National Center for Atmospheric Research,* Boulder, Colorado. (Manuscript received 2 December 1999, in final form 3 October 2000)

1MAY 2001 WU AND MONCRIEFF 1155 Long-Term Behavior of Cloud Systems in TOGA COARE and Their Interactions with Radiative and Surface Processes. Part III: Effects on the Energy Budget and SST XIAOING WU ANDMITCHELL W. MONCRIEFF National Center for Atmospheric Research,* Boulder, Colorado (Manuscript received 2 December 1999, in final form 3 October 2000) ABSTRACT Most atmospheric general circulation models (GCMs) and coupled atmosphere ocean GCMs are unable to get the tropical energy budgets at the top of the atmosphere and the surface to simultaneously agree with observations. This aspect is investigated using a cloud-resolving model (CRM) that treats cloud-scale dynamics explicitly, a single-column model (SCM) of the National Center for Atmospheric Research (NCAR) Community Climate Model that parameterizes convection and clouds, and observations made during Tropical Oceans and Global Atmosphere Coupled Ocean Atmosphere Response Experiment (TOGA COARE). The same large-scale forcing and radiation parameterizations were used in both modeling approaches. We showed that the timeaveraged top-of-atmosphere and surface energy budgets agree simultaneously with observations in a 30-day (5 December 1992 3 January 1993) cloud-resolving simulation of tropical cloud systems. The 30-day time-averaged energy budgets obtained from the CRM are within the observational accuracy of 10 W m 2, while the corresponding quantities derived from the SCM have large biases. The physical explanation for this difference is that the CRM realization explicitly represents cumulus convection, including its mesoscale organization, and produces vertical and horizontal distributions of cloud condensate (ice and liquid water) that interact much more realistically with radiation than do parameterized clouds in the SCM. The accuracy of the CRM-derived surface fluxes is also tested by using the fluxes to force a one-dimensional (1D) ocean model. The 1D model, together with the surface forcing from the CRM and the prescribed advection of temperature and salinity, simulates the long-term evolution and diurnal variation of the sea surface temperature. This suggests that the atmosphere ocean coupling requires accurate representation of cloud-scale and mesoscale processes. 1. Introduction The effect of cloud systems on air sea interaction is a key issue in climate system modeling due to the great uncertainty in parameterizing subgrid scale convection, clouds, and air sea interaction. In order to improve the simulation of seasonal and interannual variability of tropical Pacific sea surface temperature (SST) in general circulation models (GCMs) and coupled atmosphere ocean GCMs, we must understand the atmospheric and oceanic processes that control the tropical cloud systems. Cloud systems affect the surface energy budget as well as the coupling between the atmosphere and the ocean through their fundamental influence on the surface radiative, heat, moisture, and momentum fluxes. * The National Center for Atmospheric Research is sponsored by the National Science Foundation. Corresponding author address: Dr. Xiaoqing Wu, CSG/MMM NCAR, P.O. Box 3000, Boulder, CO 80307. E-mail: xiaoqing@ncar.ucar.edu Realistic cloud-scale fields have to be either observed or modeled to fully investigate the cloud ocean as well as cloud radiation interaction. Most GCMs and coupled atmosphere ocean GCMs are unable to get the tropical energy budgets at the top of the atmosphere and the surface to simultaneously agree with observations. For example, a good agreement of radiative fluxes at the top of the atmosphere between observations and large-scale models can be obtained by tuning the model cloud fraction and aerosol optical depth, but a large bias in the surface energy budget typically remains (e.g., Kiehl 1998; Kiehl et al. 1998). This is a crucial issue in coupled atmosphere ocean models. Considerable efforts have been made to understand the cause of bias in the surface energy budget and to alleviate the climate drift problem in coupled atmosphere ocean GCMs. In a sensitivity study, Ma et al. (1994) found that the surface properties of a coupled model are sensitive to the longwave radiation parameterization. The change of longwave scheme affects the surface evaporation and wind field through the impact on the longwave heating rate, atmospheric stability, and 2001 American Meteorological Society

1156 JOURNAL OF THE ATMOSPHERIC SCIENCES VOLUME 58 convective activity in the coupled model. By directly tuning the cloud cover within a convectively active grid box in the coupled model, Stockdale et al. (1994) showed that the improvement of the top-of-atmospheric shortwave fluxes and the net surface heat flux reduces the climate drift in the western Pacific. Kiehl (1998) analyzed the energy budgets over the tropical Pacific warm pool from a 15-yr climate simulation of the National Center for Atmospheric Research (NCAR) Community Climate Model version 3 (CCM3). He showed that both longwave and shortwave fluxes at the top of the atmosphere are in good agreement with observations, but the net surface heat flux is about 30 W m 2 larger than in observations. The bias in the net surface heat flux is mainly due to the bias in the shortwave