Abstract

Data from global high-resolution, nonhydrostatic simulations, covering a 1-yr period and with horizontal grid sizes of 7 and 14 km, were analyzed to evaluate the response of high cloud to global warming. The results indicate that, in a warmer atmosphere, high-cloud cover increases robustly and associated longwave (LW) cloud radiative forcing (CRF) increases on average. To develop a better understanding of high-cloud responses to climate change, the geographical distribution of high-cloud size obtained from the model was analyzed and compared with observations. In warmer atmospheres, the contribution per cloud to CRF decreases for both the LW and shortwave (SW) components. However, because of significant increases in the numbers of high clouds in almost all cloud size categories, the magnitude of both LW and SW CRF increases in the simulations. In particular, the contribution from an increase in the number of smaller clouds has more effect on the CRF change. It was also found that the ice and liquid water paths decrease in smaller clouds and that particularly the former contributes to reduced LW CRF per high cloud.

1. Introduction

Understanding changes in cloud behavior is a key element of improving the predictions of future climate change (Soden and Held 2006). The primary roles of high clouds, including cirrus, which are often defined by their altitude (e.g., Rossow and Schiffer 1999), in cooling Earth’s atmosphere through radiative processes are not fully understood, because cloud-related radiative effects depend strongly on altitude and optical depth (Manabe and Strickler 1964; Lynch et al. 2002). Although it is known that the main uncertainty associated with predictions of future change in cloud radiative forcing (CRF) is associated with low clouds (Bony and Dufresne 2005), it is also important to develop an understanding of how high clouds change, including cloud cover, and ice and water content, not only because they affect CRF in a complex way through both longwave (LW) and shortwave (SW) CRF (e.g., by covering underlying midaltitude and low clouds), but also because they are linked to active convective disturbances that greatly affect ambient environments dynamically and thermodynamically (Noda et al. 2014a).

The geometric size distributions of clouds are undoubtedly closely related to the characteristics of their convective activity and organization of convective disturbances. Extensive efforts in previous studies, using both observed and modeled data, have revealed the temporal and spatial features of cloud size distributions, including the relationship between their diurnal variation and eastward-propagating intraseasonal variability (Mapes and Houze 1993), the contribution of ambient mesoscale circulation (Machado et al. 1993; Machado and Rossow 1993), and the relationship with the homogeneity of environmental humidity (Peters et al. 2009).The size of convective systems or clouds becomes larger in warm pool regions where tropical wave disturbances become more active. In addition, satellite observations reveal that the sizes of these systems tend to be larger over oceans than over land (Yuan and Houze 2010). The shape of the size distribution function is well represented by a power law (Wood and Field 2011). Interestingly similar fractal structures are also reported in previous studies that focus on shallower clouds (Benner and Curry 1998; Siebesma and Jonker 2000; Neggers et al. 2003; Zhao and DiGirolamo 2007), although the detailed form of their size distributions differs depending on cloud type.

Wood and Field (2011) showed that the cloud size distribution closely follows a power law with a slope of −1.6, but that the relationship breaks down for radii greater than approximately 1500 km. A similar scale break was also reported by previous authors, not only at larger scales (Cahalan and Joseph 1989; Kuo et al. 1993; Mapes and Houze 1993; Benner and Curry 1998), but also over a range of much smaller scales (Pierrehumbert 1996; Davis et al. 1999; Wood and Hartmann 2006).

The power-law distribution of cloud size implies that the number of smaller clouds is much greater than that of larger clouds. However, this does not mean that larger clouds are not important. In fact, larger clouds contribute much more to the tropical radiative energy budget because of their greater coverage (Machado et al. 1993; Machado and Rossow 1993; Wilcox and Ramanathan 2001; Peters et al. 2009; Wood and Field 2011).

Although several studies have considered the nature of possible changes to high clouds, including their fractional coverage (e.g., Ramanathan and Collins 1991; Lindzen et al. 2001; Ringer et al. 2006), there is not yet a consensus on the response of high clouds to a warmer atmosphere. Recent studies using global climate models (GCMs) showed a decrease in the high-cloud cover in response to global warming (Zelinka and Hartmann 2010). However, the opposite response of the high-cloud cover was reported by studies based on a global, nonhydrostatic high-resolution model [i.e., the Nonhydrostatic Icosahedral Atmospheric Model (NICAM)] simulations (Collins and Satoh 2009; Satoh et al. 2012; Tsushima et al. 2014). It is speculated that this varying response of high clouds arises from the different treatment of cloud schemes used in each model. Conventional GCMs use cumulus parameterization, and this causes model ambiguity because their cloud activity relies entirely on assumptions underlying the parameterization (O’Brien et al. 2013), which causes inconsistency when simulating various scales of disturbance in the real atmosphere (Nastrom and Gage 1985; Wilcox and Ramanathan 2001; Lovejoy et al. 2008; Kahn and Teixeira 2009; Kahn et al. 2011; Tsuchiya et al. 2011; Pressel and Collins 2012). Satoh et al. (2012) used a cloud microphysics scheme, without cumulus parameterization, to simulate the global circulation. It is of interest to explore how the clouds behave in much higher-resolution global models, without cumulus parameterization, to understand the cause of any ambiguity.

Satoh et al. (2012) examined perpetual July simulations using a 7- and 14-km mesh NICAM, and showed that the high-cloud cover increases in warmer atmospheres. In their simulations, the vertically accumulated ice water content decreased, indicating that cirrus clouds become wider and thinner at low latitudes in a warmer atmosphere. Satoh et al. (2012) explained the response of ice clouds by changes of convective mass fluxes associated with convective updrafts in the tropics. One of the advantages of using high-resolution global nonhydrostatic models to study cloud responses is that the upper cloud configurations are comparable to observations. Inoue et al. (2008) investigated the results of 3.5- and 7-km mesh NICAM simulations performed by Miura et al. (2007) and showed that, as model resolution improves, the simulated high-cloud size statistics approach the observational data obtained from geostationary satellites. Following the approach of Inoue et al. (2008), it is possible to examine the relative importance of different sizes of high clouds on CRF. In this way, the ambiguity associated with the cloud response can be interpreted with respect to cloud size.

