Transcript Document

Current Advances in the Science of Climate
Forecasting: Prospects and Limitations
Penehuro F. Lefale
Scientific Officer
World Climate Programme (WCP)
World Meteorological Organization (WMO).
Email: [email protected]
Website: www.wmo.int
Presentation to the Regional Technical Meeting on CLIPS and Agro
meteorological Applications for the Mercosur Countries
Campinas, Brazil, 13-16 July 2005.
Credits to the following Institutions; UK Met Office Hadley Centre (Dr. R. Graham), ECMWF (Prof. D. Anderson, Dr A.
1 for
Troccoli), IRI (Dr S. Mason), Beijing Climate Center (BCC), Chinese Meteorological Administration (CMA) (Prof.Page
Chen)
some of the slides used in this presentation.
Outline of Presentation
I : Definitions
II : The Climate System
III : Seasonal to Inter-annual Forecasting
IV : Prospects and Limitations
Page 2
Preamble
“All models are wrong…” G. Box
‘…but some models are useful..’
Although they may need a bit of help…
“Current predictive capability for the onset of El Niño is still
relatively modest, particularly for the onset of weak events, due
to a number of factors related to an El Niño’s initiation”
Lyon and Barnston, IRI
US CLIVAR VARIATIONS, Spring 2005, Vol. 3. No. 2
Page 3
of parameterized
processes
in the
model
* Example
A climate model
is a very complex
system, with
manyECMWF
components….
Components
of
an
atmospheric
model
* Models must be tested at system level, i.e. by running the model and
comparing
the results with observations. Such tests can reveal problems,
Equations:
but their source is often hidden by the model’s complexity…
• momentum, hydrostatic, mass conservation,
* It is also important to test the model at the component level, i.e. by
• thermodynamic energy
isolating particular components and testing them outside the framework
• moisture
& any other trace constituents
of the complete
model….
* Here, we can make an analogy with the testing of a new aircraft. Flight
tests are
needed toterm
evaluate
theequations:
entire aircraft as a system, but
Parameterized
in these
component
tests are
also
essential…
• sub-grid
scale
organisation
e.g. convection, boundary layer fluxes
• internal diabatic heating e.g. radiative transfer and condensation
• processes involving additional variables e.g. land surface
Page 4
IPCC AR4, 2007 (in prep)
Seamless Forecasts
Climate: statistics
of the atmosphere
(average conditions
of the atmosphere)
Outlook
Prediction
Forecasts
Watches
Warnings & Alert
Coordination
GHG
Concentrations
Years
Seasons
Boundary
Conditions
Months
2 Week
1 Week
Weather: Days
State of the
atmosphere
Hours
at a given time
and
place
Minutes
Initial
Conditions
Commerce
Health
Energy
Ecosystem
Recreation
Agriculture
Hydropower
Fire Weather
Protection of
Life & Property
Benefits
Reservoir
Control
Transportation
Space
Operation
Flood Mitigation
& Navigation
Climate is traditionally viewed as the integration
of discrete weather events and variables over time and space
The corollary is that: the components of climate change should be
Page 5
manifest on all time and space scales
Environment
Threats
Assessments
Forecast Lead Time
Guidance
Forecast
Uncertainty
State/Local
Planning
Climate Change
Scenarios
Global Warming
Climate Change: change of climate
Climate
Change:
change
in
that is attributed
directly
or indirectly
to human
activity
that alters
the composition
thetoglobal
atmosphere
climate
over time,
whether of
due
natural
variability
is in of
addition
to natural
or and
as athat
result
human
activityvariability
(IPCC).
observed over comparable time periods (UNFCCC).
Part II: The Climate System
* Key Components of the Climate System
* Predictability of the Climate System
* Advances in Modeling
* Model Development – the UK Met office
GloSea and ECMWF DEMETER examples
Page 6
Schematic view of the components of the global climate system (bold),
their processes (thin arrows) and some aspects that may change (bold
arrows).
(IPCC TAR, WG I, 2001, p.88)
Page 7
Predictability of the Climate System
•
•
•
•
The feature that gives longer potential predictability is
the ocean (and maybe slow boundary changes
associated with snow cover, soil moisture, sea ice..).
The ocean has a large heat capacity and slow
adjustment times relative to the atmosphere.
If the ocean forces the atmosphere on these
timescales, then there can be longer predictability.
On the other hand if the atmosphere forces the
ocean with little or no feedback, there might be little
predictability.