radiative flux and latent heat flux, both too large compared to observations. Kiehl also found that when the CCM3 is coupled with an ocean model, the shortwave flux bias is about the same as the uncoupled model, but the net surface heat flux bias is reduced. This is due to the increase of surface latent heat flux resulting from the increase of SST gradient and zonal wind. He hypothesized that an underestimate of cloud shortwave absorption may be responsible for the surface shortwave flux bias. Recently, Yu and Mechoso (1999) discussed the surface heat flux errors in the tropical Pacific using the University of California, Los Angeles, GCM and coupled atmosphere ocean GCM. They also found that the biases are mainly in the surface shortwave flux and latent heat flux. One way or another, clouds play a key role. In present GCMs and coupled GCMs, convection and clouds are usually parameterized separately. The moist convection scheme represents the convective tendency for the temperature and moisture equations, while the cloud scheme either diagnoses or predicts cloud properties such as fractional cover, ice water, and liquid water. It is hardly surprising that a large bias in the net surface heat flux can occur considering that the multiscale nonlinear interactions among convection, clouds, radiation, surface processes, and large-scale dynamics have to be accounted for by simple parameterizations. To improve our understanding of this multifaceted interaction, we adopt a cloud-resolving model (CRM) approach. Because of the rapid increase of computer power, especially during the last decade, the CRM, which resolves cloud-scale and mesoscale dynamics, can now simulate the long-term evolution of tropical cloud systems in a large domain. A desirable degree of realism is attained by a CRM that uses objectively analyzed evolving large-scale advection and wind fields (e.g., Wu et al. 1998, 1999). Note that in a CRM, a convection parameterization is no longer needed, cloud properties are predicted by cloud condensate equations, cloud microphysical processes directly respond to the cloudscale dynamics, and the radiative flux cloud system surface flux interaction is largely controlled by explicit cloud-scale dynamics. A key point is that the large-scale forcing and vertical wind shear exert a fundamental control over the morphology of convection and its large-scale effects. This control and the attendant transport properties have a rigorous dynamical basis (Moncrieff 1981), which has often been realized in hours-long simulations using idealized initiation methods (e.g., cold pools). Only recently have evolving large-scale forcing and wind shear been shown to control weeks-long cloud system evolution in simulations started from random excitation of a horizontally uniform state. These simulations realized transitions among regimes, such as scattered convection, cloud clusters, and highly organized squall lines (Grabowski et al. 1996, 1998; Wu et al. 1998, 1999). The explicit coupling among cloud microphysics, turbulence, and radiative transfer by cloud-scale dynamics is largely responsible for the success of these simulations (Moncrieff 1995). The Tropical Ocean Global Atmosphere (TOGA) Coupled Ocean Atmosphere Response Experiment (COARE) provided a unique opportunity for a comprehensive CRM-based cloud system study because of the available long-term soundings (4 months), satellite, aircraft, radar, and surface data. As shown in Part I and Part II of this paper, the model-produced (synthetic) results were extensively evaluated against various independent datasets, such as outgoing longwave radiative (OLR) flux, albedo, cloud radiative forcing, surface sensible and latent heat fluxes, and rainfall data derived from radar measurements. These quantitative tests indirectly verify the synthetic convective mass flux and cloud properties that are crucial for improving existing (and developing new) convection and cloud parameterization schemes for large-scale models. Note that comprehensive cloud-scale data are not directly obtainable from observations, at least not in a self-consistent way. The surface radiative fluxes, heat fluxes and precipitation from the CRM provide a uniquely consistent synthetic forcing for quantifying relationships among cloud systems, surface radiative and heat fluxes, SST, and upper-ocean structure in ocean models. The objectives of this paper are as follows: (i) to examine the effects of cloud systems on both the surface energy budget and the top-of-atmosphere radiative flux using the CRM, single-column model (SCM) of CCM3, and observations and (ii) to demonstrate the importance of numerically simulated realization of cloud systems in defining an accurate surface forcing for the simulation of SST in a one-dimensional (1D) ocean model. While the focus is on the surface processes, the top-of-atmosphere radiative fluxes from the CRM are also compared with observations to better understand cloud system radiation surface interaction. Observational studies (e.g., Weller and Anderson 1996; Anderson et al. 1996) have shown that the net surface heat flux obtained during TOGA COARE was accurate to within 10 W m 2. Part II of this paper (Wu et al. 1999) showed that the difference of net surface

