The Impact of Global Warming on Marine Boundary Layer Clouds over ...

5 downloads 5464 Views 12MB Size Report
Nov 1, 2010 - The cloud–climate feedback averaged over the same east Pacific region was also .... rameter defined this way is dominated by changes to the.
5844

JOURNAL OF CLIMATE

VOLUME 23

The Impact of Global Warming on Marine Boundary Layer Clouds over the Eastern Pacific—A Regional Model Study AXEL LAUER International Pacific Research Center, University of Hawaii at Manoa, Honolulu, Hawaii

KEVIN HAMILTON AND YUQING WANG International Pacific Research Center, and Department of Meteorology, University of Hawaii at Manoa, Honolulu, Hawaii

VAUGHAN T. J. PHILLIPS Department of Meteorology, University of Hawaii at Manoa, Honolulu, Hawaii

RALF BENNARTZ Department of Atmospheric and Oceanic Sciences, University of Wisconsin—Madison, Madison, Wisconsin (Manuscript received 9 February 2010, in final form 3 June 2010) ABSTRACT Cloud simulations and cloud–climate feedbacks in the tropical and subtropical eastern Pacific region in 16 state-of-the-art coupled global climate models (GCMs) and in the International Pacific Research Center (IPRC) Regional Atmospheric Model (iRAM) are examined. The authors find that the simulation of the present-day mean cloud climatology for this region in the GCMs is very poor and that the cloud–climate feedbacks vary widely among the GCMs. By contrast, iRAM simulates mean clouds and interannual cloud variations that are quite similar to those observed in this region. The model also simulates well the observed relationship between lower-tropospheric stability (LTS) and low-level cloud amount. To investigate cloud–climate feedbacks in iRAM, several global warming scenarios were run with boundary conditions appropriate for late twenty-first-century conditions. All the global warming cases simulated with iRAM show a distinct reduction in low-level cloud amount, particularly in the stratocumulus regime, resulting in positive local feedback parameters in these regions in the range of 4–7 W m22 K21. Domain-averaged (308S–308N, 1508–608W) feedback parameters from iRAM range between 11.8 and 11.9 W m22 K21. At most locations both the LTS and cloud amount are altered in the global warming cases, but the changes in these variables do not follow the empirical relationship found in the present-day experiments. The cloud–climate feedback averaged over the same east Pacific region was also calculated from the Special Report on Emissions Scenarios (SRES) A1B simulations for each of the 16 GCMs with results that varied from 21.0 to 11.3 W m22 K21, all less than the values obtained in the comparable iRAM simulations. The iRAM results by themselves cannot be connected definitively to global climate feedbacks; however, among the global GCMs the cloud feedback in the full tropical–subtropical zone is correlated strongly with the east Pacific cloud feedback, and the cloud feedback largely determines the global climate sensitivity. The present iRAM results for cloud feedbacks in the east Pacific provide some support for the high end of current estimates of global climate sensitivity.

1. Introduction Corresponding author address: Axel Lauer, IPRC/SOEST, University of Hawaii at Manoa, 1680 East-West Rd., POST Building 401, Honolulu, HI 96822. E-mail: [email protected]. DOI: 10.1175/2010JCLI3666.1 Ó 2010 American Meteorological Society

State-of-the-art comprehensive global climate models (GCMs) display a wide range of global and regional sensitivity to imposed large-scale climate forcings. The

1 NOVEMBER 2010

LAUER ET AL.

equilibrium global surface temperature increase projected to result from a doubling of atmospheric CO2 concentration reported for the models in the latest Intergovernmental Panel on Climate Change (IPCC) intercomparison varies from 2.1 to 4.4 K. This range has not narrowed appreciably compared to that found in earlier model intercomparisons (e.g., Houghton et al. 2001). The variation in global climate sensitivity among these GCMs is largely attributable to differences in cloud feedbacks and feedbacks of low-level clouds in particular (e.g., Bony and Dufresne 2005; Stowasser et al. 2006; Solomon et al. 2007; Medeiros et al. 2008). Persistent stratocumulus decks found predominantly in the subtropical eastern ocean margins have a major impact on the radiation budget by reflecting incoming solar radiation (e.g., Randall et al. 1984; Hartmann and Doelling 1991). These clouds are prominent over cool ocean surfaces in regions where large-scale atmospheric subsidence leads to the formation of sharp temperature inversions, which trap moisture in the marine boundary layer (MBL; e.g., Albrecht et al. 1988). The simulation of these marine clouds has been a particular challenge for global and regional models (e.g., Bretherton et al. 2004; Wang et al. 2004a,b). This results in a particularly high uncertainty of the climate feedback of these low-level marine clouds (e.g., Bony and Dufresne 2005; Solomon et al. 2007). Medeiros et al. (2008) showed that trade wind cumuli might also play an important role; they are also not well captured by global climate models (Medeiros and Stevens 2009). The radiative effect of low marine clouds is dominated by their contribution to the planetary albedo as their impact on outgoing longwave radiation is limited because of the small temperature difference between cloud tops and the underlying surface. The cloud optical depth for these low-level clouds is proportional to cloud geometrical thickness, the liquid water content (LWC), and the size of the cloud droplets. There have been attempts to use empirical guidance to determine how these basic cloud properties, and hence albedo, may respond to changes in large-scale climate. The empirical relationship between cloud LWC and temperature obtained by Feigelson (1978) from aircraft measurements of midlatitude clouds for the temperature range between 2258 and 58C predicts an increase in cloud water with temperature. Somerville and Remer (1984) noted that if a constant cloud geometrical thickness for marine stratocumulus decks is assumed, a negative cloud–climate feedback would be implied (i.e., warmer temperatures would lead to more reflective clouds). On the other hand, studies analyzing satellite data from the International Satellite Cloud Climatology Project (ISCCP), the Advanced Very High Resolution Radiometer (AVHRR), and the Clouds and the Earth’s

5845

Radiant Energy System (CERES) indicate that cloud optical depth of low marine clouds might be expected to decrease with increasing temperature (Tselioudis et al. 1992; Greenwald et al. 1995; Chang and Coakley 2007; Eitzen et al. 2008). This suggests a positive shortwave cloud–climate feedback for marine stratocumulus decks. In a recent paper, Clement et al. (2009) analyzed several decades of ship-based observations of cloud cover along with more recent satellite observations, with a focus on the northeastern (NE) Pacific between 158 and 258N. They found that there is a negative correlation between cloud cover and sea surface temperature (SST) apparent on a long time scale—again suggesting a positive cloud– climate feedback in this region. Bony and Dufresne (2005) analyzed 17 yr of observed SST, top of atmosphere (TOA) net radiative flux, and 500-hPa vertical velocity (v500) from satellite data and reanalyses, respectively. Focusing only on the ocean regions between 308S and 308N, they scaled the anomalies in monthly mean TOA fluxes with the coincident SST anomalies. This analysis was done for grid points in different dynamical regimes defined by the 500-hPa vertical velocity. They showed that the cloud feedback parameter defined this way is dominated by changes to the shortwave flux and that the feedbacks are positive and in the range of 0–6 W m22 K21. The Bony and Dufresne estimate of this feedback is largest in the regions with the strongest 500-hPa subsidence (corresponding to the regions of low-level marine clouds). Of course, as emphasized by Bony and Dufresne, the cloud responses to natural interannual variations of SST may differ from the response expected to large-scale forced global warming. Cloud feedbacks have also been assessed by a number of modeling studies using a variety of models—ranging from single-column radiative–convective equilibrium models to GCMs with conventional cloud parameterizations (e.g., Somerville and Remer 1984; Xu et al. 2010; Roeckner et al. 1987; Caldwell and Bretherton 2009; Tselioudis et al. 1998; Zhang and Bretherton 2008) or to GCMs with superparameterizations in which a cloudresolving model (CRM) is run within each GCM column (e.g., Wyant et al. 2009). However, state-of-the-art GCMs display a wide range of simulated cloud–climate feedbacks in the marine stratocumulus regions. In addition, such models generally do a poor job in simulating the present-day climatology of marine stratocumulus clouds. The dynamics of marine stratocumulus clouds involve tightly coupled interactions among atmosphere, ocean, and land making them extremely challenging for climate models to capture (e.g., Bony and Dufresne 2005). Singlecolumn models and large-eddy simulations (LES) can explicitly represent detailed cloud microphysics, but the interaction between clouds and the large-scale atmospheric

5846

JOURNAL OF CLIMATE

circulation in such models has to be prescribed or determined on the basis of simplified assumptions. This paper describes a modeling study of the response of clouds in the eastern tropical and subtropical Pacific region to large-scale climate forcing. The eastern Pacific region features extensive stratocumulus decks and the cloud feedbacks in this region in climate models contribute significantly to the global mean feedbacks and climate sensitivity (e.g., Stowasser et al. 2006). For this investigation we used the version of the International Pacific Research Center (IPRC) Regional Atmospheric Model (iRAM) described in Lauer et al. (2009). Lauer et al. (2009) also showed that iRAM is able to simulate a reasonable present-day seasonal climatology of stratocumulus clouds in the eastern Pacific region. As part of the present study we will show that iRAM simulates the basic cloud climatology in the eastern Pacific better than current GCMs. We will also show that iRAM successfully simulates the main features of the observed interannual variation of clouds in this region, including the evolution of the clouds through the ENSO cycle. Section 2 describes the regional climate model used in this study and the details of the present-day and global warming simulations. This is followed by a comparison of modeled cloud properties over the east Pacific with observations in section 3. Section 4 presents the results of the cloud response to global warming. The conclusions are summarized in section 5.

