Transcript Slide 1

International efforts in Climate Modeling
Projections, Predictions and Downscaling
Coordinated by the World Climate Research Program (WCRP)
CMIP5: The 5th Coupled Model Intercomparison Project
http://cmip-pcmdi.llnl.gov
CORDEX: A Coordinated Regional Downscaling Experiment
http://wcrp.ipsl.jussieu.fr/SF_RCMTerms.html
Colin Jones
Rossby Centre, SMHI
Thanks to: Karl Taylor (PCMDI), Filippo Giorgi (ICTP), Ghassam Asrar (WCRP)
CMIP5 : a framework for climate change modeling for the next 5+ years
 Promotes a standard set of model simulations in order to :
 evaluate how realistic the models are in simulating the recent past
 provide projections of future climate change on two time scales
 understand factors responsible for model differences
Two timescales and two sets of science problems
An important input to IPCC AR5
Taylor et al. 2009, http://cmip-pcmdi.llnl.gov/cmip5/
Near-Term :
(next 1-30 years)
Long-Term :
(1860 to 2100 & beyond)
decadal climate
predictability
evaluation of climate models
(e.g. new satellite data)
ocean initialization
detection & attribution
aerosol impacts
climate change scenarios
regional climate
change (high resol)
& climate extremes
climate sensitivity, radiative
forcing and physical
feedbacks (e.g. clouds)
air quality changes
(aerosols, chemistry)
biogeochemical feedbacks
(e.g. carbon, chemistry)
CMIP5: Centennial Timescales: Earth System Modeling
Aims to improve our ability to simulate all processes that influence
the response of the climate system to increasing greenhouse gases
•
•
•
•
•
•
•
•
•
Cloud Feedbacks
Carbon-Climate Feedback (Ocean and Terrestrial)
Aerosol Feedbacks
Sea-Ice/Snow Feedbacks
Ocean circulation changes
Ice Sheets/Glacier response
Sea-level changes
Ocean acidification and ecosystem response
Permafrost and Methane Release
To better constrain the lower and upper bounds of anthropogenic
climate change requires (an accurate and complete) representation
of complex and interacting process in Earth System Models
CMIP5: centennial projections
Spin-up
• Start from a pre-industrial spin-up run (>500 yrs)
• The 20th century control run includes observed changes in GHG,
aerosol concentrations, volcanoes, and land-use from 1850-2005
• 3 RCP scenarios for the 21th century
1850
RCP8.5
20th century control
RCP4.5
RCP2.6
1900
1950
2000
2050
2100
Possibly extend with more RCP scenarios, 1% CO2 increase, …
First CMIP5 projections results now becoming available
Decadal predictability and climate prediction
Predictability we are familiar with arises from an estimate
of future changes in GHG radiative forcing, and the climate
system response to those changes.
Predictability might also arise from information contained in
the initial state of the system
- “committed warming” due to previous GHG forcing
- natural variability of the system
Tom Delworth GFDL
Assuming we can (i) observe this information (ii) assimilate it
in our models, (iii) the variability has a predictable component
and (iv) our models are good enough to simulate the
subsequent evolution of the climate system
We may be able to provide useful information about the
evolution of the climate system on a ~1-20 year timescale.
Climate Prediction as a mixed initial/boundary value problem
The Atlantic Meridional Oscillation index
A 10-yr moving average of annual North Atlantic SST anomalies
Linked to variability in the Atlantic Meridional Overturning Circulation
crucial to simulate the time evolution of AMOC for decadal prediction
There appears to be some increased skill (in a quantitative sense)
when observations are included in coupled climate model predictions
Results from the UK Met. Office DePreSys integrations
Global annual mean
surface temp (TS)
Global annual
mean ocean heat
content in upper
113m (H)
D. Smith etal. UKMO
CMIP5: decadal prediction experiments
10 yrs
• Start a 10-yr experiment every 5 years
• Initialize from observation-based
re-analysis of atmosphere and ocean
• Extend a few runs to 30 yrs
+20 yrs
10 yrs
10 yrs
10 yrs
10 yrs
10 yrs
10 yrs
10 yrs
10 yrs
+20 yrs
10 yrs
10 yrs
1960
Decadal prediction still a research
topic: Targeted for ~1-20 year
Timescale : climate services ?
• Hindcast simulations to
assess skill & uncertainty
• Test ocean and sea-ice
initialisation techniques
• Interpretation of results not
easy (potential predictability)
+20 yrs
1970
1980
1990
2000
2010
2020
Decadal climate prediction is still in an exploratory stage
It may be that much of the predictive skill for the coming
decade (and beyond) will result from the “forced” GHG
response not the initialized climate state.
Results from “initialized” climate simulations often require
(bias/drift) corrections: Application of these corrections
is not trivial, there is danger of incorrect interpretation.
For non-experts it may be safer & perhaps as informative to
use output of the first few decades of the ‘long-term’ CMIP5
experiments (uncertainty issues will need to be addressed)
In both cases an ensemble approach is an absolute necessity
CMIP5 output will be made available to everyone
Terms of use:
All output available for educational and research use
About half of all output available for unrestricted use.
Extensive documentation will be available describing
the models and the experiment conditions.
Model data accessed via a federated Earth System Grid
led by PCMDI connecting identical accessible/structured
distributed archives across the world.
