Transcript 9.22 MB

Dr. Gregory M. Flato
Canadian Centre for Climate Modelling
and Analysis.
M.S., (University of Alberta, Edmonton)
PhD (Dartmouth College)
Research Interests:
• Global coupled climate modelling
• Sea-ice dynamics and thermodynamics
• Role of the cryosphere in climate
The Arctic in Global Climate Models and
Projections of Future Change
Gregory M. Flato
Canadian Centre for Climate Modelling and Analysis
Meteorological Service of Canada
Outline
• Arctic climate and its variability
• Global climate models
• Representation of Arctic climate and
climate processes in global models
• Projections of future climate change
• Summary
Arctic Climate
Observed Surface Temperature (oC)
Winter
Annual Mean temperature anomaly
Time series: 1850-present
Jones and Moberg (2003)
Fyfe (2004)
Based on NCEP Reanalysis
John Walsh – U. Illinois
Atmospheric Circulation
Winter
Mean Sea-Level Pressure (hPa)
Summer
Fyfe (2004)
Sea-Ice Circulation
International Arctic Buoy Program
• Variations in transport, deformation, growth and melt all contribute to
observed variability and recent decline in ice coverage.
Courtesy of J. Walsh, U. Illinois.
• Transport of ice is
balanced by net growth or
melt.
• the associated salt or
freshwater fluxes impact
ocean mixing and
circulation.
Courtesy of M. Hilmer, IfM, Kiel
Ocean Circulation
Surface layer
Atlantic layer
Jones (2001)
• There are many important feedbacks and connections between
these climate components in the Arctic.
• A model provides a framework for synthesizing our
understanding of this complex system, and provides a tool for
making quantitative projections of future change.
• However, the Arctic is part of, and interacts with, the global
climate system, so it can’t be considered in isolation.
Global Climate Models
•
Based on laws of physics
•
Mathematical representation of
– 3-D atmosphere: its temperature, humidity, wind, radiative transfer, cloud
formation/dissipation, precipitation, …
– 3-D ocean: its temperature, salinity, circulation, mixing, …
– Sea-ice: its formation, melt, motion and deformation.
– Land surface: its temperature, moisture content, reflectivity,
evapotranspiration, …
The equations are solved numerically on a
discrete grid.
A problem peculiar to the Arctic is the
convergence of meridians at the North
Pole – this causes numerical difficulties,
particularly in the ocean model.
A recent trend is to
make use of
alternate grid
configurations to
better resolve
ocean (and ice)
processes in the
Arctic.
These examples are from the POP ocean code,
used in the NCAR community climate model.
http://climate.lanl.gov/Models/POP/index.htm
Model Intercomparison Projects
• There are perhaps 15 or so global climate models under
development around the world.
• Intercomparison projects provide an opportunity to:
• evaluate models in a systematic fashion;
• compare/contrast results from different models;
• and hopefully, to identify reasons for the differences.
Global Climate Models of the mid 1990s
Model Name
Reference
Flux-Adjustment
Sea-Ice Variable
Sea-Ice Processes
BMRC
Power et al. (1993)
none
thickness
thermo-only
CCCma
Flato et al. (2000)
heat, water
mass per unit area
thermo-only
CCSR
Emori et al. (1999)
heat, water
water equivalent depth
thermo-only
COLA
Schneider et al.
(1997)
none
thickness
thermo-only
CSIRO
Gordon and
O’Farrell (1997)
heat, water,
momentum
thickness
dynamicthermodynamic2
MPI_E4
Roeckner et al.
(1996)
heat, water (annual
mean)
thickness
dynamicthermodynamic
GFDL
Manabe et al.
(1991)
heat, water
thickness
drift-thermodynamic3
GISS_M
Miller and Jiang
(1996)
none
mass per unit area
thermo-only
GISS_R
Russel et al. (1995)
none
% time grid cell
occupied by ice
thermo-only
MRI
Tokioka et al.
(1996)
heat, water
thickness
drift-thermodynamic
NCAR_CSM
Boville and Gent
(1998)
none
thickness
dynamicthermodynamic
NCAR_WM
Washington and
Meehl (1996)
none
thickness
dynamicthermodynamic
UKMO
Johns et al. (1996)
heat, water
thickness
drift-thermodynamic
 Motionless ice with a prognostic equation for ice growth and melt.
2 Prognostic equations for growth/melt and ice motion, including representation of internal ice stress.
3 Prognostic equation for ice growth/melt, ice motion diagnosed as a function of ocean surface current.
Flato (2004)
One can look at ensemble mean quantities, or look at individual models …
IPCC (2001)
Intermodel standard deviation of surface air temperature (oC)
Model disagreement is
largest over area
influenced by sea ice.
Just as sea-ice feedbacks
amplify climate change, they
also amplify model errors
and contribute to
uncertainty in projections of
future climate.
based on CMIP archive data
MSLP ensemble mean error
Atmosphere-only models
Coupled models
Walsh et al. (2002)
Annual Mean Sea-Level Pressure
NCEP Reanalysis
CCCma CGCM2
Modelled ice extent in the 12 model CMIP ensemble
10% of models have less ice
than this.