2. Model and model simulations a. iRAM IPRC iRAM is based on the hydrostatic primitive equations and uses s coordinates in the vertical (Wang et al. 2004a). All model simulations presented here were conducted at a horizontal resolution of 0.58 3 0.58, with the model domain covering the tropical and subtropical eastern Pacific as well as large parts of South America (408S–408N and 1608–508W). There are 28 model levels from the surface up to about 10 hPa (;30 km) with 10 levels below 800 hPa. We used the final analysis data (FNL) from the U.S. National Centers for Environmental Prediction [(NCEP); NCEP FNL Operational Model Global Tropospheric Analyses, continuing from July 1999, are updated daily. Dataset ds083.2 is published by the Computational and Information Systems Laboratory (CISL) Data Support Section at the National Center for Atmospheric Research (NCAR), Boulder, Colorado, available online at http://dss.ucar.edu/datasets/ ds083.2/] as initial and lateral boundary conditions for the model integrations. The FNL data with a horizontal resolution of 18 3 18 and 26 vertical pressure levels

VOLUME 23

(NCEP–NCAR reanalysis with a horizontal resolution of 2.58 3 2.58 and 17 pressure levels prior to the year 2000; Kalnay et al. 1996) at 6-h time intervals are interpolated linearly in time and using cubic splines to the model grid. SSTs employed are the National Oceanic and Atmospheric Administration (NOAA) analyses (Reynolds et al. 2007), which are based on daily mean satellite observations from AVHRR and the Advanced Microwave Scanning Radiometer (AMSR) instruments. The prognostic model variables are nudged to the NCEP FNL analysis data within a 108 buffer zone along the lateral boundaries. The buffer zone is not in the analyses of the results shown here. Grid-scale cloud processes are calculated using a double-moment cloud microphysics scheme with a semiprognostic aerosol component considering six aerosol species—sulfate, sea salt, soluble and insoluble organic matter, black carbon, and mineral dust (Phillips et al. 2007, 2008, 2009)—which replaces the original singlemoment cloud microphysics module of Wang (2001). The cloud microphysics scheme predicts the mass mixing ratios of water vapor, cloud liquid water, cloud ice, rain, snow, and graupel as well as the number mixing ratios of cloud droplets and ice crystals. The size distributions of cloud and precipitation particles are assumed to follow gamma distributions. Diffusional growth of cloud particles and precipitation is predicted explicitly with a linearized supersaturation scheme from the modeled updraft and properties of cloud liquid and water vapor. The predicted supersaturation is applied to calculate the activation of aerosol particles at cloud base using the aerosol activation scheme of Ming et al. (2006) or inside the cloud when supersaturation becomes high enough. Critical droplet diameters and supersaturations as well as the equilibrium supersaturations of the droplets are obtained from the k-Ko¨hler theory using the hygroscopicity parameters k from Petters and Kreidenweis (2007). The autoconversion of cloud droplets to rainwater is parameterized after Liu et al. (2007). We chose the parameterization from Liu et al. (2007) over other schemes as it results in improved agreement for our regional model with a 50-km resolution between modeled and observed liquid water path (LWP) and cloud cover. Primary and secondary ice nucleation (Hallet and Mossop 1974), as well as homogeneous freezing of aerosols (Koop et al. 2000) and cloud droplets (Phillips et al. 2007), are included. All the known and empirically quantified mechanisms for initiation of cloud droplets and ice are represented (Phillips et al. 2007). The cloud microphysics module is coupled to the radiation scheme and provides effective radii of cloud droplets and ice crystals as well as the liquid water and ice content as input for the radiative transfer calculations.

1 NOVEMBER 2010

LAUER ET AL.

The radiation scheme is based on the radiation package of Edwards and Slingo (1996) with improvements by Sun and Rikus (1999). It considers four bands in the solar spectral range and seven bands in the thermal spectral range. Cloud amount is diagnosed from cloud liquid water/ice content and relative humidity following Xu and Randall (1996). Subgrid-scale convection including shallow, mid-level, and deep convection is parameterized following Tiedtke (1989) with modifications by Gregory et al. (2000). The average entrainment rate for shallow convection and relative mass flux at a level of zero buoyancy (overshooting cumuli) have been adjusted using results from large-eddy simulations (Wang et al. 2004a,b). Cloud water and cloud ice detrained at the cloud tops are considered as an additional source of cloud water/ice used by the grid-scale cloud microphysics (Wang et al. 2003). The iRAM results with double-moment cloud microphysics have been compared extensively to measurements from aircraft, ships, and satellites to evaluate the model performance simulating clouds over the eastern Pacific (Lauer et al. 2009). This evaluation showed that the model is able to simulate average cloud properties such as liquid water content, cloud droplet number concentration, cloud cover, and the radiative effect of clouds [also referred to as cloud radiative forcing (CRF)] that compare well with the observed climatology. Lauer et al. (2009) also showed that the diurnal cycle of cloud liquid water over the eastern Pacific is reasonably well simulated by iRAM. For additional details on iRAM, we refer to Wang (2001), Wang et al. (2004a), and the literature cited therein. Additional details on the double-moment cloud microphysics scheme and a model evaluation can be found in Phillips et al. (2007, 2008, 2009) and Lauer et al. (2009), respectively.

b. Model experiments In the present study, we performed two kinds of experiments: a simulation of January 1997–December 2008 using observed SSTs and lateral boundary conditions and a set of 10-yr integrations designed to simulate late twenty-first-century conditions. For the twenty-first-century experiments we apply what has been termed the ‘‘pseudo-global-warming’’ (PGW) method, which has been employed in other recent studies to downscale global climate change projections using a regional atmospheric model (Kimura and Kitoh 2007; Sato et al. 2007; Knutsen et al. 2008). In the PGW method, initial and lateral boundary conditions for the model integration are given by the sum of 6-h reanalysis data as used for the present-day experiment and a climate change signal based on results from a coupled global

5847

climate model (or an ensemble of such models). We based the climate change signals used in iRAM on the monthly averaged differences between present-day climate and projections for the end of the twenty-first century made by GCMs included in the IPCC Fourth Assessment Report (AR4). Specifically, the climate change signal adopted here was computed as the difference in 10-yr means for each calendar month for the late twentieth century [1990–99 in the AR4 twentieth-century forced runs (20C3M) and for the late twenty-first century 2090–99 in the Special Report of Emissions Scenarios (SRES) A1B runs]. Data were obtained from the World Climate Research Programme’s (WCRP’s) Coupled Model Intercomparison Project phase 3 (CMIP3) data archive (Meehl et al. 2007). Following the A1B scenario, we increased the CO2 concentration in iRAM from 370 ppm used in the present-day run to 720 ppm. Here, we only consider global warming perturbations of the meteorological boundary conditions [i.e., temperature, horizontal wind components, sea level pressure (SLP), and humidity] and of the SST. Concentrations of trace gases other than CO2 such as ozone or aerosols were not changed in our global warming simulations and remained at their present-day levels. Of course the global change signals differ quite significantly among the CMIP–AR4 GCMs. We performed three 10-yr experiments using 1999–2008 as the base and adding different climate change signals derived from the results of the CMIP–AR4 model simulations: 1) IPCC AR4 ensemble mean (case A): The climate change signal is averaged over all 19 AR4 models (see Table 1) that provided all data needed for specifying the climate change contribution to the boundary conditions in iRAM. All model results are interpolated to the 18 3 18 grid and 26 pressure levels of the FNL data before averaging. 2) Canadian Centre for Climate Modelling and Analysis (CCCma) Coupled General Circulation Model, version 3.1(T63) (CGCM3.1; case B): The climate change signal is obtained from simulations with CGCM3.1 (McFarlane et al. 1992; Flato 2005) of CCCma. Among the AR4 GCMs, the CGCM3.1 has one of the higher global climate sensitivities and also has a strong positive cloud–climate feedback over the eastern Pacific (see Fig. 9 below). 3) NCAR Community Climate System Model, version 3 (CCSM3; case C): The climate change signal is obtained from NCAR CCSM3 (Collins et al. 2006). Among the AR4 GCMs, the CCSM3 has one of the lower global climate sensitivities and also has a negative cloud–climate feedback over the eastern Pacific (see Fig. 9 below).

UKMO-HadGEM1

Parallel Climate Model version 1 (PCM1) UKMO-HadCM3

IPSL-CM4 MIROC3.2 high-resolution version (hires)

Instituto Nazionale di Geofisica e Vulcanologia SXG (INGV-SXG) INM-CM3.0

GFDL Climate Model version 2.0 (GFDL-CM2.0) GFDL-CM2.1 The Goddard Institute for Space Studies Model E-H (GISS-EH) GISS Model E-R (GISS-ER)

CSIRO-Mk3.0 CSIRO-Mk3.5 ECHAM5/MPI-OM Flexible Global Ocean–Atmosphere–Land System Model gridpoint version 1.0 (FGOALS-g1.0)

CNRM-CM3

CCSM3 Coupled General Circulation Model, version 2.3.2 (CGCM2.3.2) CGCM3.1(T63)

Model

1.98 3 1.98, L45

Me´te´o-France/Centre National de Recherches Me´te´orologiques, France CSIRO Atmospheric Research, Australia CSIRO Atmospheric Research, Australia Max Planck Institute for Meteorology, Germany National Key Laboratory of Numerical Modeling for Atmospheric Sciences and Geophysical Fluid Dynamics (LASG)–Institute of Atmospheric Physics, China U.S. Department of Commerce/NOAA/GFDL, United States U.S. Department of Commerce/NOAA/GFDL, United States National Aeronautics and Space Administration (NASA)–GISS, United States NASA–GISS, United States

L’institut Pierre-Simon Laplace, France Center for Climate System Research (University of Tokyo), National Institute for Environmental Studies, and Frontier Research Center for Global Change (JAMSTEC), Japan NCAR, United States Hadley Centre for Climate Prediction and Research/Met Office, United Kingdom Hadley Centre for Climate Prediction and Research/Met Office, United Kingdom

Institute for Numerical Mathematics, Russia

Instituto Nazionale di Geofisica e Vulcanologia, Italy

1.98 3 1.98, L31

CCCma, Canada

1.38 3 1.98, L38

2.88 3 2.88, L26 2.58 3 3.758, L19

2.58 3 3.758, L19 1.18 3 1.18, L56

58 3 48, L21

1.18 3 1.18, L19

48 3 58, L20

2.08 3 2.58, L24 2.08 3 2.58, L24 48 3 58, L20

1.98 3 1.98, L18 1.98 3 1.98, L18 1.98 3 1.98, L31 2.88 3 2.88, L26

1.48 3 1.48, L26 2.88 3 2.88, L30

Resolution (atmosphere)

NCAR, United States Meteorological Research Institute (MRI), Japan

Host institute

TABLE 1. The IPCC AR4 models providing data for this study (WCRP CMIP3 multimodel dataset; Meehl et al. 2007).