The aim of this article is to analyze and interpret changes in the geometric size dimensions of high clouds. In particular, CRF and its dependence on cloud size are investigated to determine how much each size of high cloud contributes to CRF and its changes. Section 2 describes the model setting and data used in this study. Section 3 presents the results, including cloud size statistics and the dependence of CRF on cloud size. Changes in cloud-associated variables [i.e., the liquid water path (LWP), ice water path (IWP), and precipitable water (PW)], are also examined. The summary and conclusions are given in section 4.

2. Data

a. Model

The nonhydrostatic GCM used in the present study was NICAM (Tomita and Satoh 2004; Satoh et al. 2008). The model configuration used was the same as Noda et al. (2012) and with mesh intervals of approximately 7 and 14 km (R7 and R14, respectively). This article examines the resolved responses to global warming derived from these mesh experiments. Although Inoue et al. (2008) showed that a mesh interval of 3.5 km generates better simulations of the cloud size distribution, R7 and R14 reproduce realistic behaviors of tropical cloud systems (Tomita et al. 2005; Miura et al. 2007; Oouchi et al. 2009a,b; Sato et al. 2009). Recently, Miyamoto et al. (2013) conducted higher-resolution NICAM simulations up to a mesh size of 870 m, and found that convective cores become resolvable at a mesh size of less than about 1.7 km. Such a dataset is useful for more quantitative evaluation and comparison with observations. Using the two datasets of differing resolution, R7 and R14, we will be able to discuss the statistical results for high clouds, at least qualitatively. In addition, because of the relatively smaller computational demands of R7 and R14, it is possible to perform 1-yr experiments, which enables a statistical analysis that includes the seasonal cycle, although these simulations are computationally much heavier than those of the conventional GCMs.

The experiment follows a time-slice approach (Bengtsson et al. 1996), consisting of a control (CTL) experiment spanning the period from June 2004 to May 2005 and a global warming experiment (GW) that was initiated in May and then time-integrated for 1 year at the end of the twenty-first century (Yamada et al. 2010; Yamada and Satoh 2013). We selected 2004 to enable us to evaluate the model’s performance, because Japan experienced typhoon landfalls during this year. We regard the year 2004 as representing present-day conditions, and examine the response to a warmer atmosphere (GW, described below).

For CTL, the boundary conditions pertaining to the sea surface temperature (SST) and the sea ice concentration (SIC) used here were the same as those in the 2004 boreal summer experiment of Noda et al. (2012). The conditions for GW were created using the dataset of the World Climate Research Program’s third phase of the Coupled Model Intercomparison Project (CMIP3) and the method of Mizuta et al. (2008). GW climate forcing was triggered by adding the SST and SIC differences between the present-day and future periods to the CTL data; the present-day condition was based on the average for the period 1979–2003, while the future condition was represented by the average for 2075–99. The carbon dioxide concentration used in GW was uniformly twice that in CTL. The initial conditions for GW were taken from a present-day National Centers for Environmental Prediction reanalysis dataset for 0000 UTC 1 May 2004, and spun up for one month. We generate 3D snapshot (2D temporal mean) data every 6 h (1 h).

b. Satellite-based data

We used the International Satellite Cloud Climatology Project (ISCCP) D2 data (Rossow and Schiffer 1999), and the Clouds and the Earth’s Radiant Energy System (CERES) Synoptic Radiative Fluxes and Clouds SYN1deg data version 2.5 (Wielicki et al. 1996) to evaluate the modeled cloud cover and CRF, respectively. We also used global infrared data (Janowiak et al. 2001) to analyze observed cloud size statistics for comparison with the modeled high clouds. The period used was the same as that in the modeled results (1 June 2004–31 May 2005). We compared the first two datasets at a resolution of 2.5° averaged over the simulation period, while the global infrared hourly data were analyzed at their original resolution in space (approximately 4 km).

c. Cloud size analysis

We defined a high-cloud region as one where the modeled outgoing LW radiation (OLR) at the top of the atmosphere (TOA) was less than 210 W m−2, because this threshold has often been used in previous studies of high clouds (e.g., Mapes and Houze 1993; Inoue et al. 2008). For the global infrared data, we used a corresponding equivalent brightness temperature of appro-ximately 253 K, following Inoue et al. (2008). Cloud size was defined based on spatially continuous grid points, the number of which was greater than unity to exclude numerical noise. Cloud sizes are denoted using the equivalent radius of a circle.

We focus here on the region between 30°S and 30°N, which will be referred to as “low latitudes” hereafter. The period of the analysis was 1 year, unless otherwise stated.

3. Results

We first compare the model simulations of the geographical distribution of cloud cover over the year in Fig. 1. We used an ISCCP simulator code (http://cfmip.metoffice.com/ISCCP.html) to evaluate cloud cover in the NICAM simulations interactively (Noda et al. 2010). Note that because we used NICAM as a cloud-resolving model, the cloud cover at every grid point is either 0% or 100%. The values were linearly interpolated onto 2.5° grids, and then the gridpoint values were averaged over a year. The modeled spatial distributions of high clouds in both R7 and R14 mostly agree with the ISCCP observations, although the cloud cover is quantitatively greater (Fig. 1). The geographical distribution of high cloud is similar to that in the previous NICAM experiments (Noda et al. 2010). Both R7 and R14 show biases that produce more high clouds especially over South America, Africa, and the middle of the Pacific Ocean. In contrast, the results show less low-cloud cover, especially over the regions off the western coasts of continents. For midaltitude clouds, the models predict fewer clouds, especially over the maritime, American, and Eurasian continents, and also at high latitudes. Although the problem of ISCCP high clouds has been recognized previously, for example, the misidentification of middle to high clouds (Marchand et al. 2010), the underestimation of the modeled middle clouds is also reported in Kodama et al. (2012). A possible reason for this ambiguity may be a lack of spatial resolution in the model, which causes underestimates of the dilution of deep convection.

Fig. 1.

Comparison of cloud cover (%) from the ISCCP data with that from the R7 and R14 runs of the CTL and GW simulations. The column on the right shows the differences between GW and CTL for R7 and R14. The modeled data were linearly interpolated in space from the original mesh size to a mesh size of 2.5°.

Fig. 1.