1MAY 2001 WU AND MONCRIEFF 1157 heat flux between the CRM and the improved meteorological (IMET) surface mooring buoy data is indeed within this level of uncertainty. The comparison of CRMs with SCMs is a natural next step to identify and isolate the cause of uncertainties, such as the bias in the net surface energy budget in full GCMs, and to eventually improve parameterizations. A key point is that convection and clouds are directly resolved in a CRM, while they are represented separately (not necessarily in a consistent way) by parameterization schemes in an SCM. It should be noted that unlike GCMs, cloud systems simulated by the CRM and parameterized by the SCM do not feed back to the large-scale advection that is prescribed from observations. 1 Therefore, any improvement in parameterization will need to be tested in full GCMs. The paper is organized as follows. Section 2 briefly describes the CRM and SCM, the experimental design, and the observed large-scale datasets. Section 3 compares the top-of-atmosphere radiative fluxes and the surface energy budgets obtained from observations, the CRM, the SCM of CCM3, and the special TOGA COARE reanalysis product of the European Centre for Medium-Range Weather Forecasts (ECMWF). The 1D ocean model is introduced in section 4, and the accuracy of surface fluxes is further evaluated by examining the ocean response to the observed and modeled surface forcing. Concluding remarks follow in section 5. 2. Cloud-resolving model, single-column model, experimental design, and observational datasets a. Cloud-resolving model The detailed description of the CRM can be found in Wu et al. (1998, 1999). Its dynamical core is a twodimensional version of the Clark Hall cloud model (Clark et al. 1996). The cloud-interactive radiation parameterization is represented by the radiation code of the NCAR Community Climate Model (CCM) (Kiehl et al. 1994), modified to distinguish between the radiative properties of liquid and ice cloud particles. The microphysics parameterization is represented by the Kessler (1969) bulk warm rain parameterization and the Koenig and Murray (1976) bulk ice parameterization. Two types of ice particles (i.e., type A ice and type B ice) are considered in the ice scheme. Type A ice particles represent particles that are initially small and are created by either heterogeneous sorption nucleation of ice crystals above 40 C or homogeneous freezing of cloud droplets at temperature below 40 C. Type B ice particles represent particles that are initially larger and 1 The feedback between simulated cloud systems and the environment is diagnosed to the extent that evolving observed large-scale forcing affects the simulated organization of convection and thereby convective heat sources, moisture sinks, and dynamical tendencies. form when a raindrop freezes after collision with a type A ice particle. Only type A ice is used in the radiation calculation. Effective radii of liquid and type A ice particles are assumed to be 10 and 30 m, respectively. The surface fluxes of sensible and latent heat are calculated using an evolving, horizontally uniform SST and a simplified version of the TOGA COARE surface flux algorithm (Fairall et al. 1996). The subgrid-scale mixing is parameterized using the first order eddy diffusion method of Smagorinsky (1963). The nonlocal boundary layer diffusion scheme (Troen and Mahrt 1986; Holtslag and Moeng 1991; Hong and Pan 1996) assumes a constant boundary layer height of 600 m. The two-dimensional x z domain is 900 km long by 40 km deep with a 3-km horizontal resolution and periodic lateral boundary conditions. A stretched grid in the vertical with 52 levels (100 m at the surface, increasing to 1500 m at the model top of 40 km) is used, requiring a time step of 15 s. Rigid, free-slip bottom and top boundary conditions are applied together with a gravity wave absorber in uppermost 14 km of the domain. b. Single-column model The SCM was developed from the CCM3 by Hack et al. (2000). The governing equations for the temperature and moisture can be expressed as F 1A 1D 1R, (1) t q F q 2A 2D, (2) t where is the potential temperature, q the water vapor R /c mixing ratio, (p/p d p o) is the Exner function (p is the pressure and p o 1000 hpa), R d is the gas constant for dry air, and c p the specific heat of dry air at constant pressure. The term 1R is the radiative heating rate; 1A and 2A are the tendencies due to the moist convection, the large-scale (stable) condensation, and the dry adiabatic adjustment for the temperature and moisture, respectively; 1D is the tendency due to the diffusion, gravity wave drag, and surface sensible heat flux; and 2D is the tendency due to the diffusion and surface latent heat flux. The terms F and F q are the large-scale forcing for the temperature and moisture, respectively, and are identical to those used in the CRM. The description of the physical parameterization schemes can be found in Kiehl et al. (1998). The radiation scheme is modified to be consistent with the one used in the CRM, that is, the tropospheric aerosol and various trace gases are turned off. Deep moist convection is represented by the Zhang and McFarlane (1995) scheme. The Hack (1994) scheme represents shallow convection and also removes moist convective instability remaining after the Zhang McFarlane scheme has been applied. Cloud amount, cloud liquid, and ice water are diagnosed by the Slingo (1987) cloud scheme, which is dependent

