Predicting Arctic Sea Ice Retreat
by Cecilia Bitz
Atmospheric Sciences
University of Washington
QuickTime™ and a
TIFF (Uncompressed) decompressor
are needed to see this picture.
Arctic
Summer 2007
Crew members of the USCG Healy, August 2007.
Siberia
Sept 15, 2007
ice extent
Median ice
edge in pink
Canada
extent = above
15% coverage
figure from NSIDC
New York Times,
9 days ago
Low
High
Sea level
pressure
Summer 2007
-20%
Cloud Cover
Summer 2007
Deviation
from mean
+8F
Surface Air
Temperature
Summer 2007
Deviation
from mean
Solar heating
trend
1979-2005
September Arctic Sea Ice Extent, 1979-2007
= successive record lows
September 1979 sea ice extent and successive September record lows
September 2007 monthly average will fall somewhere between the 9/1 and 9/16 images pictured below
1979
1985
2002
2005
1990
9/1
2007
1995
9/16
2007
1980’s
1990’s
• Increased ice advection away from the Russian coast.
• Faster export of sea ice from the pole to Fram Strait.
(Rigor et al. 2002)
1980’s
1990’s
(Rigor et al. 2002)
QuickTime™ and a
GIF decompressor
are needed to see this picture.
Sea ice has
strong postive feedback in summer reflectivity (albedo)
Stronger negative feedback in winter conduction/growth
Sea Ice in the Climate System
Repartitions shortwave radiation in the climate
system
Moderator of air-sea heat and moisture exchange
Freshwater storage and transport
Brine rejection and its influence on water-mass
formation, thermohaline circulation, etc.
The very best sea ice models in global
climate models
1) Continuum fluid-like mechanics
2) Viscous-plastic rheology with elliptical yield
curve (isotropic and scale independent)
3) Subgrid-scale parameterization for ice-thickness
distribution (pdf)
4) Account for internal melt around brine pockets
5) Two-stream, multiple-scattering radiative transfer
Upcoming series of slides summarize results
from global climate models used in the
Intergovernmental Panel on Climate Change
2007
“Forcing” (greenhouse gases and aerosols) vary
as estimated from 21st century observations
In the future, forcing is estimated according
to economic/political scenarios.
“SRES A1B” is a moderate scenario
21st century warming SRES A1B
Relevant papers:
Arctic sea ice decline: Faster than forecast, Stroeve et al 2007
Future abrupt reductions in the summer Arctic sea ice, Holland et al 2006
0.6 Correlation coefficient for linear trend and mean 1990-2020
Simulations with the CCSM3
Community Climate System Model
version 3
resolution:
2.8 deg in atmosphere and land
0.5-1 deg in ocean and sea ice
106 km2
A1B Scenario with CCSM3
September Ice Extent in
one ensemble member
September
Concentration
Holland, Bitz,
and Tremblay,
2006
1 1 8 6 2 more days
Ocean Transport
Absorbed Sunlight
1) Increase in absorbed shortwave is lead by
2) Increase in Ocean Heat Transport through Fram Strait
Two strong positive feedbacks?
Excursions from Ensemble
Mean in 106 km2
Feedback analysis applied to Earth’s temperature
Climate sensitivity = equilibrium change in
global mean temperature, ∆T
due to the reduction in outgoing terrestrial (or
longwave) radiation, ∆R
that would result from 2XCO2
Planetary Energy Balance
S/4 (S/4)a
R
S (1-a) / 4 = R
Atmosphere
Earth
S = Solar constant
a = Albedo (or reflectivity)
R = Outgoing Terrestrial (longwave) Radiation
For a blackbody Earth-like planet
∆To = o ∆R
≈ 1.2 K
R = s TE4
TE = 255K
o = (4sTeq3)-1
Now with additional physical processes
∆T = o ∆R + o C ∆T
= feedback factor
= gain
∆T as a function of latitude, global mean is 2.6 ºC
For all feedbacks
G = 2.6/1.2 = 2.2
f=0.54
G and f are functions of latitude too
Considering individual physical processes
(additive)
(not additive)
= net feedback
∆T as a function of latitude
For only sea ice albedo feedback
G = 2.6/2.0 = 1.3
f=0.23
All feedbacks
2.6 C
No sea ice
albedo feedback
2.0 C
G and f are functions of latitude too
Same thing for sea ice: H = ice thickness
f = net feedback
= gain
BUT o is fundamentally nonlinear!
