Clouds have a significant effect on the Earth's heat budget. Changes in clouds affect the temperature change in global warming. This is called cloud feedback and has posed the largest uncertainty in the study of climate sensitivity for almost twenty years (Bony et al. 2006; Soden and Held 2006). Change that occurs in low clouds represents the largest uncertainty (Webb et al. 2006; Bony and Dufresne 2005). Although several hypotheses have been proposed for the feedback stemming from high clouds (Lindzen 1990; Ramanathan and Collins 1991; Lindzen 2001), no consensus has yet been obtained (Lau et al. 1994; Lin et al. 2002). All of the hypotheses relate the changes in high clouds to those in deep convection. Thus it is essential to understand cumulus convection to discuss the high cloud feedback.

We have recognized the uncertainty of cumulus convection for a long time, but have not simulated it in detail. Spatial resolution of current global climate models (GCMs) for climate simulations is on the order of 100 km, by which we cannot resolve cumulus convection. Available computer resources present the main constraint on resolution. GCMs introduce cumulus parameterizations which try to simulate the essential role of cumulus convection on the GCM grid. We begin by describing climate change simulations based on traditional cloud and cumulus parameterizations and the open issues identifi ed in these simulations. Cloud-resolving simulations under idealistic conditions (Wu and Moncrieff 1999; Bretherton 2007; Tompkins and Craig 1999), global simulations with embedded cloud-resolving grids in a GCM grid (Grabowski 2001; Randall et al. 2003; Wyant et al. 2005), and global cloud-resolving simulations that resolve convective systems over the entire globe (Satoh et al. 2008; Miura et al. 2005, 2007) have been started and are the emerging approaches in global process-oriented cloud modeling. We discuss the prospects for new observations for evaluating cloud radiative effects and feedbacks in global models. In addition, we consider the response of cloud systems to an idealized global warming in a global cloud-resolving model (GCRM).

Projected Changes in Clouds and Cloud Effects

In global simulations of climate change with current generation models, cloud amounts are projected to decrease throughout most of the troposphere below 200 hPa and between 50°N and 50°S (IPCC 2007). Cloud cover is projected to increase in the upper extratropical troposphere with robust changes close to 200 hPa. Generally, the amplitude of the spatial patterns of cloud perturbation expand with increasing concentrations of CO2, but the magnitude of the changes remains less than 4% at 2100 for the canonical SRES A1B emissions scenario.

In part, because the vertical dipole of projected cloud amount changes, the forecast perturbations in column-integrated cloud amount are inconsistent across the multi-model ensemble for most regions on the globe. The models do not concur on either the magnitude or sign of changes in global mean cloud radiative effect, and the size of the cloud radiative effect changes relative to the present-day value is generally less than 10%; this is comparable to various forcings and current uncertainties in the top-of-atmosphere planetary energy budget. Thus, the model projections suggest that observational verification of global cloud radiative effect changes will be a significant challenge. The sign of radiative feedbacks from low clouds and deep convective systems is not consistent across the IPCC multi-model ensemble (Bony and Dufresne 2005). Tropical clouds, mixed-phase clouds and cloud phase, as well as intermodel differences in meridional shifts in storm tracks contribute the divergences in cloud responses (IPCC 2007).

Open Issues in Global Simulation of Perturbed Clouds Observational and Theoretical Tests

The necessary and suffi cient conditions for accurate global projections of clouds and cloud radiative effects are not known. The absence of a complete theory has complicated efforts to construct comprehensive observational tests of cloud parameterizations. It has also complicated efforts to derive these parameterizations through systematic simplifi cation of known cloud physics and process-oriented CRMs. It would be particularly useful to have an analog of the tests for snow, connecting the observed seasonal cycle in snow cover and the simulated snow-albedo feedbacks (Hall and Qu 2006). Some aspects of cloud perturbations from climate change project onto observed intra- and interannual cloud variations (Williams et al. 2006). Whether present-day seasonal variations in clouds should be used as a test for cloud radiative feedbacks in global models has not been definitively determined (Cess et al. 1992; Tsushima et al. 2005).

