A Mechanism for Low Cloud Response in SP-CAM

A Mechanism for Low Cloud Response in SP-CAM
Matthew C. Wyant
Christopher S. Bretherton
Peter Blossey
Department of Atmospheric Sciences
University of Washington
(thanks also to Marat Khairoutdinov and CMMAP)
Wyant et al.(2008) submitted to JAMES, July, 2008
Overview
• Why do we care about low cloud changes?
• In SP-CAM, tropical low cloudiness increases by 10-20%
as the SST is increased by 2K. What causes this
change?
• Does a CO2 increase affect low clouds differently?
• What are some future directions for research on lowcloud feedbacks in CMMAP?
What are the advantages of superparameterization in
studying cloud feedbacks?
• Many clouds are formed by turbulent circulations. These
circulations may be resolved in a superparameterized
model but must be parameterized in a GCM.
• Aerosol and microphysical processes can thereby also
be incorporated more realistically.
• Higher resolutions are possible within the subgrid cloudresolving model (CRM) of a superparameterized GCM
than in a global CRM (e.g. by using 2-D CRM and/or a
CRM domain smaller than the parent GCM column).
SP-CAM Climate Model (prototype-MMF)
• SP-CAM is “superparameterized”- contains a CRM
running in every grid column replacing convective
parameterizations.
• Uses CAM3 as its host GCM (Khairoutdinov and
Randall, 2005) with 2.8° x 2.8° grid.
• Uses System for Atmospheric Modeling (SAM) 2D CRM
(Khairoutdinov and Randall 2003).
– 32 sub-columns in each CAM column (4km horizontal
resolution)
– 28-level vertical resolution
– 5 category bulk microphysics, temperature diagnostic
for phase and ice habit
– CAM3 Radiation
Cloud Forcing
• Change in net downward radiative flux at the top of
atmosphere due to clouds:
SWCF = SW↓net - SW ↓net clear
LWCF = - (LW ↑ - LW ↑clear)
Net cloud forcing = SWCF + LWCF
• Positive net cloud forcing → Clouds warm climate
system
Climate Sensitivity ()
Ts = G
Change in global
mean surface
temperature
Global-mean change in
outgoing radiation at the top
of the atmosphere
Climate Sensitivity for +2K SST
Ts/G
Low cloud feedbacks and climate sensitivity
Δ Cloud Forcing (W m-2 K-1)
Stratocumulus
Deep Convection
High-sensitivity
Low-sensitivity
Subsidence Rate ω @ 500mb (mb day-1)
Cloud forcing sensitivity from 15 coupled GCMs in a 2xCO2 experiment binned in
30N-30S by subsidence rate ω (Bony & Dufresne, 2005). Red values are from 8
high-sensitivity models, blue are for the remaining 7 low-sensitivity models.
+2K Cloud and Cloud Forcing changes
• SWCF trends dominate
net cloud forcing because
of low-cloud response.
• Low cloud increases in
subtropics, summer highlatitude.
Lower tropospheric stability LTS = 700hPa - 2m
Correlated with
subtropical
marine stratus
cloud cover
(Klein &
Hartmann
1993) In
observations and
models
SPCAM has reasonable net CRF and low clouds
• Patterns good; not enough offshore stratocumulus; ‘bright’ trades/ITCZ.
• Excessive subtropical coastal stratofogulus (poor vertical resolution?)
• In most areas, clouds have plausible vertical distribution.
Analysis Approach
• Use 2.5 - 5 year simulations with specified SST, and
analyze monthly climatologies.
• Present-day SST, CO2 is the control experiment.
• Compare with SST +2K run.
• Compare with 4xCO2 run, with SST unchanged.
• Focus on tropical (30N – 30S) oceans.
• Sort column-months of the large-scale grid using lower
tropospheric stability.
low LTS
warm SST
ascent
high LTS
cold SST
80-90% subsidence
LTS-sorted low-latitude ocean cloud response
warm SST
low LTS
cold SST
high LTS
subsidence
low LTS
high LTS
subsidence
• 10-20% relative increase in low cld fraction/condensate
across all high-LTS (cool-SST, subsiding) regimes.
• This is responsible for SP-CAM’s negative tropical low-cloud
feedback.
Typical vertical structure in trades (SE Pac)
Inversion strengthens
and LTS increases
Subsidence changes
are location-dependent.
• Cloud fraction and inversion strength increase together.
• Net cloud liquid (not shown) proportional to cloud fraction.
• Little change in PBL depth
Other LTS-ordered fields
high SST
low SST
high SST
low SST
diverse
changes
1-2%
moister
PBL
more PBL
rad cool
low LTS
high LTS
low LTS
high LTS
Conceptual model of
SP-CAM trade ‘Cu’ feedbacks
80-90% LTS
Radiative Mechanism
Higher
SST
More
absolute
humidity
More
radiative
cooling
More
convection
More
clouds
Mechanism could be sensitive to GHG and warming scenario since
radiatively-driven.
4xCO2 experiment setup
• Increase CO2 while keeping SST constant.
• Complements +2K SST experiment by focusing on the
effects of radiative changes.
• Gregory and Webb (2008) found this approach useful in
studying the rapid response of cloud forcing to CO2
increase.
• An updated version of SP-CAM is used.