1158 JOURNAL OF THE ATMOSPHERIC SCIENCES VOLUME 58 TABLE 1. Surface energy budget (W m 2 ) from observations (OBS and SSM/I/GMS), the CRM and SCM, averaged over the period of 30 days from 5 Dec 1992 to 3 Jan 1993. is the net surface long-wave flux, the net surface short-wave flux, lat the surface latent heat flux, sen the surface sensible heat flux, and net the net surface flux. Fluxes (W m 2 ) OBS lat sen net 51.6 178.6 122.9 10.3 6.2 SSM/I GMS CRM M0 SCM ECMWF 48.2 173.2 126.9 7.7 9.6 48.2 179.3 125.2 9.5 3.6 net lat sen 57.0 227.7 139.3 11.1 20.3 57.3 206.2 145.1 11.7 7.9 on convective mass flux, relative humidity, vertical velocity, and atmospheric stability. Only one type of ice particle is considered in the cloud scheme. Effective radii of liquid and ice particle used in the SCM are the same as those used in the CRM. c. Experimental design and observational datasets The 39-day CRM simulation M0 presented in Wu et al. (1999) is used. Since this simulation is less realistic during the undisturbed period (6 12 January), arguably due to the lack of large-scale forcing for the condensed water (Grabowski et al. 1996; Wu et al. 1998), only the first 30 days (5 December 3 January) of model output is analyzed. The SCM is run for the same 30-day period with the same forcing and SST used in the CRM. Following Wu et al. (1998, 1999), the evolving large-scale forcing is calculated from the objectively analyzed sounding data (Lin and Johnson 1996) and is averaged over the intensive flux array (IFA) of TOGA COARE. The evolving SST is the average measurements from four buoys, namely the IMET and three TOGA TAO (Tropical Atmosphere Ocean) ATLAS (Automated Temperature Line Acquisition System) moorings. The horizontal momentum fields in the SCM are specified by the horizontally averaged evolving wind components obtained from observations. 3. Surface energy budget and radiative flux at the top of the atmosphere a. Comparison among observations, CRM, SCM, and ECMWF special analysis In Table 1, we compare the model-produced and observed surface energy budgets averaged over the 30- day period. The observed fluxes (OBS) are averaged from three ships (R/V Kexue 1, R/V Shiyan 3, and R/V Moana Wave) and a buoy (IMET) data within the IFA of TOGA COARE. The daily surface energy budget derived from the Special Sensor Microwave/Imager (SSM/I) and the Geostationary Meteorological Satellite (GMS) radiance measurements (Chou et al. 1997; Chou et al. 1998) is also used for the comparison. The CRMproduced flux is averaged over the 900-km domain. The surface flux from the ECMWF TOGA COARE special analysis is averaged over the IFA. Note that the ECMWF analysis incorporated the special sounding data from radiosonde deployed during the field experiment. The net surface longwave, shortwave, latent heat, and sensible heat fluxes from the CRM are in excellent agreement with the observations, each difference being smaller than 5 W m 2. The difference between the CRMproduced and observed net surface fluxes is 2.6 W m 2. The CRM-produced fluxes are also consistent with the fluxes independently derived from the SSM/I and GMS measurements. The comparison between the SCM and observations shows a large discrepancy, especially in the net shortwave radiative flux and the latent heat flux. The SCMproduced net shortwave flux and latent heat flux are 49.1 W m 2 and 16.4 W m 2, respectively, larger in magnitude than the observed fluxes. The SCM-produced net longwave radiative flux and sensible heat flux are also larger than the observed values. The larger heating of the ocean due to the SCM-produced shortwave flux is therefore partly compensated by the larger cooling of the ocean due to the SCM-produced latent heat flux, sensible heat flux, and longwave flux. The difference of net surface flux between the SCM and observations (26.5 W m 2 ) is similar to that in the full GCM (Kiehl 1998). The surface fluxes from the ECMWF special analysis, the product of a numerical weather prediction model, data assimilation, and some TOGA COARE observations, show the similar biases in the shortwave flux and latent heat flux. To quantify the bias, Fig. 1 shows the 30-day evolution of net shortwave flux and latent heat flux derived from the observations, CRM, and SCM. The CRM flux evolves in a way similar to the observed flux, albeit with some differences. The larger difference of shortwave and latent heat fluxes between the SCM and observations (Figs. 1a and 1b) occurs during the strong convection period from 11 to 26 December (Wu et al. 1998, 1999). The large difference of latent heat flux (Fig. 1b) toward the end of the 30-day period between the SCM and observations is partly due to the dry biases in the SCM moisture field (not shown). The lack of shallow convection in the SCM during the last 5 days of the 30-day period may be responsible for the dry and warm biases. The next question is how the model-produced radiative fluxes at the top of the atmosphere compare to observations. Table 2 lists the 30-day averaged TOA radiative fluxes from the CRM, SCM, and three kinds of (indirect) observational estimates. The FC (flux and cloud) data is derived from a radiative transfer model using the satellite-measured radiance and the cloud properties obtained from the International Satellite Cloud Climatology Project (ISCCP) (Zhang et al. 1995).

