Transcript 040521_Cloud_Feeback_Presentation_Wood

Cloud Feedback Reading Group
May 21, 2004
Discussion of Cess et. al. (1989)
Discussion of Cess et. al. (1996)
Cloud-Climate Feedback as Produced by different General
Circulation Models
1989 Paper
A. Summary
1. Intercomparison of 14 atmospheric GCMs
2. Climate change interpretation: 2-step process involving
forcing and then subsequent response
2. Sea-surface perturbations used as proxy for climate change
3. Threefold variation in global climate sensitivity among
models
a. mostly due to differences in models of depiction of
cloud-climate feedback
b. cloud feedback ranged from modest negative to
strong positive
B. Radiative Forcing, G
1. G = F - Q
F: Global mean emitted IR flux at TOA
Q: Global mean downwelling solar radiation
F, Q: climate change responses to radiative forcing
C. Climate Sensitivity, λ
1. TS = λG
Thus, according to definition of G,
λ = 1 / [F/ TS - Q/ TS ]
2. As λ↑, there is increased climate change for a given forcing
3. Example: Basic Temperature-Radiation Negative Feedback
a. Q/ TS = 0, F = εσT4,
F/ TS = 4F/TS = 3.3 Wm-2K-1 (using F = 240 Wm-2 and
TS = 288K
thus, λ = 0.3 Km2W-1
b. when positive water vapor feedback is included, λ is
increased to 0.5 Km2W-1 (this feedback does not include
the effect of clouds)
4. Cloud-climate feedbacks
a. Considerably more complex: i.e. a reduction in clouds
increases IR emission at TOA, but also increases
absorption of solar radiation due to decreased albedo
D. Method
1. For each model integration, λ is produced along with λC, a
clear-sky sensitivity
2. ±2 K sea-surface perturbations, with perpetual July
simulation, and fixed sea ice constraint
a. Inverse of a climate change simulation, in that climate
change is prescribed, and forcing is then determined
b. Eliminates need for models to equilibrate to the ocean
3. Grid sizes ranged from 2.8º x 2.8º to 5.0º x 7.5º
E. Reconciliation of clouds by models
1. Stratiform
a. All models (except ECMWF & ECMWF/UH) allow
stratiform clouds to be formed when the relative
humidity exceeds a prescribed value (usually ~ 90%)
b. ECMWF & ECMWF/UH use vertical velocity/lapse
rate
2. Convective Clouds
a. Much less consistency among models
i. 6 of the GCMs calculate convective clouds in much
the same way as stratiform clouds
ii. Others: parameterization of convective clouds
F. Cloud Radiative Forcing (CRF) versus Cloud Feedback
1. CRF: radiative impact of clouds on Earth’s radiation budget:
CRF = FC – F + Q – QC (subscript c: clear-sky)
CRF > 0: clouds produce warming
2. Cloud Feedback (λ/ λC)
λ/ λC = 1 + ∆CRC/G
If ∆CRC = 0, climate sensitivity parameter is the same
as the clear-sky sensitivity
λ/ λC > 1: positive feedback
(same as ∆CRC/G > 0, which is used in 1996 paper)
G. Results
1. λC for all the models is rather similar
Mean λC = 0.47 Km2W-1, Standard Deviation of 0.04
(Small Range)
2. Much larger range of λ among the models (0.39 ≤ λ ≤ 1.11)
Mean λ = 0.68 Km2W-1, Standard Deviation of 0.24
3. Consequent λ/ λC range:
0.70 (modest negative feedback) to 2.47 (strong positive)
Mean ∆CRC/G = 0.45
II.
1996 Cess et. Al. paper
A.
Main purpose
1. intercomparison of 19 GCMs (rather than 14 as in ’89)
2. Assess differences in feedbacks with revised GCMs
B.
Procedure
1. Perpetual July simulation (same as before)
2. Fixed soil moisture (to eliminate effect of excessive
continental drying)
3. Fixed sea ice/snow
C. Some Revisions made to GCMs
1. ECHAM GCM: inclusion of prognostic water content, cloud
optical properties related to cloud water content, different
radiative transfer code, updated convective scheme
a. These changes reduced cloud feedback
b. In ’90 run, clouds were too bright, leading to large
SW positive cloud feedback when climate was warmed
2. GISS GCM also incorporated interactive cloud optical
properties with ’96 run
D. General Results
1. Reduction in mean feedback (∆CRF/G) from earlier runs
a. 1989: (∆CRF/G) = 0.45
b. 1990: (∆CRF/G) = 0.38
c. 1996: (∆CRF/G) = 0.11
2. General convergence of models to less positive cloud feedback
3. However, in many cases there are different explanations
as to why the models produce similar feedbacks
a. ex. CCC: near zero feedback as both SW and LW feedback
components are very small
b. ECMWF, DNM, and CSIRO: nearly compensating
LW and SW components
E. Some Concluding Remarks
1. 1996 model runs: generally modest positive cloud feedback
a. Exception: LMD, which has a much larger LW cloud feedback
component than all the other models (similar to the ’90 run)
2. Smaller variation of net cloud feedback among models
a. This does not necessarily mean, however, that the accuracy
of the revised models is higher
3. Both 1989 and 1996 papers warn that execution of 2 K
perturbation may yield different results than would
occur with real climate change
a. But, such a study is important to attempt to quantitatively
assess importance of clouds in climate change