• 2 ½ year integrations are used with the first ½ year
discarded.
• Though the duration is short, the main results hold in
each of the final two years.
Control
Radiative
Heating
Cloud
∆ 4 x CO2
Control
ω
RH
∆ 4 x CO2
∆ Radiative Heating 80-90% LTS
Increased CO2
Reduced LW Cooling
in and above BL
Less BL Convection
Shallower BL
Reduced LWP
50-100% LTS Comparison
SST +2K
4xCO2
∆Radiative Heating
(K/day, 800-950 hPa)
-0.17 (-10%)
+0.16 (+9%)
∆ Low Cloud Fraction
+0.04 (13%)
+0.00 (0%)
∆ Liquid Water Path
(g/m2)
+6.3 (10%)
-2.0 (-3%)
∆ Shortwave Cloud
Forcing (W/m2)
-4.1 (-9%)
+0.7 (1%)
Conclusions
• Subtropical boundary-layer cloud increases
dramatically in SP-CAM simulations with +2K warmer
SST, more-so than in most other conventional GCMs
• Tropospheric warming increases the clear-sky
radiative cooling of the moist trade-cumulus layer,
driving more trade-cumulus cloud. This further
increases the radiative cooling.
• In experiment with 4xCO2, the cloud response is
weaker. With reduced clear-sky radiative cooling, cloud
height is lowered and liquid water is reduced.
• In a fully coupled CO2 experiment we speculate that
low cloud would increase, though perhaps less than
what one would expect from the SST change alone.
Using a Cloud Resolving Model (CRM) understand
and test SP-CAM
• Use regime-composite large-scale forcing from SP-CAM
output to force ‘single-column’ CRM simulations.
• We focus on high-LTS bins with suppressed deep
convection (70-80% and 80-90%) and trade-cumulus
and stratocumulus
θ
RH
CLOUD
SWCF
SP-CAM
CRM
LES
LES resolution (x=100 m, z=40 m, Nx=512)
Summary of CRM Experiments
• Steady-state CRM experiments at SP-CAM resolution
are able to reproduce many features of composite SPCAM profiles and low-cloud response.
• Better horizontal and vertical resolution leads to lower
cloud fraction and different cloud structure.
• Cloud feedbacks are reduced in LES with improved
resolution.
Future Directions
• Examine low-cloud feedback mechanism further in existing SP-CAM
runs (aquaplanet), and future SP-CAM runs with finer horizontal and
vertical resolution.
• Consider alternative model configurations (e.g. embedded mini-LES,
adaptive vertical grid (Marchand)).
• Continue work on single-column analogue CRM experiments.
– Find minimum resolution needed to accurately simulate BL-cloud
feedbacks.
– Apply method to different cloud regimes (stratocumulus, deep
cumulus) and forcings (e.g. aerosols).
– Add synoptic variability to forcing.
• Study feedbacks in future SP-CAM runs utilizing improved physics
(e.g. double-moment Morrison microphysics, RRTMG radiation,
higher-order turbulence closures).
Extra Slides
Interpretation
LES
CRM
• 4 km makes Cu clouds too weak and broad
• Excessive Cu needed to flux water up to inversion.
Comparison of regime sorting methods over
tropical (30N-30S) oceans
warm
cold
neutral
stable
ascent
subsidence
Comparison of Tropical Clouds with ISCCP
Wyant et al (2006)
Comparing GCM Feedbacks
2xCO2 experiment with 12 Coupled models
(Soden and Held 2006)
averaging period
Column Analogue for SP-CAM low-cld feedbacks
(1) Calculate SP-CAM composite for LTS decile (e.g. 8090%).
(2) Use composite , horizontal advective T/q tendencies
and SST. Nudge to composite winds. A realistic wind
direction profile is also needed (RICO).
(3) Allow mean subsidence to adjust to local diabatic
cooling to keep SCM T profile close to SP-CAM
sounding.
(4) Nudge moisture above surface layer to counteract
effects of sporadic deep convection and detraining high
cloud in SP-CAM composite forcings.
(5) Run to a statistically-steady state (average over days
20-60).
Key Assumption 1
(like Zhang&Breth 2008, Caldwell&Breth 2008)
• Regime-mean +2K cloud response can be recovered
from regime-mean profile/advective tendency changes.
Key assumption 2: Vertical Velocity Feedbacks
•
•
•
•
In low latitudes, the free-tropospheric temperature profile
is remotely forced by deep convection over the warm
parts of the tropics.
Weak temperature gradient approximation (WTG):
Stratified adjustment (compensating vertical motions)
prevents build-up of local temperature anomalies.
Our new WTG formulation for column modeling builds on
Caldwell & Bretherton (2008); related to approaches used
by Mapes (2004), Raymond & Zeng (2005),Kuang (2008).
Compared to existing approaches, it has the advantage of
a clear derivation from a relevant physical model
applicable to quasi-steady dynamics.
Vertical Velocity Feedbacks (Derivation)
•
•
•
Assume small perturbation to a reference state.
The linear, damped, hydrostatic, quasi-steady momentum and
mass conservation equations in pressure coordinates give:
amu*  fv *    * x
 * p   Rd Tv* p
am v *  fu*  0
u* x   * p  0
These equations can be combined to relate * to Tv*:
2 *
*
R

