• Annan, J. D., , J. Hargreaves, , N. Edwards, , and R. Marsh, 2005a: Parameter estimation in an intermediate complexity earth system model using an ensemble Kalman filter. Ocean Modell., 8 , 135154.

    • Search Google Scholar
    • Export Citation
  • Annan, J. D., , J. Hargreaves, , R. Ohgaito, , A. Abe-Ouch, , and S. Emori, 2005b: Efficiently constraining climate sensitivity with paleoclimate simulations. Sci. Online Lett. Atmos., 1 , 181184.

    • Search Google Scholar
    • Export Citation
  • Arakawa, A., , and W. H. Schubert, 1974: Interactions of cumulus cloud ensemble with the large-scale environment. Part I. J. Atmos. Sci., 31 , 671701.

    • Search Google Scholar
    • Export Citation
  • Boer, G. J., , and B. Yu, 2003: Climate sensitivity and response. Climate Dyn., 20 , 415429.

  • Bony, S., , and J-L. Dufresne, 2005: Marine boundary layer clouds at the heart of tropical cloud feedback uncertainty in climate models. Geophys. Res. Lett., 32 , L20806. doi:10.1029/2005GL023851.

    • Search Google Scholar
    • Export Citation
  • Bony, S., and Coauthors, 2006: How well do we understand and evaluate climate change feedback processes? J. Climate, 19 , 34453482.

  • Cess, R. D., and Coauthors, 1990: Intercomparison and interpretation of climate feedback processes in 19 atmospheric general circulation models. J. Geophys. Res., 95 , 1660116615.

    • Search Google Scholar
    • Export Citation
  • Collins, M., 2007: Ensembles and probabilities: A new era in the prediction of climate change. Philos. Trans. Roy. Soc. London, A365 , 19571970.

    • Search Google Scholar
    • Export Citation
  • Colman, R. A., 2003: A comparison of climate feedbacks in general circulation models. Climate Dyn., 20 , 865873.

  • Gordon, N. D., , J. R. Norris, , C. P. Weaver, , and S. A. Klein, 2005: Cluster analysis of cloud regimes and characteristic dynamics of midlatitude synoptic systems in observations and a model. J. Geophys. Res., 110 , D15S17. doi:10.1029/2004JD005027.

    • Search Google Scholar
    • Export Citation
  • Gregory, D., , and P. R. Rowntree, 1990: A mass flux convection scheme with representation of cloud ensemble characteristics and stability dependent closure. Mon. Wea. Rev., 118 , 14831506.

    • Search Google Scholar
    • Export Citation
  • Gregory, D., , and S. Allen, 1991: The effect of convective scale downdrafts upon NWP and climate simulations. Preprints, Ninth Conf. on Numerical Weather Prediction, Denver, CO, Amer. Meteor. Soc., 122–123.

    • Search Google Scholar
    • Export Citation
  • Gregory, D., , and S. Allen, 1996: The sensitivity of climate simulations to the specification of mixed phase cloud. Climate Dyn., 12 , 641651.

    • Search Google Scholar
    • Export Citation
  • Hargreaves, J. C., , and J. D. Annan, 2006: Using ensemble prediction methods to examine regional climate variation under global warming scenarios. Ocean Modell., 11 , 174192.

    • Search Google Scholar
    • Export Citation
  • Hargreaves, J. C., , A. Abe-Ouchi, , and J. D. Annan, 2007: Linking glacial and future climate through an ensemble of GCM simulations. Climate Past, 3 , 7787.

    • Search Google Scholar
    • Export Citation
  • Jakob, C., , and G. Tselioudis, 2003: Objective identification of cloud regimes in the tropical western Pacific. Geophys. Res. Lett., 30 , 2082. doi:10.1029/2003GL018367.

    • Search Google Scholar
    • Export Citation
  • K-1 Model Developers, 2004: K-1 coupled GCM (MIROC) description. K-1 Tech. Rep. 1, University of Tokyo, 1–34.

  • Kimoto, M., 2005: Simulated change of the East Asian circulation under global warming scenario. Geophys. Res. Lett., 32 , L16701. doi:10.1029/2005GL023383.

    • Search Google Scholar
    • Export Citation
  • Klein, S. A., , and C. Jakob, 1999: Validation and sensitivities of frontal clouds simulated by the ECMWF model. Mon. Wea. Rev., 127 , 25142531.

    • Search Google Scholar
    • Export Citation
  • Knutti, R., , G. A. Meehl, , M. R. Allen, , and D. A. Stainforth, 2006: Constraining climate sensitivity from the seasonal cycle in surface temperature. J. Climate, 19 , 42244233.

    • Search Google Scholar
    • Export Citation
  • Lambert, F. H., , and M. J. Webb, 2008: Dependency of global mean precipitation on surface temperature. Geophys. Res. Lett., 35 , L16706. doi:10.1029/2008GL034838.

    • Search Google Scholar
    • Export Citation
  • Le Treut, H., , and Z. X. Li, 1991: Sensitivity of an atmospheric general circulation model to prescribed SST changes: Feedback effects associated with the simulation of cloud optical properties. Climate Dyn., 5 , 175187.

    • Search Google Scholar
    • Export Citation
  • Medeiros, B., , B. Stevens, , 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 , 49744991.

    • Search Google Scholar
    • Export Citation
  • Murphy, J., , D. M. H. Sexton, , D. N. Barnett, , G. S. Jones, , M. J. Webb, , M. Collins, , and D. A. Stainforth, 2004: Quantification of modelling uncertainties in a large ensemble of climate change simulations. Nature, 430 , 768772.

    • Search Google Scholar
    • Export Citation
  • Murphy, J., , B. B. B. Booth, , M. Collins, , G. R. Harris, , D. M. H. Sexton, , and M. J. Webb, 2007: A methodology for probabilistic predictions of regional climate change from perturbed physics ensembles. Philos. Trans. Roy. Soc. London, A365 , 19932028.

    • Search Google Scholar
    • Export Citation
  • Nagashima, T., , H. Shiogama, , T. Yokohata, , T. Takemura, , S. A. Crooks, , and T. Nozawa, 2006: The effect of carbonaceous aerosols on surface temperature in the mid twentieth century. Geophys. Res. Lett., 33 , L04702. doi:10.1029/2005GL024887.

    • Search Google Scholar
    • Export Citation
  • Nozawa, T., , T. Nagashima, , H. Shiogama, , and S. A. Crooks, 2005: Detecting natural influence on surface air temperature change in the early twentieth century. Geophys. Res. Lett., 32 , L20719. doi:10.1029/2005GL023540.

    • Search Google Scholar
    • Export Citation
  • Numaguti, A., , M. Takahashi, , T. Nakajima, , and A. Sumi, 1997: Description of CCSR/NIES atmospheric general circulation model. Supercomputer Monograph Report, Center for Global Environmental Research, National Institute for Environmental Studies, 1–48.

    • Search Google Scholar
    • Export Citation
  • Ogura, T., , S. Emori, , M. J. Webb, , Y. Tsushima, , T. Yokohata, , A. Abe-Ouchi, , and M. Kimoto, 2008: Towards understanding cloud response in atmospheric GCMS: The use of tendency diagnostics. J. Meteor. Soc. Japan, 86 , 6979.

    • Search Google Scholar
    • Export Citation
  • Piani, C., , D. J. Frame, , D. A. Stainforth, , and M. R. Allen, 2005: Constraints on climate change from a multi-thousand member ensemble of simulations. Geophys. Res. Lett., 32 , L23825. doi:10.1029/2005GL024452.

    • Search Google Scholar
    • Export Citation
  • Pope, V. D., , M. L. Gallani, , P. R. Rowntree, , and R. A. Stratton, 2000: The impact of new physical parametrizations in the Hadley Centre climate model: HadAM3. Climate Dyn., 16 , 123146.

    • Search Google Scholar
    • Export Citation
  • Randall, D. A., and Coauthors, 2007: Climate models and their evaluation. Climate Change 2007: The Physical Science Basis, S. Solomon, et al., Eds., Cambridge University Press, 589–662.

    • Search Google Scholar
    • Export Citation
  • Rossow, W. B., , and R. A. Schiffer, 1999: Advances in understanding clouds from ISCCP. Bull. Amer. Meteor. Soc., 80 , 22612287.

  • Shiogama, H., , T. Nagashima, , T. Yokohata, , S. A. Crooks, , and T. Nozawa, 2006: Influence of volcanic activity and changes in solar irradiance on surface air temperatures in the early twentieth century. Geophys. Res. Lett., 33 , L09702. doi:10.1029/2005GL025622.

    • Search Google Scholar
    • Export Citation
  • Smith, A., 1990: A scheme for predicting layer clouds and their water content in a general circulation model. Quart. J. Roy. Meteor. Soc., 116 , 435460.

    • Search Google Scholar
    • Export Citation
  • Soden, B. J., , and I. M. Held, 2006: An assessment of climate feedbacks in coupled ocean–atmosphere models. J. Climate, 19 , 33543360.

    • Search Google Scholar
    • Export Citation
  • Soden, B. J., , R. S. Broccoli, , and R. S. Helmer, 2004: On the use of cloud forcing to estimate cloud feedback. J. Climate, 17 , 36613665.

    • Search Google Scholar
    • Export Citation
  • Stainforth, D. A., and Coauthors, 2005: Uncertainty in predictions of the climate response to rising levels of greenhouse gases. Nature, 433 , 403406.

    • Search Google Scholar
    • Export Citation
  • Taylor, K. E., , C. Doutriaux, , M. Crucifix, , C. D. Hewitt, , J. F. B. Mitchell, , M. J. Webb, , P. Braconnot, , and A. J. Broccoli, 2007: Estimating shortwave radiative forcing and response in climate models. J. Climate, 20 , 25302543.

    • Search Google Scholar
    • Export Citation
  • Tett, S. F. B., and Coauthors, 2002: Estimation of natural and anthropogenic contributions to twentieth century temperature change. J. Geophys. Res., 107 , 4306. doi:10.1029/2000JD000028.

    • Search Google Scholar
    • Export Citation
  • Webb, M., , C. Senior, , S. Bony, , and J-J. Morcrette, 2001: Combining ERBE and ISCCP data to assess clouds in the Hadley Centre, ECMWF and LMD atmospheric climate models. Climate Dyn., 17 , 905922.