1MAY 2001 WU AND MONCRIEFF 1159 FIG. 1. 30-day evolution of (a) daily net surface shortwave fluxes (W m 2 ) and (b) daily surface latent heat fluxes (W m 2 ) from the CRM, SCM, and observation (OBS). The GMS1 data is produced by the calibration of satellite instruments against collocated observations from the National Aeronautics and Space Administration (NASA) ER-2 aircraft (Collins et al. 1997). The GMS2 data is estimated by using an empirically obtained narrowband broadband relationship from simultaneous satellite measurements (Minnis et al. 1991; Doelling et al. 1990). In terms of the TOA net longwave flux, the CRMproduced flux is about 5 W m 2 less than the observations, while the SCM is about 10 W m 2 larger than the observed fluxes. The ECMWF flux is even larger, which suggests a general underprediction of cloud amounts during the 30-day period. For the net shortwave flux at the TOA, the CRM value is in a good agreement with the FC and GMS2 but smaller than the GMS1 flux. The difference between the GMS1 and GMS2 shortwave fluxes is 20 W m 2. A careful study, beyond the scope of this paper, is needed to explain this difference. The shortwave fluxes from the SCM and ECMWF are larger than all three estimates but closer to the GMS1 flux. The net radiative heating ( R ) in the atmosphere can be calculated by subtracting the net surface longwave and shortwave fluxes from the net TOA flux. Table 3 lists the 30-day averaged R, net longwave heating R (), and net shortwave heating R () from the CRM, SCM, ECMWF, and observations. The net longwave and shortwave fluxes averaged from three ships and IMET buoy (OBS in Table 1) are used to get R, R (), and R () for the FC, GMS1, and GMS2, TABLE 2. TOA radiative fluxes (W m 2 ) from observations (FC, GMS1, and GMS2), the CRM and SCM, averaged over the period of 30 days from 5 Dec 1992 to 3 Jan 1993. Flux (W m 2 ) FC GMS1 GMS2 CRM M0 SCM ECMWF 198.8 269.5 200.4 294.4 205.1 274.4 195.8 274.8 211.7 312.4 246.9 307.7

1160 JOURNAL OF THE ATMOSPHERIC SCIENCES VOLUME 58 TABLE 3. Net radiative heating R, net longwave heating R (), and net shortwave heating R () from observations (FC, GMS1, and GMS2), the CRM, SCM, and ECMWF TOGA COARE special analysis, averaged over the period of 30 days from 5 Dec 1992 to 3 Jan 1993. Heating (W m 2 ) FC GMS1 GMS2 CRM M0 SCM EMCWF R R () R () 56.3 147.2 90.9 R R () R () R () R () 33.0 148.8 115.8 57.7 153.5 95.8 52.1 147.6 95.5 70.0 154.7 84.7 88.1 189.6 101.5 respectively. The CRM-produced R is close to the heating from GMS2 and FC but is 19.1 W m 2 larger in magnitude than that from GMS1. The SCM-produced R is 17.9 W m 2 larger in magnitude than the CRMproduced R. Among all datasets listed in Table 3, the ECMWF special analysis produces the largest longwave cooling and net radiative cooling, while CMS1 has the largest shortwave heating and the smallest net radiative cooling. The SCM and the ECMWF special analysis produce a larger R ( 70.0 W m 2 and 88.1 W m 2 ) compared to that of the CRM ( 52.1 W m 2 ). Figure 2 shows the 30-day evolution of R from the CRM, SCM, GMS2, and GMS1. It is noted that the difference of R between GMS1 and GMS2 is smaller during the first half of the 30-day period than during the second half. The different kinds of cloud systems that occur as the attendant wind fields evolve explain the distinctive radiative heating during these two periods. As shown in Wu and LeMone (1999), during the period of 10 17 December, when easterly winds prevail above 2 km, both precipitating clouds and upper-tropospheric anvil clouds travel westward at about the same speed ( 10 m s 1 ). However, during the period of 20 27 December, when westerlies dominate below 5 km (westerly wind burst), precipitating clouds move eastward and anvil clouds move westward. The above results show that the CRM gets the surface energy budgets and the TOA radiative fluxes to simultaneously agree with the observations within the limits of measurement uncertainty. On the other hand, the SCM and ECMWF special analysis produce large biases, especially in the shortwave flux and latent heat flux, during the strong convection period. b. Radiative effects of cloud condensate Both atmospheric and cloud properties have impacts on the TOA and surface radiative fluxes. While temperature and moisture fields play an important role in the longwave radiative flux, cloud condensate is the most important factor for the shortwave radiative flux (e.g., Zhang et al. 1995). To find out the physical explanation responsible for the different radiative fluxes between the CRM and SCM, we are comparing the cloud and atmospheric properties from the two models. Figures 3 and 4 present the 30-day evolution of cloud ice water mixing ratio and cloud liquid water mixing ratio from the CRM and SCM, respectively. The vertical FIG. 2. 30-day evolution of daily net radiative heating (W m 2 ) from the CRM, SCM, and observations (GMS1 and GMS2).

