.

- All cells are nearly same size (within about 5% in terms of area).

- Avoids the large amount of global communication

- Data structure well suited to high efficiency on distributed memory parallel computers

Introduction of GME

<Grid generation>

-7-

CORDEX Phase I experiment design

Model Evaluation Framework

Climate Projection Framework

Multiple AOGCMs RCP4.5, RCP8.5 1951-2100 or 1980-2050

Decadal Predictions

1980-2100, 1990-2000, 2005-2035 Multiple regions (initial focus on Africa)

50 km grid spacing

Regional Analysis Regional Databanks AMIP OBS BC

1979-2009

All worldwide regions 40 km grid spacing

RCP4.5, RCP8.5 2006-2100

Decadal Analyses 2010-2100

PKNU CORDEX Experiment design

CMIP present-day BC

1979-2005

-8-

(2) Validation in Asia Monsoon Region

• Spatial Mean Precipitation during JJA, 1998-2008 (unit: mm/day)

• In the comparison with observation, the GME reflects the spatial precipitation well in JJA.

- Asia (India, Bay of Bengal, Central China, Korea, Japan)

• Comparing high-resolution observation (APHRODITE), the

GME shows the better performance as well as the detailed

features of precipitation than 4 GCMs/CMIP5.

-9-

• Annual Cycle in Zonal-averaged precipitation over Asia (75-145 o W) during 1998-2008

• The GME shows the precipitation pattern well from May to Sep during Asia Monsoon Season over the whole Asia region (75- 145

W).

• Although the GME simulated the overestimated precipitation in low latitude region (20-30

N), it captures the temporal and spatial pattern well with observation.

: 2 mm/day GPCP-1dd

(2) Validation in Asia Monsoon Region

-10-

◦ Boundary condition : SST & Sea Ice Concentration - Historical run (1979-2009) : AMIP observation data

- Present-day run (1979-2005) : Present-day based CMIP5 multi-models by IPCC AR5 - Future (2006-2100) : RCP 8.5, RCP 4.5 CMIP5 multi-models by IPCC AR5

◦ CMIP5 multi-models

- CanESM2 (Canada) : 128×64 grid - GISS-E2R (USA) : 144×90 grid - HadGEM2-CC (UK) : 192×145 grid - HadGEM2-ES (UK) : 192×145 grid

Configuration of the Experiment

Sea Surface Temperature (K)

-11-

Future Climate Projection in global

• Air 2m Temperature ( o C)

• Precipitation Change (mm/day)

RCP 8.5 : 0.41

C/decade RCP 4.5 : 0.25

C/decade

RCP 8.5 : 0.004 mm/day

RCP 4.5 : 0.002 mm/day

-12-

[Future climate change in Air 2m Temperature]

-13-

[Future climate change in Mean Precipitation]

-14-

[Future climate change of Northern Sea Ice in 01/01/YYYY 00UTC]

-15-

[Future climate change of Northern Sea Ice in 01/07/YYYY 00UTC]

-16-

Period Scenario Spring Summer Autumn Winter

Near-Future

(2010-39) RCP 8.5 0.94 0.99 1.13 1.12

RCP 4.5 0.83 0.98 0.99 1.01

Mid-Future

(2040-69) RCP 8.5 1.99 2.20 2.33 2.22

RCP 4.5 1.51 1.67 1.90 1.77

End of Future

(2070-99) RCP 8.5 2.97 3.49 3.95 3.63

RCP 4.5 1.94 2.16 2.38 2.04

Air 2m Temperature Anomaly (unit: o C)

- Near-future : 2010-2039, Mid-future : 2040-2069, End of future : 2070-2099

-17-

Period Scenario Spring Summer Autumn Winter

Near-Future

(2010-39) RCP 8.5 1.36 2.44 6.40 4.31

RCP 4.5 3.61 2.71 0.71 4.46

Mid-Future

(2040-69) RCP 8.5 3.46 6.38 9.84 10.82

RCP 4.5 4.35 3.12 8.21 10.63

End of Future

(2070-99) RCP 8.5 6.70 10.52 17.51 14.63

RCP 4.5 5.76 7.51 10.91 11.74

Precipitation (unit: mm/day)

- Near-future : 2010-2039, Mid-future : 2040-2069, End of future : 2070-2099

-18-

Analysis Domains

• The 40-km mesh long-term climate simulation data can cover the analysis of the

detailed regional features as well as the climate change patterns over the global.

-19-

Results of the Experiment

◦ Resolution : 900×451 (40km)/ 21 Layers

◦ Interval of Model output : 3 hourly

◦ Variable : Surface - 2m Air T, Precipitation, 10m U∙V, MSLP, etc. (totally 80 variables) Vertical (21 layers) - Z, T, U, V, PS, QV, QC, QI, O3

◦ Data Processing (including the total data size)

◦ The model output data and all processed data are stored in GSDC (Global Sciences experimental Data Hub Center) /KISTI.