Scaling Issues and Parametric Uncertainty

Forecasting climate change with skill at local and regional scales has become increasingly necessary to assess societal and environmental impacts. Extreme rainfall events on small temporal and spatial scales constitute some of the most important impacts. Partly in response, the climate modeling community has increased the spatial resolution of their climate forecasts by a factor of five from the first to the fourth IPCC assessment reports (IPCC 2007). However, whether the simulated hydrological cycle, cloud features, and cloud feedbacks improve and converge with increasing resolution remains open. Some models exhibit systematic and apparently nonconvergent variations in the main features of the hydrological cycle (e.g. the Intertropical Convergence Zone, ITCZ), with increasing resolution (Williamson et al. 1995). The same models exhibit large, systematic changes in climate sensitivity, cloud radiative effects and feedbacks, and the probability distributions of rainfall with increasing resolution (Kiehl et al. 2006; Williamson et al. 2007). The parameterizations of clouds and hy-drological processes are generally adjusted to produce a realistic simulation at particular model resolution or resolutions. To date, these parameterizations have not usually been designed to yield climate simulations that are invariant to horizontal and vertical truncation scales.

Current parameterizations contain large numbers of free parameters that govern the evolution of micro- and macrophysical cloud properties in relation to the meteorological environment. As is well known, simulated climate radiative feedbacks and climate responses are quite sensitive to variations in these free parameters (Senior and Mitchell 1993; Murphy et al. 2004). It has proven difficult to exclude parameter settings that produce large positive cloud feedbacks and climate sensitivities to increasing greenhouse gas concentrations. Many of the parameters are only loosely constrained by observations or process-oriented modeling and are connected to the physics of cloud formation through scaling arguments.

Global Process-oriented Simulation Global Cloud-resolving Models and Multiscale Modeling Framework

Traditional climate models cannot explicitly resolve a number of climatically important cloud processes. It would be desirable to design and apply global models that explicitly simulate cloud system organization; relationships among cloud geometry, radiation, and precipitation; and the interactions between clouds and convection. Two new frameworks have been created for process-oriented cloud simulation: a GCRM (Satoh et al. 2008) and the multiscale modeling framework (MMF), originally introduced as super-parameterization (Grabowski 2001; Randall et al. 2003). Features of the GCRM and its simulations for cloud response to idealized climate change are discussed below.

MMF is essentially an enhancement of traditional GCMs to treat interactions between clouds and their mesoscale environment. In MMF-enhanced GCMs, two- or three-dimensional CRMs replace the physical parameterizations at each grid point. The grid-point tendencies in the prognosed fields (e.g., temperature, humidity, and momentum), calculated with the traditional parameter-izations, are overwritten by corresponding tendencies computed by the CRMs. MMF is an intermediate step towards a GCRM, and MMFs can be formulated to converge smoothly to a GCRM in the limit of very high spatial resolution. The MMF simplifies the interactions between clouds and their synoptic environment by averaging the boundary conditions applied to the CRMs and the tendencies applied to the host GCM at the scale of individual GCM grid points. Early results from MMF-enhanced GCMs show significant improvements relative to the original GCM in the strength, frequency, and characteristics of the Madden-Julian Oscillation and in the diurnal cycle of convective precipitation (Khairoutdinov et al. 2007). It is not yet known which cloud-scale processes are responsible for the improved fidelity of the climate simulations.

GCRM and MMF represent some of the first attempts to simulate global clouds with process-oriented models. Perhaps the most important question is whether these models can reduce the uncertainty in simulations of climate sensitivity and of cloud radiative feedbacks relative to current GCMs. Since there is only one GCRM and only a few MMF-enhanced GCMs, the characteristics of multi-model ensembles of these systems are still unresolved. The implications of explicit simulation of cloud processes for the coupled climate system are unknown. Although GCRM and MMF eliminate many of the ad hoc parameters governing cloud interactions at the mesoscale, there are still many free parameters for unresolved processes, including turbulence and cloud mi-crophysics. Given the computational expense of the GCRM and MMF, the sensitivity of the simulated climate to these remaining free parameters has not been determined. The convergence of the MMF with increasing spatial resolution of the grid in the host GCM has yet to be determined.