Comparison of cloud cover (%) from the ISCCP data with that from the R7 and R14 runs of the CTL and GW simulations. The column on the right shows the differences between GW and CTL for R7 and R14. The modeled data were linearly interpolated in space from the original mesh size to a mesh size of 2.5°.

Figure 2 compares the observed and modeled CRF and its response to the global warming. LW (SW) CRF is defined as the difference between clear-sky outgoing LW (SW) radiation and cloudy-sky outgoing LW (SW) radiation at the top of the atmosphere. As may be speculated from Fig. 1, the model simulates greater LW CRF, especially over the tropical western and central Pacific Ocean, caused by the greater cloud cover, and less spatial contrast from the Indian Ocean to the west Pacific (Figs. 2a,b). For the SW component (Figs. 2e,f), the magnitude of the simulated CRF is less than that of the observations, mainly owing to the underestimation of the low-cloud cover. Although such a bias in low-cloud cover is common in GCMs, reproducing a more realistic distribution of low clouds by improving turbulent schemes and environmental dynamic and thermodynamic conditions remains an important issue (Noda et al. 2014a,b, manuscript submitted to J. Meteor. Soc. Japan).

Fig. 2.

Comparison of observed and modeled CRF (W m−2) in R7 and responses to global warming. The LW components from (a) CERES data, (b) CTL, (c) the difference between GW and CTL, and (d) the zonal mean of the difference are shown. (e)–(h) As in (a)–(d), but for the SW component. The modeled data were linearly interpolated in space from the original mesh size to a mesh size of 2.5°.

Fig. 2.

Comparison of observed and modeled CRF (W m−2) in R7 and responses to global warming. The LW components from (a) CERES data, (b) CTL, (c) the difference between GW and CTL, and (d) the zonal mean of the difference are shown. (e)–(h) As in (a)–(d), but for the SW component. The modeled data were linearly interpolated in space from the original mesh size to a mesh size of 2.5°.

Regarding responses to the global warming, LW CRF increases due to a slight increase in high-cloud coverage, especially over central and eastern parts of the Pacific Ocean (Fig. 2c). As a result of the decrease in low clouds at almost all low latitudes, SW CRF weakens there, but in contrast it increases at high latitudes and (only fractionally) in parts of the equatorial and East Asian regions (Figs. 2f,g). The mean change in CRF for the LW (SW) component was 0.66 (0.51) W m−2 over the global domain, and 0.12 (0.96) W m−2 at low latitudes, respectively.

Figure 3 compares the observed and modeled size distributions of the number of high clouds at low latitudes. Note that the reason that the numbers are generally not integers is that these values are means derived from the hourly data from the analysis period (section 2). Consistent with the observations, the number of high clouds monotonically decreases as size increases. However, R7 and R14 underestimate the number of high clouds in the smallest bin (0–20 km in radius), and R7 overestimates it in the larger bins by 100 km. We expect that the difference between the observed and modeled number of high clouds would be smaller when using significantly higher-resolution models. In fact, this distortion of the size distribution was somewhat reduced on improving the horizontal resolution from R14 to R7, which is consistent with the findings of Inoue et al. (2008). Although the high clouds in the smallest bin should be treated carefully, the relative ratio of the size distribution can be used for analysis in the subsequent discussion.

Fig. 3.

Frequency distribution of high-cloud size (radius) derived from the global infrared data and also from the R7 and R14 runs of the CTL and GW simulations. Cloud size (radius) is binned every 20 km.

Fig. 3.

Frequency distribution of high-cloud size (radius) derived from the global infrared data and also from the R7 and R14 runs of the CTL and GW simulations. Cloud size (radius) is binned every 20 km.

Responding to the global warming, the cloud numbers in all radius categories increased, including those greater than 400 km in radius in R7 (Fig. 3b). Such a response to the warmed atmosphere is more pronounced for the larger clouds in R7 than for those in R14. As will be investigated in more detail later, these increased cloud numbers result in an increased high-cloud cover and its associated CRF (Figs. 1 and 2).

Another interesting question is the extent to which changes to the probability density functions (PDFs) of the clouds occur in response to the global warming. Figure 4 shows size distributions similar to Fig. 3, but compares their PDFs. The global infrared data show that the size distribution of high clouds follows a trend line of r−2.5 until r ≈ 200 km, but that this power-law relationship then breaks down beyond r ≈ 500 km. Interestingly, the simulated results in the 20–200-km range show a slope that is similar to that from the satellite observations (i.e., best fit of slope, a = −2.5), although the result of R14 (a = −2.4) is closer to that of the global IR data than that of R7 (a = −2.9) in this radius range. This encouraging result is reasonable because effective resolutions are generally several times larger than the mesh size (Skamarock 2004).

Fig. 4.

PDFs of high-cloud numbers as a function of radius derived from the global infrared data and also from the R7 and R14 runs of the CTL and GW simulations. Cloud size (radius) is binned every 20 km, and the sum of all values is unity. Data plotted are thinned in geometric progression for readability. Slopes, ra, for a = −2.5, −2.4, and −2.9 are regression curbs for the results of Global IR, R14, and R7, respectively, between 20- and 200-km radii.

Fig. 4.

PDFs of high-cloud numbers as a function of radius derived from the global infrared data and also from the R7 and R14 runs of the CTL and GW simulations. Cloud size (radius) is binned every 20 km, and the sum of all values is unity. Data plotted are thinned in geometric progression for readability. Slopes, ra, for a = −2.5, −2.4, and −2.9 are regression curbs for the results of Global IR, R14, and R7, respectively, between 20- and 200-km radii.

In response to the global warming (Fig. 5), the relative frequency of clouds in the 0–20-km size range decreases, but it increases for clouds in the 20–40-km range. The occurrence frequency of clouds larger than 40 km decreased (except for R14 that showed a small increase in the 40–60-km size class). Note that numbers of almost all categories of cloud size increase as shown by Fig. 3, so that Fig. 5 implies that the relative increase is larger for clouds smaller than 40 km than for the larger clouds. The increase in the numbers of small clouds reflects the reduced organization of moderately strong deep convective systems in a warmer atmosphere (Noda et al. 2014b, manuscript submitted to J. Meteor. Soc. Japan). Later, and to avoid unnecessary repetition, we will focus mainly on the results from R7, because the results from R14 are very similar to those of R7.