    • Search Google Scholar
    • Export Citation
  • Webb, M., and Coauthors, 2006: On the contribution of local feedback mechanisms to the range of climate sensitivity in two GCM ensembles. Climate Dyn., 27 , 1738.

    • Search Google Scholar
    • Export Citation
  • Wetherald, R. T., , and S. Manabe, 1988: Cloud feedback processes in a general circulation model. J. Atmos. Sci., 45 , 13971451.

  • Williams, K. D., , and G. Tselioudis, 2007: GCM intercomparison of global cloud regimes: Present-day evaluation and climate change response. Climate Dyn., 29 , 231250.

    • Search Google Scholar
    • Export Citation
  • Williams, K. D., , and M. J. Webb, 2008: A quantitative performance assessment of cloud regimes in climate models. Climate Dyn., 33 , 141157. doi:10.1007/s00382-008-0443-1.

    • Search Google Scholar
    • Export Citation
  • Williams, K. D., , C. A. Senior, , and J. F. B. Mitchell, 2001: Transient climate change in the Hadley Centre models: The role of physical processes. J. Climate, 14 , 26592674.

    • Search Google Scholar
    • Export Citation
  • Yokohata, T., , S. Emori, , T. Nozawa, , Y. Tsushima, , T. Ogura, , and M. Kimoto, 2005a: Climate response to volcanic forcing: Validation of climate sensitivity of a coupled atmosphere–ocean general circulation model. Geophys. Res. Lett., 32 , L21710. doi:10.1029/2005GL023542.

    • Search Google Scholar
    • Export Citation
  • Yokohata, T., , S. Emori, , T. Nozawa, , Y. Tsushima, , T. Ogura, , and M. Kimoto, 2005b: A simple scheme for climate feedback analysis. Geophys. Res. Lett., 32 , L19703. doi:10.1029/2005GL023673.

    • Search Google Scholar
    • Export Citation
  • Yokohata, T., and Coauthors, 2008: Comparison of equilibrium and transient responses to CO2 increase in eight state-of-the-art climate models. Tellus, 60 , 946961.

    • Search Google Scholar
    • Export Citation
  • Yoshimori, M., , T. Yokohata, , and A. Abe-Ouchi, 2009: A comparison of climate feedback strength between CO2 doubling and LGM experiments. J. Climate, 22 , 33743395.

    • Search Google Scholar
    • Export Citation
  • View in gallery

    Total climate feedback parameter (Λ), and climate feedback by SW cloud, surface, clear-sky atmosphere, and LW cloud and clear-sky atmosphere (unit except for SAT anomaly is W m−2 K−1), and the globally averaged surface air temperature anomaly (i.e., equilibrium climate sensitivity). Shaded bars represent the ensemble means of MIROC3.2 (slash), HadSM3 (gray), and their difference (MIROC3.2-HadSM3, white), and the error bars represent the two ensemble standard deviations for the different quantities.

  • View in gallery

    (first row) Global map of SW cloud, (second row) SW clear-sky atmosphere, (third row) SW surface, (fourth row) LW cloud, and (bottom row) LW clear-sky atmosphere feedback for (left) the ensemble mean of MIROC3.2, (middle) HadSM3, and (right) their difference. Feedback is calculated as the local change in radiative flux at the TOA divided by the global mean SAT change. Unit is W m−2 K−1.

  • View in gallery

    Fractional contribution of the strength of climate feedback processes shown in Fig. 1 to the variance of their sum, for (a) MIROC3.2 and (b) HadSM3. See Eq. (9) and section 4a(2).

  • View in gallery

    Scatterplot of the total climate feedback (Λ) and globally averaged climate feedback by (a) SW cloud, (b) SW clear-sky atmosphere, (c) SW surface, (d) LW cloud, and (e) LW clear-sky atmosphere. Each point represents an ensemble member of MIROC3.2 (red) and HadSM3 (blue), and straight lines represent the linear regressions. See section 4a(2) for more details.

  • View in gallery

    Globally averaged values of SW feedback by clouds with different cloud-top pressure (ptop). Contributions from the cloud at high level (50–440 hPa), middle level (440–680 hPa), and low level (680–1000 hPa), and contributions from the seven categories of ptop are shown. Bars represent the ensemble mean (slash: MIROC3.2; gray: HadSM3; white: MIROC3.2-HadSM3), and error bars represent the two standard deviation. Unit is W−2 m−2 K−1.

  • View in gallery

    Ensemble mean global map of the SW feedback by the cloud at (top) high level, (middle) middle level, and (bottom) low level for (left) MIROC3.2, (middle) HadSM3, and (right) their difference.

  • View in gallery

    Globally averaged values of the SW and LW feedback by the cloud classes defined by Webb et al. (2006). Bars represent the ensemble mean (slash: MIROC3.2; gray: HadSM3; white: MIROC3.2-HadSM3), and error bars represent the two standard deviation. (a) Sum of the SW and LW feedback by the cloud classes A (S+LN) and E (S LN); B (S+L) and F (SL+); C (SNL) and G (SNL+); D (SL) and H (S+L+); (b) sum of the SW and LW feedback by the eight cloud classes; (c) same as (b) but for the SW feedback; and (d) same as (b) but for the LW feedback. Unit is W−2 m−2 K−1.

  • View in gallery

    Fractional contributions of the globally averaged SW feedback by the clouds with different ptop to the within-ensemble variation of their sum, (a) MIROC3.2 and (b) HadSM3. Categories of ptop are the same as those in Fig. 5.

  • View in gallery

    Fractional contributions of the globally averaged feedback by the cloud classes defined by Webb et al. (2006), to the within-ensemble variation of their sum. Components of the cloud classes are same as those of Fig. 7, but for the (a),(c),(e),(g) MIROC3.2 ensemble and (b),(d),(f),(h) HadSM3 ensemble.

  • View in gallery

    Global map of (a) cloud albedo anomaly, (b) first term in Eq. (10), (c) second term in Eq. (10), (d) cloud cover anomaly, (e) 1XCO2 cloud cover, and (f) 1XCO2 in-cloud albedo, shown as the differences between the MIROC3.2 and HadSM3 ensemble means.

  • View in gallery

    Relationship between (a) the SW cloud feedback at the second lowest level (ptop = 680–800 hPa) and 1XCO2 cloud albedo at the second lowest level, (b) the SW feedback by the change in in-cloud albedo at the second lowest level and 1XCO2 cloud cover at the second lowest level, (c) the cloud SW feedback at the second lowest level and 1XCO2 cloud albedo at the lowest level (ptop = 800–1000 hPa), and (d) the SW cloud feedback by the lowest level and 1XCO2 cloud albedo at the lowest level. The lines are linear regressions of the scatterplots.

  • View in gallery

    The difference in (top) cloud albedo, (middle) cloud cover, and (bottom) in-cloud albedo at the low level (680–1000 hPa) between the ensemble means of model simulation (left is MIROC3.2 and right is HadSM3) and the observation (ISCCP-D2; Rossow and Schiffer 1999).

  • View in gallery

    Same as in Fig. 11, but for the scatterplot of total climate feedback (Λ) and the cloud albedo at the low level (680–1000 hPa). Each point represents an ensemble member of MIROC3.2 (red) and HadSM3 (blue). Values of horizontal axis are shown as difference from ISCCP-D2 (Rossow and Schiffer 1999). The lines represent the linear regression for the scatterplots.

All Time Past Year Past 30 Days
Abstract Views 0 0 0
Full Text Views 76 76 5
PDF Downloads 28 28 1

Structural Similarities and Differences in Climate Responses to CO2 Increase between Two Perturbed Physics Ensembles

View More View Less
  • 1 National Institute for Environmental Studies, Tsukuba, Japan
  • | 2 Met Office, Hadley Centre, Exeter, United Kingdom
  • | 3 Center for Climate System Research, University of Tokyo, Tokyo, Japan
  • | 4 Frontier Research Center for Global Change, Japan Agency for Marine–Earth Science and Technology, Yokohama, Japan
© Get Permissions
Full access

Abstract

The equilibrium climate sensitivity (ECS) of the two perturbed physics ensembles (PPE) generated using structurally different GCMs, Model for Interdisciplinary Research on Climate (MIROC3.2) and the Third Hadley Centre Atmospheric Model with slab ocean (HadSM3), is investigated. A method to quantify the shortwave (SW) cloud feedback by clouds with different cloud-top pressure is developed. It is found that the difference in the ensemble means of the ECS between the two ensembles is mainly caused by differences in the SW low-level cloud feedback. The ensemble mean SW cloud feedback and ECS of the MIROC3.2 ensemble is larger than that of the HadSM3 ensemble. This is likely related to the 1XCO2 low-level cloud albedo of the former being larger than that of the latter. It is also found that the largest contribution to the within-ensemble variation of ECS comes from the SW low-level cloud feedback in both ensembles. The mechanism that causes the within-ensemble variation is different between the two ensembles. In the HadSM3 ensemble, members with large 1XCO2 low-level cloud albedo have large SW cloud feedback and large ECS; ensemble members with large 1XCO2 cloud cover have large negative SW cloud feedback and relatively low ECS. In the MIROC3.2 ensemble, the 1XCO2 low-level cloud albedo is much more tightly constrained, and no relationship is found between it and the cloud feedback. These results indicate that both the parametric uncertainties sampled in PPEs and the structural uncertainties of GCMs are important and worth further investigation.

* Current affiliation: Japan Agency for Marine-Earth Science and Technology, Yokohama, Japan.