These frameworks could be used as benchmark or reference models for improving GCMs by quantifying the systematic tendency errors caused by conventional cloud parameterizations. The models have not been applied in this manner, and therefore the utility of GCRM and MMF for the improvement of conventional parameterizations has not been established.

Near-term Prospects for Advances in Global Numerical Modeling

The currently available GCRM is a non-hydrostatic icosahedral atmospheric model (NICAM) (Satoh et al. 2008). NICAM runs at horizontal mesh intervals of a few kilometers (3.5km, at most) over the globe using the Earth Simulator1 high performance computing system. Under these resolutions, deep convective clouds are marginally permitted with representing updraft cores explicitly. Thus far, NICAM has been used for many purposes. A realistic simulation of a Madden-Julian Oscillation event was produced with its internal cloud-system structures (Miura el al. 2007). In addition, NICAM produces realistic climatology close to that observed (Iga et al. 2007). NICAM is now coupled with an aerosol transport radiation model, SPRINTARS (Takemura et al. 2005), to study aerosol direct and indirect effects. Using 7 km mesh global cloud-resolving simulation by NICAM-SPRINTARS, Suzuki et al. (2008) examined vertical profiles of effective cloud particle radius as functions of cloud-top temperature in different regions, an approach similar to that by Rosenfeld (2000). Suzuki and Stephens (submitted) compared timescales of warm rain formation by comparing the GCRM results and CloudSat and MODIS combined data. Suzuki et al. (2008) further analyzed the simulated data to identify aerosol indirect effects on different cloud types and compared with satellite observations. Since the structures of shallow clouds are not resolved by the current GCRM simulations, aerosol indirect effects on shallow clouds depend on subgrid pa-rameterizations, as in GCMs. The advantage of using results of GCRMs is that convective movements are resolved. However, concern remains about the representation of ice phase microphysical processes. Evaluations of indirect effects on deep clouds from satellite observations are hard because contributions of cloud liquid and ice particles are difficult to resolve. This can partly be overcome by using an analysis of GCRM results to guide retrieval algorithms of satellite observations.

In principle, numerical data produced by GCRMs can be used to improve conventional cloud and cumulus parameterizations. For example, shapes of PDFs of cloud fraction using GCRM results can be evaluated and used to improve large-scale cloud schemes in GCMs (Watanabe et al., submitted). By analyzing the spectrum of momentum fluxes, we see resolution dependency and intend to use this to improve cumulus parameterization and momentum transport.

At present, GCRMs do not pose an alternative to current GCMs, but they should be viewed as a complementary tool. GCRMs are not yet coupled in the Earth system model and cannot be used for simulations on a temporal scale of hundreds of years. However, within the coming decade, we should be able to use an Earth system model by replacing current atmospheric GCMs with a

GCRM and using the results to study climate for time-slice experiments of a few years.

Related to GCMs, the MMFs replace the conventional cloud param-eterizations with a CRM in each grid column of a GCM (Grabowski 2001; Khairoutdinov and Randall 2001; Randall et al. 2003). The MMF can explicitly simulate deep convection, cloudiness and cloud overlap, cloud radiation interaction, surface fluxes, and surface hydrology at the resolution of a CRM. In addition, MMFs provide global coverage and two-way interactions between the CRMs and GCMs. MMF could be a natural extension of current cloud-resolving modeling activities, as they simulate reasonable features in the tropics, such as the diurnal cycle of precipitation and intraseasonal oscillations (Randall et al. 2003). They are also used to examine the cloud-aerosol interactive processes by implementing more sophisticated microphysics and coupling with global aerosol transport models. In the future, we expect MMFs to bridge the gap between traditional CRM simulations and current as well as future GCRMs.