Geodesic grid in 3-h interval

GR IB for ma t (Al l v ar ia b le s)

Regular grid

GRI B f orm at (Al l v ar ia b le s)

Regular grid with daily/month/seaso

nal/annual interval Bi na ry form at (Ana lysi s va ri ab le s)

Analysis

Product Fi g ur e ( jpg , gi f, …) form at

271 TB 190 TB 160 TB 10.5 TB

-20-

1. Standard output data set (variables, frequency etc.)

2. Format of output (most likely follow CMIP5 protocol)

3. Location of ‘online’ RCM storage + mechanism for data access/distribution.

Issues not yet fully resolved

-21-

CORDEX – Current Status

Model Evaluation

Framework Climate Projection Framework

ERA-Interim BC 1989-2007

Multiple AOGCMs RCP4.5, RCP8.5

1951-2100

1981-2010, 2041-2070, 2011-2040

Multiple regions (Initial focus on Africa) 50 km grid spacing

Regional Analysis

Regional Databanks

Three Main Functions of KISTI

Super Computing Center

Supercomputing Management and Operation Supercomputing

High Speed Research Network National Grid Infrastructure

Knowledge

Information Center Developing National Portal Systems for Information Resources

Developing Next Generation Technology in Information Services

Information Analysis Center

Core Technologies Analysis

Core Technologies Feasibility study Foreign Information Trend Analysis

Information Analysis System Development

22

KISTI Super Computing Center Mission

Test-bed Center Value-adding Center Resource Center Technology Center

Help Korea research

communities to be equipped with proper knowledge of super computing

Keep securing and providing world-class super computing systems

Make the best use of

What the center has to create the new values

Validate newly emerging concepts, ideas, tools

and system

23

Global Research Networks

■ GLORIAD

 GLObal RIng Network for Advanced

Applications Development) with 10/40Gbps Optical lambda networking

 Consortium of 11 Nations: Korea, USA, China, Russia, Canada, the Netherlands and 5 Nordic Countries

 Supporting Advanced Application Developments such as HEP, Astronomy, Earth System, Bio-Medical, HDTV etc.

 Funded by MEST (Ministry of Education, Science and Technology) of KOREA

■ KREONET

 Korea Research Environment Open NETwork

 National Science & Research Network of Korea, Funded by Government since 1998

 20Gbps Backbone, 1 ~ 20Gbps Access Networks

24

25

GLORIAD

KREONET

3. High Performance Network

MISSION

Physics 2009 ~ Next ~

ALICE STAR Belle

CDF

Various

Bio-informatics CORDEX Astrophysics

26

Overall Projects

27 Ministry of Education,

Science and Technology (Korean Government)

GSDC

CERN ALICE Tier-1 Test- bed

CDF/FNAL Data Handling

CERN-KISTI-FNAL/BNL Global Pipeline

Asian-Pacific CAF and Training Program

High Quality System

Store Various Science Data

Virtualized Data Farm and Global Sites Establishment

Integrated Monitoring System

Mobile User Portal

Cyber Laboratory

Advanced Cyber

Research and Training Technology

Support Scientist

Community

-28-

GSDC Mission

(CDF/FNAL) (Alice/CERN)

(Belle /KEK)

2PB/2,000CPU 10PB/10,000CPU

(LIGO /LLO) (STAR

/BNL)

Phase 1

(2009~2011)

Phase 2

(2012~2014)

Provide Global Science Data Analysis Environment

National Data Center

Asian-Pacific Hub Center

Expand Supporting Fields: Earth

Environment, Biometrics, Nano-tech, etc.

Cyber Research and Training Environment

28

IMGW2011 at Pyeongchang

-29-

Data Center for CORDEX

Computing and Storage Infrastructure Technology Development

Apply Grid Technology to legacy app.

support Korean Gov.

29

-30-

CORDEX Data Center

-31-

-32-

-33-

GME(20km)

Gyeongsangnam-do

(18UTC 09AUG 2010 ~ 18UTC 11AUG 2010)

Hourly accumulated precipitation(mm)

Gyeongsangnam-do Jeollanam-do

QPM(1km)

QPM: Comparison with Radar Obs.

-34-

Regional Climate Model Evaluation System (RCMES) version 2.1

(http://rcmes.jpl.nasa.gov: Powered by Apache Open Climate Workbench )

Raw Data:

Various sources, formats, Resolutions,

Coverage

RCMED

(Regional Climate Model Evaluation Database) A large scalable database to store data from

variety of sources in a common format

RCMET

(Regional Climate Model Evaluation Tool) A library of codes for extracting data fro m RCMED and model and for calculating

evaluation metrics

Metadata

Data Table Data Table

Data Table Data Table Data Table Data Table Common Format,

Native grid, Efficient architect

ure

MySQL

Extractor for variou s data for

mats TRMM

MODIS

AIRS

CERES

ETC Soil moist

ure

Extract OBS dat

a Extract model d

ata

User input

Regridder

(Put the OBS & model data o n the same time/space grid)

Analyzer

Calculate evaluation metrics

& assessment model input da ta

Visualizer

(Plot the metrics)

URL

Use the r e-gridde d data fo

r user’s own anal yses and

VIS.

Data extractor (Binary or netCDF)

Model data

Other Data Centers

(e.g., ESGF, DAAC, ExArch Netw ork)

Assess.

modelin

g

-35-

Summary

1. The Regional Climate Downscaling community is getting better organized.

2. Probabilistic assessments of regional change are emerging from coordinated ensemble simulations.

3. CORDEX is building on prior experiences to

provide a global framework for assessing,

advancing and utilizing regional-climate

downscaling.

-36-

Proposed CORDEX Data Center may play as a bridge to connect among sub-CORDEX communities.

It is necessary to be approved by CORDEX and sub-

CORDEX Community to recognize it as one of official

CORDEX Data Centers!