1MAY 2001 WU AND MONCRIEFF 1161 FIG. 3. 30-day evolution of 6-h vertical profiles of (a) cloud ice water mixing ratio (g kg 1 ) and (b) cloud liquid water mixing ratio (g kg 1 ) from the CRM. profiles of cloud condensate are distinctly different between the two models. For the cloud ice field, the peak value for the CRM lies between 400 and 500 hpa (Fig. 3a), while the peak for the SCM is at 300 hpa (Fig. 4a). The CRM produces much more cloud ice throughout a much deeper layer than the SCM. The depleted cloud ice in the SCM allows more longwave emission to space and less shortwave flux reflection to space than those in the CRM. The result is that the net longwave and shortwave fluxes at the TOA are large compared to the CRM (Table 2). The respective cloud liquid water fields are also very different (Figs. 3b and 4b). The CRM-produced cloud liquid water has one major peak around 600 hpa, while the SCM tends to produce two peaks at about 500 and 800 hpa. Overall, the CRM produces more cloud liquid water below 500 hpa than the SCM. There are more shallow clouds in the CRM, which have a larger impact on the shortwave radiative flux but a relatively small impact on the longwave radiative flux. The respective cloud fractions are also compared in Figs. 5a and 5b (time series) and Fig. 6 (30-day mean). The cloud fraction from the CRM is calculated as follows. At each level, a grid box is defined to be cloudy if the sum of liquid and ice water paths exceeds threshold 0.2 g m 2. This threshold is obtained from offline radiation calculations. The condensate with the cloud water path smaller than 0.2 g m 2 has negligible effect on the TOA net longwave flux (Fig. 7), the TOA net shortwave flux, and the surface radiative flux (not shown). The cloud fraction from the CRM is larger during the second half of the 30-day period than in the first half (Fig. 5a), which is a result of different cloud systems occurring in these two periods, as shown in Wu and LeMone (1999). The SCM produces generally smaller cloud fractions (Fig. 5b) than the CRM for these two periods. The peak of cloud fraction for the SCM (300 hpa) is lower than that for the CRM (200 hpa) (Fig. 6). The SCM has much less cloud than the CRM in the layer between 400 and 700 hpa. To further identify the impact of convection and clouds on the radiative flux, the net radiative flux is separated into clear-sky and cloudy-sky fluxes, as shown in Table 4. The clear-sky radiative flux for the CRM and the SCM is obtained from an offline calculation of

1162 JOURNAL OF THE ATMOSPHERIC SCIENCES VOLUME 58 FIG. 4. Same as Fig. 3 but from the SCM. the radiative transfer model by setting the cloud condensate fields to zero. The clear-sky flux is thereby more precisely defined than the method used in Parts I and II of this paper (Wu et al. 1998, 1999), especially when clouds occupy most of the domain. The cloudy-sky flux, or cloud radiative forcing, is the difference between the net flux and the clear-sky flux. The largest difference between the CRM and the SCM is in the shortwave cloud radiative forcing ( ). The surface and TOA from the CRM are 45.0 and 41.1 Wm 2, respectively, larger than the SCM. Note that the ratio of / from both the CRM and SCM is close to 1, which is a value consistent with fundamental derivations (e.g., Stephens and Tsay 1990). The TOA and surface clear-sky longwave radiative fluxes of the SCM are significantly larger in magnitude than those of the CRM. An offline calculation of clear-sky radiative fluxes, which used the SCM-produced temperature profile and the CRM-produced moisture profile, is performed to examine the relative contributions of temperature and moisture bias to the difference between the SCM and CRM. The TOA and surface from this offline calculation are 281.6 and 66.7 W m 2, respectively, compared with 292.4 and 80.8 W m 2 of the SCM. Another offline calculation of clear-sky radiative fluxes that used the SCM-produced moisture profile and the CRM-produced temperature profile gives 289.5 W m 2 for the TOA and 83.9 W m 2 for the surface, which are close to the values given by the SCM. This demonstrates that the moisture bias (Fig. 8b) is most important in determining the differences in clear-sky longwave radiative fluxes at the surface and TOA, while the temperature bias (Fig. 8a) is secondary. As the SCM is integrated forward in time, the biases produced by the convection and cloud parameterization schemes as well as other physical parameterizations will be reflected in the temperature and moisture fields. Because the SCM represents only one grid point in a GCM, the biases could amplify with time, unlike in the full GCM, wherein the large-scale dynamics may smooth these biases. To illustrate the radiative impacts of cloud condensate, the clear- and cloudy-sky radiative fluxes from two cloud-resolving simulations (i.e., M0 and E0) are compared in Table 5. Simulation E0 is presented in Part I (Wu et al. 1998) and M0 in Part II (Wu et al. 1999). M0 differs from E0 mainly in the ice parameterization scheme. In M0, the representation of ice fall speed used

1MAY 2001 WU AND MONCRIEFF 1163 FIG. 5. 30-day evolution of 6-h vertical profiles of cloud fraction from (a) the CRM and (b) the SCM. FIG. 6. 30-day mean vertical profiles of cloud fraction from the CRM (solid) and SCM (dashed). FIG. 7. Threshold cloud water path vs 30-day mean TOA net longwave flux during TOGA COARE.