Simulated Present Day Equilibrium Ice Thickness, H
Equilibrium runs are computed without a dynamical
ocean model!
Equilibrium Ice Thickness Change for all Feedbacks, ∆H
all Feedbacks - ∆H
GAIN from
Ice-Albedo
Feedback
G=2.2 (on
average)
No Ice-Albedo
Feedback - ∆H0
feedback factor
f=0.33 (on average)
Uncertainty in climate sensitivity
Spread in f can give a very long
tail in ∆T - Roe and Baker (2007) soon
to appear in Science
Especially as f approaches 1
f isn’t that close to 1 for ice-albedo feedback
If f is normally distributed
with f = 0.65 and sf=0.1
f = 0.33 and sf=0.1 for ∆H
f = 0.65 and sf=0.1 for ∆T
Initial
Thickness - H
No Ice-Albedo
Feedback - ∆H0
The “no-feedback”, or “reference”, ∆H mirrors H!
Ultimately stems from insulating effect of sea ice,
which depends on 1/H
•Thin ice is a poor insulator, so it can grow fast
•Thick ice is a good insulator, so it grows slowly
Knowing H is key for predicting ∆H,
rather than knowing f to very high accuracy
Bitz and Roe (2003)
showed that
o ~ a + b H2
Here I have shown that
f~0.3 for sea ice albedo feedback
Summary
September 2007 Arctic sea ice cover was 20% lower than
any time in the satellite era. The cause was from anomalous
high pressure, warm air advection, low cloud cover, and ice
transport.
Sea ice in climate models can be complex. Generally the
models appear to either have too little variability and/or
trend, though two models are consistent with 1979-2006
observations.
Rapid retreat appears to be larve variability on top of a
trend, not an instability.
Sea Ice albedo feedback causes sea ice thickness to
decrease about 50-100% faster. Although positive, the
feedback is not enough to cause much uncertainty in
thickness prediction, instead errors are probably more a
function of error in the mean state.
GAIN from adding
Ocean Circulation
Feedback - ∆H
Next series of slides present the governing
equations for state of the art sea ice model
used for climate studies (i.e., appropriate for
basin scale or larger and for full seasonal
cycle or longer)
Specialty models exists with greater or
lesser detail. Some cannot meet the spatial
or temporal requirements. Others may, but
have not yet been implemented in climate
models (to my knowledge).
1st Governing Equations
Ice thickness distribution g(x,y,h,t) evolution equation
from Thorndike et al. (1975)
g(h)dh
A PDF of ice thickness h
in a region, such as a
grid cell
h
1
2
3
4
5
1. Lagrangian time derivative of g following “parcel”
2. Convergence of parcel
3. Y = Mechanical redistribution
4. Ice growth/melt results in “advection of g in thickness
space”
5. L = Reduction of g from lateral melt
h = ice thickness
u = ice velocity
ƒ = growth rate
Y = Mechanical redistribution
g(h)dh
h
Advection in thickness space from growth
g(H)dH
g(h)dh
H
h
2nd Governing Equation
Conservation of momentum, see for example Hibler (1979)
Impact of Arctic Oscillation:
Residence Time of Sea Ice On Arctic Ocean
Low AO (1980’s)
High AO (1990’s)
• Increased ice advection away from the Russian coast.
• Faster export of sea ice from the pole to Fram Strait.
(Rigor et al. 2002)
2007 First Time Ice-Free Area
(http://NSIDC.ORG)