Corresponding author address: Tokuta Yokohata, 3173-25, Showamachi, Kanazawa-ku, Yokohama, 236-0001, Japan. Email: yokohata@jamstec.go.jp

Abstract

The equilibrium climate sensitivity (ECS) of the two perturbed physics ensembles (PPE) generated using structurally different GCMs, Model for Interdisciplinary Research on Climate (MIROC3.2) and the Third Hadley Centre Atmospheric Model with slab ocean (HadSM3), is investigated. A method to quantify the shortwave (SW) cloud feedback by clouds with different cloud-top pressure is developed. It is found that the difference in the ensemble means of the ECS between the two ensembles is mainly caused by differences in the SW low-level cloud feedback. The ensemble mean SW cloud feedback and ECS of the MIROC3.2 ensemble is larger than that of the HadSM3 ensemble. This is likely related to the 1XCO2 low-level cloud albedo of the former being larger than that of the latter. It is also found that the largest contribution to the within-ensemble variation of ECS comes from the SW low-level cloud feedback in both ensembles. The mechanism that causes the within-ensemble variation is different between the two ensembles. In the HadSM3 ensemble, members with large 1XCO2 low-level cloud albedo have large SW cloud feedback and large ECS; ensemble members with large 1XCO2 cloud cover have large negative SW cloud feedback and relatively low ECS. In the MIROC3.2 ensemble, the 1XCO2 low-level cloud albedo is much more tightly constrained, and no relationship is found between it and the cloud feedback. These results indicate that both the parametric uncertainties sampled in PPEs and the structural uncertainties of GCMs are important and worth further investigation.

* Current affiliation: Japan Agency for Marine-Earth Science and Technology, Yokohama, Japan.

Corresponding author address: Tokuta Yokohata, 3173-25, Showamachi, Kanazawa-ku, Yokohama, 236-0001, Japan. Email: yokohata@jamstec.go.jp

1. Introduction

In order for policy makers to have mitigation and adaptation strategies for future climate changes, it is important for climate science to quantify and reduce uncertainties in future climate predictions made by numerical models. Complex general circulation models (GCMs) are the most powerful tool for this purpose, but major uncertainties in GCMs remain. Recently, several groups have developed methods for sampling uncertainties in the parameter values in GCMs (e.g., Murphy et al. 2004; Stainforth et al. 2005; Annan et al. 2005b; Piani et al. 2005; Knutti et al. 2006). The resulting ensembles are sometime referred to as “Perturbed Physics Ensembles” (PPEs).

Although PPEs sample the uncertainties in model parameters, one limitation is that they do not sample uncertainties in model structure, such as the specification of physical parameterization schemes or the dynamical formulation (Murphy et al. 2007). The results obtained are likely to depend on which model is used to generate the ensemble, and there is no guarantee that the ranges of uncertainty sampled by a PPE and by other methods are comparable. In generating projections in terms of probability distribution functions, Murphy et al. (2007) suggests that the magnitude of the impact of structural uncertainties can be approximated by combining the output from the Coupled Model Intercomparison Project phase three (CMIP3) with results from larger PPEs generated by a single model. As well as this statistical approach, understanding features of structural difference between climate models is also important in order to reduce uncertainties in climate predictions. Detailed comparison of the PPEs generated by the structurally different GCMs is crucial for this purpose (Collins 2007), but it has not been extensively performed so far.

When we consider uncertainties in climate prediction, equilibrium climate sensitivity (ECS), defined as the equilibrium global mean surface temperature change following a doubling of atmospheric CO2 concentrations, is a useful measure because many aspects of climate response scale well with global average surface air temperature (e.g., Randall et al. 2007). Although ECS may not be a good predictor of the transient response of atmosphere–ocean-coupled GCMs to doubled CO2 concentration in some models (e.g., Yokohata et al. 2008), ECS remains a widely used measure of the climate response to increased carbon dioxide, which can be calculated directly by the relatively computationally efficient atmosphere–slab ocean GCMs used to create the larger PPEs. Uncertainty in ECS is mainly determined by uncertainty in climate feedback processes, especially those caused by clouds (e.g., Cess et al. 1990; Bony et al. 2006; Soden and Held 2006). Previous studies pointed out the importance of low-level cloud feedbacks for the spread in cloud feedback among climate models (e.g., Bony and Dufresne 2005; Webb et al. 2006, hereafter W06; Medeiros et al. 2008). Clustering techniques, applied to quantify the contribution of different cloud types, reveals some of the details in the cloud feedback (e.g., Jakob and Tselioudis 2003; Gordon et al. 2005; Williams and Tselioudis 2007; Williams and Webb 2008, hereafter WW08). However, such methods require data with high frequency, and thus it is not always suitable for the analysis of the model simulations with a large number of ensemble members where storage of model output is a limiting factor.

In the present study, features of ECS of the two PPEs generated by the structurally different GCMs are investigated. We perform a climate feedback analysis called approximate partial radiative perturbation (APRP), in which the problem in the conventional cloud radiative forcing (CRF) method that the noncloud component is undesirably included in the cloud component is fixed in the diagnosis of the shortwave (SW) radiative feedback (Taylor et al. 2007; Yokohata et al. 2008). In addition, we develop a method to quantify the contribution of SW feedback by clouds with different cloud-top pressure. The method uses climatological monthly mean variables of the model output, and thus can be applied to a large GCM ensemble. We also investigate the relationship between the cloud feedback processes and the present-day climate states and compare the latter with observations.

2. Model and experiment

Two PPEs created from structurally different GCMs are used for the analysis. In both ensembles, preindustrial control (1XCO2) and doubled CO2 (2XCO2) experiments are performed. One of the ensembles is of the slab ocean version of the MIROC3.2 GCM, which was jointly developed by the Center for Climate System Research, University of Tokyo, the National Institute for Environmental Studies, and the Frontier Research Center for Global Change in Japan (K-1 Model Developers 2004). The atmospheric component of MIROC3.2 used for the ensemble has a reduced resolution of 5.6 longitude by 5.6 latitude (T21) with 20 levels compared to the standard T42 version. This is coupled to a motionless 50-m-depth slab ocean. Ocean heat transport is represented by the so-called q-flux term, which is calculated in the usual way in a calibration phase with imposed observed seasonally varying sea surface temperature and sea ice distributions. The nonconvective cloud parameterization is based on Le Treut and Li (1991), who calculate the cloud condensate content from a mass budget equation. The mass ratio of liquid and ice cloud water is diagnosed from atmospheric temperature (Ogura et al. 2008). Moist convective parameterization is based on Arakawa and Schubert (1974) with several modifications by Numaguti et al. (1997).

The MIROC3.2 ensemble is generated using an Ensemble Kalman Filter (EnKF) method for parameter estimation, details of which are described in Annan et al. (2005a,b). The numerical experiments were performed within the Japan Uncertainty Modeling Project (JUMP). In this method, the EnKF is used to assimilate observational data into the model, thereby generating an ensemble of runs with a range of values for the uncertain parameters, all with relatively low mean error compared to present-day climatology. A previous application to MIROC3.2 is discussed in Annan et al. (2005a), and there are only three differences between that ensemble and the one considered here. First, the model is very slightly changed, with one bug being fixed, which relates to the calculation of the temperature over the sea ice. Second, the number of parameters allowed to vary was reduced from 25 to 13 (see Table 1 for the description), as some of the original parameters were found to have little influence on either the simulated past and future climate changes or the goodness of fit to present-day climate (Hargreaves et al. 2007). Third, one additional constraint was added to the assimilation, which is explained below. Since the q-flux term represents the ocean heat transport, the annual mean of globally averaged q-flux should be zero (the real ocean is not a significant net source or sink of heat). It was found that for the ensemble discussed in Annan et al. (2005b) the annual mean of globally averaged q-flux was significantly different from 0, up to around 10 W m−2, indicating a significant net radiation imbalance. We addressed this problem by adding an additional constraint on the net q-flux, which is easily undertaken within the EnKF method (Hargreaves and Annan 2006). In this case, we simply augment the model state with the average q-flux, and include an “observational estimate” of zero (with some uncertainty) on the net q-flux as one of the constraints on the present-day climate state. To quantify the difference between the ensemble members and the present-day climate, 15 data based variables—some from reanalyses—were used that describe the main features of the climate system (Annan et al. 2005b) including temperature, moisture, precipitation, radiation balance, and wind field.

The goodness of fit to the data of this new ensemble is consistent with that obtained in Annan et al. (2005b); however, there is a trend toward a higher ECS as the value of the average q-flux is decreased. The number of ensemble members generated was 40, but the ensemble members that exhibit a warming drift and do not reach steady states during their 2XCO2 simulation (70 years) are excluded from the analysis. The reason for the drift should be investigated in future work. We used 32 members without the warming drift for the analysis.

Although ECS of the standard version of MIROC3.2 is high when compared to the CMIP3 multimodel ensemble dataset (Yokohata et al. 2008), it has been shown that MIROC3.2 successfully simulates the present climate states as well as climate changes in the twentieth century (e.g., Kimoto 2005; Nozawa et al. 2005; Nagashima et al. 2006; Shiogama et al. 2006; Yokohata et al. 2005a).

The other ensemble is performed using HadSM3, the Third Hadley Centre Atmospheric Model (HadAM3) general circulation model coupled to a 50-m mixed layer ocean, developed at the Met Office (Pope et al. 2000; Williams et al. 2001). The atmospheric component has a resolution of 2.5° latitude by 3.75° longitude (N48) with 19 vertical levels. The cloud scheme is based on Smith (1990) and Gregory and Allen (1996), who solve a prognostic equation for total water (vapor, liquid, and ice). The moist and dry convection scheme is based on Gregory and Rowntree (1990) and Gregory and Allen (1991), and the boundary layer parameterization is formulated by Smith (1990). A simple thermodynamic sea ice scheme is included. The q-flux is calculated in the calibration phase of each perturbed experiment in the same way as in the MIROC3.2 ensemble.

Numerical experiments are performed as part of the Met Office Hadley Centre Quantifying Uncertainty in Model Predictions (QUMP) project. The ensemble members of HadSM3 are produced by making simultaneous perturbations to multiple parameters in the atmosphere component of the model, which includes the sea ice scheme (W06; Murphy et al. 2007; Rougier et al. 2008, manuscript submitted to Climate Dyn.). Ensemble members used for this analysis are essentially those members described in W06. The number of parameters perturbed is 31 (description given in Table 2). From this set of ensemble members (128), we exclude 1 member with a cooling drift in the 1XCO2 and 2XCO2 experiment, which is generated by an unrealistic positive feedback caused by the interaction between negative SST anomalies and low cloud cover, which only occurs with the slab ocean (Stainforth et al. 2005, supplementary information http://dx.doi.org/10.1175/2010JCLI2917.s1). In W06, parameters were chosen in order to sample a wide range of uncertainty in model parameters and climate change feedbacks while minimizing root-mean-squared errors in mean climate variables. Although the methods are different between the MIROC3.2 and HadSM3 ensembles, the basic principle of both experiments is that multiple model parameters are allowed to vary simultaneously such that the resulting model climate states reasonably reproduce real observations. The HadSM3 experiment additionally targets a wide range of feedbacks and parameter values, such that this ensemble may cover a broader spread of results than the MIROC3.2 experiment achieves.