1164 JOURNAL OF THE ATMOSPHERIC SCIENCES VOLUME 58 TABLE 4. Clear-sky () and cloudy-sky () longwave () and shortwave () radiative fluxes at the TOA and at the surface from the CRM and SCM, averaged over the period of 30 days from 5 Dec 1992 to 3 Jan 1993. Fluxes (W m 2 ) CRM M0 SCM 277.2 81.4 394.1 119.3 70.8 22.6 304.2 124.9 292.4 80.7 390.6 78.2 80.8 23.8 307.6 79.9 TABLE 5. Clear-sky () and cloudy-sky () longwave () and shortwave () radiative fluxes at the TOA and at the surface from two CRM simulations (M0 and E0), averaged over the period of 30 days from 5 Dec 1992 to 3 Jan 1993. Fluxes (W m 2 ) CRM M0 CRM E0 277.2 81.4 394.1 119.3 70.8 22.6 304.2 124.9 275.9 103.4 394.1 149.1 72.6 29.5 304.2 156.8 in Koenig and Murray s ice scheme was modified based on microphysical measurements. Less extensive ice cloud was produced in M0 than in E0. The 30-day mean ice water path is 0.18 kg m 2 for M0 and 0.39 kg m 2 for E0. The 30-day mean liquid water path is about the same (0.20 kg m 2 for M0 and 0.19 kg m 2 for E0). The cloud condensate is the primary factor that produces the differences in radiative fluxes. The clear-sky radiative fluxes are basically the same for M0 and E0; the major difference lies in the cloud radiative forcing. The TOA and surface cloud radiative forcing from E0 are 29.8 and 31.9 W m 2, respectively, larger than those from M0. More extensive ice clouds in the upper troposphere from E0 also result in larger TOA longwave cloud radiative forcing, 22.0 W m 2 more than that from M0. The effect of cloud condensate and cloud fraction on the radiative flux can be quantified by two offline radiation calculations (MS1 and MS2). In MS1, the SCMproduced temperature, moisture, and cloud fraction profiles are used with the CRM-produced cloud condensate profile. In MS2, the SCM-produced temperature, moisture, and condensate profiles are used with the CRMproduced cloud fraction profile. The net TOA and surface radiative fluxes of MS1 and MS2 are presented in Table 6 together with those of SCM. The differences between SCM and MS1 are larger than those between SCM and MS2 except for the net surface long-wave flux. This shows that cloud condensate is more important than cloud fraction in determining the radiative impact of cloud condensate. Larger cloud condensate in MS1 allows less shortwave flux into the surface and more shortwave flux reflected into space (therefore smaller net TOA shortwave flux) compared with those in SCM. Larger cloud condensate in MS1 allows less longwave flux emitting to space and results in smaller net TOA longwave flux. Larger cloud fraction in MS2 emits more downwelling longwave flux at the surface and produces smaller net surface longwave flux. c. Radiative effects of cloud distribution The comparison between the CRM and the SCM in the last section shows large differences in the profiles of cloud condensate. Naturally, the next question is can the radiation scheme in the SCM produce the right flux FIG. 8. 30-day mean profiles of differences of (a) temperature ( C) and (b) water vapor mixing ratio (g kg 1 ) between models and observations. The CRM is in solid lines, and the SCM in dashed lines.

