ATMOSPHERIC AND OCEANIC SCIENCE LETTERS, 2012, VOL. 5, NO. 3, 258−263
Preliminary Evaluation of Cloud Fraction Simulations by GAMIL2 Using
COSP
DONG Li1, LI Li-Juan1, HUANG Wen-Yu2, WANG Yong1, and WANG Bin1,2
1
State Key Laboratory of Numerical Modeling for Atmospheric Sciences and Geophysical Dynamics (LASG), Institute of Atmospheric Physics (IAP), Chinese Academy of Sciences, Beijing 100029, China
2
Ministry of Education Key Laboratory for Earth System Modeling, and Center for Earth System Science, Tsinghua University, Beijing 100084, China
Received 16 February 2012; revised 20 March 2012; accepted 26 March 2012; published 16 May 2012
Abstract The Cloud Feedback Model Intercomparisons
Project (CFMIP) Observation Simulator Package (COSP)
is adopted in the Grid-point Atmospheric Model of IAP
LASG (GAMIL2) during CFMIP at Phase II to evaluate
the model cloud fractions in a consistent way with satellite observations. The cloud simulation results embedded
in the Atmospheric Model Intercomparison Project (AMIP)
control experiment are presented using three satellite
simulators: International Satellite Cloud Climatology
Project (ISCCP), Moderate Resolution Imaging Spectroradiometer (MODIS), and Cloud-Aerosol Lidar with Orthogonal Polarization (CALIOP) lidar onboard the CloudAerosol Lidar and Infrared Pathfinder Satellite Observations (CALIPSO). Overall, GAMIL2 can produce horizontal distributions of the low cloud fraction that are
similar to the satellite observations, and its similarities to
the observations on different levels are shown in Taylor
diagrams. The discrepancies among satellite observations
are also shown, which should be considered during evaluation.
Keywords: COSP, GAMIL, cloud fraction, satellite simulator
Citation: Dong, L., L.-J. Li, W.-Y. Huang, et al., 2012:
Preliminary evaluation of cloud fraction simulations by
GAMIL2 using COSP, Atmos. Oceanic Sci. Lett., 5, 258–
263.
1
Introduction
Clouds play an important role in the climate system
and its change, but they are among the most uncertain
components in climate simulations, such as cloud fraction
simulation. In the Intergovernmental Panel on Climate
Change (IPCC) Fourth Assessment Report (AR4), the
cloud feedbacks are identified as the primary source of
inter-model differences and model biases (Randall et al.,
2007). An indispensable step to improving cloud representation in an atmospheric global circulation model
(AGCM) is to evaluate it with other models and observations. Unfortunately, the cloud definitions among AGCMs
are different because of the differences of the cloud
parameterizations and assumptions that are used. Hence,
direct intercomparisons of cloud simulations among
AGCMs and among climate models are difficult and inaccurate.
Corresponding author: LI Li-Juan, ljli@mail.iap.ac.cn
Satellites provide a global or near-global view of the
Earth, so they are a natural source of cloud observations.
Several satellite cloud products are available for model
evaluation purposes, such as the International Satellite
Cloud Climatology Project (ISCCP; Rossow and Schiffer,
1999), the Moderate Resolution Imaging Spectroradiometer (MODIS; King et al., 2003), the Multiangle Imaging Spectroradiometer (MISR; Marchand and Ackerman, 2010), CloudSat (Haynes et al., 2007), and the
Cloud-Aerosol Lidar with Orthogonal Polarization
(CALIOP) lidar onboard the Cloud-Aerosol Lidar and
Infrared Pathfinder Satellite Observations (CALIPSO;
Winker et al., 2010). However, these satellites do not directly measure the geophysical quantities of interest (e.g.,
wind field, temperature), and even for the retrievable
quantities (e.g., liquid water path), there are some inevitable assumptions and limitations in the retrieval techniques.
Thus, direct comparisons between models and satellite
retrievals are also difficult because of the different representations or definitions in the models and the satellite
observations.
To facilitate direct intercomparisons and comparisons
between the models and observations, satellite simulators
have been introduced (Yu et al., 1996; Klein and Jakob,
1999; Webb et al., 2001; Bodas-Salcedo et al., 2011) and
are widely used (Lin and Zhang, 2004; Zhang et al., 2005).
These simulators mimic the way that satellites observe the
Earth by converting the model variables into a consistent
form that can be compared with the satellites directly.
