Transcript Separating forcing from feedback

Climate forcing, adjustment and feedback
Christopher S. Bretherton
Atmospheric Sciences and Applied Mathematics
University of Washington
with help from Matt Wyant
Motivation: Mechanisms of PBL cloud response to climate change
Photo courtesy Rob Wood
The climate sensitivity problem
IPCC models match 20C climate, but vary on 21C warming
IPCC 2007
2.54.3°C
Aerosols partly cancel GHGs
2xCO2
3xCO2
Climate feedbacks: a simple view
CO2
atm
surf
T
f
C dΔT/dt = N = G – λ-1ΔT,
λ = λ0-1(1-f)
G (2xCO2) ~ 4 W m-2
λ0-1 ~ 3.3 W m-2 K-1
f ~ 0.6 ± 0.1 (water vapor+LR, snow/ice, clouds)
C ~ 3.3x108 J m-2 K-1 for 100 m deep ocean mixed layer
covering ¾ of the planet
 ΔTeq = G/λ ~ 3 K
teq = C/λ ~ 8 yrs
Complications to the simple view
• Multiple response timescales
internal atmospheric adjustment (days-weeks)
land surface physical adjustment (weeks-months)
ocean mixed-layer (yrs-decades)
biological/carbon-cycle (decades-100s yrs)
deep ocean/thermohaline (100s-1000s yrs)
ice sheets (100s-1000s yrs)
 Scope of feedback analysis depends on the timescale
• GHGs and aerosols don’t just change climate through
their effects on global average surface ΔT
Linear scaling of gross climate change with ΔT
IPCC 2007
For a given model, many aspects of climate change do
scale with ΔT in IPCC scenario runs.
However, not all perturbations fit this mold
• Seasonal cycle (large changes in hemispheric T with
little change in global-mean T)
• ENSO
• Aerosol forcing (mainly in NH, so induces hemispheric
cooling anomaly)
 Thus, even global-mean changes in clouds, rainfall, etc.
may scale differently with ΔT for these perturbations than
in greenhouse warming scenarios.
Example: Response to step increase in CO2
Traditional AGCM climate sensitivity method:
1. Run AGCM to equilibrium over climatological SSTs
2. Calculate implied net energy flux (Qflux) into the ocean
3. Construct a slab ocean model with this Qflux and
climatological ocean mixed layer depth
4. Suddenly double CO2 and run AGCM+SOM to
equilibrium (20-50 yrs). Mean surface air temperature
increase ΔTeq is the climate sensitivity
Net Rad
ΔT
t
ΔT
Transient evolution of such runs also relevant
ΔT
slow CO2
increase
2xCO2
SOM
G = log(CO2 increase)
Fast response of clouds to radiative changes
• Transient behavior of coupled GCM runs in which CO2 is suddenly
doubled shows a quick response of the clouds before the climate
warms which explains some of the ultimate 2xCO2 cloud feedback.
• Quick increase in SWCF (less cloud) after 2xCO2.
Gregory and Webb 2008
Adjustment cartoon
weeks
years
Adjustment
Temperature feedback
time
Cloud adjustment to step 2xCO2 is significant
(Andrews and Forster 2008)
Suggests quick decrease in both low and (to lesser extent)
high cloud. Fast adjustments are about 20% of
equilibrated cloud response to 2xCO2. After subtracting
them out, the T-modulated cloud feedback remains positive
in all models and exhibits somewhat reduced spread.
Geographical pattern is fairly complex and model-dependent
Physical mechanism?
4xCO2 cloud response in superparameterized GCM
• Superparameterization - a climate
model with a small cloud-resolving
(large-eddy simulation) model in place
of the normal physical
parameterizations in every grid column.
• Computationally expensive, but may
simulate turbulent clouds (especially
deep convection) more realistically.
• SP-CAM (Khairoutdinov and Randall 2005) uses 2D CRMs with 32x30
gridpoints, x = 4 km and z ~ 200 m in PBL.
• Investigate low cloud response to instantaneous 4xCO2 increase with
fixed SST.
4xCO2 fixed SST experiment with SP-CAM
• 2½ year integrations are used with the first ½ year discarded...short,
but results hold in each of the 2 years.
∆4xCO2
Radiative
Heating
Cloud
SWCF = 0.7 W m-2
less PBL-top cooling
warm
SST
LTS
cold
SST
4xCO2  ~10 W m-2 more LWdown at low cloud top.
 Less PBL radiative cooling, less turbulence
 Shallower inversion, less boundary layer cloud
…something we weren’t expecting…
Land heats up due to enhanced greenhouse effect
despite fixed SST
The result… with 4xCO2 but fixed SST
Quantitatively…
Over land: More rain, clouds, ascent
Over ocean: Less rain, clouds, ascent
Over full tropics: Less atmos. radiative cooling, thinner low cld
Conclusion
• CO2 increase causes fast as well as temperaturemediated adjustment that significantly affect clouds and
land/ocean interaction.
• A careful analysis of GHG feedbacks on temperature
needs to account for this adjustment as part of the
forcing. This appears to make cloud feedbacks less
positive.
• In thinking about climate forcing and feedback it is critical
to keep in mind the timescale of interest and whether a
given control parameter (e.g. global mean surface ΔT)
adequately characterizes the evolving system.
Fixed Anvil Temperature (FAT) hypothesis
• Substantial clear-sky radiative cooling requires water vapor
• Upper tropospheric water vapor is temperature-limited.
• Clear-sky radiative cooling is weak for T < 200 K.
 Convective anvil tops will occur near T = 200 K, regardless
of surface temperature.
Hartmann and Larson 2002
AR4 models, A1B scenario
(Zelinka and Hartmann 2010)
Vertical cloud profile changes with Ts
Kuang and Hartmann 2007: Radiative-convective equilibrium over fixed SST in
a cloud-resolving model (CRM).
• Entire cloud profile above the freezing level rises 350
m/KSST when plotted vs. z but collapses when plotted
vs. T. GCMs do similarly (Zelinka and Hartmann
2010).