Challenges in Observations Future Prospects for Remote Sensing

One of the principal predictions of climate models is that clouds and cloud effects will evolve in response to anthropogenic climate change (IPCC 2007). However, the projected changes in cloud radiative effects are small compared to the climatological mean cloud radiative effect. Depending on the rate of change, it may take decades for the signals in cloud radiative effect to become sufficiently large for robust detection relative to unforced natural variability and other secular trends in the climate system (e.g. greater forcing by CO2). Together, the small magnitude and long integration times imply that very stable, and preferably absolutely calibrated, satellite instruments are required to detect cloud radiative effect feedbacks. Although, to date, such instruments have not yet been deployed, stable and accurate observations of cloud radiative effects over several decades would be valuable in the evaluation of global simulations of clouds and climate. Direct measurement of the Earth's radiation field is preferable to calculations of cloud radiative effects based on retrieved cloud properties (Loeb et al. 2007). Spectrally resolved measurements are especially useful to separate and classify the radiative effects from climate forcing and climate response (Goody et al. 1998). None of the proposed satellites offers a stand-alone solution for better quantification of planetary albedo for tests of climate models, since albedo requires adequate sampling over eight independent parameters (Wielicki et al. 1996).

CLARREO (CLimate Absolute Radiance and Refractivity Observatory) offers one possibility, as it is designed to detect the radiative forcing, thermal response, and radiative feedbacks in the Earth's climate system (Space Studies Board 2007). The spectral radiometers in CLARREO are made for absolute calibration against traceable standards to ensure that trends observed in the observations are as free as possible of instrumental artifacts (Keith et al. 2001; Anderson et al. 2004). The payload of CLARREO will include infrared as well as ultraviolet, visible, and near-infrared (UV/VIS/NIR) radiometers. It could be used either as an orbiting calibration facility ("NIST in space") or as an Earth-observing platform in its own right. Although the feasibility for detection of infrared greenhouse gas forcing has been amply illustrated in modeling and satellite studies (e.g., Haskins et al. 1997), it is important to recognize that the utility of the infrared measurements or detection of longwave cloud feedbacks remains unproven (Leroy et al. 2006). The advantages of the UV/VIS/ NIR data for detection and estimation of shortwave forcings and feedbacks have not been studied or demonstrated in detail. Open questions include:

1. To what extent is it possible to isolate forcings and feedbacks associated with changes in specific species and processes in the CLARREO measurements?

2. Can the indirect shortwave forcings from aerosol-cloud interactions and the feedbacks from clouds be detected and quantified using CLARREO data?

3. Can changes in and longwave feedbacks from low, middle, and high clouds be detected and quantified using the CLARREO infrared data?

DSCOVR offers another possibiltiy. It has been designed for deployment at Lagrange point L1 from which it would measure the radiation emitted by the sunlight side of the Earth (Valero et al. 1999). DSCOVR would include several single-pixel NIST-advanced radiometers (NISTARs) with a ground-based calibration chain tied directly to primary national standards. These instruments would measure the total solar, near infrared, and infrared radiance field emitted by the Earth in the direction of L1. The inherent stability and traceable calibration of these instruments are ideally suited for the detection of secular trends in the Earth's short- and longwave radiation. Attribution of any observed changes to perturbations in clouds and cloud radiative effects will depend on auxiliary coincident measurements from multichannel imagers onboard DSCOVR.

Upcoming satellite experiments designed to connect clouds, aerosols, and radiative processes include the EarthCARE mission (ESA 2001). EarthCARE is designed to characterize the vertical and horizontal distributions of clouds and aerosols in the Earth's atmosphere. This data will be particularly useful for constraining (a) simulations of radiative energy divergence, (b) the interactions among radiation, aerosols, and clouds, and (c) the spatial distributions and interactions of cloud condensate and precipitation.