This process can take into account the assumptions and
limitations in the satellite retrieval techniques so that the
differences between models and observations can be
mainly attributed to model defects (Bodas-Salcedo et al.,
2011).
This paper presents the preliminary results of a cloud
simulation in the Atmospheric Model Intercomparison
Project (AMIP) control experiment of the Grid-point Atmospheric Model of the Key Laboratory of Numerical
Modeling for Atmospheric Sciences and Geophysical
Fluid Dynamics, Institute of Atmospheric Physics
(GAMIL2), by using the Cloud Feedback Model Intercomparisons Project (CFMIP) Observation Simulator
Package (COSP, see Bodas-Salcedo et al., 2011). The
observations and simulators of three satellites (ISCCP,
MODIS, and CALIPSO) are used. The first two are both
passive imagers, but they have different instruments and
NO. 3
DONG ET AL.: EVALUATION OF GAMIL CLOUD FRACTION USING COSP
retrieval strategies. The third is an active instrument that
measures backscatter lidar beam signals. For more information about those satellites, readers can refer to Bodas-Salcedo et al. (2011). The results demonstrate that the
multiple satellite simulators are useful for evaluating the
modeled clouds and the large uncertainties existing in
both the observations and simulations, which must be
well understood before drawing any conclusions. The
used version of GAMIL and its experimental design are
briefly described in section 2. In section 3, three satellites
are selected to present the simulated results for evaluating
the clouds simulated by GAMIL2. Section 4 presents the
conclusion and discussion.
2
2.1
Model description and experimental design
GAMIL2 description
GAMIL employs a hybrid horizontal grid, with a
Gaussian grid of 2.8° between 65.58°S and 65.58°N and
weighted equal-area grid poleward of 65.58°. GAMIL’s
dynamical core includes a finite difference scheme that
conserves mass and effective energy while solving the
primitive hydrostatic equations of baroclinic atmosphere
(Wang et al., 2004), and it has a two-step shape-preserving advection scheme for the moisture equation (Yu,
1994). The cloud fraction is based on a diagnostic Slingotype scheme (Slingo, 1987), and it depends on relative
humidity, atmospheric stability, water vapor and convective mass fluxes (Collins et al., 2003). The main differences between the two versions, GAMIL1 (Li and Wang,
2010) and GAMIL2, are the upgraded cloud-related processes, for example, the deep convective parameterizations
(Zhang and McFarlane, 1995; Zhang and Mu, 2005),
convective cloud fraction (Rasch and Kristjánsson, 1998;
Xu and Krueger, 1991) and microphysical schemes
(Rasch and Kristjánsson, 1998; Morrison and Gettelman,
2008). For the energy balance at the top of atmosphere
(TOA), some parameters in the convection (including
shallow and deep convection), cloud microphysical and
boundary layer schemes are tuned in GAMIL2 compared
with GAMIL1 (Li et al., 2012). Furthermore, the large
impact of parameter changes on cloud simulations and
climate sensitivity is being explored by many groups and
scientists (e.g., Tannahill et al., 2010; http://www.cesm.
ucar.edu/working_groups/Atmosphere/Presentations/2010/
lucas.pdf; Yang et al., 2011).
2.2
Observation data and experimental design
The simulation was conducted following the standard
settings for AMIP at Phase II with the solar constant, the
orbital parameters, and the greenhouse gas concentrations
from the fifth phase of the Coupled Model Intercomparison Project (CMIP5) website. The lower boundary conditions of the modeled atmosphere are the sea surface temperature (SST) and sea ice from HadIsst (Rayner et al.,
2003) interpolated into the model grids using the Karl
Taylor procedure (Taylor et al., 2000). The experiment
was run for 30 years, starting from 1 January 1979 to 31
December 2008, and the monthly mean variables were
259
analyzed. The COSP is adopted in GAMIL2 off-line for
the time being. The needed input quantities are outputted
from GAMIL2 at a frequency of three hours. All the satellite simulators are switched on. For the basic performance
of GAMIL2, readers can refer Li et al. (2012)①.
Several developers of satellite simulators in COSP
have converted the original satellite data into a form that
can be compared with the output of COSP. The processed
data can be accessed from http://climserv.ipsl.Polytechnique.fr/cfmip-obs/. The time period covered is from July
1983 to June 2008 for ISCCP (Rossow et al., 1996; Rossow and Schiffer, 1999), from July 2002 to April 2011 for
MODIS (Pincus et al., 2012), and from June 2006 to December 2010 for CALIPSO (Chepfer et al., 2010).