Washington et al. (2000) Gordon et al. (2000); Pope et al. (2000) Johns et al. (2006); Martin et al. (2006)

Alekseev et al. (1998); Galin et al. (2003) Hourdin et al. (2006) Hasumi and Emori (2004)

Anderson et al. (2004) Anderson et al. (2004) Schmidt et al. (2006); Bleck (2002) Schmidt et al. (2006); Russell et al. (1995) Roeckner et al. (1996)

Gordon et al. (2002) Gordon et al. (2002) Roeckner et al. (2003) Yu et al. (2004)

McFarlane et al. (1992); Flato (2005) Salas-Me´lia et al. (2005)

Collins et al. (2006) Yukimoto et al. (2006)

References

5848 JOURNAL OF CLIMATE VOLUME 23

1 NOVEMBER 2010

LAUER ET AL.

5849

FIG. 1. Annual average TOA shortwave cloud forcing for present-day conditions from 16 IPCC AR4 models (see Table 1) and iRAM compared with CERES satellite observations (Loeb et al. 2009).

The PGW method has some obvious limitations, notably the variability from daily to interannual periods in the boundary conditions is necessarily the same in the warming simulations and in the present-day simulation (e.g., Hara et al. 2008). We would also like to note that this study examines cloud response to a given climate change signal only. The usage of prescribed SSTs does not allow for possible atmosphere–ocean feedbacks. We expect that an interactive coupling of atmosphere and ocean could modify the climate change signals and may thus result in a different cloud response.

3. Comparison with observations In this section, we compare the cloud fields in our present-day iRAM simulation with observations. We compare 10-yr mean simulated values with observed climatology. We also evaluate the interannual variations in the simulation, which provides an opportunity to see how realistically the simulated clouds respond to changes in large-scale meteorological forcing. We also evaluate correlations between simulated low-level cloud amount and, for instance, sea surface temperatures, lowertropospheric stability (LTS), or 500-hPa vertical velocities.

These correlations are then compared to corresponding correlations obtained from observations.

a. Shortwave cloud forcing Shortwave cloud forcing (SCF) at the top of the atmosphere is calculated as the difference between all-sky and clear-sky shortwave radiation at the top of the atmosphere. Figure 1 shows a comparison of the multiyear annual average SCF from iRAM as well as from 16 IPCC AR4 models with multiyear (2000–05) satellite observations from CERES (Loeb et al. 2009). Observations show a large area of small (absolute) SCF south of the intertropical convergence zone (ITCZ) and between the western domain boundary at 1508W and about 1008W. These weakly negative SCF values correspond to a low average cloud amount over the warm-pool region. The size and extent of this area are reproduced by iRAM reasonably well, although the maximum SCF values are overestimated by about 5 W m22 compared with the CERES observations. The satellite data show the most negative SCF in our domain in the ITCZ and in the two stratocumulus decks off the coasts of North and South America. Here, iRAM overestimates the

5850

JOURNAL OF CLIMATE

VOLUME 23

TABLE 2. Multiyear annual averages of cloud properties from the iRAM present-day simulation and from satellite observations. Values given are domain averages (308S–308N, 1508–608W) over the ocean. The iRAM data used to calculate the averages are matched to the observational periods. Low cloud amount refers to clouds below 680 hPa.

iRAM Observations

SCF (W m22)

CFnet (W m22)

Total cloud amount (%)

Low cloud amount (%)

LWP (g m22)

Rain rate (mm day21)

253 246a

233 223a

48 58b

35 33b

63 62c

3.4 2.0d

a

CERES Energy Balanced and Filled climatology (EBAF) for the years 2000–05 (CERES EBAF TOA Terra Edition 1a dataset; Loeb et al. 2009). b ISCCP observations for the years 1999–2007 (Rossow et al. 1996). c University of Wisconsin climatology based on data from SSM/I, TMI, and AMSR-E for the years 1999–2007 (O’Dell et al. 2008). d TMI level 3 monthly 0.58 3 0.58 profiling (3A12) for the years 1999–2008 (Kummerow et al. 2001).

magnitude of the SCF in the ITCZ by about 25%, but the modeled SCF of the stratocumulus decks agrees well with the observations. However, the stratocumulus deck in iRAM over the southeastern (SE) Pacific is shifted by about 98 (;1000 km) northwestward compared with the observed, also seen in the simulated cloud liquid water and cloud cover. Lauer et al. (2009) speculated that this deficiency could be related to the model horizontal resolution, which leads to an overly smooth representation of the steep Andes. Table 2 summarizes the multiyear mean cloud properties averaged over the ocean region of the model domain. Specifically cloud forcing, cloud amount, liquid water path, and rain rate from iRAM are compared with satellite observations. The details of the satellite datasets used for comparison are given in the table along with relevant references. The model overpredicts the average SCF over the ocean by 7 W m22 and underpredicts the magnitude of the longwave cloud forcing (LCF) by 4 W m22 compared with satellite measurements. This overestimation in the magnitude of SCF is mainly caused by a too large (absolute) SCF in the ITCZ as well as an underprediction of the extent of the region of weak SCF associated with the warm pool. The small values of domain-averaged LCF in the model mainly reflect an underestimation of cirrus clouds in, and north of, the ITCZ. The difference between domain-averaged modeled and observed total cloud forcing (CFnet) is about 210 W m22. Figure 1 also shows that the 16 IPCC AR4 models investigated here, with the possible exception of the Met Office (UKMO) Hadley Centre Global Environmental Model version 1 (HadGEM1) model, fail to adequately reproduce the large area of small magnitude of SCF over the warm-pool region. Among the better models in reproducing observed SCF in the ITCZ and the two stratocumulus regions are the two Geophysical Fluid Dynamics Laboratory (GFDL) models that, however, fail to reproduce the extent and position of the stratocumulus deck in the southeastern Pacific.

b. Cloud amount, liquid water path, and rainfall The results in Table 2 show that the domain-averaged iRAM simulated low cloud cover (35%) is in good agreement with satellite observations (33%). Here, lowlevel cloud amount refers to clouds below 680 hPa. While low-level cloud amount can be averaged over all time steps in the model, the satellite data cover only periods not obstructed by high-level clouds. This is, however, not expected to be a problem as the satellite data show only small differences between low-level and total cloud amount in the stratocumulus regions, which are the main focus of this study. The domain-averaged total cloud cover in iRAM is 48%, rather less than the 58% from the satellite climatology, a difference that reflects the underprediction of the cirrus clouds in and north of the ITCZ in iRAM. The modeled multiyear annual average LWP over the ocean, 63 g m22, is in good agreement with 62 g m22 determined from satellite observations. Figure 2 shows the geographical distribution of the iRAM simulated LWP compared with observations and the 20C3M simulations in the IPCC GCMs. The iRAM captures the observed overall pattern reasonably well, whereas the IPCC models vary widely among themselves and none reproduces all the major features in the observed LWP. The domain-average rainfall rate in iRAM over the ocean is 3.4 mm day21, which is considerably higher than the 2.0 mm day21 in the satellite-based climatology. Li and Fu (2005) showed that the rainfall climatology from the Tropical Rainfall Measuring Mission (TRMM) satellite data has lower average rain rates than the Global Precipitation Climatology Project (GPCP; Huffman et al. 1997; Adler et al. 2003), particularly over the ocean. A comparison of the geographical pattern of long-term mean rain rate with satellite observations (not shown) reveals that iRAM captures the location and intensity of the rain rate over most of the ocean reasonably well, but the belt of heavy rainfall in the ITCZ is overestimated by iRAM in both its meridional extent and its intensity.

1 NOVEMBER 2010

LAUER ET AL.

