Observational and model evidence for positive low-level cloud feedback

Robert J. Burgman and Amy C. ClementRosenstiel School of Marine and Atmospheric Sciences,

University of Miami

Joel R. NorrisScripps Institution of OceanographyUniversity of California, San Diego

“Cloud feedbacks remain the largest source of uncertainty [of equilibrium climate sensitivity].”

-Summary for Policy Makers, IPCC AR4 WG1

“…simulation of the sensitivity of marine boundary layer clouds to changing environmental conditions constitutes, currently, the main source of uncertainty in tropical cloud feedbacks simulated by GCMs.”

-Bony and Dufresne, GRL (2005)

Here we address this issue by

(1) Examining low-frequency fluctuations in observations of low-level cloud and regional meteorology, and

(2) Using these observations to evaluate climate models.

Atmospheric response to Pacific Decadal Variability

Burgman et al (2008) use multiple EOF (9.6% variance) of 5 NCEP/NCAR reanalysis variables (t, u, v, omega, and sh) for the lowest 8 pressure levels (1000 hPa–300 hPa) in the 60S-60N region over the time period 1970 to 2003. ENSO signal is ‘‘removed’’ by extreme cross correlation at least lag (+/- 12 months).

Recent shift to cooler tropical Pacific in late 1990’s allows incorporation of satellite observations to study atmospheric response to decadal changes in SST.

Passive Microwave era

Atmospheric response to Pacific Decadal Variability

• 1990’s shift strengthening in the overturning circulation– increased subsidence – drying in the upper and middle

troposphere in NE Subtropical Pacific– shoaling of the marine boundary layer– increased lower tropospheric stability.

• Co-location of increased cloud with persistent cool SSTs indicate a possible feedback between Sc and the underlying SST on decadal timescales.

mb/stdv

Wm^2/stdv

mm/stdv

okta/stdvC/stdv

Black- total cloud

Bars- low cloud

Trend noted by Stevens et al. (2007)

ISCCP cloud data corrected for satellite view angle, and inter-calibration errors (Norris, personal communication)

1955 1960 1965 1970 1975 1980 1985 1990 1995 2000 2005

NESP (145W-115W, 15N-25N)

Black- total cloud

Bars- low cloud

Trend noted by Stevens et al. (2007)

COADS surface based total cloud (1952-2007) and marine stratiform cloud (1952-1997) from

Hahn and Warren (1999)

1955 1960 1965 1970 1975 1980 1985 1990 1995 2000 2005

NESP (145W-115W, 15N-25N)

HW99 msc comprises ordinary stratocumulus, cumulus under stratocumulus, fair-weather stratus, and bad-weather stratus

Black- total cloud

Bars- low cloud

• Cloud fluctuations coincide with known climate shifts (1976, and late 1990’s)

• When SST is high, cloud cover is reduced.

1955 1960 1965 1970 1975 1980 1985 1990 1995 2000 2005

NESP (145W-115W, 15N-25N)

Spatial structure of decadal cloud changes

• SURFACE OBS

• Hahn and Warren data available from (1952-1997)

• Dominated by 1976 shift

• SATELLITE RETRIEVAL + ALGORTIHM

• Available from 1983-2006

• Dominated by 1990’s shift

Regression of low clouds on NE Pacific SST

Regional changes are related to large-scale: Regression of meteorology on NE Pacific SST

SST SLP (colors); Surface winds

500 mb subsidence Lower tropospheric stability (t700-tsurf)

Both thermal structure and circulation changes are consistent with weaker stratus deck (Klein and Hartmann 1993; Norris and Leovy 1994; Klein et

al. 1995; Norris and Klein 2000; Stevens et al. 2007)

Familiar PDV

pattern

Regression of ISCCP ‘cloud radiative effect’ on NE Pacific SST

8 W/m2 cloud shortwave ~balances the upward IR due to 1K increase in SST

Cloud change maintains SST anomaly

• When SST is high, LTS is weak, SLP is low, surface winds and subsidence are weak

Reduced low level cloud

• Reduced low level cloud maintains SST

Observations show that:

Models that simulate the wrong sign r(cloud,SST)

Wrong sign or no r(cloud,LTS)

Wrong sign or no r(cloud,SLP)

Wrong sign r(cloud,w500)

Is this feedback present in IPCC AR4 models?

Physically-based parameterization & correct correlations

Observed correlation between NE Pacific cloud and meteorology

20c3m

Is this feedback present in IPCC AR4 models?

The INM-CM3.0 adopts a more empirical approach that parameterizes low-level cloud cover as a linear function of relative humidity with coefficients that depend on temperature, altitude, land/ocean, and stratification

HadGEM1 has higher spatial resolution, more explicit cloud microphysics, interactive parameterization of cloudiness as a function of local variability in humidity, and a sophisticated planetary boundary layer mixing scheme

HADGEM1 simulates reduced NE Pacific cloud cover under doubled CO2 change

WHY?

• Model simulates regional changes in SST (increase) and LTS (increase) that are consistent with all other models

• Circulation changes are also consistent with the multi-model average

• Weaker overturning circulation (Held and Soden 2006, Vecchi and Soden 2007)

• Already observed (Zhang and Song 2006)

Conclusions• It is possible to identify decadal fluctuations in low-level cloud in the NE

Pacific in multiple, independent cloud datasets.

• Cloud changes are physically consistent with local meteorology changes: Cloud cover decreases when SST is high, LTS is weak, SLP is low, and meridional advection and subsidence are weak.

• When put to the test of simulating the relationship between cloud and thermal structure as well as circulation, only one of the current state of the art model with a physically-based parameterization of clouds passes

• Under doubled CO2, that model simulates a decrease in cloud cover in the NE Pacific along with robust changes in thermal structure and circulation

Observed decadal variability in low-level clouds as an analogue to how these clouds will change with global warming

ISCCP Correction• Satellite view angle changes removed by linearly regressing out that portion of cloud

variability associated with local changes in satellite view angle. • Satellite intercalibration error removed by regressing out from each individual grid box

time series the time series of standardized cloud cover anomalies averaged over the entire view area of successive satellites.

• This procedure will remove any real cloud cover variability occurring on near-hemispheric spatial scales but should have little impact on our regression patterns that focus on differences between regions.

Global context

Decadal Cloud Changes

Changes in CRE

SLP Changes