A climate sensitivity test using a global cloud resolving model under an aqua planet condition



[1] A global Cloud Resolving Model (CRM) is used in a climate sensitivity test for an aqua planet in this first attempt to evaluate climate sensitivity without cumulus parameterizations. Results from a control experiment and an experiment with global sea surface temperature (SST) warmer by 2 K are examined. Notable features in the simulation with warmer SST include a wider region of active convection, a weaker Hadley circulation, mid-tropospheric moistening in the subtropics, and more clouds in the extratropics. Negative feedback from short-wave radiation reduces the climate sensitivity parameter compared to a result in a more conventional model with a cumulus parameterization.

1. Introduction

[2] Present-day Atmospheric General Circulation Models (AGCMs) suffer from several shortcomings, such as cloud parameterizations, which can be regarded as ambiguities in climate study. Current AGCMs also lack two-way interactions for sub-grid scale parameterizations. Three strategies have been proposed to overcome these problems using Cloud Resolving Models (CRMs). One strategy is the Multi-scale Modeling Framework (MMF) [Grabowski, 2001; Khairoutdinov and Randall, 2001]. A second strategy, implemented here, is a global cloud resolving simulation. A third strategy is the Diabatic Acceleration and REscaling (DARE) proposed by Kuang et al. [2005].

[3] Tomita et al. [2005] developed “NICAM” (Nonhydrostatic ICosahedral Atmosphere Model), a global CRM on a hexagonal–pentagonal grid [Tomita et al., 2001], and reported global cloud-resolving simulations for an aqua planet condition [Neale and Hoskins, 2000]. The highest resolution used in their study, about 3.5 km, resolved the updraft cores of deep convection in the tropics. Eastward propagating cloud systems and the diurnal variation in surface precipitation were reasonably reproduced.

[4] Application of a global CRM to study climate can largely reduce ambiguities related to clouds and aid in the understanding of the response of clouds to a global warming, even though model dependencies remain because of cloud microphysics schemes, horizontal and vertical resolutions, and numerical diffusion [Khairoutdinov and Randall, 2003]. This first attempt reports on a climate sensitivity test using the global CRM in a setting like that given by Cess et al. [1990] for an aqua planet.

2. Experimental Settings

[5] The model in this study was the same as that of Tomita et al. [2005]. Model descriptions are given by Tomita and Satoh [2004] and Tomita et al. [2005]. Two experiments were performed to diagnose climate sensitivity. The first experiment was the same aqua planet experiment as in the work by Tomita et al. [2005] and is denoted by “CTL” (control). The surface boundary condition was changed in the second experiment: SSTs were 2 K warmer than those in CTL over the entire globe (referred to as “+2K”). The setup followed Cess et al. [1990], who detailed systematic comparisons of AGCMs with a focus on climate sensitivity. Cess et al. [1990] included cases with 2 K warmer and 2 K cooler SSTs that were computed with a realistic surface boundary condition. Computer resource limitations in the present study allowed only a +2K SST case and these results are compared to CTL results. This sensitivity study for an idealized aqua planet would reveal a possible response of clouds to a global warming.

[6] The strategy for the series of simulations was the same as in the work by Tomita et al. [2005]. Initial conditions were given by the CCSR/NIES/FRCGC AGCM [Numaguti et al., 1997; K-1 Model Developers, 2004] with T42 truncation (referred to as “T42”). The simulation period was 3.5 years and climatology for the last 3 simulated years was used. From these initial conditions, 90-day simulations were then run on the 14-km grid. Simulation results at day 60 were interpolated to the 7-km grid, at which time 30-day simulations were run. Model results from these 30-day simulations were denoted “CTL-7km” and “+2K-7km” for the control and +2K SST case, respectively. Tomita et al. [2005] argued that the 7-km experiment almost reached statistical equilibrium. Physical processes in the 14-km and 7-km model runs were the same and included an explicit microphysics scheme [Grabowski, 1998]. Model behavior in these two runs was unaffected by cloud parameterizations, which were not used. Results from NICAM were compared to results from the CCSR/NIES/FRCGC AGCM with T42 resolution that was used as initial states for the NICAM calculations. Results from the CTL and +2K experiments of the T42 run were denoted “CTL-T42” and “+2K-T42,” respectively. A 30-day average was assumed to yield a representative climate state for the NICAM 7-km experiments because a seasonal cycle did not exist on the aqua planet.

[7] This study mainly describes results from the 7-km grid, but this grid size cannot sufficiently resolve deep tropical convection. However, Tomita et al. [2005] showed that the primary structures of the global circulation simulated with a 7-km grid closely resembled those simulated with a 3.5-km grid. This similarity in results suggests that climate sensitivity can be determined using results from the 7-km grid with no cumulus parameterization. The main difference between the 7-km and 3.5-km experiments was the intensity of precipitation on the equator and the width of the Intertropical Convergence Zone (ITCZ). These differences may have resulted from the misrepresentation of mesoscale circulation and shallow clouds. Bony et al. [2004] noted that differences between AGCMs were greater for low-level clouds than for high clouds. Uncertainties remained for low cloud feedbacks even on the 7-km grid, and the moist circulation had not yet converged numerically [Stevens et al., 2002]. Nevertheless, the premise remains that qualitative sensitivity could be discussed at least in terms of statistics of deep clouds and stratiform clouds in the middle to higher latitudes.