5851

FIG. 2. As in Fig. 1, but for liquid water path. Satellite observations are from the University of Wisconsin climatology (UWisc) based on SSM/I, TMI, and AMSR-E (O’Dell et al. 2008).

c. Low-level cloud amount and lower-tropospheric stability Observations show that the low-level cloud cover in tropical and subtropical regions is strongly correlated with LTS, defined as the difference in potential temperatures u at 700 hPa and the surface (LTS 5 u700hPa – us) (Slingo 1980, 1987; Klein and Hartmann 1993). For the tropical and subtropical clouds over the eastern Pacific in our model domain we find very similar correlations between observed low-level cloud amount and LTSestimated inversion strength (Wood and Bretherton 2006). In this study, we use LTS for our analysis. We combined monthly mean ISCCP satellite observations of low-level cloud amount (Rossow et al. 1996) with LTS values calculated from monthly mean NCEP final analysis data (NCEP–NCAR reanalysis before the year 2000; Kalnay et al. 1996). The ISCCP data (2.58 3 2.58) are interpolated to the 18 3 18 grid of the FNL data. For comparison with results from iRAM we averaged the model data onto the 18 3 18 FNL grid. The black and blue dots in Fig. 3a show low-level cloud amount binned by LTS values from the combined ISCCP–NCEP dataset and from iRAM for the years 2000–07. The mean low-level

cloud amount and its standard deviation are calculated for all grid cells in the whole domain (308S–308N and 1508–608W) and all individual monthly means in the time period 2000–07 within the same LTS bin. The vertical bars show 61 standard deviation. The standard deviation within each LTS bin will reflect spatial, annual, and interannual variability. Observations and model show an approximately linear increase in low cloud amount with increasing LTS. Such a linear relationship between seasonal mean LTS and low-level cloud amount for regions in the subtropics has also been found by Klein and Hartmann (1993). A linear fit to the observations in Fig. 3a has a slope of 0.031 K21, which is close to the 0.030 K21 obtained for the present-day iRAM simulation. The range of variability of cloud amount for any given LTS value is larger in the model than in the observations. The difference even persists when the iRAM data are averaged to 2.58 resolution to be directly comparable to the ISCCP data. The corresponding probability density functions (PDFs) of monthly mean LTS from the present-day iRAM simulation and NCEP FNL data are shown as the blue and black curves in Fig. 3b. For both model and FNL data the most common LTS values are found to be in the range of 12–14 K. Maximum LTS densities are found

5852

JOURNAL OF CLIMATE

VOLUME 23

FIG. 3. (a) Low-level cloud cover vs LTS as derived from monthly mean ISCCP satellite observations (Rossow et al. 1996) and NCEP final analysis data for the years 2000–07, respectively, compared with results from iRAM for present-day and the global warming case A (IPCC AR4 ensemble mean). Filled circles depict mean values and bars show standard deviation. (b) PDF of the monthly mean LTS in the model domain (308S–308N, 1508–608W) from iRAM and NCEP final analysis data. (c) Changes in the LTS–low-level cloud amount relationship between the (normal) year 2005 and El Nin˜o year 1997 from iRAM and NCEP/ISCCP data. (d) Changes in the LTS–low-level cloud amount relationship between present-day and global warming case A. For details see text.

to be at 12.8 K for NCEP FNL data and 13.2 K for the iRAM present-day simulation. The model reproduces the LTS PDF from NCEP FNL data for the east Pacific region reasonably well, although the modeled distribution is wider than the observed one for small LTS values (LTS 5 10–12 K).

Figure 3c shows the changes in the LTS–low-level cloud amount relationship between the strong El Nin˜o year 1997 and the year 2005 (no El Nin˜o or La Nin˜a events) from iRAM and NCEP/ISCCP data. Average LTS and low-level cloud amount in the El Nin˜o year were calculated for the same grid cells and months that

1 NOVEMBER 2010

LAUER ET AL.

5853

FIG. 4. Time series of anomalies in (a) monthly mean low-level cloud amount, (b) liquid water path, (c) sea surface temperature, (d) lower-tropospheric stability, and (e) 500-hPa vertical velocity obtained from satellite observations and NCEP final analysis data (NCEP–NCAR reanalysis before 2000; Kalnay et al. 1996) compared with results from iRAM. The figure shows monthly means as well as 1-yr running means averaged over the southeastern Pacific (258–58S, 1008–758W). Warm and cold ENSO episodes are shaded in light red and light blue, respectively. For details see text.

were used to calculate low-level cloud amount for each of the year 2005 LTS bins. LTS decreases in the El Nin˜o year for all bins with year 2005 LTS values larger than 14 K in the NCEP data and larger than 12.5 K in the model data, respectively, while corresponding low-level cloud amount decreases particularly in the LTS range most relevant for stratocumulus clouds (LTS . 15 K). The displacement arrows from NCEP/ISCCP and iRAM are almost parallel within this LTS range, indicating that the model simulates the observed changes in the LTS– low-level cloud amount relationship during El Nin˜o reasonably well.

d. Interannual variations in low-level cloud amount and liquid water path To evaluate the response of the modeled low-level clouds to interannual variations in local thermal structure and circulation, we compare monthly mean cloud

anomalies with other observed properties in the stratocumulus regions off the coasts of North and South America. Cloud properties of primary interest are lowlevel cloud amount and liquid water path as these are key parameters determining the cloud radiative properties. SST and LTS reflect the local thermal structure, while 500-hPa vertical velocities are indicative of largescale circulation changes. Figure 4 shows time series of anomalies in monthly mean low-level cloud amount, liquid water path, SST, LTS, and 500-hPa vertical velocities from iRAM in comparison with observations. Monthly mean anomalies are calculated by subtracting the average seasonal cycle calculated over the entire period. Observed low-level cloud amounts are obtained from ISCCP satellite data (Rossow et al. 1996); liquid water path from the Special Sensor Microwave Imager (SSM/I), TRMM Microwave Imager (TMI), and AMSR for Earth Observing System (AMSR-E; O’Dell

5854

JOURNAL OF CLIMATE

et al. 2008); SSTs are taken from NOAA daily highresolution blended analyses (Reynolds et al. 2007); and LTS as well as 500-hPa vertical velocity are calculated from NCEP FNL data (NCEP–NCAR reanalysis before 2000; Kalnay et al. 1996). The time series cover the period 1997–2007 (for which we have satellite data for liquid water path and cloud cover) and are averaged over the southeastern Pacific stratocumulus region (258– 58S, 1008–758W, Fig. 4). The warm and cold ENSO episodes denoted by shading in Fig. 4 are based on observed Nin˜o-3.4 SST anomalies. Even though they are averaged over a large domain, the monthly mean anomalies still show significant variability. To reduce the noise introduced by the subseasonal variability we calculate 1-yr running means shown as thick curves in Fig. 4. SSTs in the southeastern Pacific region show strong positive deviations from average values during the 1997/98 El Nin˜o event and negative deviations from the average in the subsequent cold ENSO episode (Fig. 4c). From 2001 through 2006 SST anomalies are fairly small and it seems that the weak El Nin˜os of 2002/03, 2004/05, and late 2006 have at most small effects on low clouds in the southeastern Pacific. Both model and observations show a strong negative low-level cloud amount anomaly (28%) during the strong El Nin˜o event in 1997/98 and a small positive anomaly (2%–3%) in the years 2002–04 (Fig. 4a). During the rest of the time period 1997–2007, the 1-yr running mean anomalies of observed low-level cloud amount are small. This behavior is reproduced by iRAM except for a 15-month period following the strong El Nin˜o event of 1997/98 where the model predicts a positive low-level cloud amount anomaly of 4%. Anomalies in the liquid water path show a high positive correlation with low-level cloud amount anomalies (Fig. 4b). The top part of Table 3 shows correlations of the time series of the southeastern Pacific region mean value of low cloud amount with other quantities averaged over the same region (using the 1-yr running mean of all quantities). The correlations of observed and modeled low-level cloud amount and LWP are 0.85 and 0.89, respectively. SST is strongly anticorrelated with observed and modeled low-level cloud amount with correlations coefficients of 20.81 and 20.75, respectively. Figure 4d shows a comparison of modeled LTS anomalies with those calculated from NCEP data. The 1-yr running mean from iRAM agrees reasonably well with the FNL data, showing smaller than average LTS values between 1997 and 2001 and larger than average values thereafter. Earlier observational studies have shown that LTS is strongly correlated with subtropical low-level stratocumulus cloud fraction (Slingo 1980, 1987; Klein and Hartmann 1993). Correlation coefficients for the modeled and FNL LTS with low-level cloud amount over the

VOLUME 23

TABLE 3. Correlations of the 1997–2007 time series for low-level cloud amount from observations and calculated by iRAM with liquid water path, sea surface temperature, lower-tropospheric stability, 500-hPa vertical velocity, and sea level pressure. Quantities have been averaged over the southeast and northeast Pacific, and a 1-yr running mean has been applied to the time series. For details see text. Parameter

Observations

Southeast Pacific (258–58S, 1008–758W) LWP 0.85 SST 20.81 LTS 0.80 v500 0.37 SLP 0.69 Northeast Pacific (208–308N, 1208–1308W) LWP 0.66 SST 20.27 LTS 0.34 v500 0.23 SLP 0.27

iRAM 0.89 20.75 0.81 0.21 0.19 0.60 20.62 0.49 0.14 0.58

southeastern Pacific stratocumulus region are 0.81 and 0.80, respectively. By contrast, 500-hPa vertical velocities (Fig. 4e) have a much weaker correlation with low-level cloud amount anomalies over the east Pacific stratocumulus regions during the period studied here. Table 3 also gives the correlation coefficient with domain-averaged SLP, which is 0.69 in observations but much smaller (0.19) in the iRAM simulation. We repeated this analysis for the averages over the northeastern Pacific stratocumulus region (208–308N, 1208–1308W). In this region as well, there is a correlation of the SST with the ENSO state, notably with anomalously warm surface waters during the 1997/98 El Nin˜o and cold water during 1999. The correlation coefficients of low cloud amounts averaged over 208–308N, 1208– 1308W, with SST, LWP, LTS, SLP, and midtropospheric vertical velocity, are given in the bottom part of Table 3. These correlations are similar in the observations and in the iRAM simulation. The ability of iRAM to reproduce the interannual variations of cloud properties in stratocumulus regions through the ENSO cycle is much better than that of typical current coupled GCMs. Clement et al. (2009) showed that many global models have difficulties even in reproducing the correct sign of the correlations between cloud properties and meteorological quantities that we show in Table 3. Of course, the results in Fig. 4 and Table 3 involve iRAM run with prescribed SSTs and lateral boundary conditions and might be more directly comparable to prescribed SST GCM simulations than free-running coupled GCMs. However, the reality is that climate change experiments in AR4 have been performed with models that have poor representation of

1 NOVEMBER 2010

5855

LAUER ET AL.