The satellite data above are available for only a few
years, whereas the model climatology compared with
such data are computed from 30 simulated years. In this
study, we assume that the impacts of uncertainties associated with the mismatching time periods of observations
and simulations are small for the climatological mean
state when compared with the magnitude of current model
errors and the different instrument sensitivities and errors
of retrieval algorithms (Gleckler et al., 2008; Kay et al.,
2012; Hillman, 2011; Pincus et al., 2012).
3
Cloud simulation evaluation
By utilizing COSP, the diagnostic information of the
modeled clouds has been enriched. More satellite observations can be used to evaluate modeled clouds, not only
the ISCCP as in previous researches (e.g., Zhang et al.,
2005). In the following, the results of ISCCP, MODIS,
and CALIPSO simulators are examined. Figure 1 shows
the zonal averages of the annual mean cloud fraction at
different levels, including the total cloud fraction. Several
characteristics can be observed from this figure: 1) The
observed total cloud fractions show that the maximum
cloud fractions are located in the Intertropical Convergence Zone (ITCZ) and mid-latitudes storm tracks, and
the minima occur over the subtropical zone in Fig. 1a.
GAMIL2 reproduces these fractions well through different simulators despite some degree of underestimations
that are also found in other climate models, such as the
Community Atmospheric Model (CAM) and the Geophysical Fluid Dynamics Laboratory (GFDL) AM (Kay et
al., 2012; Pincus et al., 2012). 2) The middle level cloud
fractions simulated by GAMIL2 are better than the other
levels, even with some overestimation in the MODIS
simulator (Fig. 1c). 3) The clouds observed and simulated
by CALIPSO (active instruments) are more than ISCCP
and MODIS; these results are consistent with other studies (Kay et al., 2012). 4) There are large uncertainties
between ISCCP and MODIS, which are both passive instruments. For example, the total cloud fraction is 66.1%
as retrieved by ISCCP, but only 49.9% by MODIS. This
difference can be mainly attributed to the different strate①
Li, L.-J., B. Wang, L. Dong, et al., 2012: Development and evaluation
of Grid-point Atmospheric Model of IAP LASG, Version 2 (GAMIL2),
Adv. Atmos. Sci., to be submitted.
260
ATMOSPHERIC AND OCEANIC SCIENCE LETTERS
gies toward the pixel-scale retrievals and aggregation as
well as instrument sensitivities (Marchand and Ackerman,
2010; Pincus et al., 2012). MODIS adopts a more conservative strategy that excludes the partly cloudy pixels (e.g.,
cloud-edge and inhomogeneous pixels on the scales observed by MODIS) and pixels where the retrievals of
cloud properties (e.g., cloud top pressure, optical depth,
and particle size) fail, which can account for the fewer
clouds detected by MODIS (retrieval fraction) than by
ISCCP to some extent (see the black and red dashed lines
in Fig. 1a).
Figure 2 shows the geographical distribution of the
annual mean low clouds. The different satellite retrievals
resemble each other in their spatial patterns: there are
more marine boundary layer clouds over the east coastal
regions (California, Peru, Canary, Angola, and Australia),
more low-level clouds in the mid-latitude storm tracks,
and a striking land-sea contrast. In addition, there are few
low clouds detected in the warm pools, where thick, high
clouds may have attenuated the instrument signal
(Chepfer et al., 2010). The basic geographical distributions of low clouds are reproduced by GAMIL2, including shallow cumulus clouds over the eastern coasts and
the clouds in the storm tracks, only with smaller magnitudes. The globally averaged low cloud amount is 22.5%,
15.9%, and 27% by the ISCCP, MODIS, and CALIPSO
simulators, respectively, which is less than 25.2%, 19.1%,
and 37.9% of the corresponding observations, respectively.
Although there are large differences among the satellites,
their combination will provide more realistic cloud fields
than any one of them by itself, as stated in Marchand and
Ackerman (2010) (e.g., the detection of multilayer clouds
by consulting the ISCCP, MODIS, and MISR retrievals).
We also note some weak signs of the double ITCZ over
the central Pacific Ocean, which will be discussed in another study (Li et al., 2012①).
An overview of the discrepancies among observations
VOL. 5
or among observations and simulations can be clearly
presented by the Taylor diagram (Taylor, 2001), as depicted in Fig. 3 for the different cloud fraction types. The
region is restricted to within 60°S–60°N, following the
settings in Hillman (2011) and Kay et al. (2012), because
the three satellite data sets have larger discrepancies at
high latitudes. For example, a large part of the ISCCP data
sets comes from geostationary satellites, and therefore, “a
dependence in ISCCP cloud fractions with latitude has
long been recognized” (Marchand and Ackerman, 2010).