3. Method of analysis

To investigate the processes that determine the features of the climate feedback in the two ensembles, we decompose the shortwave and longwave (LW) radiative feedback into the contributions from clouds, clear-sky atmosphere, and surface. For the SW radiative feedback, these contributions are diagnosed by the approximate partial radiative perturbation (APRP) method described by Taylor et al. (2007). The APRP method produces values for the SW clouds and surface (i.e., ice–albedo) feedback that are very close (less than 10%; Taylor et al. 2007) to those obtained using the partial radiative perturbation (PRP) method (e.g., Wetherald and Manabe 1988) and thus more accurate when compared to the conventional cloud radiative forcing methods (e.g., Cess et al. 1990), which can erroneously include the noncloud feedback component in the cloud feedback (e.g., Colman 2003; Soden et al. 2004; W06).

The planetary albedo A is described as a function of seven parameters
i1520-0442-23-6-1392-e1
where c is cloud cover, αclr and αoc are surface albedo (reflectivity) under clear-sky atmosphere (clr) and overcast (oc) regions, 1 − μclr and 1 − μcld represent the absorptivity of clear-sky atmosphere (clr) and cloud (cld), and γclr and γcld are the albedo of clr and cld (Taylor et al. 2007). These seven parameters can be diagnosed with a straightforward calculation using the model output of SW radiative fluxes of full-sky and clear-sky atmosphere at the surface and the top of the atmosphere (TOA) of each grid point. By perturbing the components of surface (αclr and αoc), clear-sky atmosphere (μclr and γclr), and cloud (c, μcld, and γcld) separately, these contribution to planetary albedo anomaly (ΔAsfc, ΔAclr, and ΔAcld) can be calculated. Here, anomaly, Δ, is defined as the difference between 2XCO2 and 1XCO2 experiments. For the perturbation calculation, we adopt a forward and backward substitution as described by Colman (2003), and shown in Eq. (12b) in Taylor et al. (2007).

The SW feedbacks by the surface, clear-sky atmosphere, and cloud are calculated by multiplying the downward SW radiation at the TOA and −ΔAsfc, −ΔAclr, and −ΔAcld, respectively (minus is applied because we define downward as positive), dividing by the globally averaged SAT anomaly (units are W m−2 K−1).

In this study, we advance the APRP analysis so that we can diagnose the contributions from clouds at different altitudes. The method uses the GCM output consistent with the satellite data from International Satellite Cloud Climatology Project (ISCCP; Rossow and Schiffer 1999), which is produced by the ISCCP simulator (Klein and Jakob 1999; Webb et al. 2001). By using the ISCCP simulator output, we divide the cloud cover, in-cloud albedo, and absorptivity in Eq. (1) into contributions from clouds with different cloud-top pressures (ptop).

The ISCCP simulator can diagnose cloud cover (areal fraction of cloud, cij) as a function of seven categories of cloud-top pressure (ptop, i = 1, 2, … , 7) and optical depth (τ, j = 1, 2, … , 7) at each grid point in GCMs. In the ISCCP simulator, areal fractions of clouds with different ptop are diagnosed by subdividing each grid column into 100 subcolumns and simulating the cloud overlapping under some assumptions (Klein and Jakob 1999; Webb et al. 2001). As described in WW08, τ in the ISCCP simulator can be converted to in-cloud albedo, aj (reflectivity over the cloud; value of aj is shown in Table 2 in WW08). Therefore, we can derive a total in-cloud albedo (γttl) and total cloud cover (cttl) by summing up seven categories of ptop and τ as
i1520-0442-23-6-1392-e2
i1520-0442-23-6-1392-e3
where Σi and Σj represents the summation of the variables with the seven categories of ptop and τ, respectively. The contribution from the clouds with seven categories of ptop to the total in-cloud albedo and total cloud cover can be described as
i1520-0442-23-6-1392-e4
i1520-0442-23-6-1392-e5
As for the absorptivity, we cannot evaluate the contributions from different cloud type from the ISCCP data. Therefore, the contributions of different cloud-top pressure to μcld are scaled by using and as follows:
i1520-0442-23-6-1392-e6
where μcld is the variable used in Eq. (1).
The variables and cttl in Eqs. (2) and (3) are essentially same as those in γcld and c in Eq. (1). Therefore, using Eqs. (4), (5), and (6), the planetary albedo given in (1) can be described as
i1520-0442-23-6-1392-e7
In the same way as the APRP analysis [Eq. (16b) in Taylor et al. 2007], the contribution from the ith categories of ptop cloud can be calculated by perturbing ci, , and using the forward and backward substitution as follows:
i1520-0442-23-6-1392-e8
From Eq. (8), we can find the contributions from clouds with different ptop to the total SW cloud feedback.

It should be noted that the in-cloud albedo (γcld) and cloud cover (c) used in APRP as shown in Eq. (1) are not exactly the same as those calculated from ISCCP simulator output as shown in Eqs. (2) and (3). This is mainly because the in-cloud albedo, aj, which determines the τ in the ISCCP simulator, is not exactly the same as the in-cloud albedo calculated in GCMs. Here, we multiply the ratio of and c/cttl by the and ci in Eqs. (4), (5), and (6) at each grid point of the GCMs, so that the sum of and ci become the same as γcld and c of APRP method. We confirmed that the results are not significantly different in cases where the ratio of and c/cttl is multiplied or not. The spatial correlation of between the two cases is more than 0.98 in the ensemble means of the MIROC3.2 and HadSM3 ensembles. In addition, the error—difference between the two cases—of ensemble mean of the globally averaged is only 10% both in MIROC3.2 and HadSM3. Note that from the definition of ptop, there is no cloud above clouds with ptop. Therefore, there is no interaction between the SW reflected by the clouds with ptop at the low level and the clouds with ptop at the higher levels. Because of this feature, the sum of the cloud cover of the clouds with different ptop is the total cloud cover as described in Eq. (5). In addition, the sum of albedo of the clouds with different ptop is also the total cloud albedo as described in Eq. (4). Because of this linearity, the residuals—differences between the SW cloud feedback and the sum of SW feedback by clouds with different ptop—is less than 1% in the MIROC3.2 and HadSM3 ensembles.

For the LW radiative feedback, the conventional CRF method (e.g., Cess et al. 1990; Boer and Yu 2003) is used for the analysis because we can use the model output to consistently compare the LW feedback between the MIROC and HadSM3 ensembles. In addition, the CRF method is a standard method used frequently in the literature (e.g., Bony and Dufresne 2005; W06; WW08), and thus by adopting this method we can compare our analyses to previous results obtained using the same method. Although Yokohata et al. (2005b) tried to develop an LW APRP method for LW feedbacks, but the results were no different from the LW CRF method. We note that the cloud and noncloud component of LW feedback cannot be cleanly separated in the present analysis, and the cloud-masking effect may cause a negative bias (Soden et al. 2004).

For the LW cloud feedback, we calculate the anomaly of the difference between LW full-sky and clear-sky radiative flux at the TOA, the radiative forcing component is removed and then divided by the globally averaged SAT anomaly (W m−2 K−1). The radiative forcing component for LW cloud feedback is calculated as the difference between full-sky and clear-sky component of the adjusted LW radiative forcing at the tropopause (Flw), calculated using the method of Tett et al. (2002) from the standard version of both MIROC3.2 (Yokohata et al. 2005a) and HadSM3 (W06). For the LW clear-sky atmosphere feedback we calculate the anomaly of LW clear-sky radiative flux at the TOA, the radiative forcing component (the clear-sky component of Flw) is removed and then divided by the globally averaged SAT anomaly.

We also perform the cloud class analysis used in W06. The cloud feedback is classified into eight types by the sign and relative strength of the SW and LW cloud feedback at each grid box. For example, the cloud feedback classified as positive or negative SW feedback (S+ or S) with neutral LW feedback (LN) can be regarded as the response of low clouds. The positive or negative LW feedback (L+ or L) with neutral SW feedback (SN) can be regarded as the response of high clouds. In W06, the ninth class is defined as the region with high surface reflectance mainly because the SW cloud feedback was diagnosed by CRF method and contaminated by the ice–albedo feedback. We do not consider this class because the SW cloud feedback in this study is clearly separated from the ice–albedo feedback in the APRP method.

Monthly mean climatological values are used for the calculation. Anomalies are calculated as the difference in 20-years’ average between the 2XCO2 and 1XCO2 experiments after the climate system reached equilibrium. The output of HadSM3 (N48) is interpolated to that of MIROC3.2 (T21) before the calculation in order to compare the two ensembles at the same resolution.

4. Results and discussion

a. Climate feedback analysis

1) Difference between the ensemble means

The ensemble mean of the ECS, the total climate feedback (Λ), and the strength of each climate feedback are shown in Fig. 1; Λ is calculated by dividing the globally averaged adjusted radiative forcing at the tropopause (F, details of which are described in section 3) by the ECS. Here, the sum of the individual climate feedbacks is almost the same as Λ. The residuals, the differences between the ensemble means of Λ and the sum of the individual climate feedback, are less than 5% for the MIROC3.2 and HadSM3 ensembles. The ensemble mean of the ECS in MIROC3.2 is larger than that of HadSM3. Since the radiative forcing in the standard T42 version of MIROC3.2 (3.1 W m−2) is smaller than that of HadSM3 (3.8 W m−2), the difference in mean ECS must be caused by the difference in climate feedback processes.