FIG. 5. The 10-yr average change in (middle) low-level cloud amount (DCA), (bottom) cloud feedback parameters (l), and (top) the underlying global warming signals in sea surface temperatures (DSST) for (left to right) the three global warming cases A–C compared with present-day conditions (1999–2008). For details see text.

the mean cloud climatology in the eastern Pacific stratocumulus regions and do not reproduce the connections between tropical and subtropical clouds and large-scale meteorological variables (e.g., Stowasser and Hamilton 2006; Clement et al. 2009). The much better cloud representation for current climate in iRAM provides the motivation for conducting the climate change experiments described in the next section.

means for each case [i.e., including only the last 10 yr (1999–2008) of the control run]. Figure 5 shows changes in low-level cloud amount as well as the imposed changes in SST for all three global warming cases. Also shown is the local cloud feedback parameter l calculated as the change in net cloud forcing (CFnet) divided by the change in surface temperature Ts: l5

4. Global warming results a. iRAM global warming simulations We estimate the response of clouds to global warming by calculating the differences between each of the three global warming cases A–C (see section 2) and our present-day reference experiment. We compare 10-yr

DCFnet . DT s

(1)

The net cloud forcing is calculated as the sum of SCF and LCF, where SCF and LCF are calculated as the difference between the all-sky and clear-sky shortwave and longwave radiation at the top of the atmosphere, respectively. Negative values correspond to a cooling effect on the climate system. Although this definition

5856

JOURNAL OF CLIMATE

[Eq. (1)] of the cloud feedback parameter depends on changes in both cloud and clear-sky properties, such as changes in water vapor, temperature, or surface albedo (Soden et al. 2004), it is commonly used to diagnose global climate simulations because its calculation is straightforward and the cloud forcing defined in this way can be estimated in a fairly direct way from observations (Bony et al. 2006). The spatial structure of the late twenty-first-century SST warming patterns taken from the multimodel ensemble (case A), CGCM3.1 (case B), and CCSM3 (case C) are rather similar, but the overall magnitude of the warming differs quite significantly among the cases (largest for CGCM3.1, smallest for CCSM3). In each case, the largest warming occurs in the equatorial east Pacific and the smallest warming occurs in the southernmost part of our model domain between 208 and 308S. Changes in the amount of low-level marine clouds calculated by iRAM in response to the global warming signals (cases A–C) have similar geographical patterns showing a strong decrease of 5%–10%, particularly in the two stratocumulus regions, and an increase in lowlevel cloud amount in the range of 2%–8% over the equatorial Pacific between 1508 and 1008W. Consistent with the amplitudes of the imposed global warming signals, case B shows the largest decrease in low-level cloud amount in both horizontal extent and amplitude, whereas case C has the smallest cloud response. The local cloud feedback parameters [Eq. (1)] are shown in the bottom panels of Fig. 5 and basically scale the cloud changes (specifically in shortwave cloud forcing) by the imposed SST changes. The feedback parameters are quite similar in cases A–C and are in the range of 4–7 W m22 K21 in the stratocumulus regions, 22 to 24 W m22 K21 over the equatorial Pacific between 1508 and 1008W, and about 1 W m22 K21 over much of the rest of the Pacific. Clement et al. (2009) estimate a warming effect from changes in net cloud forcing because of changes in SST in the northeast Pacific stratocumulus region (158–258N, 1158–1458W) of about 6 W m22 K21. This observation-based estimate compares reasonably well to results from iRAM ranging between 4.2 and 5.9 W m22 K21 averaged over the same region (global warming cases A–C). As noted above, there is a strong similarity in the feedback parameters among the cases A–C, despite the different warming increments imposed in SST. However, it is possible that the close agreement in l may depend on the overall geographic pattern of SST warming being similar among the three cases. This issue was investigated in a fourth experiment in which a uniform 2-K warming was applied to the sea surface throughout the domain and through the depth of the atmosphere

VOLUME 23

on the lateral boundaries. The l distribution in that experiment (not shown) was indeed rather different from that seen in cases A–C (the domain-average feedback in this uniform warming case was 3 W m22 K21). The red dots in Fig. 3a show the dependence of lowlevel cloud amount on LTS in the global warming simulation case A. The mean relation between LTS and low-level cloud amount is significantly different in the perturbed climate from that in the control run. Specifically, in the warmer climate there are systematically smaller low-level cloud amounts for any given LTS value, except for a narrow region around LTS 5 14.5 K. Also the slope of the linear fit to results from the global warming scenario (0.025 K21) is somewhat smaller than that for the present-day model results or from observations. The model results suggest that the average relation between LTS and low-level cloud amount obtained from present-day observations over the east Pacific can change significantly in an altered climate. Application of simple models (e.g., Miller 1997) or parameterizations of the boundary layer cloud amount based on the observed present-day relation between LTS and low-level cloud cover may not be appropriate for climate change scenarios. The red curve in Fig. 3b shows the LTS PDF in the global warming simulation case A—compared with the present-day result there is a shift toward higher LTS values. Figure 3d shows the changes in the LTS–lowlevel cloud amount relationship between our presentday simulation and global warming case A. As for the 1997/98 El Nin˜o case discussed above (Fig. 3c), average LTS and low-level cloud amount in the global warming case were calculated for the same model grid cells and months that were used to calculate low-level cloud amount for each of the present-day LTS bins. In other words, the blue dots for the present-day simulation in Fig. 3d are identical with the ones in Fig. 3a, whereas the red dots show LTS and low-level cloud amount values averaged over the same model grid cells but for the global warming case A. LTS increases in the global warming scenario for all bins while low-level cloud amount decreases for bins with present-day LTS values smaller than 13 K or larger than 16 K. However, lowlevel cloud amount remains close to its present-day level for the bins in the LTS range 13–16 K. Here, the increase in low-level cloud amount in the equatorial region between 1508 and 1008W (see Fig. 5) balances approximately the decrease in low-level cloud amount within the same LTS range over other parts of the ocean. The general increase in average LTS is consistent with the reduction in mean tropospheric lapse rates and increased dry stability, which are robust predictions from current GCMs in a warmed climate.

1 NOVEMBER 2010

LAUER ET AL.

5857

FIG. 6. The 10-yr mean diurnal cycle of cloud-base and cloud-top height averaged over NE and SE Pacific for the present-day simulation and global warming case A (IPCC AR4 ensemble mean).

Figure 6 shows the mean diurnal cycles of cloudbottom and cloud-top heights for the core regions of the stratocumulus regimes over the northeastern (208– 308N, 1208–1308W) and southeastern Pacific (258–58S, 858–958W). We define cloud-bottom height as the lowest level between the surface and 4 km at which the monthly mean cloud liquid water content exceeds 0.025 g kg21 and cloud-top height as the highest level at which LWC falls below this threshold value. The results shown have been averaged over the whole 10-yr period of the presentday simulation and the global warming scenario case A. While the cloud-bottom heights in the stratocumulus regions change only modestly, the average cloud-top heights in the global warming case (A) are about 50– 100 m lower compared with those of the present-day scenario. The vertical model resolution in the vicinity of the cloud tops over the northeastern Pacific is about 200 m and over the southeastern Pacific about 300 m. The overall thinning of the stratus cloud in the global warming case is consistent with the reduction in cloud shortwave forcing. To provide the thermal structure context for the cloud changes, an analysis of the vertical temperature profile and lapse rate in the iRAM experiments is conducted. Conventional averaging with the vertical coordinate fixed in time and horizontal space will blur any marked feature of the vertical structure that is strongly variable in time (and horizontal space; Birner 2006). This blur effect makes the temperature inversion atop the MBL in the east Pacific region hard to see in a multiyear climatology. Following the approach adopted by Birner (2006) to characterize behavior near the tropopause, we use the inversion layer base height as a common reference level to composite all temperature profiles. This is done by introducing a modified vertical coordinate defined as z 2 zB with z as altitude and zB as inversion layer base height. The data are interpolated from model levels onto vertical levels in z 2 zB with a cubic spline interpolation.

