2007 Overview - Ensembles

advertisement
ENSEMBLES
Overview, progress, plans
EMS 2007, El Escorial, October 2007
Chris Hewitt, Met Office Hadley Centre
Project Office can be contacted on ensemblesfp6@metoffice.gov.uk
Web site is http://www.ensembles-eu.org
Outline
 Overview of the Ensembles project
 Progress
 Plans
© Crown copyright 2007
Page 2
Motivation
Predictions of natural climate variability on seasonal to
decadal to centennial timescales, and the human impact on
climate are inherently probabilistic
due to uncertainties in:
 initial conditions
 representation of key processes within models
 climatic forcing factors
Reliable seasonal to decadal forecasts, and estimates of
climatic risk can only be made through ensemble integrations
of Earth-System Models in which these uncertainties are
explicitly incorporated.
The ENSEMBLES project will provide these probabilistic
estimates.
© Crown copyright 2007
Page 3
Motivation: Recent Status of Climate
Change Prediction
We can produce a small number of different predictions with little idea of
how reliable they might be
© Crown copyright 2007
Page 4
Probabilistic Climate Predictions
required position
current position
Probability
Probability
100%
0%
20%
40%
60%
2080s SE England winter rainfall
© Crown copyright 2007
0%
20%
40%
60%
2080s SE England winter rainfall
Page 5
Sources of uncertainty
Effects of natural variability
Future emission scenarios
Modelling of
Earth system
processes
© Crown copyright 2007
Page 6
Climate Prediction Modelling
From Murphy et al, Nature 2004
© Crown copyright 2007
Page 7
Motivation (continued) Seasonal forecast case
study: Winter 1997 El-Niño, Precipitation
DEMETER multi-model
Flooding: East Africa
Uruguay / Argentina
Verification
Rainforest fires: Amazonia, Indonesia
Storm damage: California
Tornados: Florida
courtesy
Francisco Doblas-Reyes: verification of a multi-model system
©Slide
Crown
copyrightof2007
Page 8
Better performance of multi-model for seasonal fc
Test of individual and multi-model probabilistic
skill (area-averaged ROC) over 8 regions, 4
seasons, 4 events, 2 lead times
ECMWF
DEMETER
multi-model
16%
UKMO
MPI CNRM
13%
56%
9%
courtesy
Francisco Doblas-Reyes
©Slide
Crown
copyrightof2007
6%
Page 9
Impact of ensemble size on seasonal fc
Ranked probability skill score for tropical (region in blue in the inset)
precipitation as a function of the ensemble size for a multi-model (red) and a
single-model (blue) with the same ensemble size. The multi-model has
superior performance than the single model for every ensemble size.
courtesy
Francisco Doblas-Reyes
©Slide
Crown
copyrightof2007
Page 10
Motivation: Ensemble Climate Prediction
 Run ensembles of different climate models to sample
uncertainties
 Measure variations in reliability between models using
hindcasts
 Produce probabilistic predictions of climate change
 Do this for seasonal to decadal and longer timescales,
and global, regional and local spatial scales, for use in a
wide range of applications

ENSEMBLES
© Crown copyright 2007
Page 11
The ENSEMBLES Project
 5-year Integrated Project supported by EC FP6 funding
coordinated by Met Office Hadley Centre
 67 partners: EU, Switzerland, Australia, US
we welcome requests from new groups to participate on an unfunded
basis – currently 19 such groups worldwide affiliated to the project
 Builds upon EC FP5 projects
e.g. DEMETER, MICE, PRUDENCE, STARDEX
 Integrates a wide range of research communities
 Work carried out in Ten Research Themes
© Crown copyright 2007
Page 12
Strategic Objectives
1. Develop an ensemble prediction system based on global
and regional Earth System models, validated against
observations and analyses, to produce for the first time, a
probabalistic estimate of uncertainty in future climate at the
seasonal, decadal and longer timescales
2. Quantify and reduce uncertainty in the representation of
physical, chemical, biological and human-related feedbacks in
the Earth System
3. Exploit the results by linking the outputs to a range of
applications, including agriculture, health, food security,
energy, water resources, insurance
We are aiming to increase availability of scientific knowledge and provision
of relevant information related to the impacts of climate change
© Crown copyright 2007
Page 13
Research Themes (RTs) summary
System
development
and assembly
© Crown copyright 2007
Model ‘engine’:
hindcasts, climate
integrations
Understanding,
evaluation
Impacts,
Scenarios and
policy
Page 14
Progress: GCM seasonal to decadal
3 different s2d forecast systems to estimate model uncertainty:
 Multi-model (9) system, installed at ECMWF built from EUROSIP
operational activities and DEMETER experience
 Perturbed parameter system, built from the decadal prediction system at
the Met Office
 Stochastic physics system, from the CASBS system developed for
medium-range forecasting at ECMWF
Design of a set of common experiments to determine the
benefits of each approach
Ensemble of ocean analyses available for initialisation
Improved ocean data assimilation systems for s2d prediction
Publications demonstrating additional skill in annual-decadal projections
by initialising the ocean component of CGMS
© Crown copyright 2007
Page 15
Progress: GCM centennial
Multi-model ACC simulations (contributed to IPCC 4AR)
 Conducted historical runs (1860-2000)
 and scenario runs (IPCC A1B, A2, B1)
 including Hadley Centre perturbed
parameter runs
© Crown copyright 2007
Page 16
Progress: regional scenarios
 Defined RCM domain
 Central model data archive available at DMI
 Contains 14 RCM hindcasts (1961-2000 driven by ERA40) at 50km and 25km
 Matrix of driving GCMs/RCMs devised
 Transient scenario runs underway
 25km RCMs, 1950-2050/2100 driven by GCMs, A1B scenario from 2001
 Regional scenario web portal launched in June 2007
DRAFT – TO BE FINALISED
RCM’s
GCM’s
METOMPIMET CNRM
HC
DMI
METO- 1950HC
2100
MPIME 1950T
2100
ETH
KNMI
ICTP
19502050
19502100
19502050
19502050
SMHI
UCLM
19502050
19502050
19502050
C4I
GKSS
MetNo
CHMI
19502050
19502050
FUB
IPSL
CNRM
NERSC
© Crown copyright 2007
0.22º (25km) grid mesh (courtesy of Burkhardt Rockel)
19502050
19502050
19502050
19502050
19502050
19502050
Page 17
Progress (continued)
 Scientific analyses (e.g. cloud feedbacks, carbon, sea-ice, …)
 Coordinated experiments (web), especially for land-sea
warming contrasts
 Linking impact models to probabilistic projections
 Publicly available Climate Explorer http://climexp.knmi.nl/
further developed as integrated diagnostic tool
 Producing daily gridded datasets for Europe with uncertainty
estimates. Particularly of use for evaluating extremes
OLD (ECA daily dataset)
© Crown copyright 2007
NEW
Page 18
Progress (continued)
New emissions scenario developed
 A1B baseline, stabilise towards
475ppmv CO2eq
 provides information towards EU goal
of limiting warming to less than 2°C
above pre-industrial levels
 Uses proposed IPCC “AR5” design
 ESMs including the carbon cycle will be driven by GHG concentrations,
rather than emissions. Carbon fluxes give implied emissions
 Will inform details of AR5 design and how to exploit the runs
© Crown copyright 2007
Page 19
Pre AR4 used forward approach
AR5 likely to use reverse approach
Forward approach: start with socio-economic variables
Socio-economic variables
Emissions
Concentrations
Surface
temperature
Reverse approach: start with stabilization scenario concentrations
Socio-economic variables
mitigation costs
Emissions