Fig. 5.

Difference between PDFs of high-cloud numbers between GW and CTL for R7 and R14 (shown in Fig. 4). The sum of values in all size categories is zero for both R7 and R14.

Fig. 5.

Difference between PDFs of high-cloud numbers between GW and CTL for R7 and R14 (shown in Fig. 4). The sum of values in all size categories is zero for both R7 and R14.

Before examining the regional characteristics of the changes in cloud size, we will compare the simulated frequency of occurrence of each size category with its occurrence under present atmospheric conditions using the satellite observations in Fig. 6. The frequency of occurrence is defined so that the sum of the frequency in the low-latitude domain shown equals unity for each size category. The spatial distributions in the satellite data were similar among the categories, except for r < 50 km (e.g., those high clouds that frequently develop along the tropics). A more detailed comparison among the size categories in the global infrared data reveals that larger clouds tend to occur more frequently over the Amazon, Africa, the Atlantic, and the Maritime Continents, but, in contrast, less frequently over the most of the Australian continent and the middle of the Pacific Ocean. Comparing the R7 result with the observations, there are several ambiguities in the simulated result that should be improved, such as the underestimation of relative frequency of occurrence of, for example, medium-sized clouds along the intertropical convergence zone (ITCZ) over the Pacific Ocean, although the characteristics of systematic evolution along the ITCZ are similar to the observations. In the warmer atmosphere (Fig. 7), changes to the spatial distribution of the frequency of occurrence are similar among the categories. That is, the values tend to decrease over the middle of the Pacific Ocean, the northern parts of the Maritime Continent, the tropical Atlantic, eastern South America, and the South Pacific convergence zone (SPCZ), but they increase around the Atlantic, parts of the Maritime Continents, and the southern part of the Indian Ocean (e.g., 16°S, 70°E).

Fig. 6.

Spatial comparison of the frequency of occurrence of high clouds for each radius category for global infrared and the R7 run of CTL. Note that the sum of grid values over the region shown in each panel is unity. For clarity, the resolution of the results was reduced to a mesh size of 5° by linear interpolation.

Fig. 6.

Spatial comparison of the frequency of occurrence of high clouds for each radius category for global infrared and the R7 run of CTL. Note that the sum of grid values over the region shown in each panel is unity. For clarity, the resolution of the results was reduced to a mesh size of 5° by linear interpolation.

Fig. 7.

Spatial distribution of the frequency of occurrence of high clouds for each radius category. The differences for R7 between GW and CTL are shown. Note that the sum of the grid values over the region shown in each panel is zero. For clarity, the resolution of the results was reduced to a mesh size of 5° by linear interpolation.

Fig. 7.

Spatial distribution of the frequency of occurrence of high clouds for each radius category. The differences for R7 between GW and CTL are shown. Note that the sum of the grid values over the region shown in each panel is zero. For clarity, the resolution of the results was reduced to a mesh size of 5° by linear interpolation.

To aid our understanding of the reasons behind the spatial changes in cloud frequency, we also show the horizontal field of midlevel vertical velocity and its response to the global warming in Fig. 8. As expected, changes in the cloud distribution agree well with that of midlevel updraft (Figs. 7 and 8a). Because of global warming, the updraft over the Maritime Continent (the eastern part of the Pacific Ocean) decreases (increases), indicating a weakening of the Walker circulation (Figs. 8a,b), which is frequently reported from GCM simulations and is driven by changes in the energy balance (Held and Soden 2006; Tokinaga et al. 2012).

Fig. 8.

Spatial distribution of the vertical velocity (10−3 m s−1) averaged between altitudes of 4 and 7 km for R7 in (a) CTL and (b) the difference between GW and CTL. The modeled data were linearly interpolated in space from the original mesh size to a mesh size of 2.5°.

Fig. 8.

Spatial distribution of the vertical velocity (10−3 m s−1) averaged between altitudes of 4 and 7 km for R7 in (a) CTL and (b) the difference between GW and CTL. The modeled data were linearly interpolated in space from the original mesh size to a mesh size of 2.5°.

In the analysis presented above, we have confirmed the size distributions of modeled high clouds, and their responses to global warming, and it would be interesting to determine how these size statistics relate to changes in CRF. For this reason, we analyzed the relationship between CRF and radius. Figure 9 shows the contributions of clouds of different sizes to LW and SW CRF. This shows values of LW CRF, FLW(r, y), and SW CRF, FSW(r, y), of integration in r and the vertical (y) axes that satisfy the following:

 
formula

and

 
formula

For the CTL result, LW CRF is most likely to be around 70 W m−2 for the smallest clouds, and gradually increases as the clouds become bigger; for the largest clouds shown (r = 500 km), LW CRF is most likely to be around 100 W m−2. Clouds with a radius of less than 40 km contribute strongly to the net changes (Figs. 9a,c). In response to the global warming (Fig. 9b), the LW component distributes a donut-shaped pattern; clouds with CRF around 65 W m−2 and r < 20 km are strongly reduced, while those with CRF around 50 and 80 W m−2 are strongly increased. In total, the change in CRF is positive (cf. Fig. 2a). For the SW component (Figs. 9c,d), the contributions of clouds with CRF stronger than −60 W m−2 are enhanced (i.e., they become increasingly negative); however, the contribution from clouds with weaker CRF prevails and, as a result, negative CRF is enhanced. Note that this result only shows the contribution from high clouds, and the net decrease of SW CRF at low latitudes (Fig. 2b) is mainly caused by the reduction in the occurrence of low cloud.

Fig. 9.

Contributions of each cloud size to CRF of the top of the atmosphere in R7 and their responses to global warming. The LW component in (a) CTL and (b) the difference between GW and CTL for the major regions are shown. (c),(d) As in (a),(b), but for the SW component.

Fig. 9.

Contributions of each cloud size to CRF of the top of the atmosphere in R7 and their responses to global warming. The LW component in (a) CTL and (b) the difference between GW and CTL for the major regions are shown. (c),(d) As in (a),(b), but for the SW component.