3. Results

[8] Figure 1a shows that temperatures in the boundary layer were about 2 K warmer in +2K-7km than in CTL-7km, perhaps expected given the warmer SSTs. Temperature changes in the middle and upper troposphere exceeded 2 K, reflecting the moist adiabatic temperature changes throughout the tropics. Changes in the water vapor mixing ratio were distinct in the middle troposphere on both sides of the ITCZ (Figure 1b), suggesting a widening of the convectively active region. Relative humidity (Figure 1c) increased in the lower to middle troposphere in the subtropics in +2K-7km, but decreased in the upper troposphere. This change in relative humidity could have occurred as cloud tops shifted upward, i.e., as the detrainment level moved to higher altitudes in +2K-7km. Upward velocities were stronger on both sides of the ITCZ, and weaker on the equator (Figure 1d). The positive response in the +2K-7km run in the subtropics suggests that downward motion there was weaker, indicating that the Hadley circulation was weaker in the +2K-7km model.

Figure 1.

Latitude–height cross section of differences between +2K-7km and CTL-7km for time- and zonally-averaged (a) temperature, (b) water vapor mixing ratio, (c) relative humidity, and (d) vertical velocity.

[9] Cloud amount, albedo, and precipitation rate in Figures 2a–2c also show that the ITCZ widened in the +2K-7km run. Cloudy grids were defined following Tompkins and Craig [1998], i.e., by using 0.005 g kg−1 as a critical value of total condensate mixing ratio. Cloud amount and albedo noticeably increased in the extratropics in NICAM, but decreased in T42. Changes in the subtropics were not obvious.

Figure 2.

Time- and zonally-averaged (a) cloud amount, (b) albedo, (c) precipitation rate, and (d) high cloud amount. Anomalies of cloud amount of the +2K from CTL; (e) monthly means of the 7-km grid and the second and third months of the 14km grid of NICAM (14km-2, 3); (f) monthly means of the 18th, 30th, and 42nd months of CCSR/NIES/FRCGC AGCM T42 (T42-18, 30, 42).

[10] Changes in tropical upper-level cloudiness affect the water vapor feedback in global warming [e.g., Lindzen et al., 2001]. Figure 3 shows a snapshot of outgoing long wave radiation (OLR) from CTL-7km. Deep convection developed near the equator and clear regions occurred in the subtropics. Cloud tops were not very high at middle to higher latitudes. Therefore, upper-level cloud amounts in the latitudes between 15°S and 15°N were computed. Grids were assumed to contain upper-level cloud if the OLR was less than 130 W m−2. Figure 2d shows that this criteria yields upper-level cloud amounts near the equator of about 15%, a value roughly consistent with Hartmann and Michelsen [2002, Figure 4], who used data from Japan's Geostationary Meteorological Sattellite-5. Convectively active regions were more confined to near the equator in Figure 2d compared to the actual atmosphere.

Figure 3.

A snapshot of outgoing long-wave radiation of CTL-7km.

[11] Upper-level cloud amounts in results from +2K-7km were larger on both sides of the ITCZ, but smaller on the equator. The wider region of convection is associated with the broadening of the equatorial lower tropospheric humidity maximum in Figures 1b and 1c. Upper-level cloud averages within the latitude range in Figure 2d were 3.19% for CTL-7km and 3.50% for +2K-7km.

[12] Tables 1a and 1b include the global average radiation budgets. Cloud amount was greater in response to +2K SST forcing in NICAM, and cloud amount was smaller in T42. NICAM result was also contradictory to those given by Cess et al. [1990], although a direct comparison would be inappropriate. The cloud amount increase was largely due to an increase in the extratropics (Figure 2a). There remains a question whether this increase is significant or not in view of the short averaging period, 30 days. To answer the question, changes of cloud between the 14-km simulations CTL-14km and +2K-14km are examined as a substitute. In Figure 2e, average cloud amount changes between the 14-km simulations over two 30-day periods are plotted together with the cloud changes between the 7 km runs. From Figure 2e, we may judge that the increase is significant, at least in the extratropics. On the contrary, in the T42 simulation, changes of cloud amount for three samples of 30 day averages show large scatter (Figure 2f), but by close inspection we can recognize a small decreasing tendency as verified by 3 year average (Figure 2a).