Latif et al., 2002, Timmermann 2005, Hasselmann,
1976, Anderson., D., 2005, ECMWF, June 2005.
Page 8
This plot shows where the skill in predicting SST is highest. The tropical
Pacific is highest but the tropical Atlantic and Indian ocean could have
skill as well. The skill currently realised by models is lower.
Page 9
Predictability of the Climate System
•
Estimating anthropogenic climate change on times
much longer than the predictability time-scale of natural
climate fluctuations does not, by definition, depend on
the initial state.
•
Predicting climate change is one of estimating
changes in the probability distribution of climatic
states (e.g., cyclonic/anticyclones weather, El Niño,
global temperature, etc) as atmospheric composition is
altered in some prescribed manner.
IPCC TAR, 2001; IPCC AR4, 2007 (in prep)
Page 10
Part III: Climate Forecasting
* The basis for seasonal forecasting
* Atmospheric Modeling
* Ocean Modeling
* Coupled Modeling
Page 11
Demand driven Climate Forecasting
Traditional Methods:
• Bio-indicators
• The Farmer’s Almanac
Coupled Models:
• AOGCM
• the ocean (memory)
• SSTs (ENSO)
Statistical/Empirical Models:
• univariate
• multivariate
Page 12
Courtesy: IRI with modification by Lefale.
Sources of Seasonal Prediction
– Known
to be important:
– El Nino variability (biggest single signal)
– Other tropical ocean SST ( important, but multifarious)
– Climate change (all forms) (especially important in mid-latitudes)
– Local land surface conditions (e.g. soil moisture in spring)
– Other possible factors:
– Mid-latitude ocean temperatures (always controversial)
– Remote soil moisture/ snow/ice cover (not well established)
– Volcanic eruptions (important for large events)
– Stratospheric
- possible tropospheric impact
– Dynamic memory of atmosphere
- most likely on one or two month
– Solar cycle, stratosphere
- questionable statistical connections
Page 13
Methods of Seasonal Prediction
•
•
•
Empirical/Statistical method
– Use past observational record and statistical methods
+
– Works with reality instead of error-prone numerical models
+
– Limited number of past cases means that it works best when observed
variability is dominated by a single source of predictability
– A non-stationary climate is problematic
Dynamical method (Single-tier GCM forecasts)
– Include comprehensive range of sources of predictability
+
– Predict joint evolution of SST and atmosphere flow
+
– Includes indeterminacy of future SST, important for prob. forecasts +
– Model errors are an issue!
Combination method (Two-tier forecast systems)
– First predict SST anomalies (ENSO or global; dynamical or statistical)
– Use ensemble of atmosphereice GCMs to predict global response
– Use El Nino index to statistically predict a local variable of interest
Page 14
Page 15
IPCC TAR, 2001.
Hindering Prediction
 Pin table analogy
 Small differences at the start
are amplified by chaos effects
 Individual “plays” are
unpredictable
 Probability of where a ball
ends up is given by “climate”
statistics
Frequency
Extreme
seasons
Normal
seasons
Page 16
Courtesy: Graham, R., 2005, UK Met Office.
Helping Prediction: role of SST
• SST influences the
frequency statistics
• SST can be predicted
• Thus we can predict
how the probability of
particular outcomes is
enhanced/diminished.
• “Ensemble” forecasting
Frequency
Page 17
Courtesy:
Page
16
© WMO copyright 2005
Graham, R., 2005, UK Met Office.