Profiles that do not contain a temperature inversion in the lower troposphere (0–3 km) are not included in the average. The inversion layer base height is taken as the minimum temperature in this altitude range calculated from daily mean temperature profiles. Figure 7 (left panel) shows the results for the composited temperature profiles. We added the average of the inversion layer base height to the vertical coordinates shown in Fig. 7. The right panel of Fig. 7 shows the corresponding lapse rates (2dT/dz). The mean inversion heights in the presentday iRAM simulation are about 1.4 km in the southeast Pacific region and 0.7 km in the northeast Pacific region. These are somewhat lower than the mean cloud-top heights presented in Fig. 6 (1.6 km and 1.0 km) as the mean heights for the clouds were computed including occasions when there is no well-defined inversion and also depend on the threshold value for monthly mean LWC (see above). The drop in diurnal mean cloud heights by about 100 m in the southeast Pacific and 50 m in the northeast Pacific (Fig. 6) in the global warming simulation is paralleled by the very similar reductions in the mean inversion heights (Fig. 7). In contrast, our sensitivity experiment with a uniform 2 K increase in SST throughout the domain and in atmospheric temperatures on the lateral boundaries shows only a little change or slight increase in inversion layer base heights in the stratocumulus regions. This suggests that the reduction in mean tropospheric lapse rates predicted by the GCMs in a warmed climate (cases A–C) is important for the shallowing of the marine boundary layer in these regions. Analysis of the entrainment rates at the top of the boundary layer shows that entrainment in the global warming run is reduced by 9% in the northeast and by 12% in the southeast Pacific stratocumulus region compared with the present-day simulation. Reduced entrainment could be a reason for the reduction in boundary layer height (e.g., Stevens 2006), causing the inversion

5858

JOURNAL OF CLIMATE

VOLUME 23

FIG. 7. The 10-yr mean vertical profiles of (left) T and (right) lapse rate (2dT/dz) averaged over NE and SE Pacific for the present-day simulation and global warming case A (IPCC AR4 ensemble mean). Temperature profiles have been aligned at the inversion layer base height prior to averaging. The arrows indicate the average inversion layer base height.

to drop and the clouds to thin. Consistent with Caldwell and Bretherton’s (2009) hypothesis that a decreased radiative cooling of the boundary layer in an enhanced greenhouse case could cause the inversion to drop, we find average turbulent kinetic energy (TKE) is less in our global warming run in both stratocumulus regions. In addition to less turbulence, entrainment could also be reduced by greater inversion strength in the global warming case particularly in the southeastern Pacific stratocumulus region (see also Fig. 7).

b. Comparison with IPCC model results We calculated the local feedback parameters for all 16 IPCC AR4 models that provided both TOA clear-sky and all-sky fluxes needed to compute TOA cloud forcings [Eq. (1)]. Just as for the other aspects of the global warming signal (see section 2), we calculate the change in cloud forcing due to global warming by subtracting 10-yr averages for the present-day simulation (experiment 20C3M, years 1990–99) from projections for the end of the twenty-first century (SRES scenario A1B years 2090–99). Figure 8 shows a comparison of l from IPCC AR4 models with the results from iRAM for global warming case A. The geographical patterns as well as the amplitudes of the local feedback parameters vary widely among the IPCC AR4 models. Of the 16 IPCC models, six of them [Centre National de Recherches Me´te´orologiques Coupled Global Climate Model, version 3 (CNRM-CM3); Institute of Numerical Mathematics Coupled Model, version 3.0 (INM-CM3.0); L’Institut Pierre-Simon Laplace Coupled Model, version 4

(IPSL CM4); ECHAM5–Max Planck Institute Ocean Model (MPI-OM); the third climate configuration of the Met Office Unified Model (UKMO HadCM3); and UKMO HadGEM1] simulate fairly strong positive local feedback parameters throughout most of the stratocumulus regions. By contrast, the Commonwealth Scientific and Industrial Research Organisation Mark version 3.5 (CSIRO-Mk3.5) and CCSM3 simulate fairly strong negative local feedback parameters in the two stratocumulus regions. The other IPCC models have feedback parameters in the stratocumulus areas that are either quite small (CGCM3) or that vary in sign through these regions. In other parts of the domain shown in Fig. 8 the simulated local feedback parameter differs greatly among the IPCC models. While the IPCC models disagree widely among themselves, none of the GCM simulated patterns of l compare well with that in the iRAM simulation. The domain-averaged l from iRAM is 1.8 W m22 K21 for global warming case A (1.9 W m22 K21 when averaged over ocean grid cells only). This positive feedback parameter mainly reflects the decrease in shortwave cloud forcing resulting from decreased low-level cloud amount and liquid water path. The response of clouds to global warming in cases B and C gives similar domain-averaged local feedback parameters of 1.8 and 1.9 W m22 K21 (2.1 and 2.0 when averaged over ocean grid cells only), respectively, even though the amplitude of the global warming signals varies significantly among these cases. The domain-averaged changes in shortwave and net cloud forcing as well as the local feedback parameters for all global warming cases are summarized in Table 4.

1 NOVEMBER 2010

5859

LAUER ET AL.

FIG. 8. Local feedback parameter l [Eq. (1)] from 16 IPCC AR4 models and iRAM calculated from the changes in shortwave TOA cloud forcing and surface temperatures between 10-yr means for present-day conditions and SRES A1B end of the twenty-first-century simulations. For details see text.

The light gray bars in Fig. 9 compare the local feedback parameters from the IPCC AR4 models averaged over the domain of the regional model (308S–308N, 1508–608W) with iRAM. The domain-averaged feedback parameter simulated by iRAM is higher than that simulated by any of the 16 IPCC AR4 models. Out of the 16 IPCC models, 10 simulate positive feedback parameters for the east Pacific region, and 6 predict a negative domain-averaged feedback parameter. The dark gray bars in Fig. 9 show the feedback parameter for each of the IPCC models averaged over the entire tropical– subtropical belt (308S–308N, 08–3608). The mean feedbacks in the entire tropical–subtropical belt in each model are fairly closely related to those for the east Pacific

domain (the correlation coefficient over the 16 models is 0.95). The east Pacific cloud feedbacks in the GCMs also correlate reasonably well with the equilibrium global climate sensitivities given in Table 8.2 of Solomon et al. (2007). The GCMs that have the highest east Pacific cloud feedback (and hence are closest to the iRAM result) are the Model for Interdisciplinary Research on Climate 3.2 (MIROC3.2), IPSL-CM4, and UKMO-HadGEM1. These (along with the medium-resolution version of MIROC3.2 not considered in this paper) are the GCMs with the highest global climate sensitivity according to the IPCC Table 8.2. It may also be noted that UKMO-HadGEM1 was identified by Clement et al. (2009) as the GCM that

TABLE 4. Changes in the 10-yr averages of SST, SCF, CFnet, and l for the individual global warming studies from iRAM compared with the present-day run. The values given are averaged over all ocean grid cells within the model domain (308S–308N, 1508–608W).

Global warming experiment

DSST (K)

DSCF (W m22)

DCFnet (W m22)

l (W m22 K21)

Case A—Global warming, IPCC AR4 ensemble mean Case B—Global warming, CCCma CGCM3.1(T63) (positive cloud feedback) Case C—Global warming, NCAR CCSM3 (negative cloud feedback)

2.0 2.2 1.7

5.0 5.2 5.4

4.1 4.6 3.6

1.9 2.0 2.1

5860

JOURNAL OF CLIMATE

VOLUME 23

FIG. 9. Annual average local feedback parameter l [Eq. (1)] in W m22 K21 for the east Pacific region and the latitude belt 308S–308N from 16 IPCC AR4 models and calculated by iRAM.

had the most realistic cloud responses to variations in the large-scale environment.

5. Summary and conclusions We have examined the cloud simulations and cloud– climate feedbacks in the tropical and subtropical eastern Pacific region in 16 state-of-the-art coupled GCMs and in the regional atmospheric model iRAM using prescribed boundary conditions. We find that the simulation of the mean cloud climatology for this region in the GCMs is very poor. The cloud feedbacks to imposed climate forcings vary widely among the GCMs in the east Pacific and in the 308N–308S band in general. These variations account for a large fraction of the uncertainty in global climate sensitivity. Following Lauer et al. (2009), we have found that iRAM forced with observed boundary conditions simulates rather realistic mean cloud fields in the east Pacific domain. Going beyond the earlier analysis of Lauer et al., we have also shown that the iRAM reproduces the observed interannual variations in cloud fields (as well as LTS), notably correctly simulating the response of the clouds through the 1997–99 El Nin˜o to La Nin˜a transition. By contrast, Clement et al. (2009) note that low clouds in GCMs generally do not respond realistically through the ENSO cycle. To investigate cloud feedbacks in iRAM, three global warming scenarios have been run with SSTs and horizontal boundary conditions meant to be appropriate for late twenty-first-century conditions; specifically, warming signals based on IPCC AR4 SRES A1B simulations from 1) an ensemble mean of 19 GCMs, 2) the CGCM3.1 model, and 3) the CCSM3 model. All three global warming cases simulated with iRAM show a distinct reduction in low-level cloud amount particularly in the stratocumulus regime, resulting in positive

local feedback parameters in these regions in the range of 4–7 W m22 K21. The model results suggest that the reduction in stratocumulus clouds because of global warming is caused by a drop in average inversion layer base height and a consequential decrease in cloud-top height. As the cloud-base height remains approximately unchanged the decrease in cloud-top height causes the stratocumulus clouds to thin and liquid water path to decrease. This results in a less efficient reflection of solar radiation and a reduction in shortwave cloud forcing— domain-averaged feedback parameters from iRAM range between 1.8 and 1.9 W m22 K21 (cases A–C). We have analyzed the relation between monthly mean low-level cloud cover and LTS in our iRAM simulations. The present-day simulation reproduces quite well the long-term mean relation in observations (NCEP data and satellite cloud retrievals). In both present-day iRAM simulation and observations, the El Nin˜o perturbations in cloud cover are largely accounted for by the reduction in LTS. By contrast, in the global warming simulation the clouds and thermal structure change in such a way that the cloud cover versus LTS relation is significantly different from the present-day simulation. This suggests that the decrease in low-level cloud amount during the 1997/98 El Nin˜o, and the decrease because of global warming by doubled CO2, is controlled by different physical processes as proposed by Zhu et al. (2007). Furthermore, this shows rather dramatically the inadequacy of cloud parameterization schemes based purely on present-day empirical relations between cloud cover and large-scale environmental fields. The cloud–climate feedback averaged over the east Pacific region has also been calculated from SRES A1B simulations for 16 AR4 GCMs. The GCM feedbacks vary from 21.0 to 11.3 W m22 K21, which are all less than the 11.8 to 11.9 W m22 K21 obtained in the comparable iRAM simulations. The iRAM results by themselves