As shown in Fig. 1, the difference in Λ between the means of the two ensembles can be explained largely by the SW cloud feedback (61% of the difference). The LW clear-sky atmosphere feedback (21%) and SW clear-sky atmosphere feedback (11%) also contribute significantly to the difference in Λ. The absolute values of the LW clear-sky atmosphere feedback (which includes the water-vapor, lapse-rate feedback and Stephan–Boltzmann damping) are large, but the differences between the two ensemble means and within-ensemble variations are relatively small compared to those in SW cloud feedback. This is possibly because uncertainties in water-vapor and lapse-rate feedback tend to cancel each other out (e.g., Bony et al. 2006). The MIROC3.2 ensemble has larger SW clear-sky atmosphere feedback than the HadSM3 ensemble, which suggests that the former has the larger SW component of water vapor feedback. It is likely that this is caused by greater absorption of SW by water vapor in MIROC3.2.

Figure 2 shows the ensemble mean spatial map of SW and LW feedback of MIROC3.2, HadSM3 and their difference (MIROC3.2 minus HadSM3). The SW cloud feedback in MIROC3.2 is positive over the entire tropical ocean which causes the main difference between the two ensembles. This feature is consistent with that found in Yoshimori et al. (2009) in which the climate feedback analysis by the PRP method was performed using the standard T42 slab ocean version of MIROC3.2. In contrast, HadSM3 has negative SW cloud feedback in the equatorial Pacific and Atlantic. Negative SW cloud feedback in the Southern Ocean is a common feature of both the ensemble means. The difference in the SW cloud feedback between the two ensembles is discussed in section 4b. The value of the SW clear-sky atmosphere feedback is not particularly large in any one place, but has a significant global mean value in both ensembles. The SW surface feedback, namely the ice–albedo feedback, is larger in MIROC3.2 than in HadSM3, especially in the Southern Ocean. This is possibly because the initial (1XCO2) sea ice concentration in MIROC3.2 is larger than that of HadSM3 (not shown), as discussed in Yokohata et al. (2008). Features of the LW cloud feedback of the two ensemble means are generally similar, although differences are seen in the equatorial Pacific. The negative LW cloud feedback at the low latitudes would correspond to the decrease in high cloud over the Southern Pacific convergence zone (SPCZ; e.g., W06), or to the cloud-masking effect (Soden et al. 2004). The difference in the global LW clear-sky atmosphere feedback between the two ensembles mainly comes from the ocean.

2) Variance of the ensemble members

In Fig. 3, the fractional contributions of the globally averaged SW and LW climate feedback processes to the variance of the total feedback in each ensemble are shown. The fractional contribution can be calculated in the same way as W06 and Boer and Yu (2003),
i1520-0442-23-6-1392-e9
where k is the index identifying the ensemble members (k = 1, n), l is the index identifying the components whose fractional contribution is calculated (l = 1, 5), Λl,k is the strength of lth climate feedback of kth ensemble member, and Λk is the sum of the climate feedbacks in kth member. The overbar denotes the average of ensemble members, and σΛ2 is the variance of the total climate feedback in the ensemble in question. If the value of Vl is positive and large, the contribution of lth climate feedback to the variance in Λ is large. On the other hand, the negative value of Vl means that lth climate feedback is anticorrelated with Λ.

As shown in Fig. 3, the largest contribution to the ensemble variance of total climate feedback is the SW cloud feedback in both MIROC3.2 and HadSM3. In HadSM3, the LW cloud feedback also contributes to the total spread. The LW clear-sky atmosphere feedback has a large contribution in both the ensembles. It is interesting to note that in the MIROC3.2 ensemble, the SW surface (i.e., ice–albedo) feedback has negative value of the fractional contribution, which means that variations in these feedbacks are anticorrelated with variations in the total climate feedback.

Scatterplots (Fig. 4) give information about the correlation between the total climate feedback (Λ) and components of climate feedback, as well as identifying overlapping regions of the strength of climate feedbacks in the two ensembles. The SW cloud and LW clear-sky atmosphere feedback are highly correlated with Λ in both ensembles (correlation coefficient of 0.72 and 0.77 in MIROC3.2, 0.58, and 0.92 in HadSM3). For the SW cloud feedbacks, the region of overlap of the two ensembles is small, hence the difference in the ensemble mean climate responses. Members with large Λ in the MIROC3.2 ensemble tend to have large values of the LW clear-sky atmosphere feedback, which is generally larger than that the majority of ensemble members in HadSM3. In HadSM3, the LW cloud feedback is also well correlated with Λ (correlation coefficient of 0.66), while it is not in MIROC3.2 (−0.04).

b. Radiative feedback by clouds with different altitudes

Next, we focus our attention on the SW and LW feedback by clouds at different altitudes, because the cloud feedbacks are the most important factors that determine the difference in climate response between the two ensemble means, as well as the within-ensemble variation (Figs. 1 and 3). For this purpose, we perform analysis of the SW feedback by clouds with different ptop as outlined in section 3, and the cloud class analysis presented in W06.

1) Difference between the ensemble means

Figure 5 shows globally averaged SW cloud feedback by the clouds within seven categories of ptop (Rossow and Schiffer 1999). The seven categories are hereafter called prs1 (<180 hPa), prs2 (180–300 hPa), prs3 (300–440 hPa), prs4 (400–560 hPa), prs5 (560–680 hPa), prs6 (680–800 hPa), and prs7 (800–1000 hPa). In accordance with previous work (e.g., Rossow and Schiffer 1999), these categories are further summarized into three: high level (<440 hPa), middle level (440–680 hPa), and low level (680–1000 hPa). In addition, we call the cloud with ptop at the high (middle, low) level as “high- (middle-, low-) level cloud.”

As shown in Fig. 5a, the globally averaged SW cloud feedback by high-level cloud is negative in the two ensembles, while the feedback by the middle- and low-level clouds is positive. The largest contribution to the difference between the two ensembles is from the low-level cloud. This result is consistent with previous works (e.g., Bony and Dufresne 2005; W06; Williams and Tselioudis 2007; Medeiros et al. 2008; WW08), which demonstrated that low-level cloud is important for determining differences in the cloud feedback among many different climate models. Our results indicate that the contribution to the difference in the SW cloud feedback between the MIROC3.2 and HadSM3 ensemble means from the high-, middle-, and low-level cloud is −34%, 14%, 120% (when we define the total difference in the SW cloud feedback to be 100%), respectively.

Figure 6 shows the ensemble mean map of SW feedback by the high-, middle-, and low-level clouds. The mean SW feedback by the high- and middle-level clouds is similar for the two ensembles, but the strengths of the SW feedback by the low-level cloud are substantially different between the two ensembles. In MIROC3.2, the value of low-level cloud feedback is positive over the large areas across Pacific and Atlantic, which corresponds to the stratocumulus/cumulus transition or stratocumulus regime in WW08. The negative feedback by the low-level cloud over the SPCZ in MIROC3.2 corresponds to the cumulus regime in WW08. Similarly, positive feedback by the low-level cloud in HadSM3 over the eastern Pacific corresponds to the stratocumulus/cumulus transition or stratocumulus regime in WW08. In addition, the SW feedback by the low-level cloud is negative over the equatorial Pacific and the Southern Ocean, which corresponds to the stratocumulus regime and cumulus regime in WW08, respectively.

Figure 7 shows the globally averaged values of SW and LW feedback separated into the cloud feedback classes defined by W06. Contributions from classes A and E (S+LN and SLN) are primarily due to changes in low clouds. Similarly B and F (S+L and SL+) relate to optically thick high clouds, and C and G (SNL+ and SNL) to optically thin high clouds. Classes D and H (SL and S+L+) are indicative of situations where changes in low and high clouds combine to give SW and LW feedbacks of the same sign; for example, where increases in cirrus coincide with reductions in low-level clouds.

As shown in Fig. 7a, the largest contribution to the difference in SW + LW cloud feedback between the MIROC3.2 and HadSM3 ensembles is from classes A + E (67%, when we define the total difference in the SW + LW cloud feedback to be 100%), which we attribute to low clouds. Smaller differences are also seen in (C + G) and (D + H) (13% and 22%, respectively), which we attribute to optically thin high cloud changes in the former case and a combination of these with low cloud changes in the latter. Classes (B + F) contribute very little to the net (−2%).

Figures 7b,c,d shows that these differences are mainly due to positive SW feedback from low clouds (classes A and H) and to a lesser extent from a positive LW feedback from thin cirrus (classes G and H). For classes B and F, the feedback by optically thick clouds is small because the SW and LW feedback cancels each other out.

2) Variance of the ensemble members

Figure 8 shows the fractional contribution of the clouds with different ptop to the within-ensemble variation of the global average of the SW cloud feedback. Consistent with previous research (e.g., W06; Bony and Dufresne 2005; Williams and Tselioudis 2007; Medeiros et al. 2008; WW08), the SW feedback by the low-level cloud plays dominant role in determining the within-ensemble variation of the total SW cloud feedback in the two ensembles. A feature common to both ensembles is that the contribution from the low-level cloud is the largest, and the one from the middle-level cloud is the smallest. It is interesting to note that the fractional contributions from the low-, middle-, and high-level clouds are very similar for the two ensembles.

However, while the largest contribution is from the cloud with ptop at the lowest level (800–1000 hPa) in MIROC3.2, it is that at the second lowest (680–800 hPa) in HadSM3. Note that both ensembles underestimate the 1XCO2 middle-level cloud compared to observations (WW08), which might cause an underestimation of the contribution from middle-level cloud.

Figures 9a,b shows that the sum of classes A and E (S+LN and SLN, attributable low cloud changes) dominates the within-ensemble variation of the cloud feedback in both the MIROC3.2 and HadSM3 ensembles. It is also worth noting that the fractional contributions are very similar for the two ensembles. Figures 9c,d show that class E, the negative SW low cloud feedback (SLN), dominates the total variance of the cloud feedback in both the MIROC3.2 and HadSM3 ensembles.