Example of Dynamical seasonal forecasting
system: The Hadley Centre GloSea
• Enhanced version of the Hadley Centre Climate model HadCM3
• 41 member ocean-atmosphere global forecast ensemble
• 5 ocean analyses from perturbed wind stresses
• Ocean analyses further perturbed with instantaneous SST perturbations
• hindcast period, 1987-present (1987-2002 used)
• Real-time system, run to 6 months ahead from 1st day of each month
Retrospective Forecasts - 15 member ensemble
Atmosphere
NWP/re- analyses
Real - Time
Forecast
41 member
ensemble
15 member
Ocean Analysis - 5 member ensemble
1987
1988
Page 17
2004
© WMO copyright 2005
Page 18
Example: Glosea product (SSTs)
Page 19
European multi-model: research and operations
•
EU research project DEMETER
– multi-models represent forecast uncertainties due to model
formulation, as well as initial condition uncertainty
– 7 coupled CGCMs from European institutes
– retrospective period, up to 43 years (1959-2001)
•
Result: multi-models improve skill and reliability
•
Real-time operational European multi-model prediction system
– Currently Met Office (GloSea) and ECMWF (system2)
– Meteo-France system will soon join operations
– Other European models may join
Page 20
Single Model against multi-model comparison
Page 21
Single Model against multi-model comparison
probability of well-above temperatures, Feb-Mar-Apr
GloSea
GloSea+ECMWF multi-model
P(well above)
P(well below)
Page 22
Example combined dynamical/statistical products:
North East Brazil: March-May 2005 rainy season
dry
avge
wet
Probabilities for
3 categories
Risk of ‘extreme’
Reference
climatological
data
Page 23
Development of Met Office climate models
ATMOSPHERE LAND OCEAN ICE SULPHUR
CARBON
1999 ATMOSPHERE LAND OCEAN ICE SULPHUR
CARBON
2000
CHEMISTRY
1997 ATMOSPHERE LAND OCEAN ICE SULPHUR
1992 ATMOSPHERE LAND OCEAN ICE
Component models
ATMOSPHERE LAND OCEAN
are constructed off-line
and coupled in to the
1985 ATMOSPHERE LAND
climate model when
sufficiently developed
1960s ATMOSPHERE
Page 24
HADLEY CENTRE CLIMATE MODELS
HadCM2 HadCM3
1994
1998
HadGEM1
2003
Atmosphere
2.5 x 3.75 2.5 x 3.75
19 levels 19 levels
1.25 x 1.875
38 levels
Ocean
2.5 x 3.75 1.25 x 1.25
20 levels 20 levels
1x1
40 levels
Flux adjust?
Radiation
Yes
No
No
CO2 equiv separate ghg separate ghg
Sulphur cycle
Carbon cycle
Chemistry
Computing
No
No
No
1
Yes
No
No
4
Yes
No
No
40
Page 25
Numerical models of the
atmosphere
Hor. scales
• Climate models
• Global weather prediction
• Limited area weather pred.
• Cloud resolving models
• Large eddy models
500 km
50 km
10 km
500 m
50 m
Vert. Scales time range
1000 m
500 m
500 m
500 m
50 m
100 years
10 days
2 days
1 day
5 hours
Different models need different processes parameterized and
different complexities of parameterization
Page 26
Evaluation of the performance of
atmospheric models
• Routine weather forecasting*
• “AMIP”-style runs: seasonal or interannual
with specified SSTs and sea ice
• Comparison of processes and phenomena
with observed/analysed data
* Initial value prediction is dependent on several factors beyond the numerical model
itself – e.g. data assimilation techniques, ensemble size, ensemble generation method)
Page 27
Typical data use to evaluate climate models
Re-analyses of the
global circulation
Synthesized climatologies
e.g. Precipitation
Satellite
Observations
In situ
Measurements
Page 28
Part III: The Calibration and Verification
* Observations
* Data Assimilation
* Downscaling
* Errors in the Models
Page 29
Observations (new data)
• A wide range of remote and in situ systems
being used to expand the observational data
flow from the oceans
• Signals from the tropical subsurface (e.g. TAO
array) are good indicators of where process is
tending
• Considerable improvements from satellite
systems
• New observations add additional information
Page 30
The Observational network
• Mooring (TAO, PIRATA AND TRITON) (T)
• XBTs (dropped from ships of opportunity (T)
• CTDs (High quality but very few-contained by research
ships) T & S
• SST from satellite (IR, MW), ship and buoys.
• SSS from a few ships, from a few moorings, from
satellite in future (e-g- SMOS Aquarius)
• Sea level from altimetry (ERS, TOPEX, Jason)
• Current meters (very few ~5 along the equatorial
pacific)
• Subsurface temperature and salinity from ARGO
Page 31
The Observational network
Atlas moorings are the
backbone of the equatorial
ocean observing system.
They measure T at 10
depths from the surface to
500m. The data are
transmitted via satellite and
are on the GTS within a few
hours.
Page 32
The Observational network
Page 33
The Observational network
Operating
method of
ARGO floats
Page 34
Data coverage for June 1982
Page 35
Data coverage for March 2002
Page 36
Build up of ARGO February 2005
Page 37
Dynamic downscaling (DD) versus
statistical downscaling (SD)
• DD builds on physically based models for both global and regional
scales and has a great potential.
• SD relies on GCM for large scale and statistical models for regional
and/or local scales.