In addition to Fig. 9, we compare the mean CRF per cloud in Fig. 10. This shows the average CRF of one contiguous cloud in each size category. CRF increases with radius for both the LW and SW components. In contrast to the result for the SW component, the rate of increase in the LW component is reduced. In response to the global warming, CRF in both the LW and SW components decreases in all clouds, regardless of cloud radius. Therefore, the explanation for the increases of net CRF in both the LW and SW components (Figs. 2a and 9) is not the increase in CRF from individual clouds (cf. Fig. 9), but the increase in the number of clouds, particularly small clouds (r < 40 km in Fig. 9a; see also Fig. 3).

Fig. 10.

Size distribution of (a) mean LW and (b) mean SW components of CRF at the top of the atmosphere per cloud in R7. Lines with closed circles and open squares show the results for CTL and GW, respectively. Cloud size (radius) is binned every 20 km.

Fig. 10.

Size distribution of (a) mean LW and (b) mean SW components of CRF at the top of the atmosphere per cloud in R7. Lines with closed circles and open squares show the results for CTL and GW, respectively. Cloud size (radius) is binned every 20 km.

The decrease in LW and SW CRF might be related to a decrease in cloud optical thickness. To develop a deeper understanding of changes in CRF, we also examined changes in cloud properties, and Fig. 11 shows the contributions of clouds to PW, LWP, and IWP as a function of radius. Here we also mention LWP not only because it is also a component that constitutes condensates of deep clouds, but also, as will be mentioned later, their response to global warming differs from that of IWP qualitatively. Although IWP increases with radius (e.g., IWP < 0.03 kg m−2 and r < 430 km in Fig. 11c), these quantities correlate less well with radius when compared with LW CRF (cf. Fig. 9a). Rather, they can take a wide range of values even for similar radii, as shown by the almost flat distributions in the y axis direction for r < 60 km. In response to the global warming, PW shows positive and negative signs across approximately 60 kg m−2, meaning that almost all high-cloud regions possess a larger amount of PW. A similar response was found for IWP and LWP, which showed changes of sign across approximately 0.02 kg m−2. However, in contrast to IWP, LWP correlates less well with cloud radius. At each size, both IWP and LWP generally decrease because of a greater reduction in the larger values of IWP or LWP.

Fig. 11.

Contributions of each cloud size to water substances in R7 and their response to global warming. PW in (a) CTL and (b) the difference between GW and CTL in the major regions are shown. Cloud size (radius) is binned every 20 km. Note that the integrals of all values on the x and y axes are the mean for the area 30°S–30°N. (c),(d) As in (a),(b), but for IWP. (e),(f) As in (a),(b), but for LWP.

Fig. 11.

Contributions of each cloud size to water substances in R7 and their response to global warming. PW in (a) CTL and (b) the difference between GW and CTL in the major regions are shown. Cloud size (radius) is binned every 20 km. Note that the integrals of all values on the x and y axes are the mean for the area 30°S–30°N. (c),(d) As in (a),(b), but for IWP. (e),(f) As in (a),(b), but for LWP.

Figure 12 shows values similar to those in Fig. 10, but for IWP and LWP. For all high clouds, the mean IWP and LWP per cloud gradually increase with increasing radius. For clouds larger than approximately 100 km, the IWP per cloud becomes less sensitive to radius. In response to the global warming, IWP decreases in all size categories. For LWP, there is a change in size dependency across 100 km; LWP decreases (increases) in smaller (larger) clouds. The change in (especially) IWP reflects changes in cloud thickness, and this leads to the decrease in LW and SW CRF per cloud as shown in Fig. 10.

Fig. 12.

As in Fig. 10, but for (a) IWP and (b) LWP.

Fig. 12.

As in Fig. 10, but for (a) IWP and (b) LWP.

A question that naturally arises is why such different responses in IWP and LWP occur even in the same categories of cloud size (Fig. 11). One possible explanation may be the different regional occurrences of these high clouds. To investigate this hypothesis, we show the spatial distributions of the relative occurrence of high clouds that generates the increase or decrease in IWP (Fig. 11d) in Fig. 13. This figure indicates that the negative response is attributed to high clouds that are distributed over all regions at low latitudes, while the positive response is caused by high clouds that are distributed relatively closer to the tropics. However, the changes in the two categorized high-cloud regions showed no systematic differences (e.g., a tendency to prefer land or ocean) and thereby they increase (decrease) almost homogeneously.

Fig. 13.

Spatial comparison of the frequency of occurrence of high clouds between increased and reduced regions of IWP. Contributions of clouds in (a) (0 ≤ x ≤ 60 km, 0 ≤ y ≤ 0.04; kg m−2) of Fig. 11d, and (c) the difference between GW and CTL are shown. (b),(d) As in (a),(c), but for (0 ≤ x ≤ 60 km, 0.04 < y; kg m−2). For clarity, the resolution of the results was reduced to a mesh size of 2.5° by linear interpolation.

Fig. 13.

Spatial comparison of the frequency of occurrence of high clouds between increased and reduced regions of IWP. Contributions of clouds in (a) (0 ≤ x ≤ 60 km, 0 ≤ y ≤ 0.04; kg m−2) of Fig. 11d, and (c) the difference between GW and CTL are shown. (b),(d) As in (a),(c), but for (0 ≤ x ≤ 60 km, 0.04 < y; kg m−2). For clarity, the resolution of the results was reduced to a mesh size of 2.5° by linear interpolation.

4. Summary and discussion

We analyzed data from the first year-long runs of a global, nonhydrostatic model (NICAM) and used horizontal grid sizes of R7 and R14 to evaluate the responses of high clouds to global warming conditions. Statistical analysis of clouds generated by these 1-yr NICAM simulations is presented for the first time in this paper. The results from both resolutions showed similar spatial distributions of cloud cover, associated CRF distribution, and responses to global warming. In GW, the high (low) cloud cover increases (decreases) robustly and, as a result, the annual mean LW (SW) CRF increases (decreases) both globally and at low latitudes.