MODIS and CALIPSO, in contrast, are both aboard
Sun-synchronized satellites. ISCCP is chosen as the reference in Fig. 3a. For the pattern correlation, it is exciting
that the three data sets agree well for all cloud types (approximately 0.9 and above), except for the CALIPSO
middle clouds (approximately 0.8). However, when the
focus is shifted to the spatial variations, the results are
puzzling in regard to which observation should be believed. Compared with ISCCP, there are large uncertainties of the standard deviations, from 0.75 for the MODIS
middle clouds to 1.5 for the CALIPSO low clouds. Which
factor is responsible for these uncertainties? Is it instrument sensitivity or perhaps the retrieval algorithm? A
similar analysis of ISCCP, MODIS, and MISR can be
found in Hillman (2011). The performance of GAMIL2
can be discerned from Fig. 3b. Overall, GAMIL2 performs well in the cloud spatial patterns, and the correlations of most cloud fractions are approximately 0.75 and
above, except for the low cloud fraction by the ISCCP
simulator (0.55). There are smaller spatial variations in
GAMIL2 compared with the corresponding observations,
which may be related to the lower simulated cloud amounts and other unknowns. Moreover, the statistical significance of the apparent differences in the Taylor diagram should be evaluated, but for the time being, there is
only one AMIP simulation that is processed by COSP, and
only its climatological mean state is evaluated in this
Figure 1 Zonal averages of the annual mean cloud fractions (%): (a) total cloud fraction, (b) high level cloud fraction, (c) middle level cloud fraction, and (d) low level cloud fraction. The dashed lines are the satellite (black: ISCCP, red: MODIS, blue: CALIPSO) observations and the solid lines
are the simulator results from GAMIL2 using COSP for the corresponding satellites.
NO. 3
DONG ET AL.: EVALUATION OF GAMIL CLOUD FRACTION USING COSP
261
Figure 2 Geographic distributions of annual mean low cloud fractions. The left column is the observation data and the right is the simulated results
from GAMIL2 using COSP. The area-weighted averages are indicated on the top right of each figure.
Figure 3 Taylor diagrams comparing the geographic distribution of near global annual mean cloud fractions between (a) observations with ISCCP
as a reference and (b) observations and their corresponding simulators in GAMIL2 with observations as a reference. All comparisons are restricted to
the latitudes from 60°S to 60°N.
262
ATMOSPHERIC AND OCEANIC SCIENCE LETTERS
study because of the substantial data storage cost. Thus,
the statistical significance will be evaluated in the future.
In the meantime, Fig. 3b likely gives some indications
of the source of the uncertainties. Given the same input
from GAMIL2, the three simulators or retrieval algorithms present different results in the standard deviations,
suggesting that those algorithms make important contributions to the uncertainties.
4
Summary and discussion
In this study, preliminary results from GAMIL2 using
COSP are presented. In the first step, the package of satellite simulators, COSP, is incorporated to generate clouds
that can be compared with satellite observations consistently, even though the discrepancies among the observations or between the observations and the simulator will
make the analysis of the results more complex, such as
the different retrieval strategies adopted by ISCCP and
MODIS. On the whole, GAMIL2 performs well at presenting the spatial distributions of the cloud fractions at
different levels, such as the maximum cloud amounts in
the ITCZ and storm tracks and the minima in the subtropical regions. These good results also indicate the reasonable temperature and humidity structure simulated by
the model. The bias of the middle cloud simulation is
relatively smaller than of the others clouds (Figs. 1c and
3b). However, there is some degree of underestimation for
all clouds, which is also a common phenomenon for the
climate models when using the simulators (Zhang et al.,
2005; Kay et al., 2012; Pincus et al., 2012).
It should also be noted that there are large uncertainties
among the observation data sets, the simulators (or retrieval algorithms), the parameters in the models, and
other unknowns. For example, the low cloud biases result
from the high cloud masking, and the simulators are limited to the different strategies for the pixel-scale retrievals
and aggregation and the instrument sensitivities. Additionally, the inconsistencies between model and simulator
assumptions and the mismatches between meteorological
regimes in the observations and models due to positional
errors are not entirely understood yet (Bodas-Salcedo et
al., 2011; Kay et al., 2012). Under these conditions, care
should be taken in evaluating the model performances
qualitatively.