DD still has problems with today’s climate!
SD can deal with non-standard or difficult (e.g. Sea ice)
variables.
SD can handle a variety of different scales.
SD is lless problematic with bias (because of data-based).
SD is fast ->large number of predictions
However, more risky with extrapolations!
Needs extensive data!
Page 38
Improvement in Forecast Skills
Courtesy: ECMWF
Page 39
Part IV: Prospects and Limitations
Page 40
Prospects for improving atmospheric models
• Role of controlled
sensitivity experiments
in both simplified and full
models
• Use of ultra high
resolution to improve
parameterizations for
use in longer run lower
resolutions
• Importance of vertical
resolution and what is
going on at the top of
the lower boundary layer
and tropopause surfaces
• Correct scaling of and
new parameterization
schemes
• Diversity of models
• Incorporation of
processes such as
inertia instability e.g.
especially as air and
water crosses the
equator
• Random errors due to
the cumulative impact of
unresolved impact of
unresolved processes
Page 41
Conventional vs. Dateline
El Niño Impacts
• The NOAA definition for El Niño was recently adopted by the WMO
Region IV (May 2005)
• The definition identifies as El Niño many more Autumn and Winter
seasons than has been conventional
• These additional seasons show warming in NINO 3.4 and out
towards the Dateline, but not in the eastern tropical Pacific
• El Niño impact associations need to be re-evaluated given this new
definition
 These additional Dateline seasons have substantially different
seasonal average impact association structures compared to
conventionally identified impacts
Larkin and Harrison, GRL, 2005a&b (in press)
Page 42
Conventional vs. Dateline
El Niño Impacts – Autumn and winter
temperature Anomaly Composites
Page 43
Larkin and Harrison, GRL, 2005a&b (in press)
Conventional vs. Dateline
El Niño Impacts – Autumn and winter
precipitation Anomaly Composites
Larkin and Harrison, GRL, 2005a&b (in press)
Page 44
Conventional vs.. Dateline
El Niño Impacts – Frequency of Extreme
seasons (top 20%)
Larkin and Harrison, GRL, 2005a&b (in press)
Page 45
Conventional vs. Dateline
Global Impacts
Page 46
Larkin and Harrison, GRL, 2005a&b (in press)
Conventional vs. Dateline
Global Impacts
Page 47
Larkin and Harrison, GRL, 2005a&b (in press)
International Climate Activities
Global Earth Observation Systems of System (GEOSS) –
Group of Earth Observation (GEO)
UNESCO
ISSC
ICSU
UN Educational
International International
Scientific and
Social
Council for
Science
Science Cultural Organization
Council
International
Non-government
Organization
IOC
WMO
WORLD
BANK
UN
UNEP
MAB
IPCC
FAO
UNDP
UN FCCC/COP
GEF
CCA
Coordination activities within the Climate Agenda
WORLD CLIMATE PROGRAMME (WCP)
IHDP
International
Human
Dimensions
Programme
IGBP
International
GeosphereBiosphere
Programme
GOOS
(IOC, WMO, UNEP, ICSU)
WCRP
(WMO, ICSU)
WHO
WCDMP
(WMO)
WCAC
(WMO)
CCl
CAgM
WCIRP
(UNEP)
THE CLIMATE AGENDA
GCOS
(WMO, IOC, UNEP, ICSU)
GTOS
(UNEP, FAO, UNESCO, WMO, ICSU)
Page 48
INTERNATIONAL SCIENTIFIC PROGRAMMES
Concluding Remarks
•
•
Uncertainty embedded in climate forecasting - ‘chaotic’ processes inherent
in the atmospheric system
Skill of the SI Forecast varies by geographic region, by climate parameter
and by time-scale
Operational
•
Improved Climatology
•
6 to 12-month forecast in every month
•
Probability forecast
•
Real-time monitoring of verification (WMO-SVS)
Major Research Topics
•
Model improvement
•
climate system model (more component of climate system,
ice/land/…)
•
land surface initialization
•
direct couple scheme/coupler
•
sstatistical-dynamical combination (downscaling…)
•
multi-model ensemble
Page 49
Concluding Remarks
•
•
•
•
Coupled models appear to be mature enough to be used in decision making
Further model developments as well as increase in model resolution, will
however be beneficial in advancing science of SI
Improvements in the underlying science must be matched by improvements
in communications between providers and users
Forecasting should be viewed as a more integrated system rather than an
end to end process.
Page 50