The PDFs of the size distributions of the modeled high clouds in R14 mostly follow a power law with a slope of −2.4, which is similar to the result from the global infrared data (−2.5), while that in R7 is somewhat steeper (−2.9). We also showed that the PDFs for GW in the R7 and R14 runs change slightly from those in CTL: the proportion of smaller clouds (0–20 km) decreases, while that of larger clouds (20–40 km) increases. The proportions of much larger clouds (>60 km) decrease significantly in both R7 and R14. There is a resolution dependency in an intermediate radius range (40–60 km). The relatively lower contribution of much larger clouds to the total number of clouds suggests less convective organization in the warmer tropical atmosphere.

In terms of absolute numbers, the number of high clouds increases for almost all sizes in response to global warming. The spatial distribution of their frequency of occurrence corresponds well to that of updraft in the middle troposphere. In the warmer atmosphere, the frequency of occurrence changes, primarily as a result of the reduced intensity of the Walker circulation. Additional analysis regarding tropical cyclones also revealed reduced numbers of high clouds in tropical cyclones (not shown). Few notable differences were found across those spatial size categories in response to global warming. For all radius categories, the frequency of occurrence of high clouds decreases over the northwest Pacific, Pacific ITCZ, and SPCZ, while it increases in the southwest Pacific, south Indian Ocean, the tropical Atlantic, and the eastern part of South America.

We also examined the dependence and contributions of different sizes of high clouds to CRF. In the warmer atmosphere, the contributions per high cloud to both the LW and SW components decreased. However, because of an increase in the numbers of high clouds, the net change in CRF was positive. Overall, the contribution of changes to small high clouds to the total changes was significant because of their large numbers. In an attempt to develop a more in-depth understanding of the response of CRF to the global warming, we also investigated PW, IWP, and LWP for each radius category. In particular, the decrease of IWP and LWP for all cloud sizes showed a reduced CRF per cloud.

In contrast to many GCMs, which predict fewer high clouds in a warmer atmosphere, the simulations with the nonhydrostatic, high-resolution global model predict more high clouds in general. The response of the increased number of high clouds has been reported in previous case studies using NICAM, although the magnitude of their changes depends strongly on the physical schemes used and related unknown parameters. The high-cloud response is one of the most notable and interesting differences.

This paper reveals that a likely reason for these differences arises from the response of small high clouds, which can be explicitly computed using a nonhydrostatic, high-resolution GCM. It is readily understood that the contribution of a single small cloud to net CRF is smaller than that of a large cloud. However, the number of small high clouds is much larger; consequently, their net contribution becomes important. In addition to the well-known ambiguity associated with cumulus parameterization, the spatial resolution in conventional GCMs is too coarse to resolve such small clouds, and they are represented by subgrid cloud schemes in GCMs. On the other hand, we also showed that a 7-km mesh is still too coarse to simulate the exact number of small high clouds (Fig. 3). Thus, the net impacts of small high clouds would be more pronounced in higher-resolution models that could better simulate such small clouds quantitatively (Miyamoto et al. 2013).

Acknowledgments

We thank two anonymous reviewers for helpful comments. This work was partly supported by the Program for Risk Information on Climate Change from MEXT, Japan, and the Research Program on Climate Change Adaptation. The simulations in this study were performed using the K computer and the Earth Simulator. The CERES data were obtained from the NASA Langley Research Center Atmospheric Science Data Center.