1 NOVEMBER 2010

LAUER ET AL.

cannot be connected definitively to global climate feedbacks, but we have shown that among the GCMs the cloud feedbacks averaged over 308S–308N and the equilibrium global climate sensitivity are both correlated strongly with the east Pacific cloud feedback. To the extent that iRAM results for cloud feedbacks in the east Pacific are credible, they provide support for the high end of current estimates of global climate sensitivity. Acknowledgments. This research was supported by the Japan Agency for Marine-Earth Science and Technology (JAMSTEC), by NASA through Grant NNX07AG53G, and by NOAA through Grant NA09OAR4320075, which sponsor research at the International Pacific Research Center. This research was also supported by NOAA/ CPPA Grant NA07OAR4310257 and DOE Regional and Global Climate Modeling (RCGM) Program Grant ER64840. NCEP FNL data for this study are from the Research Data Archive (RDA), which is maintained by CISL at NCAR. NCAR is sponsored by the National Science Foundation (NSF). NCEP–NCAR reanalysis data have been provided by the NOAA/OAR/ESRL PSD, Boulder, Colorado, available online from their Web site at http://www.cdc.noaa.gov/. We acknowledge the modeling groups, the Program for Climate Model Diagnosis and Intercomparison (PCMDI) and the WCRP’s Working Group on Coupled Modelling (WGCM) for their roles in making available the WCRP CMIP3 multimodel dataset. Support of this dataset is provided by the Office of Science, DOE. REFERENCES Adler, R. F., and Coauthors, 2003: The version-2 Global Precipitation Climatology Project (GPCP) monthly precipitation analysis (1979–present). J. Hydrometeor., 4, 1147–1167. Albrecht, B. A., D. A. Randall, and S. Nicholls, 1988: Observations of marine stratocumulus during FIRE. Bull. Amer. Meteor. Soc., 69, 618–626. Alekseev, V. A., E. M. Volodin, V. Ya. Galin, V. P. Dymnikov, and V. N. Lykossov, 1998: Modelling of the Present-Day Climate by the RAS Atmospheric Model ‘‘DNM GCM.’’ Institute of Numerical Mathematics, 200 pp. Anderson, J. L., and Coauthors, 2004: The new GFDL global atmosphere and land model AM2-LM2: Evaluation with prescribed SST simulations. J. Climate, 17, 4641–4673. Birner, T., 2006: Fine-scale structure of the extratropical tropopause region. J. Geophys. Res., 111, D04104, doi:10.1029/ 2005JD006301. Bleck, R., 2002: An oceanic general circulation model framed in hybrid isopycnic-Cartesian coordinates. Ocean Modell., 4, 55–88. Bony, S., and J.-L. Dufresne, 2005: Marine boundary layer clouds at the heart of tropical cloud feedback uncertainties in climate models. Geophys. Res. Lett., 32, L20806, doi:10.1029/ 2005GL023851. ——, and Coauthors, 2006: How well do we understand and evaluate climate change feedback processes? J. Climate, 19, 3445–3482.

5861

Bretherton, C. S., and Coauthors, 2004: The EPIC 2001 stratocumulus study. Bull. Amer. Meteor. Soc., 85, 967–977. Caldwell, P., and C. S. Bretherton, 2009: Response of a subtropical stratocumulus-capped mixed layer to climate and aerosol changes. J. Climate, 22, 20–38. Chang, F. L., and J. A. Coakley, 2007: Relationships between marine stratus cloud optical depth and temperature: Inferences from AVHRR observations. J. Climate, 20, 2022– 2036. Clement, A. C., R. Burgmann, and J. R. Norris, 2009: Observational and model evidence for positive low-level cloud feedback. Science, 325, 460–464. Collins, W. D., and Coauthors, 2006: The Community Climate System Model: CCSM3. J. Climate, 19, 2122–2143. Edwards, J. M., and A. Slingo, 1996: Studies with a flexible new radiation code. I: Choosing a configuration for a large-scale model. Quart. J. Roy. Meteor. Soc., 122, 689–719. Eitzen, Z. A., K.-M. Xu, and T. Wong, 2008: Statistical analyses of satellite cloud object data from CERES. Part V: Relationships between physical properties of boundary layer clouds. J. Climate, 21, 6668–6688. Feigelson, E. M., 1978: Preliminary radiation model of a cloudy atmosphere. Part I: Structure of clouds and solar radiation. Beitr. Phys. Atmos., 51, 203–229. Flato, G. M., cited 2005: The third generation Coupled Global Climate Model (CGCM3). [Available online at http://www.ec. gc.ca/ccmac-cccma/default.asp?lang=En&n=1299529F-1.] Galin, V. Ya., E. M. Volodin, and S. P. Smyshliaev, 2003: Atmosphere general circulation model of INM RAS with ozone dynamics. Russ. Meteor. Hydrol., 5, 13–22. Gordon, C., C. Cooper, C. A. Senior, H. T. Banks, J. M. Gregory, T. C. Johns, J. F. B. Mitchell, and R. A. Wood, 2000: The simulation of SST, sea ice extents and ocean heat transports in a version of the Hadley Centre coupled model without flux adjustments. Climate Dyn., 16, 147–168. Gordon, H. B., and Coauthors, cited 2002: The CSIRO Mk3 climate system model. [Available online at http://www.cmar. csiro.au/e-print/open/gordon_2002a.pdf.] Greenwald, T. J., G. L. Stephens, S. A. Christopher, and T. H. Vonder Haar, 1995: Observations of the global characteristics and regional radiative effects of marine cloud liquid water. J. Climate, 8, 2928–2946. Gregory, D., J.-J. Morcrette, C. Jakob, A. C. M. Beljaars, and T. Stockdale, 2000: Revision of convection, radiation and cloud schemes in the ECMWF integrated forecasting system. Quart. J. Roy. Meteor. Soc., 126, 1685–1710. Hallet, J., and S. C. Mossop, 1974: Production of secondary ice particles during the riming process. Nature, 249, 26–28. Hara, M., T. Yoshikane, H. Kawase, and F. Kimura, 2008: Estimation of the impact of global warming on snow depth in Japan by the pseudo-global-warming method. Hydrol. Res. Lett., 2, 61–64. Hartmann, D. L., and D. Doelling, 1991: On the net radiative effectiveness of clouds. J. Geophys. Res., 96, 869–891. Hasumi, H., and S. Emori, Eds., 2004: K-1 coupled model (MIROC) description. Center for Climate System Research K-1 Tech. Rep. 1, 34 pp. Houghton, J. T., Y. Ding, D. J. Griggs, M. Noguer, P. J. van der Linden, X. Dai, K. Maskell, and C. A. Johnson, Eds., 2001: Climate Change 2001: The Scientific Basis. Cambridge University Press, 881 pp. Hourdin, F., and Coauthors, 2006: The LMDZ4 general circulation model: Climate performance and sensitivity to parameterized

5862

JOURNAL OF CLIMATE

physics with emphasis on tropical convection. Climate Dyn., 27, 787–813. Huffman, G. J., and Coauthors, 1997: The Global Precipitation Climatology Project (GPCP) combined precipitation dataset. Bull. Amer. Meteor. Soc., 78, 5–20. Johns, T. C., and Coauthors, 2006: The new Hadley Centre climate model HadGEM1: Evaluation of coupled simulations. J. Climate, 19, 1327–1353. Kalnay, E., and Coauthors, 1996: The NCEP/NCAR 40-Year Reanalysis Project. Bull. Amer. Meteor. Soc., 77, 437–471. Kimura, F., and A. Kitoh, 2007: Downscaling by pseudo–global warming method. ICCAP, 43–46. Klein, S. A., and D. L. Hartmann, 1993: The seasonal cycle of lowcloud properties and meteorological parameters in the subtropical trade wind boundary layer. J. Climate, 6, 1588–1606. Knutsen, T. R., J. J. Sirutus, S. T. Garner, G. A. Vecchi, and I. M. Held, 2008: Simulated reduction in Atlantic hurricane frequency under twenty-first century warming conditions. Nat. Geosci., 1, 359–364. Koop, T., B. Luo, A. Tsias, and T. Peter, 2000: Water activity as the determinant for homogeneous ice nucleation in aqueous solutions. Nature, 406, 611–614. Kummerow, C., and Coauthors, 2001: The evolution of the Goddard profiling algorithm (GPROF) for rainfall estimation from passive microwave sensors. J. Appl. Meteor., 40, 1801– 1820. Lauer, A., Y. Wang, V. T. J. Phillips, C. S. McNaughton, R. Bennartz, and A. D. Clarke, 2009: Simulating marine boundary layer clouds over the eastern Pacific in a regional climate model with double-moment cloud microphysics. J. Geophys. Res., 114, D21205, doi:10.1029/2009JD012201. Li, R., and Y. Fu, 2005: Tropical precipitation estimated by GPCP and TRMM PR observations. Adv. Atmos. Sci., 22, 852–864. Liu, Y., P. H. Daum, R. L. McGraw, M. A. Miller, and S. Niu, 2007: Theoretical expression for the autoconversion rate of the cloud droplet number concentration. Geophys. Res. Lett., 34, L16821, doi:10.1029/2007GL030389. Loeb, N. G., B. A. Wielicki, D. R. Doelling, G. L. Smith, D. F. Keyes, S. Kato, N. Manalo-Smith, and T. Wong, 2009: Towards optimal closure of the earth’s top-of-atmosphere radiation budget. J. Climate, 21, 748–766. Martin, G. M., M. A. Ringer, V. D. Pope, A. Jones, C. Dearden, and T. J. Hinton, 2006: The physical properties of the atmosphere in the new Hadley Centre Global Environmental Model, HadGEM1. Part 1: Model description and global climatology. J. Climate, 19, 1274–1301. McFarlane, N. A., G. J. Boer, J.-P. Blanchet, and M. Lazare, 1992: The Canadian Climate Centre second-generation general circulation model and its equilibrium climate. J. Climate, 5, 1013–1044. Medeiros, B., and B. Stevens, 2009: Revealing differences in GCM representations of low clouds. Climate Dyn., doi:10.1007/ s00382-009-0694-5, in press. ——, ——, I. M. Held, M. Zhao, D. L. Williamson, J. G. Olson, and C. S. Bretherton, 2008: Aquaplanets, climate sensitivity, and low clouds. J. Climate, 21, 4974–4991. Meehl, G. A., C. Covey, T. Delworth, M. Latif, B. McAvaney, J. F. B. Mitchell, R. J. Stouffer, and K. E. Taylor, 2007: The WCRP CMIP3 multimodel dataset: A new era in climate change research. Bull. Amer. Meteor. Soc., 88, 1383–1394. Miller, R. L., 1997: Tropical thermostats and low cloud cover. J. Climate, 10, 409–440.