Tv
 1 2


2
d
am f  am

p
p
p x 2

•

Assuming sinusoidal pertubations in x of wavenumber k:
2
R
k
 1 2



am f  am2
 d Tv
p
p
p


A horizontal length scale , where k=/(2), and momentum-damping rate am
are needed. We choose =650km and am=1/(2 days) w/ am vertically uniform.
LTS80-90 forcings and profiles
,q profiles; SST
 0 , 0   
Hor. advection
ctrl
+2K
u s
u  q
winds
+ q nudging
1 d 1 , p  550 hPa
 1
0 d , at surface
Results
CRM
SP-CAM
• CRM has deeper moist layer, but similar +2K cloud response.
• Mean and +2K cloud response depend a bit on setup details,
wind shear.
CRM Vertical Velocity Feedback
•Vertical velocity feedback  is small compared to SPCAM 0, has little change in +2K run.
CRM Cu-layer forcing/nudging
Rad
Heating
T Vertical
Advection
Q Vertical
Advection
Q
Nudging
• Q nudging small compared to Q vertical advection
Radiative Heating
Radiative cooling also stronger in +2K CRM (though
less so than SP-CAM)
+2K cloud/CRF changes
• SWCF trends dominate net
 low cloud response.
• Low cloud increases in
subtropics, summer highlatitude.
• LTS increases over all
ocean regions.