Despite the large uncertainties in the cloud fractions
among observations and simulations, it is meaningful to
evaluate the model performance using the observations.
This new approach for evaluating the performance of
models on clouds and their feedbacks will definitely
enlarge the perspective of modelers by introducing more
sophisticated analysis approaches (e.g., the combination
of several satellites will provide more realistic cloud
fields), and it will promote further understanding of the
uncertainties from the models and observations. For example, the same model inputs could help us distinguish
the contributions to the uncertainties from these algorithms and others.
Acknowledgements. This work was jointly supported by the Na-
VOL. 5
tional High Technology Research and Development Program of
China (863 Program) (Grant No. 2010AA012304), the National
Basic Research Program of China (973 Program) (Grant No.
2010CB951904), the China Meteorological Administration R &D
Special Fund for Public Welfare (meteorology) (Grant No.
GYHY201006014), and the National Natural Science Foundation of
China (Grant Nos. 41023002 and 41005053).
References
Bodas-Salcedo, A., M. J. Webb, S. Bony, et al., 2011: COSP: Satellite simulation software for model assessment, Bull. Amer. Meteor. Soc., 92(8), 1023–1043.
Chepfer, H., S. Bony, D. Winker, et al., 2010: The GCM-Oriented
CALIPSO Cloud Product (CALIPSO-GOCCP), J. Geophys. Res.,
115, D00H16, doi:10.1029/2009JD012251.
Collins, W. D., J. J. Hack, B. A. Boville, et al., 2003: Description of
the NCAR Community Atmospheric Model (CAM2), NCAR,
Boulder, 190pp.
Gleckler, P. J., K. E. Taylor, and C. Doutriaux, 2008: Performance
metrics for climate models, J. Geophys. Res., 113, D06104,
doi:10.1029/2007JD008972.
Haynes, J. M., R. Marchand, Z. Luo, et al., 2007: A multipurpose
radar simulation package: QuickBeam, Bull. Amer. Meteor. Soc.,
88(11), 1723–1727.
Hillman, B. R., 2011: Evaluating Clouds in Global Climate Models
Using Instrument Simulators, M.S. thesis, University of Washington, 93pp.
Kay, J. E., B. R. Hillman, S. A. Klein, et al., 2012:
Exposing global
cloud biases in the Community Atmosphere Model (CAM) using
satellite observations and their corresponding instrument simulators, J. Climate, in press, doi:10.1175/JCLI-D-11-00469.1.
King, M. D., W. P. Menzel, Y. J. Kaufman, et al., 2003: Cloud and
aerosol properties, precipitable water, and profiles of temperature
and water vapor from MODIS, IEEE Trans. Geosci. Remote
Sens., 41(2), 442–458.
Klein, S. A., and C. Jakob, 1999: Validation and sensitivities of
frontal clouds simulated by the ECMWF model, Mon. Wea. Rev.,
127, 2514–2531.
Li, L.-J., and B. Wang, 2010: Influences of two convective schemes
on the radiative energy budget in GAMIL1.0, Acta Meteor.
Sinica, 24(3), 318–327.
Li, L.-J., X. Xie, B. Wang, et al., 2012: Evaluating the performances
of GAMIL1.0 and GAMIL2.0 during TWP-ICE with CAPT,
Atmos. Oceanic Sci. Lett., 5, 38–42.
Lin, W. Y., and M. H. Zhang, 2004: Evaluation of clouds and their
radiative effects simulated by the NCAR Community Atmospheric Model CAM2 against satellite observations, J. Climate,
17, 3302–3318.
Marchand, R., and T. Ackerman, 2010: An analysis of cloud cover
in multiscale modeling framework global climate model simulations using 4 and 1 km horizontal grids, J. Geophys. Res., 115,
D16207, doi:10.1029/2009JD013423.
Morrison, H., and A. Gettelman, 2008: A new two-moment bulk
stratiform cloud microphysics scheme in the community atmosphere model, version 3 (CAM3). Part I: Description and numerical tests, J. Climate, 21(15), 3642–3659.
Pincus, R., S. Platnick, S. A. Ackerman, et al., 2012: Reconciling
simulated and observed views of clouds: MODIS, ISCCP, and
the limits of instrument simulators, J. Climate, in press,
doi:10.1175/JCLI-D-11-00267.1.