REFERENCES

REFERENCES
Bengtsson
,
L.
,
M.
Botzet
, and
M.
Esch
,
1996
:
Will greenhouse gas-induced warming over the next 50 years lead to higher frequency and greater intensity of hurricanes?
Tellus,
48A
,
57
73
, doi:.
Benner
,
T. C.
, and
J. A.
Curry
,
1998
: Characteristics of small tropical cumulus clouds and their impact on the environment. J. Geophys. Res.,103, 28 753–28 767, doi:.
Bony
,
S.
, and
J.-L.
Dufresne
,
2005
:
Marine boundary layer clouds at the heart of tropical cloud feedback uncertainties in climate models
.
Geophys. Res. Lett.
,
32
,
L20806
, doi:.
Cahalan
,
R. F.
, and
J. H.
Joseph
,
1989
:
Fractal statistics of cloud fields
.
Mon. Wea. Rev.
,
117
,
261
272
, doi:.
Collins
,
W. D.
, and
M.
Satoh
,
2009
: Simulating global clouds: Past, present, and future. Clouds in the Perturbed Climate System: Their Relationship to Energy Balance, Atmospheric Dynamics, and Precipitation, J. Heintzenberg and R. J. Charlson, Eds., Vol. 2, The MIT Press,
469
486
.
Davis
,
A. B.
,
A.
Marshak
,
H.
Gerber
, and
W. J.
Wiscombe
,
1999
:
Horizontal structure of marine boundary layer clouds from centimeter to kilometer scales
.
J. Geophys. Res.
,
104
,
6123
6144
, doi:.
Held
,
I.
, and
B.
Soden
,
2006
:
Robust responses of the hydrological cycle to global warming
.
J. Climate
,
19
,
5686
5699
, doi:.
Inoue
,
T.
,
M.
Satoh
,
H.
Miura
, and
B.
Mapes
,
2008
:
Characteristics of cloud size of deep convection simulated by a global cloud resolving model over the western tropical Pacific
.
J. Meteor. Soc. Japan
,
86A
,
1
15
, doi:.
Janowiak
,
J. E.
,
R. J.
Joyce
, and
Y.
Yarosh
,
2001
:
A real-time global half-hourly pixel-resolution infrared dataset and its applications
.
Bull. Amer. Meteor. Soc.
,
82
,
205
217
, doi:.
Kahn
,
B. H.
, and
J.
Teixeira
,
2009
:
A global climatology of temperature and water vapor variance scaling from the atmospheric infrared sounder
.
J. Climate
,
22
,
5558
5576
, doi:.
Kahn
,
B. H.
, and Coauthors
,
2011
:
Temperature and water vapor variance scaling in global models: Comparisons to satellite and aircraft data
.
J. Atmos. Sci.
,
68
,
2156
2168
, doi:.
Kodama
,
C.
,
A. T.
Noda
, and
M.
Satoh
,
2012
:
An assessment of the cloud signals simulated by NICAM using ISCCP, CALIPSO, and CloudSat satellite simulators
.
J. Geophys. Res.
,
117
,
D12210
, doi:.
Kuo
,
K.-S.
,
R. M.
Welch
,
R. C.
Weger
,
M. A.
Engelstad
, and
S. K.
Sengupta
,
1993
:
The three-dimensional structure of cumulus clouds over the ocean. I. Structural analysis
.
J. Geophys. Res.
,
98
,
20 685
20 711
, doi:.
Lindzen
,
R. S.
,
M.-D.
Chou
, and
A. Y.
Hou
,
2001
:
Does the earth have an adaptive infrared iris?
Bull. Amer. Meteor. Soc.
,
82
,
417
432
, doi:.
Lovejoy
,
S.
,
D.
Schertzer
, and
V. C.
Allaire
,
2008
:
The remarkable wide range spatial scaling of TRMM precipitation
.
Atmos. Res.
,
90
,
10
32
, doi:.
Lynch
,
D. K.
,
K.
Sassen
,
D.
Starr
, and
G.
Stephens
,
2002
: Cirrus. Oxford University Press, 480 pp.
Machado
,
L. A. T.
, and
W. B.
Rosso
,
1993
:
Structural characteristics and radiative properties of tropical cloud clusters
.
Mon. Wea. Rev.
,
121
,
3234
3260
, doi:.
Machado
,
L. A. T.
,
J.-P.
Duvel
, and
M.
Desbois
,
1993
:
Diurnal variations and modulation by easterly waves of the size distribution of convective cloud clusters over West Africa and the Atlantic Ocean
.
Mon. Wea. Rev.
,
121
,
37
49
, doi:.
Manabe
,
S.
, and
R. F.
Strickler
,
1964
:
Thermal equilibrium of the atsmosphere with a convective adjustment
.
J. Atmos. Sci.
,
21
,
361
385
, doi:.
Mapes
,
B. E.
, and
R. A.
Houze
Jr.
,
1993
:
Cloud clusters and superclusters over the oceanic warm pool
.
Mon. Wea. Rev.
,
121
,
1398
1416
, doi:.
Marchand
,
R.
,
T.
Ackerman
,
M.
Smyth
, and
W. B.
Rossow
,
2010
:
A review of cloud top height and optical depth histograms from MISR, ISCCP, and MODIS
.
J. Geophys. Res.
,
115
,
D16206
, doi:.
Miura
,
H.
,
M.
Satoh
,
T.
Nasuno
,
A. T.
Noda
, and
K.
Oouchi
,
2007
:
A Madden–Julian Oscillation event realistically simulated by a global cloud-resolving model
.
Science
,
318
,
1763
1765
, doi:.
Miyamoto
,
Y.
,
Y.
Kajikawa
,
R.
Yoshida
,
H.
Yashiro
, and
H.
Tomita
,
2013
:
Deep moist atmospheric convection in a subkilometer global simulation
.
Geophys. Res. Lett.
, 40, 4922–4926, doi:.
Mizuta
,
R.
,
Y.
Adachi
,
S.
Yukimoto
, and
S.
Kusunoki
,
2008
: Estimation of the future distribution of sea surface temperature and sea ice using the CMIP3 multi-model ensemble mean. Meteorlogical Research Institute Tech. Rep. 56, Tsukuba, Japan, 28 pp.
Nastrom
,
G.
, and
K.
Gage
,
1985
:
A climatology of atmospheric wave number spectra of wind and temperature observed by commercial aircraft
.
J. Atmos. Sci.
,
42
,
950
960
, doi:.
Neggers
,
R. A. J.
,
H. J. J.
Jonker
, and
A. P.
Siebesma
,
2003
:
Size statistics of cumulus cloud populations in large-eddy simulations
.
J. Atmos. Sci.
,
60
,
1060
1074
, doi:.
Noda
,
A. T.
,
K.
Oouchi
,
M.
Satoh
,
H.
Tomita
,
S.
Iga
, and
Y.
Tsushima
,
2010
:
Importance of the subgrid-scale turbulent moist processes: Cloud distribution in global cloud-resolving simulations
.
Atmos. Res.
,
96
,
208
217
, doi:.
Noda
,
A. T.
,
K.
Oouchi
,
M.
Satoh
, and
H.
Tomita
,
2012
:
Quantitative assessment of diurnal variation of tropical convection simulated by a global nonhydrostatic model without cumulus parameterization
.
J. Climate
,
25
,
5119
5134
, doi:.
Noda
,
A. T.
,
K.
Nakamura
,
T.
Iwasaki
, and
M.
Satoh
,
2014a
:
Responses of subtropical marine stratocumulus to perturbed lower atmospheres
.
SOLA
,
10
,
34
38
, doi:.
