Understanding Climate Feedbacks
Using Radiative Kernels
Brian Soden
Rosenstiel School for Marine and Atmospheric Science
University of Miami

Overview of “radiative kernels”

Recent advances in understanding cloud feedback:

Cloud feedback is not negative
 High cloud feedback is positive and robust
 Low cloud feedback is neutral to positive and highly variable
Change in Global Surface Temperature
Radiative
Feedbacks
IPCC
1990
1995
2001
2007
The likely range of
climate sensitivity is:
1.5 to 4.5 C
1.5 to 4.5 C
1.5 to 4.5 C
2.0 to 4.5 C
Radiative
Forcing
Global Mean Surface Temperature
IPCC AR4 GCMs
IPCC Assessments
Water Vapor Feedback
Cloud Feedback
1990:
“The best understood feedback
mechanism is water vapor feedback, and
this is intuitively easy to understand”
“Feedback mechanisms related to
clouds are extremely complex”
1992:
“There is no compelling evidence that
water vapor feedback is anything other
than positive—although there may be
difficulties with upper trop. water vapor”
“The effects of clouds remain a
major area of uncertainty in the
modeling of climate change”
1995:
“Feedback from the redistribution of
water vapor remains a substantial source
of uncertainty in climate models”
2001:
“The balance of evidence favours a
positive clear-sky water vapour feedback
of magnitude comparable to that found
in (model) simulations“
“In previous IPCC reports cloud
feedback was identified as a major
source of uncertainty. Considerable
research efforts have further
reinforced this conclusion.”
2007:
“Observational and modelling evidence
provide strong support for a combined
water vapour/lapse rate feedback of
around the strength found in GCMs”
“… there has been no apparent
narrowing of the uncertainty range
associated with cloud feedbacks“
“Cloud feedback has been
confirmed as a primary source of
uncertainty.”
Climate Feedbacks: Kernel Method
Ts 
G = radiative forcing
R = net radiation at TOA
= climate sensitivity parameter
(rate of radiative damping)
G

dR dT dR dW dR dC dR d




dT dTs dW dTs dC dTs d dTs
Temperature
Feedback
Climate
Feedback
=
Water Vapor
Feedback
dR/dX
Radiative
Transfer
Cloud
Feedback
X
Sfc Albedo
FeedbacK
dX/dTs
Climate
Response
Water Vapor Feedback using Kernels
Water Vapor Kernel (from RT code)
Water Vapor Response to 2xCO2 (from GCM)
x
dR
dW
dW
dTs
Radiation is most sensitive to upper
Fractional changes in water
troposphere because clouds mask
vapor are also largest in upper
contributions from lowerWater
levels Vapor Feedback = Kernel x troposphere
Response due to C-C.
=
Ensemble Mean Feedbacks: IPCC AR4 GCMs
Climate Feedbacks in IPCC Models
Bony et al. 2006
• Water vapor provides the strongest positive feedback in GCMs.
• Water vapor and lapse-rate are strongly correlated.
• There is no model with a negative cloud feedback.
Lapse Rate and Water Vapor Feedbacks:
IPCC AR4 GCMs
Water vapor feedback is larger in
models with greater warming of
upper troposphere
Models with greater lowlatitude warming have larger
lapse-rate feedback.
Climate Feedbacks in IPCC Models
Bony et al. 2006
• Water vapor provides the strongest positive feedback in GCMs.
• Water vapor and lapse-rate are strongly correlated.
• There is no model with a negative cloud feedback.
Intercomparison of Climate Feedbacks
Cess et al. (1989)
Cloud feedback is primary cause of uncertainty
Causes of Intermodel Spread in Cloud Feedback
High (0.07)
Mixed (0.18)
Low (0.75)
• High cloud feedback
is positive and robust
• Consistent with FAT
hypothesis of Zelinka
and Hartmann (2010)
• Low cloud feedback
is highly variable
Soden and Vecchi (2011)
Local contribution to intermodel spread
in cloud feedback
• Most of intermodel spread arises from low stratocumulus/cumululs regions
• Possible links to weakening of atmospheric circulation
Soden and Vecchi (2011)
Remaining Challenges

Why is low cloud feedback positive in models?

What role do changes in the large-scale circulation
(subsidence) regulate low cloud changes in the tropics?

What other factors besides subsidence are important?

Is there observational evidence to support any of this?
Extra Slides
Why is High Cloud Feedback Positive?
High cloud changes in GCMs follow a nearly constant
temperature (rather than constant altitude).
This behavior is supported by observations (Z&H 20011)
Zelinka and Hartmann (2010)
Water Vapor Kernel
(zonal, annual mean)
Change in OLR due
to constant RH
increase in WV
Total Sky
Upper Tropospheric
amplification due to C-C:
des/es ~dT/T2
~6%/K @ T=300
~14%/K @ T=200
Clear Sky
Largest feedback comes
from upper troposphere
because that is where the
fractional change in water
vapor is greatest.
W/m2/K/100 mb
Ensemble Mean Cloud Feedback
W/m2/K
W/m2/K
Why is High Cloud Feedback Positive?
Zelinka and
Hartmann (2010)
As climate warms, there is an upward shift in the level
of divergence (and QR) due to increased water vapor
Observational Evidence for PHAT
Observed interannual changes in tropical high clouds follow FAT/PHAT.
Zelinka and Hartmann (2011)
How Good Are Model Simulations of Water Vapor?
Multi Model Ensemble Mean Specific Humidity
IPCC AR4 GCMs - AIRS
Pierce et al. 2006
John et al. 2006
The biases in current GCM simulations are large:
• Moist biases of up 100% in upper troposphere.
• Dry biases of ~25% in lower troposphere.
Model Simulations of Water Vapor
The biases in water vapor vary substantially from model to model
… yet all have very similar wv feedbacks.
Change in Total Column Water Vapor (%)
The Consistency of Water Vapor Feedback
Change in Temperature (C)
• The absorption by water vapor increases in proportion to the logarithm
of its concentration.
• Consistent fractional changes = consistent feedback from water vapor.
Importance of Water Vapor Feedback
0
0.4
0.8
1.2
1.6 W/m2/K
• Positive feedbacks mutually amplify their impact on climate sensitivity.
Satellite-Observed and Model-Simulated
Changes in Atmospheric Water Vapor
?
La Nina
El Nino
La Nina
(cold)
El Nino
(warm)
Pinatubo
Radiative Forcing: 2xCO2 vs. A1b
Uncertainty in Aerosol
~3 W/m2
Uncertainty in CO2:
~0.5 W/m2
Climate Sensitivity and Climate Feedbacks
Ts 
G

G = radiative forcing
R = net radiation at TOA
= climate sensitivity parameter
(rate of radiative damping)
dR dT dR dW dR dC dR d




 ...
dT dTs dW dTs dC dTs d dTs
Temperature
Feedback
4.2
Water Vapor
Feedback
-2.0
Cloud
Feedback
-0.8
Sfc Albedo
FeedbacK
-0.3
W/m2/K