Table 1a. Global cloud amount (Ac, %), and net downward solar (Q, W m−2) and emitted infrared (F, W m−2) fluxes for CTL-7km and CTL-T42
Table 1b. Differences in Ac, Q, and F in +2K from those in CTL for the 7 km run in NICAM and for the T42 run in CCSR/NIES/FRCGC AGCM
NICAM 7km0.96−1.822.68
AGCM T42−1.460.594.17

[13] Net downward solar flux was smaller in the +2K-7km run than in the CTL-7km run. The solar feedback was −0.92 W m−2 K−1 in NICAM. Results showed a negative feedback for the short-wave radiation in contrast to the positive feedback for short-wave radiation in T42, which is corroborated by Cess et al. [1990]. Positive feedback for the long-wave radiation was larger than in the results from T42 and Cess et al. [1990]. The infrared feedback was 1.35 W m−2 K−1 in NICAM. The middle troposphere in the subtropics in NICAM was more humid in the +2K-7km run. This positive feedback was consistent with the relative importance of the mid-troposphere in the subtropics for water vapor feedback, as argued by Held and Soden [2000]. The climate sensitivity parameter was 0.44 K m2 W−1 in NICAM, which was smaller than 0.57 K m2 W−1 in T42 and the mean of 19 AGCMs in the work by Cess et al. [1990].

[14] Table 1a shows an imbalance between the net downward solar and emitted infrared fluxes of about 30 W m−2 both for CTL-7km and CTL-T42, which was a natural consequence of the fixed SST condition in the aqua planet experiment. The global average of surface albedo was smaller than the surface boundary condition of the earth because there was no land or snow/ice cover.

4. Discussion and Summary

[15] The present study showed that cloud amount increased in middle to high latitudes in response to a +2K SST forcing in NICAM. Consequently, global average cloud amount increased and a negative feedback in short-wave radiation occurred. Temkin et al. [1975] used a simple energy balance model and obtained qualitatively similar results, finding larger cloud amounts with warmer surface temperature and a negative feedback to the short-wave radiation that increased with latitude.

[16] In the present study, we had contradictory results that cloud amount increased in NICAM while it decreased in T42 by the same increase of SST. On the other hand, precipitation increased both in NICAM and T42 clearly in the extratropics (Figure 2c). From these results we may make the following conjecture: In the extratropics, precipitation and cloudiness associated with large-scale weather systems (e.g. extratropical cyclone) increase and decrease, respectively, in response to an increase of SST, as obtained by many coarser resolution models. However, smaller-scale and less-precipitating clouds which can be resolved in NICAM but not in T42 and other models tend to increase with increasing SST. An example of such cloud systems is cellular or stripe-shaped convection in the lower troposphere which often appears in the rear (west side) of cold fronts by cold air outbreak over warm water as seen in Figure 3.

[17] If the above conjecture is correct, there could be negative cloud-radiation feedback caused by smaller-scale clouds in the extratropics which can never be represented properly in coarse resolution models.

[18] Results in this paper do not support the “Iris effect” proposed by Lindzen et al. [2001], although a detailed discussion is beyond the scope of the present study. In this study, the change in cloud amount in the tropics simply followed the water vapor increase in the +2K-7km run. However, the change in the water vapor may not be a true response to an equilibrium state for several reasons. Tomita et al. [2005] showed that the ITCZ width did not converge in the 7-km experiment. The middle to upper troposphere may be drier in CTL-7km because of insufficient horizontal resolution. Also, if shallow clouds were properly modeled, the lower troposphere would be more humid.

[19] This paper reports on the first climate sensitivity test that used a global CRM. The cloud response to the +2K SST forcing was directly represented without cumulus parameterizations. The term “CRM” in this study is used to mean the 7-km grid model which marginally resolves deep convections. However, the experiments gave insight into changes in deep clouds and stratiform clouds that may accompany global warming, despite the limitations in experimental design that include the aqua planet condition, the numerical convergence of moist circulation, uncertainties in physical processes of CRMs, and the neglect of processes related to chemical and aerosol species.

[20] The ITCZ was broader and the Hadley circulation was weaker in +2K-7km compared to CTL-7km. The middle troposphere in the subtropics was more humid in +2K-7km. A negative feedback onto the short-wave radiation and a large positive feedback for the long-wave radiation occurred. The climate sensitivity parameter was smaller than in results from T42 and in the work by Cess et al. [1990]. The negative feedback of the short-wave radiation was in an opposite sense to the feedback in AGCMs with cumulus parameterizations (T42 and Cess et al. [1990]). However, these results were related to one aspect of climate sensitivity and cannot explain the whole processes of the actual atmosphere. An obvious next step to address these limitations would be an experiment that included realistic surface boundary conditions and the proper representation of sub-grid processes.


[21] We thank Teruyuki Nakajima and Miho Sekiguchi for useful discussion. We appreciate the cooperation of Masahide Kimoto and Tomoo Ogura on the use of CCSR/NIES/FRCGC AGCM. Two anonymous reviewers provided helpful comments. The simulations were all performed with the Earth Simulator in the Earth Simulator Center of the Japan Agency for Marine–Earth Science and Technology.