implied emissions 
Concentrations
Surface
temperature
concentrations  sensitivity  impacts
Requires interpolating and scaling
© Crown copyright 2007
Page 20
Progress (continued)

West Africa to be the non-European RCM domain

Publications of ENSEMBLES results appearing in journals.
ENSEMBLES Technical Reports series created

Expanding the “affiliated partners” (19 currently)
FAO, Rene Gommes, RT5&6
Univ. Zurich, Nadine Salzmann, RT6
Nat. Univ. of Ireland, Kieran Hickey, RT5
FRGCG, Michio Kawamiya, RT1&2A
Nat. Inst. Earth Sciences, Seita Emori, RT2A
University of Copenhagen, Eigil Kaas, RT2A
Nat.Acad.Sci. Ukraine, Alexander Palonski, RT6
OURANOS group, Daniel Caya, RT2B
Climate Analysis Group, Philippe Gachon, RT2B
IBIMET Institute, Massimiliano Pasqui, RT1, RT2A
© Crown copyright 2007
WHO, Bettine Menne, RT5
ESSC, John Christy, RT5
NCAR, Guy Brasseur, Jerry Meehl, Linda Mearns, RT2A&2B
CCSR (Uni Tokyo), Masahide Kimoto, RT4
SINTEF Energy Research, Atle Harby, RT6
University of Exeter, David Stephenson, RT4
University of Newcastle, Hayley Fowler
Canadian Reg. Clim. Modelling and Diag. Network (CRCMD), Colin Jones
Proudman Oceanographic Laboratory, Roger Proctor, RT6
Page 21
Plans for year 4 of the project (starts Sep 07)
 Start “Stream 2” simulations:
 s2d hindcasts 1960 onwards (previously 1991-2001)
 1860-2000 simulations using updated models
 21st Century scenarios using updated models and E1 (previously A1B,
A2, B1)
 RCM A1B Europe@25km 1950-2050/2100
 Plan RCM simulations for West African domain
 Develop databases
 s2d @ ECMWF building on DEMETER database
 RCM @ DMI building on PRUDENCE database
 GCM @ MPIMET building on IPCC WCDC activities
© Crown copyright 2007
Page 22
Plans for year 4 of the project (starts Sep 07)
 Develop impacts models (e.g. crops, water resources, energy)
 Develop statistical downscaling tools
 Improved estimates for changes in extreme events
 Workshops
 General Assembly, Prague, 12-16Nov07
 Socio-Economic Drivers of Climatic Change, Paris, Dec07
 Climate Change, Impacts and Adaptation in the Mediterranean, Greece, late 2008
 others to be announced
© Crown copyright 2007
Page 23
Concluding remarks – innovative work
 Brings together largely separate communities and integrates worldleading European research:
s2d, anthropogenic climate change, global modellers,
regional modellers (dynamical and statistical downscaling),
scientific understanding, evaluation with observations, training programmes,
application modellers to deliver climate impacts, emission scenario developers,
 Multi-disciplinary approach allows exchange of knowledge, ideas and
techniques, e.g. extensive work on extremes
 Multi-model ensemble-based probability approach will quantify
uncertainty, increase understanding, influence the development of the
next generation of models, leading to uncertainty reduction in the future
 Examples of new products:







multi-model RCM system at 25km resolution
probabilistic methods for use for GCMs, RCMs, impact models
probabilistic predictions from s2d2c timescales to explore impacts
gridded observations for Europe with estimate of uncertainty
public availability of large datasets
developments to the publicly available Climate Explorer
on-line tools for users to downscale Ensembles simulations
© Crown copyright 2007
Page 24
Questions
© Crown copyright 2007
Page 25
Download