Although classes B and F (S+L and SL+) do contribute to the within-ensemble variation of LW cloud feedback (Figs. 9g,h), these classes do not contribute to the variance of the SW + LW cloud feedback (Figs. 9c,d). As shown in Fig. 7, this is because SW and LW feedback cancel each other out. These results indicate that even though the LW cloud feedback does contribute to the within-ensemble variation of climate feedback, especially in the HadSM3 ensemble (Fig. 3), the contribution comes mainly from classes B and F, and thus the LW cloud feedback from these classes are canceled by the SW cloud feedback when the two are combined.

c. Relationship between climate feedback and the present climate states

Next, we investigate the relationship between the SW low-level cloud feedback and the 1XCO2 climate states, because the SW low-level cloud feedback determines the difference in climate response between the two ensemble means as well as the within ensemble variation (Figs. 5, 7, 8, 9).

The SW cloud feedback can be mainly caused by changes in the cloud albedo (ratio of SW radiation reflected by the cloud to the insolation for each grid box) because the SW absorption by clouds is generally small compared to the reflection [e.g., Lambert and Webb (2008), who demonstrated this using HadSM3 ensembles]. Using Eqs. (4) and (5), the albedo of clouds with different ptop at each grid box can be represented as . Here, denotes in-cloud albedo (reflectivity over the cloud). Therefore, an anomaly in cloud albedo can be described as
i1520-0442-23-6-1392-e10
where Δ denotes the difference between the 2XCO2 and 1XCO2 experiments, and variables without Δ denotes those in the 1XCO2 experiment.

Figure 10a shows the difference in the ensemble means of of the low-level (680–1000 hPa) cloud between the MIROC3.2 and HadSM3 ensembles, which is essentially the same as the difference in the SW feedback by the low-level cloud as shown in Fig. 6 (bottom right). As discussed in section 4b, the SW feedback by the low-level cloud in the MIROC3.2 ensemble mean is larger than that in the HadSM3 ensemble mean. This is because the decrease in the low-level cloud albedo of the MIROC3.2 ensemble mean is larger than that of the HadSM3 ensemble mean, which is evident in Fig. 10a. Comparing the spatial pattern of Fig. 10a with that of Figs. 10b,c, the first and second term in rhs of Eq. (10), it is found that the first term is dominant in determining the magnitude of the lhs of Eq. (10) between the two ensembles.

Figure 10d shows the difference in Δci of the low-level cloud between the two ensembles, which indicates that the decrease in cloud cover of the MIROC3.2 ensemble mean is larger than that of HadSM3 ensemble mean. In addition, Fig. 10e shows the difference in 1XCO2 state of cloud cover (ci) of the low-level cloud between the two ensembles. From the spatial pattern of Figs. 10d,e, it is found that the difference in the change in cloud cover (Δci) between the two ensembles, especially over the mid-to-low latitude region, is opposite to the difference in 1XCO2 cloud cover (ci); relative to HadSM3, MIROC3.2 has the larger 1XCO2 cloud cover over the region where the decrease in cloud cover under climate change is larger. This result suggests that the decrease in cloud cover of the MIROC3.2 ensemble mean is larger than the HadSM3 ensemble mean because 1XCO2 cloud cover of the former is larger than that of the latter. Figure 10f shows the difference in the 1XCO2 states of in-cloud albedo of the low-level cloud. MIROC3.2 has larger value than HadSM3 almost over the entire world, which enhances the difference in cloud albedo change by the first term in Eq. (10). These results suggest that the decrease in cloud albedo of MIROC3.2 is larger because the 1XCO2 and ci are larger than those of HadSM3. It is plausible, therefore, that the difference in the ensemble means of the SW feedback by the low-level cloud between MIROC3.2 and HadSM3 can be explained by the difference in cloud properties in the 1XCO2 experiment. This is consistent with the result obtained by WW08.

In contrast, the variance of the SW cloud feedback within the ensemble members cannot be explained simply by this mechanism. Figure 11 shows the relationship between the SW feedback by the clouds with ptop at the second lowest (680–800 hPa) and lowest (800–1000 hPa) level and the 1XCO2 cloud properties with the same ptop. As shown in Fig. 11a, the correlation between the SW feedback and 1XCO2 albedo of the clouds with ptop at the second lowest level is statistically significant in HadSM3 (correlation coefficient of −0.76; note that the 99% significant correlation coefficient from the Student’s t test, assuming independent samples, is ∼0.24 for the 125 member HadSM3 ensemble, and ∼0.45 for the 32 member MIROC3.2 ensemble).

The HadSM3 ensemble members with large cloud albedo in the 1XCO2 experiment have large negative feedback. This is possibly because the cloud cover of the second-lowest-level cloud in the 1XCO2 experiment (ci) is large and the increase in optical depth (Δγi) of the second-lowest-level cloud causes large negative feedback. As shown in Fig. 11b, the SW feedback by the change in in-cloud albedo [ciΔγi, the second term in Eq. (10) of the second lowest level] is strongly anticorrelated (−0.75) with the 1XCO2 cloud cover (ci), which suggests that the cloud coverage in the 1XCO2 experiment plays an important role. In addition, the SW feedback by the second-lowest-level cloud (Δγici) is also anticorrelated with the 1XCO2 cloud albedo (ciγi) of the lowest level (−0.87), as shown in Fig. 11c. This suggests that the cloud top at the lowest level in the 1XCO2 experiment is lifted to the second lowest level, which caused a negative feedback at the second lowest level.

On the other hand, these relationships cannot be seen in the MIROC3.2 ensemble. The SW feedback by the second lowest cloud in MIROC3.2 is almost zero, and the spread in 1XCO2 cloud albedo is also very small compared to HadSM3 (Figs. 11a,b,c). However, it is interesting to note that the value of feedback by the second lowest cloud in MIROC3.2 is similar to that in ensemble members of HadSM3, which have similar value of 1XCO2 cloud albedo.

In Fig. 11d, the scatterplot of the feedback in the lowest level (800–1000 hPa) cloud and corresponding 1XCO2 cloud albedo is also shown. At the lowest level, the relationship between the SW cloud feedback and 1XCO2 cloud albedo is not similar to that of the second lowest level, and the correlation is not significant for either the MIROC3.2 or HadSM3 ensembles (0.064 and 0.22).

d. Comparison of model simulation with observation

Results obtained in section 4c suggest that the differences in the cloud feedback between the two ensembles and the within-ensemble variation are related to the difference in the 1XCO2 cloud properties. Therefore, the reality of the 1XCO2 cloud properties should be investigated by comparing the model simulation with observations. Here, the observational data averaged over five years 1986–1990 (ISCCP-D2; Rossow and Schiffer 1999) is used. It should be noted that there are some inconsistencies between the algorithm used to retrieve cloud-top pressure in the ISCCP retrievals and the equivalent algorithm in the version of the ISCCP simulator used here. This may result in clouds with the same physical top height being diagnosed in the second lowest (680–800 hPa) category of the ISCCP retrievals, but the lowest (800–1000 hPa) in the models (S. A. Klein 2008, personal communication). To avoid unfairly penalizing the models, we compare model simulations and observations after combining the two lowest categories (680–1000 hPa).

Global maps of the difference in optical properties of the low-level cloud between the ensemble means and the observations are shown in Fig. 12. The cloud albedo (γcldc) is overestimated in both MIROC3.2 and HadSM3. The overestimation of γcldc mainly comes from the overestimation of in-cloud albedo (γcld). Conversely, the cloud cover (c) is underestimated in the both ensemble means, especially over the stratocumulus regions.

Our results in Fig. 11 indicate that the optical properties of the low-level cloud in the 1XCO2 experiment are possibly responsible for the strength of cloud feedback there. As shown in sections 4a and 4b, the low-level cloud feedback determines a large fraction of the mean and variance of the ECS. These results suggest that the 1XCO2 optical properties of the low-level cloud play an important role in determining mean and spread of ECS. The relationship between the total climate feedback (Λ) and the difference in the low-level cloud cover between the 1XCO2 model simulations and the observation are shown in Fig. 13. The correlation between these two variables is significant in HadSM3 (−0.81), but not in MIROC3.2 (−0.16).

Of course, the cloud feedback and ECS are not only determined by the optical properties of the low-level cloud in the 1XCO2 states. However, Figs. 11 and 13 clearly demonstrate that some factors in the 1XCO2 experiment can be related to the strength in climate response. In addition, this kind of relationship is not necessarily universal, but depends on an ensemble of model used for analysis. Our results illustrate that it is important for us to reproduce the cloud optical properties in the 1XCO2 experiment as realistically as possible, and also to understand the mechanism of the relationship between the cloud feedback and the 1XCO2 climate.

5. Conclusions

In this study, the equilibrium climate sensitivity (ECS) of two perturbed physics ensembles generated using structurally different GCMs, MIROC3.2 and HadSM3, are investigated. We find that the difference in the ensemble means of the ECS between the two ensembles is mainly caused by differences in the ensemble mean SW low-level cloud feedback. These differences occur in similar spatial patterns to that of the differences in the 1XCO2 low-level cloud albedo between the two ensembles. Therefore, the reason that the SW cloud feedback and ECS of the MIROC3.2 ensemble is larger than that of the HadSM3 ensemble is likely because the 1XCO2 low-level cloud albedo of the former is larger than that of the latter. These results are consistent with previous results from multimodel ensemble studies (Webb et al. 2006; Williams and Webb 2008).

We also find that the largest contribution to the within-ensemble variation of ECS comes from within-ensemble variations in the SW low-level cloud feedback in both ensembles. Although the global LW cloud feedback does make some contribution to the within-ensemble variation of ECS, the cloud class analysis shown in Fig. 9 indicates that the class with negative LW cloud feedback cancels out the corresponding positive SW cloud feedback (class B), and that cloud class with positive LW feedback cancels out the negative SW feedback (class F). Although the factors that cause the within-ensemble variation of ECS are the same in the two ensembles, the mechanism that causes the within-ensemble variation is subtly different. In HadSM3, ensemble members with large 1XCO2 low-level cloud albedo have large negative SW cloud feedback and small ECS. This is possibly because the ensemble members with large 1XCO2 cloud cover have large negative SW cloud feedback in the HadSM3 ensemble. In the MIROC3.2 ensemble, however, no relationship is found between the cloud feedback and 1XCO2 low-level cloud albedo. The spread of the 1XCO2 low-level cloud albedo in the MIROC3.2 ensemble is much smaller than that of the HadSM3 ensemble.