VOLUME 23

Ming, Y., V. Ramaswamy, L. J. Donner, and V. T. J. Phillips, 2006: A new parameterization of cloud droplet activation applicable to general circulation models. J. Atmos. Sci., 63, 1348–1356. O’Dell, C. W., F. J. Wentz, and R. Bennartz, 2008: Cloud liquid water path from satellite-based passive microwave observations: A new climatology over the global oceans. J. Climate, 21, 1721–1739. Petters, M. D., and S. M. Kreidenweis, 2007: A single parameter representation of hygroscopic growth and cloud condensation nucleus activity. Atmos. Chem. Phys., 7, 1961–1971. Phillips, V. T. J., L. J. Donner, and S. T. Garner, 2007: Nucleation processes in deep convection simulated by a cloud-systemresolving model with double-moment bulk microphysics. J. Atmos. Sci., 64, 738–761. ——, P. J. DeMott, and C. Andronache, 2008: An empirical parameterization of heterogeneous ice nucleation for multiple chemical species of aerosol. J. Atmos. Sci., 65, 2757–2783. ——, and Coauthors, 2009: Potential impacts from biological aerosols on ensembles of continental clouds simulated numerically. Biogeosciences, 6, 987–1014. Pope, V., M. L. Gallani, P. R. Rowntree, and R. A. Stratton, 2000: The impact of new physical parameterizations in the Hadley Centre climate model: HadAM3. Climate Dyn., 16, 123–146. Randall, D. A., J. A. Coakley, C. W. Fairall, R. A. Kropfli, and D. H. Lenschow, 1984: Outlook for research on subtropical marine stratiform clouds. Bull. Amer. Meteor. Soc., 65, 1290– 1301. Reynolds, R. W., T. M. Smith, C. Liu, D. B. Chelton, K. S. Casey, and M. S. Schlax, 2007: Daily high-resolution-blended analyses for sea surface temperature. J. Climate, 20, 5473–5496. Roeckner, E., U. Schlese, J. Biercamp, and P. Loewe, 1987: Cloud optical depth feedbacks and climate modeling. Nature, 329, 138–140. ——, and Coauthors, 1996: The atmospheric general circulation model ECHAM4: Model description and simulation of presentday climate. Max Planck Institute Rep. 218, 90 pp. ——, and Coauthors, 2003: The atmospheric general circulation model ECHAM5. Part I: Model description. Max Planck Institute Rep. 349, 127 pp. Rossow, W. B., A. W. Walker, D. E. Beuschel, and M. D. Roiter, 1996: International Cloud Climatology Project (ISCCP) documentation of new cloud datasets. World Meteorological Organization Rep. WMO/TD-No. 737, 115 pp. Russell, G. L., J. R. Miller, and D. Rind, 1995: A coupled atmosphere–ocean model for transient climate change studies. Atmos.–Ocean, 33, 683–730. Salas-Me´lia, D., F. Chauvin, M. De´que´, H. Douville, J. F. Gueremy, P. Marquet, S. Planton, J. F. Royer, and S. Tyteca, 2005: Description and validation of the CNRM-CM3 global coupled model. Centre National de Recherches Me´te´orologiques Working Note 103, 36 pp. Sato, T., F. Kimura, and A. Kitoh, 2007: Projection of global warming onto regional precipitation over Mongolia using a regional climate model. J. Hydrol., 333, 144–154. Schmidt, G. A., and Coauthors, 2006: Present-day atmospheric simulations using GISS ModelE: Comparison to in-situ, satellite and reanalysis data. J. Climate, 19, 153–192. Slingo, J. M., 1980: A cloud parameterization scheme derived from GATE data for use with a numerical model. Quart. J. Roy. Meteor. Soc., 106, 747–770. ——, 1987: The development and verification of a cloud prediction scheme for the ECMWF model. Quart. J. Roy. Meteor. Soc., 113, 899–927.

1 NOVEMBER 2010

LAUER ET AL.

Soden, B. J., A. J. Broccoli, and R. S. Hemler, 2004: On the use of cloud forcing to estimate cloud feedback. J. Climate, 17, 3661– 3665. Solomon, S., D. Qin, M. Manning, M. Marquis, K. Averyt, M. M. B. Tignor, H. L. Miller Jr., and Z. Chen, Eds., 2007: Climate Change 2007: The Physical Science Basis. Cambridge University Press, 996 pp. Somerville, R. C. J., and L. A. Remer, 1984: Cloud optical thickness feedbacks in the CO2 climate problem. J. Geophys. Res., 89, 1419–1427. Stevens, B., 2006: Bulk boundary-layer concepts for simplified models of tropical dynamics. Theor. Comput. Fluid Dyn., 20, 279–304. Stowasser, M., and K. Hamilton, 2006: Relationship between shortwave cloud radiative forcing and local meteorological variables compared in observations and several global climate models. J. Climate, 19, 4344–4359. ——, ——, and G. J. Boer, 2006: Local and global climate feedbacks in models with different climate sensitivities. J. Climate, 19, 193–209. Sun, Z., and L. Rikus, 1999: Improved application of exponential sum fitting transmissions to an inhomogeneous atmosphere. J. Geophys. Res., 104, 6291–6303. Tiedtke, M., 1989: A comprehensive mass flux scheme for cumulus parameterization in large-scale models. Mon. Wea. Rev., 117, 1779–1800. Tselioudis, G., W. B. Rossow, and D. Rind, 1992: Global patterns of cloud optical thickness variation with temperature. J. Climate, 5, 1484–1495. ——, A. D. Del Genio, W. Kovari Jr., and M.-S. Yao, 1998: Temperature dependence of low cloud optical thickness in the GISS GCM: Contributing mechanisms and climate implications. J. Climate, 11, 3268–3281. Wang, Y., 2001: An explicit simulation of tropical cyclones with a triply nested movable mesh primitive equation-model— TCM3. Part I: Model description and control experiment. Mon. Wea. Rev., 129, 1370–1394. ——, O. L. Sen, and B. Wang, 2003: A highly resolved regional climate model (IPRC-RegCM) and its simulation of the 1998 severe precipitation events over China. Part I: Model description and verification of simulation. J. Climate, 16, 1721–1738.

5863

——, S.-P. Xie, H. Xu, and B. Wang, 2004a: Regional model simulations of marine boundary layer clouds over the southeast Pacific off South America. Part I: Control experiment. Mon. Wea. Rev., 132, 274–296. ——, H. Xu, and S.-P. Xie, 2004b: Regional model simulations of marine boundary layer clouds over the southeast Pacific off South America. Part II: Sensitivity experiments. Mon. Wea. Rev., 132, 2650–2668. Washington, W. M., and Coauthors, 2000: Parallel climate model (PCM) control and transient simulations. Climate Dyn., 16, 755–774. Wood, R., and C. S. Bretherton, 2006: On the relationship between stratiform low cloud cover and lower-tropospheric stability. J. Climate, 19, 6425–6432. Wyant, M. C., C. S. Bretherton, and P. N. Blossey, 2009: Subtropical low cloud response to a warmer climate in a superparameterized climate model. Part I: Regime sorting and physical mechanisms. J. Adv. Model. Earth Syst., 1, doi:10.3894. JAMES.2007.1.7. Xu, K.-M., and D. A. Randall, 1996: A semiempirical cloudiness parameterization for use in climate models. J. Atmos. Sci., 53, 3084–3102. ——, A. Cheng, and M. Zhang, 2010: Cloud-resolving simulation of low-cloud feedback to an increase in sea surface temperature. J. Atmos. Sci., 67, 730–748. Yu, Y., X. Zhang, and Y. Guo, 2004: Global coupled ocean– atmosphere general circulation models in LASG/IAP. Adv. Atmos. Sci., 21, 444–455. Yukimoto, S., and Coauthors, 2006: Present-day climate and climate sensitivity in the Meteorological Research Institute Coupled GCM Version 2.3 (MRI-CGCM2.3). J. Meteor. Soc. Japan, 84, 333–363. Zhang, M., and C. S. Bretherton, 2008: Mechanisms of low cloud– climate feedback in idealized single-column simulations with the Community Atmospheric Model, version 3 (CAM3). J. Climate, 21, 4859–4878. Zhu, P., J. J. Hack, J. T. Kiehl, and C. S. Bretherton, 2007: Climate sensitivity of tropical and subtropical marine low cloud amount to ENSO and global warming due to doubled CO2. J. Geophys. Res., 112, D17108, doi:10.1029/2006JD008174.