Global Cloud-resolving Simulations Models and Experiments

Here, we discuss the first global warming experiment results using NICAM with a realistic topography. To summarize, a significant increase in cirrus has been seen over the entire subtropics, which has strongly intensified temperature increase (see Figure 20.1). Centralized deep convection supplies more moisture to the tropopause. Such an increase in cirrus has never been found in the same tests by using current GCMs (Ringer et al. 2006). Our results suggest the vital role of cumulus convection for cloud feedback mechanism. Most of the results in this section are based on Tsushima et al. (submitted).

For the experimental setting, we choose the procedure of perpetual July and a perturbed sea surface temperature (SST) of +2K simulation (Cess and Potter 1988). Detailed experimental design and the reproducibility of the control climate follow Iga et al. (2007). For comparison with the current GCM, we conducted the same simulations using an atmospheric GCM of MIROC3.2 (Hasumi and Emori 2004) with a horizontal resolution of 2.8°, 20 levels in the vertical in the atmosphere. The model's climate sensitivity to a doubling of atmospheric CO2 concentration in the slab ocean experiments is 3.8°C (Ogura et al. 2008). In both models, we incorporated the ISCCP simulator (Schiffer and Rossow 1983; Webb et al. 2001), which diagnoses values of cloud's optical thickness and cloud-top pressure from the models in a manner consistent with the view of satellites from space. Cloud height and optical thickness are important metrics for cloud radiative effects, and it is useful to analyze clouds according to the ISCCP categories when we discuss the change in clouds and the relevance to the radiative field.

We examined the response of cloud fractions and cloud forcing to the idealized global warming experiment, and not cloud feedback directly. Generally, cloud feedback means positive or negative feedback on surface temperature through changes in cloud properties as a response to global warming (Stephens 2005; Bony et al. 2006). To understand cloud feedback mechanism in models,

Figure 20.1 Schematic view of cloud responses to the +2K and control experiments using the global cloud-resolving model, showing increase in high clouds detrained from deep convection under warmer condition.

an important intermediate step is to know how cloud changes in idealized warming experiments, such as this study. Changes in cloud fractions of different types of cloud categories simulated by the ISCCP simulator can be used as proxies of various cloud properties. Values of cloud radiative forcing of model results are also used to evaluate cloud feedback parameters, although the difference between cloud feedbacks and cloud radiative forcing poses a limitation (Soden et al. 2004).

Changes in Clouds

Figure 20.2a-c show global distributions of the difference in the frequency of high-, mid-, and low-level clouds determined by the ISCCP simulator between the +2K simulation and control simulation in NICAM. Here we find a significant increase in high-level clouds, especially over the tropics and the subtropics. The same is shown for MIROC (Figure 20.2d-f). The common characteristics of the changes in global mean cloud fractions categorized by optical thickness and cloud-top pressure of the ISCCP simulator clouds in MIROC (Table 20.l and Figure 20.3) are the same as those in GCM results listed in IPCC AR4 (IPCC 2007; Ringer et al. 2006). Compared to MIROC, the magnitude of the changes in cloud fraction is much larger in NICAM, and the

Figure 20.2 Difference of cloud fractions classified by the ISCCP simulator between the control experiment and the +2K experiment. Panels (a), (b) and (c) are high, middle and low clouds for NICAM. Panels (d), (e) and (f) are those for MIROC.

Table 20.1 ISCCP cloud amount changes for the +2K experiment for NICAM and MIROC.

ISCCP cloud categories


























most significant differences in NICAM are its change in high clouds. Both the magnitude and the spatial patterns of the changes in cloud cover in high clouds are quite different between the two models. Spatial patterns of the changes in mid- and low-level clouds are similar. As summarized by Figure 20.3, the signs of changes in global mean cloud fraction are similar between NICAM and MIROC, except for high thin clouds.