This study reveals similarities between the two ensembles; for example, the SW low-level cloud feedback and LW clear-sky feedback determine the within-ensemble variation of ECS as shown in Figs. 3, 8, and 9. However, mechanisms that determine the within-ensemble variation are different between the two ensembles, as shown in Figs. 4, 11, and 13. These results indicate that parametric uncertainties sampled in a single PPE may not cover the range of possible uncertainties in all possible climate models. In addition, we also find that mechanisms that determine the difference in ECS between the ensemble means are not necessarily the same as those that determine the within-ensemble variation. For understanding the uncertainties in climate prediction, therefore, it is important to investigate the structural uncertainties of climate models demonstrated in the present study.

Acknowledgments

TY, JCH, and JDA were supported by Innovative Program of Climate Change Projection for 21st Century by MEXT, Japan. MC, MJW, and KDW were supported by the Joint DECC, Defra and MoD Integrated Climate Programme–DECC/Defra (GA01101), MoD (CBC/2B/0417 Annex C5). MY, JCH, and JDA were supported by the Global Environment Research Fund (S-5) of the MoE, Japan. The simulations of the MIROC ensemble were performed on the Earth Simulator. We wish to thank anonymous reviewers for giving valuable comments and suggestions.

REFERENCES

  • Annan, J. D., , J. Hargreaves, , N. Edwards, , and R. Marsh, 2005a: Parameter estimation in an intermediate complexity earth system model using an ensemble Kalman filter. Ocean Modell., 8 , 135154.

    • Search Google Scholar
    • Export Citation
  • Annan, J. D., , J. Hargreaves, , R. Ohgaito, , A. Abe-Ouch, , and S. Emori, 2005b: Efficiently constraining climate sensitivity with paleoclimate simulations. Sci. Online Lett. Atmos., 1 , 181184.

    • Search Google Scholar
    • Export Citation
  • Arakawa, A., , and W. H. Schubert, 1974: Interactions of cumulus cloud ensemble with the large-scale environment. Part I. J. Atmos. Sci., 31 , 671701.

    • Search Google Scholar
    • Export Citation
  • Boer, G. J., , and B. Yu, 2003: Climate sensitivity and response. Climate Dyn., 20 , 415429.

  • Bony, S., , and J-L. Dufresne, 2005: Marine boundary layer clouds at the heart of tropical cloud feedback uncertainty in climate models. Geophys. Res. Lett., 32 , L20806. doi:10.1029/2005GL023851.

    • Search Google Scholar
    • Export Citation
  • Bony, S., and Coauthors, 2006: How well do we understand and evaluate climate change feedback processes? J. Climate, 19 , 34453482.

  • Cess, R. D., and Coauthors, 1990: Intercomparison and interpretation of climate feedback processes in 19 atmospheric general circulation models. J. Geophys. Res., 95 , 1660116615.

    • Search Google Scholar
    • Export Citation
  • Collins, M., 2007: Ensembles and probabilities: A new era in the prediction of climate change. Philos. Trans. Roy. Soc. London, A365 , 19571970.

    • Search Google Scholar
    • Export Citation
  • Colman, R. A., 2003: A comparison of climate feedbacks in general circulation models. Climate Dyn., 20 , 865873.

  • Gordon, N. D., , J. R. Norris, , C. P. Weaver, , and S. A. Klein, 2005: Cluster analysis of cloud regimes and characteristic dynamics of midlatitude synoptic systems in observations and a model. J. Geophys. Res., 110 , D15S17. doi:10.1029/2004JD005027.

    • Search Google Scholar
    • Export Citation
  • Gregory, D., , and P. R. Rowntree, 1990: A mass flux convection scheme with representation of cloud ensemble characteristics and stability dependent closure. Mon. Wea. Rev., 118 , 14831506.

    • Search Google Scholar
    • Export Citation
  • Gregory, D., , and S. Allen, 1991: The effect of convective scale downdrafts upon NWP and climate simulations. Preprints, Ninth Conf. on Numerical Weather Prediction, Denver, CO, Amer. Meteor. Soc., 122–123.

    • Search Google Scholar
    • Export Citation
  • Gregory, D., , and S. Allen, 1996: The sensitivity of climate simulations to the specification of mixed phase cloud. Climate Dyn., 12 , 641651.

    • Search Google Scholar
    • Export Citation
  • Hargreaves, J. C., , and J. D. Annan, 2006: Using ensemble prediction methods to examine regional climate variation under global warming scenarios. Ocean Modell., 11 , 174192.

    • Search Google Scholar
    • Export Citation
  • Hargreaves, J. C., , A. Abe-Ouchi, , and J. D. Annan, 2007: Linking glacial and future climate through an ensemble of GCM simulations. Climate Past, 3 , 7787.

    • Search Google Scholar
    • Export Citation
  • Jakob, C., , and G. Tselioudis, 2003: Objective identification of cloud regimes in the tropical western Pacific. Geophys. Res. Lett., 30 , 2082. doi:10.1029/2003GL018367.

    • Search Google Scholar
    • Export Citation
  • K-1 Model Developers, 2004: K-1 coupled GCM (MIROC) description. K-1 Tech. Rep. 1, University of Tokyo, 1–34.

  • Kimoto, M., 2005: Simulated change of the East Asian circulation under global warming scenario. Geophys. Res. Lett., 32 , L16701. doi:10.1029/2005GL023383.

    • Search Google Scholar
    • Export Citation
  • Klein, S. A., , and C. Jakob, 1999: Validation and sensitivities of frontal clouds simulated by the ECMWF model. Mon. Wea. Rev., 127 , 25142531.

    • Search Google Scholar
    • Export Citation
  • Knutti, R., , G. A. Meehl, , M. R. Allen, , and D. A. Stainforth, 2006: Constraining climate sensitivity from the seasonal cycle in surface temperature. J. Climate, 19 , 42244233.

    • Search Google Scholar
    • Export Citation
  • Lambert, F. H., , and M. J. Webb, 2008: Dependency of global mean precipitation on surface temperature. Geophys. Res. Lett., 35 , L16706. doi:10.1029/2008GL034838.

    • Search Google Scholar
    • Export Citation
  • Le Treut, H., , and Z. X. Li, 1991: Sensitivity of an atmospheric general circulation model to prescribed SST changes: Feedback effects associated with the simulation of cloud optical properties. Climate Dyn., 5 , 175187.

    • Search Google Scholar
    • Export Citation
  • Medeiros, B., , B. Stevens, , 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 , 49744991.

    • Search Google Scholar
    • Export Citation
  • Murphy, J., , D. M. H. Sexton, , D. N. Barnett, , G. S. Jones, , M. J. Webb, , M. Collins, , and D. A. Stainforth, 2004: Quantification of modelling uncertainties in a large ensemble of climate change simulations. Nature, 430 , 768772.

    • Search Google Scholar
    • Export Citation
  • Murphy, J., , B. B. B. Booth, , M. Collins, , G. R. Harris, , D. M. H. Sexton, , and M. J. Webb, 2007: A methodology for probabilistic predictions of regional climate change from perturbed physics ensembles. Philos. Trans. Roy. Soc. London, A365 , 19932028.

    • Search Google Scholar
    • Export Citation
  • Nagashima, T., , H. Shiogama, , T. Yokohata, , T. Takemura, , S. A. Crooks, , and T. Nozawa, 2006: The effect of carbonaceous aerosols on surface temperature in the mid twentieth century. Geophys. Res. Lett., 33 , L04702. doi:10.1029/2005GL024887.

    • Search Google Scholar
    • Export Citation
  • Nozawa, T., , T. Nagashima, , H. Shiogama, , and S. A. Crooks, 2005: Detecting natural influence on surface air temperature change in the early twentieth century. Geophys. Res. Lett., 32 , L20719. doi:10.1029/2005GL023540.

    • Search Google Scholar
    • Export Citation
  • Numaguti, A., , M. Takahashi, , T. Nakajima, , and A. Sumi, 1997: Description of CCSR/NIES atmospheric general circulation model. Supercomputer Monograph Report, Center for Global Environmental Research, National Institute for Environmental Studies, 1–48.

    • Search Google Scholar
    • Export Citation
  • Ogura, T., , S. Emori, , M. J. Webb, , Y. Tsushima, , T. Yokohata, , A. Abe-Ouchi, , and M. Kimoto, 2008: Towards understanding cloud response in atmospheric GCMS: The use of tendency diagnostics. J. Meteor. Soc. Japan, 86 , 6979.

    • Search Google Scholar
    • Export Citation
  • Piani, C., , D. J. Frame, , D. A. Stainforth, , and M. R. Allen, 2005: Constraints on climate change from a multi-thousand member ensemble of simulations. Geophys. Res. Lett., 32 , L23825. doi:10.1029/2005GL024452.

    • Search Google Scholar
    • Export Citation
  • Pope, V. D., , M. L. Gallani, , P. R. Rowntree, , and R. A. Stratton, 2000: The impact of new physical parametrizations in the Hadley Centre climate model: HadAM3. Climate Dyn., 16 , 123146.

    • Search Google Scholar
    • Export Citation
  • Randall, D. A., and Coauthors, 2007: Climate models and their evaluation. Climate Change 2007: The Physical Science Basis, S. Solomon, et al., Eds., Cambridge University Press, 589–662.

    • Search Google Scholar
    • Export Citation
  • Rossow, W. B., , and R. A. Schiffer, 1999: Advances in understanding clouds from ISCCP. Bull. Amer. Meteor. Soc., 80 , 22612287.

  • Shiogama, H., , T. Nagashima, , T. Yokohata, , S. A. Crooks, , and T. Nozawa, 2006: Influence of volcanic activity and changes in solar irradiance on surface air temperatures in the early twentieth century. Geophys. Res. Lett., 33 , L09702. doi:10.1029/2005GL025622.

    • Search Google Scholar
    • Export Citation
  • Smith, A., 1990: A scheme for predicting layer clouds and their water content in a general circulation model. Quart. J. Roy. Meteor. Soc., 116 , 435460.

    • Search Google Scholar
    • Export Citation
  • Soden, B. J., , and I. M. Held, 2006: An assessment of climate feedbacks in coupled ocean–atmosphere models. J. Climate, 19 , 33543360.