O’Brien
,
T. A.
,
F.
Li
,
W. D.
Collins
,
S. A.
Rauscher
,
T. D.
Ringler
,
M.
Taylor
,
S. M.
Hagos
, and
L. R.
Leung
,
2013
:
Observed scaling in clouds and precipitation and scale incognizance in regional to global atmospheric models
.
J. Climate
,
26
,
9313
9333
, doi:.
Oouchi
,
K.
,
A. T.
Noda
,
M.
Satoh
,
H.
Miura
,
H.
Tomita
,
T.
Nasuno
, and
S.
Iga
,
2009a
:
A simulated preconditioning of typhoon genesis controlled by a boreal summer Madden–Julian Oscillation event in a global cloud-resolving model
.
SOLA
,
5
,
65
68
, doi:.
Oouchi
,
K.
,
A. T.
Noda
,
M.
Satoh
,
B.
Wang
,
S.-P.
Xie
,
H.
Takahashi
, and
T.
Yasunari
,
2009b
:
Asian summer monsoon simulated by a global cloud-system resolving model: Diurnal to intra-seasonal variability
.
Geophys. Res. Lett.
,
36
,
L11815
, doi:.
Peters
,
O.
,
J. D.
Neelin
, and
S. W.
Nesbitt
,
2009
:
Mesoscale convective systems and critical clusters
.
J. Atmos. Sci.
,
66
,
2913
2924
, doi:.
Pierrehumbert
,
R. T.
,
1996
:
Anomalous scaling of high cloud variability in the tropical Pacific
.
Geophys. Res. Lett.
,
23
,
1095
1098
, doi:.
Pressel
,
K.
, and
W.
Collins
,
2012
:
First-order structure function analysis of statistical scale invariance in the AIRS observed water vapor field
.
J. Climate
,
25
,
5538
5555
, doi:.
Ramanathan
,
V.
, and
W.
Collins
,
1991
:
Thermodynamic regulation of ocean warming by cirrus clouds deduced from observations of the 1987 El Niño
.
Nature
,
351
,
27
32
, doi:.
Ringer
,
M. A.
, and Coauthors
,
2006
:
Global mean cloud feedbacks in idealized climate change experiments
.
Geophys. Res. Lett.
,
33
,
L07718
, doi:.
Rossow
,
W. B.
, and
R. A.
Schiffer
,
1999
:
Advances in understanding clouds from ISCCP
.
Bull. Amer. Meteor. Soc.
,
80
,
2261
2287
, doi:.
Sato
,
T.
,
H.
Miura
,
M.
Satoh
,
Y. N.
Takayabu
, and
Y.
Wang
,
2009
:
Diurnal cycle of precipitation over the tropics simulated by a global cloud resolving model
.
J. Climate
,
22
,
4809
4826
, doi:.
Satoh
,
M.
,
T.
Matsuno
,
H.
Tomita
,
H.
Miura
,
T.
Nasuno
, and
S.
Iga
,
2008
:
Nonhydrostatic Icosahedral Atmospheric Model (NICAM) for global cloud-resolving simulations
.
J. Comput. Phys.
,
227
,
3486
3514
, doi:.
Satoh
,
M.
,
S.
Iga
,
H.
Tomita
,
Y.
Tsushima
, and
A. T.
Noda
,
2012
:
Response of upper clouds in global warming experiments obtained using global nonhydrostatic model with explicit cloud processes
.
J. Climate
,
25
,
2178
2191
, doi:.
Siebesma
,
A. P.
, and
H. J. J.
Jonker
,
2000
:
Anomalous scaling of cumulus cloud boundaries
.
Phys. Rev. Lett.
,
85
,
214
217
, doi:.
Skamarock
,
W. C.
,
2004
:
Evaluating mesoscale numerical weather prediction models using kinetic energy spectra
.
Mon. Wea. Rev.
,
132
,
3019
3032
, doi:.
Soden
,
B. J.
, and
I. M.
Held
,
2006
:
An assessment of climate feedbacks in coupled ocean–atmosphere models
.
J. Climate
,
19
,
3354
3360
, doi:.
Tokinaga
,
H.
,
S.-P.
Xie
,
C.
Deser
,
Y.
Kosaka
, and
Y. M.
Okumura
,
2012
:
Slowdown of the Walker circulation driven by tropical Indo-Pacific warming
.
Nature
,
491
,
439
443
, doi:.
Tomita
,
H.
, and
M.
Satoh
,
2004
:
A new dynamical framework of nonhydrostatic global model using the icosahedral grid
.
Fluid Dyn. Res.
,
34
,
357
400
, doi:.
Tomita
,
H.
,
H.
Miura
,
S.
Iga
,
T.
Nasuno
, and
M.
Satoh
,
2005
:
A global cloud-resolving simulation: Preliminary results from an aqua planet experiment
.
Geophys. Res. Lett.
,
32
,
L08805
, doi:.
Tsuchiya
,
C.
,
K.
Sato
,
T.
Nasuno
,
A. T.
Noda
, and
M.
Satoh
,
2011
:
Universal frequency specta of surface meteorological fluctuations
.
J. Climate
,
24
,
4718
4732
, doi:.
Tsushima
,
Y.
,
S.
Iga
,
H.
Tomita
,
M.
Satoh
,
A. T.
Noda
, and
M. J.
Webb
,
2014
:
High cloud increase in a perturbed SST experiment with a global nonhydrostatic model including explicit convective processes
.
J. Adv. Model. Earth Syst.
,
6
, doi:.
Wielicki
,
B. A.
,
B. R.
Barkstrom
,
E. F.
Harisson
,
R. B.
Lee
III
,
G. L.
Smith
, and
J. E.
Cooper
,
1996
:
Clouds and the Earth’s Radiant Energy System (CERES): An Earth Observing System experiment
.
Bull. Amer. Meteor. Soc.
,
77
,
853
868
, doi:.
Wilcox
,
E. M.
, and
V.
Ramanathan
,
2001
:
Scale dependence of the thermodynamic forcing of tropical monsoon clouds: Results from TRMM observations
.
J. Climate
,
14
,
1511
1524
, doi:.
Wood
,
R.
, and
D. L.
Hartmann
,
2006
:
Spatial variability of liquid water path in marine low cloud: The importance of mesoscale cellular convection
.
J. Climate
,
19
,
1748
1764
, doi:.
Wood
,
R.
, and
P. R.
Field
,
2011
:
The distribution of cloud horizontal sizes
.
J. Climate
,
24
,
4800
4816
, doi:.
Yamada
,
Y.
, and
M.
Satoh
,
2013
:
Response of ice and liquid water paths of tropical cyclones to global warming simulated by a global nonhydrostatic model with explicit cloud microphysics
.
J. Climate
,
26
,
9931
9945
, doi:.
Yamada
,
Y.
,
K.
Oouchi
,
M.
Satoh
,
H.
Tomita
, and
W.
Yanase
,
2010
:
Projection of changes in tropical cyclone activity and cloud height due to greenhouse warming: Global cloud-system-resolving approach
.
Geophys. Res. Lett.
,
37
,
L07709
, doi:.
Yuan
,
J.
, and
R. A.
Houze
Jr.
,
2010
:
Global variability of mesoscale convective system anvil structure from A-Train satellite date
.
J. Climate
,
23
,
5864
5888
, doi:.
Zelinka
,
M. D.
, and
D. L.
Hartmann
,
2010
:
Why is longwave cloud feedback positive?
J. Geophys. Res.
,
115
,
D16117
, doi:.
Zhao
,
G.
, and
L.
Di Girolamo
,
2007
:
Statistics on the macrophysical properties of trade wind cumuli over the tropical western Atlantic
.
J. Geophys. Res.
,
112
,
D10204
, doi:.