Median ice edge.
10% of models have more ice
than this.
Interestingly, median
model ice edge agrees
well with observations.
Flato, 2004
Snow cover ‘error’ in AMIP1 models (early 1990s)
Frei and Robinson, 1998
Snow cover ‘error’ in AMIP2 models (late 1990s)
Frei et al., 2003
Projections of future change
Concentration (ppmv)
• Coupled models are forced with GHG and aerosol forcing as
observed from 1850 to the present, then increasing as per
some prescribed future scenario.
1600
1400
1200
1000
IS92a
IPCC A2
IPCC B2
21
00
20
50
20
00
19
50
19
00
800
600
400
200
0
Projected Surface Air Temperature Change – 2050 vs 1980
CCCma CGCM2 -- Mean = 1.92oC
One can compare the evolution of temperature anomalies over
time …
Observations
1946-56
1986-96
CCCma Model
Projected climate warming is enhanced over sea ice; as in the case
of ‘control’ climate, this is also the location of largest disagreement.
(But all models predict warming)
NH ensemble mean temperature change (C)
NH intermodel standard deviation (C)
Based on CMIP archive
Predictions of sea-ice changes likewise vary from model to model.
Here we show NH annual mean ice extent from CCCma and Hadley Centre models.
Walsh observations
Both models underestimate
ice extent somewhat.
CCCma model indicates more
rapid historical and future
decline – not inconsistent
with observed decline.
NH Ice Extent and its Change – CMIP2 model ensemble
(CO2 increased at 1% per year for 80 years – the time of doubling
Initial Ice Extent
10
5
N
R
KM L
O
U
KM 2
Av O
er 3
ag
e
U
IA
P
LM
D
M
R
N I
C
AR
C
C
BM
RC
m
C
C a1
C
m
a
C 2
C
C
ER SR
FA
C
C S
SI
EC RO
H
A
M
G 3
FD
L
G
IS
S
0
No obvious connection between
error and ice model
characteristics
But all models predict a decline
Average
UKMO3
UKMO2
NRL
NCAR
MRI
LMD
IAP
GISS
GFDL
ECHAM3
CSIRO
CCSR
CERFACS
NH Ice Extent Change (10^6 km^2)
15
CCCma2
-1
20
CCCma1
0
C
NH Initial Ice Extent (10^6 km^2)
25
BMRC
Ice Extent Change
-2
-3
-4
-5
-6
-7
-8
-9
no flux-adj. flux adj.
thermo-only
dynamic (rheology)
dynamic (drift)
diagnostic
Flato, 2004
Summary
– Feedbacks involving the cryosphere lead to amplification of
projected climate warming in the Arctic.
– These feedbacks also amplify model errors
– Although global climate models are improving, the Arctic
remains a challenge.
– Model errors tend to be larger than elsewhere.
– Nevertheless, models universally agree that climate change will be
larger in the Arctic than at lower latitudes.
– The last decade has seen an increased focus on modelling
Arctic climate.
– Various intercomparison projects yield quantitative evaluation of
model shortcomings.
– Representation of snow in climate models has improved
demonstrably.
– More sophisticated sea-ice models are being employed, and
alternative grid configurations are being used to improve resolution
of Arctic ice and ocean processes.
The End
Model projection
of change in
permafrost
Stendel and Christensen, 2002
Sea-Level Pressure
The ‘Arctic Oscillation’
Figures courtesy J. Fyfe
Temperature
CCCma CGCM1
Fyfe et al., 1999
Observed
Fyfe, 2004
Arctic Oscillation
Climate change scenario
CCCma model
Fyfe et al., 1999
Observations to 2002
GISS model
Stratosphere included
Shindell et al., 1999
No Stratosphere
900-year time series of NH ice extent from CCCma climate model
Likelihood of observed trends based on 5000 yr control run of GFDL model
Recent trend is not likely a result of natural variability …
1978-98 trend
(p < 2%)
1953-98 trend
(p <0.1%)
Vinnikov et al. (1999)
•
There is substantial
interannual variability
in Fram Strait
outflow, but no
obvious trend.
•
Correlation with NAO
is strong (r=0.7) for
the period 1978-1997,
but weak (r=0.1) for
‘58-’77 period.
Hilmer and Jung (2000).
•
Observations are
insufficient to say much
about ice thickness
variability, but model
results give some indication.
•
Variability is expected to
be large near coastlines,
due to wind-driven
deformation.
•
Submarine observations
do provide some
evidence for long-term
change in thickness.
•
40% decrease between
1958-1976 and 19931997.
Rothrock et al., 1999
•
However, model
results indicate that
wind-driven changes in
thickness build-up
pattern, and limited
sampling, may be
important.
Holloway and Sou, 2002
Thickness change by middle of 21st century
CCCma
Hadley
CCCma
Hadley
CCCma model projects seasonal Arctic ice cover by mid century.
March
September
1971-1990
2041-2060
Ensemble mean thickness
Intermodel standard deviation
Composite results for Southern Hemisphere.
10 model ensemble
Ensemble mean thickness
JJA
Intermodel standard deviation