    • Search Google Scholar
    • Export Citation
  • Soden, B. J., , R. S. Broccoli, , and R. S. Helmer, 2004: On the use of cloud forcing to estimate cloud feedback. J. Climate, 17 , 36613665.

    • Search Google Scholar
    • Export Citation
  • Stainforth, D. A., and Coauthors, 2005: Uncertainty in predictions of the climate response to rising levels of greenhouse gases. Nature, 433 , 403406.

    • Search Google Scholar
    • Export Citation
  • Taylor, K. E., , C. Doutriaux, , M. Crucifix, , C. D. Hewitt, , J. F. B. Mitchell, , M. J. Webb, , P. Braconnot, , and A. J. Broccoli, 2007: Estimating shortwave radiative forcing and response in climate models. J. Climate, 20 , 25302543.

    • Search Google Scholar
    • Export Citation
  • Tett, S. F. B., and Coauthors, 2002: Estimation of natural and anthropogenic contributions to twentieth century temperature change. J. Geophys. Res., 107 , 4306. doi:10.1029/2000JD000028.

    • Search Google Scholar
    • Export Citation
  • Webb, M., , C. Senior, , S. Bony, , and J-J. Morcrette, 2001: Combining ERBE and ISCCP data to assess clouds in the Hadley Centre, ECMWF and LMD atmospheric climate models. Climate Dyn., 17 , 905922.

    • Search Google Scholar
    • Export Citation
  • Webb, M., and Coauthors, 2006: On the contribution of local feedback mechanisms to the range of climate sensitivity in two GCM ensembles. Climate Dyn., 27 , 1738.

    • Search Google Scholar
    • Export Citation
  • Wetherald, R. T., , and S. Manabe, 1988: Cloud feedback processes in a general circulation model. J. Atmos. Sci., 45 , 13971451.

  • Williams, K. D., , and G. Tselioudis, 2007: GCM intercomparison of global cloud regimes: Present-day evaluation and climate change response. Climate Dyn., 29 , 231250.

    • Search Google Scholar
    • Export Citation
  • Williams, K. D., , and M. J. Webb, 2008: A quantitative performance assessment of cloud regimes in climate models. Climate Dyn., 33 , 141157. doi:10.1007/s00382-008-0443-1.

    • Search Google Scholar
    • Export Citation
  • Williams, K. D., , C. A. Senior, , and J. F. B. Mitchell, 2001: Transient climate change in the Hadley Centre models: The role of physical processes. J. Climate, 14 , 26592674.

    • Search Google Scholar
    • Export Citation
  • Yokohata, T., , S. Emori, , T. Nozawa, , Y. Tsushima, , T. Ogura, , and M. Kimoto, 2005a: Climate response to volcanic forcing: Validation of climate sensitivity of a coupled atmosphere–ocean general circulation model. Geophys. Res. Lett., 32 , L21710. doi:10.1029/2005GL023542.

    • Search Google Scholar
    • Export Citation
  • Yokohata, T., , S. Emori, , T. Nozawa, , Y. Tsushima, , T. Ogura, , and M. Kimoto, 2005b: A simple scheme for climate feedback analysis. Geophys. Res. Lett., 32 , L19703. doi:10.1029/2005GL023673.

    • Search Google Scholar
    • Export Citation
  • Yokohata, T., and Coauthors, 2008: Comparison of equilibrium and transient responses to CO2 increase in eight state-of-the-art climate models. Tellus, 60 , 946961.

    • Search Google Scholar
    • Export Citation
  • Yoshimori, M., , T. Yokohata, , and A. Abe-Ouchi, 2009: A comparison of climate feedback strength between CO2 doubling and LGM experiments. J. Climate, 22 , 33743395.

    • Search Google Scholar
    • Export Citation

Fig. 1.
Fig. 1.

Total climate feedback parameter (Λ), and climate feedback by SW cloud, surface, clear-sky atmosphere, and LW cloud and clear-sky atmosphere (unit except for SAT anomaly is W m−2 K−1), and the globally averaged surface air temperature anomaly (i.e., equilibrium climate sensitivity). Shaded bars represent the ensemble means of MIROC3.2 (slash), HadSM3 (gray), and their difference (MIROC3.2-HadSM3, white), and the error bars represent the two ensemble standard deviations for the different quantities.

Citation: Journal of Climate 23, 6; 10.1175/2009JCLI2917.1

Fig. 2.
Fig. 2.

(first row) Global map of SW cloud, (second row) SW clear-sky atmosphere, (third row) SW surface, (fourth row) LW cloud, and (bottom row) LW clear-sky atmosphere feedback for (left) the ensemble mean of MIROC3.2, (middle) HadSM3, and (right) their difference. Feedback is calculated as the local change in radiative flux at the TOA divided by the global mean SAT change. Unit is W m−2 K−1.

Citation: Journal of Climate 23, 6; 10.1175/2009JCLI2917.1

Fig. 3.
Fig. 3.

Fractional contribution of the strength of climate feedback processes shown in Fig. 1 to the variance of their sum, for (a) MIROC3.2 and (b) HadSM3. See Eq. (9) and section 4a(2).

Citation: Journal of Climate 23, 6; 10.1175/2009JCLI2917.1

Fig. 4.
Fig. 4.

Scatterplot of the total climate feedback (Λ) and globally averaged climate feedback by (a) SW cloud, (b) SW clear-sky atmosphere, (c) SW surface, (d) LW cloud, and (e) LW clear-sky atmosphere. Each point represents an ensemble member of MIROC3.2 (red) and HadSM3 (blue), and straight lines represent the linear regressions. See section 4a(2) for more details.

Citation: Journal of Climate 23, 6; 10.1175/2009JCLI2917.1

Fig. 5.
Fig. 5.

Globally averaged values of SW feedback by clouds with different cloud-top pressure (ptop). Contributions from the cloud at high level (50–440 hPa), middle level (440–680 hPa), and low level (680–1000 hPa), and contributions from the seven categories of ptop are shown. Bars represent the ensemble mean (slash: MIROC3.2; gray: HadSM3; white: MIROC3.2-HadSM3), and error bars represent the two standard deviation. Unit is W−2 m−2 K−1.

Citation: Journal of Climate 23, 6; 10.1175/2009JCLI2917.1

Fig. 6.
Fig. 6.

Ensemble mean global map of the SW feedback by the cloud at (top) high level, (middle) middle level, and (bottom) low level for (left) MIROC3.2, (middle) HadSM3, and (right) their difference.

Citation: Journal of Climate 23, 6; 10.1175/2009JCLI2917.1

Fig. 7.
Fig. 7.

Globally averaged values of the SW and LW feedback by the cloud classes defined by Webb et al. (2006). Bars represent the ensemble mean (slash: MIROC3.2; gray: HadSM3; white: MIROC3.2-HadSM3), and error bars represent the two standard deviation. (a) Sum of the SW and LW feedback by the cloud classes A (S+LN) and E (S LN); B (S+L) and F (SL+); C (SNL) and G (SNL+); D (SL) and H (S+L+); (b) sum of the SW and LW feedback by the eight cloud classes; (c) same as (b) but for the SW feedback; and (d) same as (b) but for the LW feedback. Unit is W−2 m−2 K−1.

Citation: Journal of Climate 23, 6; 10.1175/2009JCLI2917.1

Fig. 8.
Fig. 8.

Fractional contributions of the globally averaged SW feedback by the clouds with different ptop to the within-ensemble variation of their sum, (a) MIROC3.2 and (b) HadSM3. Categories of ptop are the same as those in Fig. 5.

Citation: Journal of Climate 23, 6; 10.1175/2009JCLI2917.1

Fig. 9.
Fig. 9.

Fractional contributions of the globally averaged feedback by the cloud classes defined by Webb et al. (2006), to the within-ensemble variation of their sum. Components of the cloud classes are same as those of Fig. 7, but for the (a),(c),(e),(g) MIROC3.2 ensemble and (b),(d),(f),(h) HadSM3 ensemble.

Citation: Journal of Climate 23, 6; 10.1175/2009JCLI2917.1

Fig. 10.
Fig. 10.

Global map of (a) cloud albedo anomaly, (b) first term in Eq. (10), (c) second term in Eq. (10), (d) cloud cover anomaly, (e) 1XCO2 cloud cover, and (f) 1XCO2 in-cloud albedo, shown as the differences between the MIROC3.2 and HadSM3 ensemble means.

Citation: Journal of Climate 23, 6; 10.1175/2009JCLI2917.1

Fig. 11.
Fig. 11.

Relationship between (a) the SW cloud feedback at the second lowest level (ptop = 680–800 hPa) and 1XCO2 cloud albedo at the second lowest level, (b) the SW feedback by the change in in-cloud albedo at the second lowest level and 1XCO2 cloud cover at the second lowest level, (c) the cloud SW feedback at the second lowest level and 1XCO2 cloud albedo at the lowest level (ptop = 800–1000 hPa), and (d) the SW cloud feedback by the lowest level and 1XCO2 cloud albedo at the lowest level. The lines are linear regressions of the scatterplots.

Citation: Journal of Climate 23, 6; 10.1175/2009JCLI2917.1

Fig. 12.
Fig. 12.

The difference in (top) cloud albedo, (middle) cloud cover, and (bottom) in-cloud albedo at the low level (680–1000 hPa) between the ensemble means of model simulation (left is MIROC3.2 and right is HadSM3) and the observation (ISCCP-D2; Rossow and Schiffer 1999).

Citation: Journal of Climate 23, 6; 10.1175/2009JCLI2917.1

Fig. 13.
Fig. 13.

Same as in Fig. 11, but for the scatterplot of total climate feedback (Λ) and the cloud albedo at the low level (680–1000 hPa). Each point represents an ensemble member of MIROC3.2 (red) and HadSM3 (blue). Values of horizontal axis are shown as difference from ISCCP-D2 (Rossow and Schiffer 1999). The lines represent the linear regression for the scatterplots.

Citation: Journal of Climate 23, 6; 10.1175/2009JCLI2917.1

Table 1.

Varied parameters in JUMP experiment (MIROC3.2 ensemble).

Table 1.
Table 2.

Varied parameters in QUMP experiment (HadSM3 ensemble).

Table 2.

Supplementary Materials

Save