In Figure 20.4a, four colored solid lines show the zonal mean change in the frequency of high clouds in NICAM with different optical thickness categories (thin, medium, thick, and total) classified by the ISCCP simulator. High thin clouds and high thick clouds in the ISCCP are considered to correspond to cirrus clouds and deep convective clouds in their meteorological classification (Rossow and Schiffer 1999), respectively. Hereafter, we will refer to them as cirrus and deep convective clouds. We see that the significant increase in high clouds mostly comes from the change in cirrus. In MIROC (Figure 20.4b), the changes in cirrus are much smaller than those in NICAM. Changes













Isccp Cloud Classification

Figure 20.3 Matrix showing difference of global mean cloud fractions classified by the ISCCP simulator between the control experiment and the +2K experiment for NICAM (a) and MIROC (b).

Figure 20.3 Matrix showing difference of global mean cloud fractions classified by the ISCCP simulator between the control experiment and the +2K experiment for NICAM (a) and MIROC (b).


/ \





. i



V ;



Bj ||

r /




V i

tr r

Figure 20.4 Zonally averaged difference of high clouds with different optical thickness (thin: red line, medium: green, thick: blue, total: black, invisible: dashed black) classified by the ISCCP simulator between the control experiment and the +2K experiment for NICAM (a) and MIROC (b).

Figure 20.4 Zonally averaged difference of high clouds with different optical thickness (thin: red line, medium: green, thick: blue, total: black, invisible: dashed black) classified by the ISCCP simulator between the control experiment and the +2K experiment for NICAM (a) and MIROC (b).

in other clouds are similar, and most of the change in high clouds comes from the change in deep convective clouds.

A common change in the two models is that the temperature profile (isothermal line) shifts upward because of the temperature increase; the mixed-phase level, where liquid and solid condensate coexist, and the tropopause shifts upward. Together with this upward shift, the distributions of clouds change. A GCM intercomparison of the change in cloud condensate (Tsushima et al. 2006) revealed that the change is predominantly in the liquid condensates' increase in the mixed-phase layer, because the liquid condensate increases according to the phase change with the temperature increase in this level. In NICAM, the magnitude of the increase in ice condensate in the upper troposphere up to the tropopause is also large, and this rise in ice condensate corresponds to the significant increase in cirrus.

It is known that invisible clouds exist throughout the sky; their optical thickness falls below the threshold of the sensor and they are thus not detectable

through observations. The threshold thickness is considered to be 0.3 in the ISCCP (Schiffer and Rossow 1983). Although they are invisible in the observational data, the ISCCP simulator in a model can classify these invisible clouds. The black dashed lines in Figure 20.3a, b show the differences in high invisible clouds classified by the ISCCP simulator between the control experiment and the +2K experiment using NICAM and MIROC, respectively. In NICAM, invisible clouds decrease significantly, especially over the subtropics, and the decrease in invisible clouds corresponds well to the increase in cirrus (Figure 20.4a). In MIROC, the variation in invisible clouds does not correspond to changes in high clouds.

One of our interests in the climate sensitivity test using the GCRM is the changes in deep convection and precipitation. Figure 20.5a, b shows these changes over the tropics in NICAM. The spatial pattern of the variation in

Figure 20.5 Differences between the control experiment and the +2K experiment for NICAM, high thick clouds (a) and precipitation (b), as well as MIROC, high thick clouds (c) and precipitation (d).

Figure 20.5 Differences between the control experiment and the +2K experiment for NICAM, high thick clouds (a) and precipitation (b), as well as MIROC, high thick clouds (c) and precipitation (d).

precipitation is highly correlated to the change in the deep convective clouds. Deep convective clouds and precipitation decrease over the central and eastern part of the Indian Ocean, but increase over the western part. Over the Pacific and the Atlantic oceans, both deep convective clouds and precipitation increase in the center of the convergence zone, and decrease in the subsidence regions on both sides of the convergence region with increasing precipitation in NICAM. These differences show that the area of the convergence zone becomes centralized; ITCZ and SPCZ become closer.

In MIROC (Figure 20.5c, d), we do not find a clear link between the spatial pattern of the change in precipitation and deep convective clouds. In addition, centralization of the convergence zone is not clear.

0 0

Post a comment