Actual location of model data invisible to the user
How will users access CMIP5 model output? ESG
Data Providers (modeling groups)
Node 1
Node 3
Node 2
Node 5
Node 4
Data
Archive
Model & expt.
documentation
Copy of
heavilyused output
ESG Gateway
(PCMDI)
DOI
catalog
Users (climate model analysts)
CORDEX
A Coordinated Regional Downscaling Experiment
Sponsored by the World Climate Research Programme
http://wcrp.ipsl.jussieu.fr/SF_RCMTerms.html
Generating an ensemble of high-resolution regional climate
projections for the majority of land regions of the globe,
based on a suitable sample of CMIP5 GCM projections.
Strong orientation towards user needs: impacts and adaptation
General Aims and Plans for CORDEX
Provide an ensemble of coordinated Regional Climate projections for
1950-2100 (core 1980-2050), for most land-regions of the globe
based on CMIP5 simulations following RCPs 4.5, 8.5 and 2.6
Make this data available and useable to users, with a common
diagnostic set and format (following CMIP5) at CORDEX archives
Provide a framework for testing Regional Climate Models and
Downscaling techniques for the recent past and future scenarios.
Foster coordination between downscaling efforts around the world &
encourage local participation in this process esp .developing nations
With CMIP5 provide climate simulation data to support IPCC AR5
and impact-adaptation-vulnerability research on longer timescales
International emphasis on African climate & impacts coming 2 years:
START/WCRP analysis, training & capacity building activity 2011-12
Sampling the sources of uncertainty in
RCD-based Regional climate projections
RCD Configuration
(Multiple models)
Region
AOGCM Configuration
(Multiple AOGCMs)
Uncertainty in
regional climate
projection
Internal variability
(Multiple realizations)
Emission/
Concentration
Scenarios
RCD approach
(Multiple RCD methods)
CORDEX Phase I experiment design
Model Evaluation
Framework
Climate Projection
Framework
Multiple regions (Initial focus on Africa)
50km resolution (higher in some regions, Europe: 10km)
ERA-Interim BC
1989-2008
RCP4.5, RCP8.5
some RCP 2.6 runs
Multiple AOGCMs
Regional Analysis
Regional Databanks
Europe, Korea, S.Africa
Regional Projections 1950-2100
CORDEX DOMAINS
(plus Arctic & Antarctica)
•12 domains with a resolution of 0.44° (approx. 50x50km²)
•Focus on Africa : 11 groups committed to run Africa projections
•High resolution ~0.11°x0.11° for Europe (~6 institutions)
What has been decided in CORDEX
1. 6-hourly 3D model level fields will be saved by CMIP5 GCMs
making climate projections as boundary forcing for RCMs
At least 1 RCP4.5 and 1 RCP8.5 member (1950-2100) per GCM.
Many GCMs will also save an RCP2.6, plus > 1 RCP4.5 member
This data will become available on the CMIP5 data nodes in
May to October 2011
2. The standard resolution is 50km (many groups plan to also run
higher resolution for selected domains, e.g. ~10km Europe ensemble)
50km base resolution to include as many groups as possible
3.Before GCM forced runs for a given region RCMs must be
run with ERA-interim (1989-2008) for the same region
4.An initial (international) focus for climate projections will be
Africa with an aim to provide input to the IPCC AR5 process
Seasonal Mean Precipitation JAS 1998-2008
Seasonal Mean Precipitation Bias: Land GPCC
Annual Cycle spatially averaged precipitation
Annual Cycle spatially averaged precipitation
Annual Cycle spatially averaged precipitation
Annual Cycle of West African Monsoon
5-day mean rainfall averaged between 10W to 10E
Health impact examples: Malaria Incidence over Africa: 2000-2008
Mean annual malaria
Incidence (%) based on
the Liverpool Malaria
Model (LMM) for the
period 2000-2008.
The LMM has been driven
by different observations
(NCEP, ERAINTERIM and
a hybrid run using GPCP
rainfall and ERAINT
temperatures) and one
RCM from the CORDEX
project (SMHI-RCA35,
ERAINT control exp).
The RCM fairly well
reproduces the mean
annual distribution of
malaria incidence with
respect to the GPCPERAINT run.
.
A.Morse
U. Liverpool
Health impact examples: Malaria Prevalence over Africa: 2000-2008
Mean annual malaria
Prevalence (%) based
on the Liverpool
Malaria Model (LMM)
for 2000-2008.
The LMM has been driven
by different observations
(NCEP, ERAINTERIM and
a hybrid run using GPCP
rainfall and ERAINT
temperatures) and one
RCM from the CORDEX
project (SMHI-RCA35,
ERAINT control exp).
The RCM reproduces
well the mean annual
distribution of malaria
prevalence with respect
to the GPCP-ERAINT run
(best estimate).
.
A.Morse
U. Liverpool
Annual Cycle of Central African Rainfall
Monthly mean rainfall averaged between 10E to 25E
Summary
CMIP5 & CORDEX will deliver an unprecedented set of coordinated
Global and Regional climate simulations over the coming ~1-4 years
These data cover both the historical past, near-term predictions
and a range of GHG/land-use scenario forced future projections
CORDEX data will provide 50km (higher in some regions) ensembles
of downscaled regional climate projections for most land regions of
the world for use in impact-adaptation-vulnerability research
CORDEX has developed regionally-specific, locally-led evaluation and
analysis teams, with coupled capacity building and training activities
e.g. Africa CORDEX WCRP/START training and analysis workshops.
Both CMIP5 and CORDEX can provide important input to
the evolving Climate Service sector for regions worldwide