Randall, D. A., R. A. Wood, S. Bony, et al., 2007: Climate models
and their evaluation, in: Climate Change 2007: The Physical
Science Basis. Contribution of Working Group I to the Fourth
Assessment Report of the Intergovernmental Panel on Climate
Change, S. Solomon et al. (Eds.), Cambridge University Press,
Cambridge and New York, 996pp.
Rasch, P. J., and J. E. Kristjánsson, 1998: A comparison of the
CCM3 model climate using diagnosed and predicted condensate
NO. 3
DONG ET AL.: EVALUATION OF GAMIL CLOUD FRACTION USING COSP
parameterizations, J. Climate, 11(7), 1587–1614.
Rayner, N. A., D. E. Parker, E. B. Horton, et al., 2003: Global
analyses of sea surface temperature, sea ice, and night marine air
temperature since the late nineteenth century, J. Geophys. Res.,
108(D14), 4407, doi:10.1029/2002JD002670.
Rossow, W. B., and R. A. Schiffer, 1999: Advances in understanding
clouds from ISCCP, Bull. Amer. Meteor. Soc., 80, 2261–2287.
Rossow, W. B., A. W. Walker, D. E. Beuschel, et al., 1996: International Satellite Cloud Climatology Project (ISCCP) Documentation of New Cloud Datasets, WMO/TD-No. 737, World Meteorological Organization, 115pp.
Slingo, J. M., 1987: The development and verification of a cloud
prediction scheme for the ECMWF model, Quart. J. Roy. Meteor.
Soc., 113, 899–927.
Tannahill, J., S. T. Brandon, C. C. Covey, et al., 2010: The Climate
Uncertainty Quantification Project at Lawrence Livermore National Laboratory: I. Initial Analysis of the Sensitivities and Uncertainties in the Community Atmosphere Model, in: AGU fall
meeting 2010, American Geophysical Union.
Taylor, K. E., 2001: Summarizing multiple aspects of model performance in a single diagram, J. Geophys. Res., 106(D7), 7183–
7192, doi:10.1029/2000JD900719.
Taylor, K. E., D. Williamson, and F. Zwiers, 2000: The sea surface
temperature and sea-ice concentration boundary conditions for
AMIP II simulations, PCMDI Report No. 60, Program for Climate Model Diagnosis and Intercomparison, Lawrence Livermore National Laboratory, Livermore, California, 25pp.
Wang, B., H. Wan, J. Ji, et al., 2004: Design of a new dynamical
core for global atmospheric models based on some efficient numerical methods, Sci. China Ser. A-Math., 47, 4–21.
Webb, M. J., C. Senior, S. Bony, et al., 2001: Combining ERBE and
263
ISCCP data to assess clouds in the Hadley Centre, ECMWF and
LMD atmospheric climate models, Climate Dyn., 17(12), 905–
922.
Winker, D. M., J. Pelon, J. A. Coakley, et al., 2010: The CALIPSO
mission: A global 3D view of aerosols and clouds, Bull. Amer.
Meteor. Soc., 91, 1211–1229.
Xu, K. M., and S. K. Krueger, 1991: Evaluation of cloudiness
parameterizations using a cumulus ensemble model, Mon. Wea.
Rev., 119, 342–367.
Yang, B., Y. Qian, G. Lin, et al., 2011: Some issues in uncertainty
quantification and parameter tuning: A case study of convective
parameterization scheme in the WRF regional climate model,
Atmos. Chem. Phys. Discuss., 11, 31769–31817.
Yu, R., 1994: A two-step shape-preserving advection scheme, Adv.
Atmos. Sci., 11(4), 479–490.
Yu, W., M. Doutriaux, G. Seze, et al., 1996: A methodology study of
the validation of clouds in GCMs using ISCCP satellite observations, Climate Dyn., 12, 389–401.
Zhang, G. J., and N. A. McFarlane, 1995: Sensitivity of climate
simulations to the parameterization of cumulus convection in the
Canadian Climate Centre general circulation model, Atmos.—
Ocean, 33, 407–446.
Zhang, G. J., and M. Mu, 2005: Effects of modifications to the
Zhang-McFarlane convection parameterization on the simulation
of the tropical precipitation in the National Center for Atmospheric Research Community Climate Model, version 3, J. Geophys. Res., 110, D09109, doi:10.1029/2004JD005617.
Zhang, M. H., W. Y. Lin, S. A. Klein, et al., 2005: Comparing clouds
and their seasonal variations in 10 atmospheric general circulation models with satellite measurements, J. Geophy. Res., 110,
D15S02, doi:10.1029/2004JD005021.