1MAY 2001 WU AND MONCRIEFF 1165 TABLE 6. Radiative fluxes (W m 2 ) from the SCM and the offline calculations (MS1 and MS2, see the test for the description) of radiative transfer model, averaged over the period of 30 days from 5 Dec 1992 to 3 Jan 1993. Fluxes (W m 2 ) SCM MS1 (SCM MS1) 211.7 57.0 312.4 227.7 185.6 ( 26.1) 55.9 ( 1.1) 262.3 (50.1) 171.7 (56.0) MS2 (SCM MS2) 204.1 ( 7.6) 49.1 ( 7.9) 288.6 (23.8) 202.1 (25.6) TABLE 7. Radiative fluxes (W m 2 ) from the CRM (M0) and the offline calculation (M1) of radiative transfer model, averaged over the period of 30 days from 5 Dec 1992 to 3 Jan 1993. Fluxes (W m 2 ) CRM M0 M1 195.8 48.2 274.8 179.3 171.2 39.3 238.4 141.1 with the 30-day time series of domain-averaged profiles of CRM cloud condensate. The answer to this question is no because the treatment of cloud distribution or the cloud radiation interaction in the radiation scheme is another factor that contributes to the different radiative fluxes between the CRM and the SCM. To demonstrate this point, an offline calculation (M1) of the radiative transfer model was performed using the domain-averaged temperature and moisture profiles, cloud properties, and cloud fraction from the CRM. In the full CRM (M0), the so-called binary clouds (i.e., completely overcast or clear skies) are used for 300 columns. The radiative flux is calculated for each column using the radiative transfer model, and the 300 columns are then averaged to get the mean radiative flux. The radiative effect of cloud distribution (the cloud geometric association and inhomogeneity) is thereby explicitly included in the CRM-produced radiative flux. However, the effect of cloud distribution has to be treated by vertical cloud overlap assumptions in M1. A random overlap assumption is currently used in the CCM3 radiation scheme. Because the mean cloud properties are the same for M1 and M0, the different radiative fluxes shown in Table 7 is due to the random overlap assumption used. The random overlap assumption tends to overestimate the total cloud cover (e.g., Tian and Curry 1989), which results in smaller short-wave and long-wave fluxes in M1 than in M0. The effects of the cloud geometric association and inhomogeneity account for the differences between M1 and M0, that is, 36.4, 38.2, 24.6, and 8.9 W m 2 in,,, and, respectively. In summary, the above results demonstrate that an accurate cloud condensate field and the representation of vertical and horizontal cloud distributions are important for getting the TOA flux and surface energy budget correct simultaneously. In other words, the parameterization of cloud condensate and cloud fraction in the cloud scheme and the representation of cloud geometric association and inhomogeneity in the radiation scheme need to be improved in order to achieve an accurate energy budget. Moreover, it is very likely that a more physically based convection scheme in the SCM is required in order to improve the simulation of the atmospheric state. 4. Ocean response in a 1D model The analyses in the last section demonstrate that the CRM is able to produce the time-averaged TOA and surface energy budget that simultaneously agree with observations. The accuracy of the model-produced surface energy budget or surface forcing can be further evaluated using a 1D ocean model. The ocean response to the CRM-produced, SCM-produced, and observed surface forcing is expected to lead to different ocean atmosphere interaction in the coupled GCM. Several previous modeling studies (e.g., Anderson et al. 1996; Sui et al. 1997) have shown that 1D ocean models can well simulate the ocean response to the observed surface forcing and advection during TOGA COARE. The 1D ocean model used in this paper is from Large et al. (1994). The governing equations can be expressed as u(z, t) [ w u ] f, (3) t z (z, t) [ w ] fu, (4) t z T(z, t) [ w T ] n A T, (5) t z z s(z, t) [ w s ] F n A s, (6) t z z where u,, T,and s are horizontal (zonal and meridional) components, temperature, and salinity, respectively. The term w a is the vertical turbulent fluxes of either momentum, temperature, or salinity depending on the identity of a (e.g., u,, T,or s); fu and f are the Coriolis terms; and n and F n the nonturbulent heat and freshwater fluxes. The terms A T and A s are the sum of horizontal and vertical advection for the temperature and salinity, respectively. For the 30-day simulation presented in the next section, 30 W m 2 and 3 10 7 kg m 2 s 1 are used for A T and A s, respectively (Feng et al. 1998; Antonissen 2000). A nonlocal K-profile parameterization is used for the vertical mixing in the oceanic surface boundary layer. The oceanic boundary layer depth depends on the surface forcing, buoyancy, and velocity. The vertical mixing in the interior is due to internal waves, shear instability, and double diffusion. The model used in this

1166 JOURNAL OF THE ATMOSPHERIC SCIENCES VOLUME 58 This indicates a cooling effect by horizontal advection in the ocean during this 30-day period. The impact of horizontal advection on the warm pool heat budget during the westerly wind burst has been discussed by Cronin and McPhaden (1997). FIG. 9. 30-day evolution of 6-h observed SST (solid), modeled SST using CRM fluxes (dashed), modeled SST using observed fluxes (dotted), and modeled SST using SCM fluxes (dot dashed). Units: C. study has a depth of 200 m with 0.5-m resolution, and the time step is 900 s. The reader is referred to Large et al. (1994) for a detailed description of the model physics. Figure 9 shows the 30-day evolution of observed SST and the SST produced by the 1D ocean model using the observed, CRM-produced, and SCM-produced surface forcing. Note that the 1D ocean model also requires the surface momentum flux or wind stress besides the surface heat flux. The observed surface wind stress is the average estimates derived from four buoys. The wind stress from the CRM and SCM is similar to the observed value because the modeled mean wind field is relaxed to the observed wind field. The 30-day ocean model produced SSTs using the CRM and observed forcing are in reasonable agreement with the observed SST. The long-term trend and diurnal variation are each well simulated by the ocean model using the CRM-produced and observed surface forcing. Three factors play the role in these successful simulations. The first factor is the surface forcing. As shown in Fig. 9, the ocean model produced SST using the SCM forcing is much too warm compared to the observed SST, although its diurnal variation is well represented. The large surface shortwave flux produced by the SCM is responsible for this failure. The second factor is the upper oceanic mixing and entrainment of the oceanic deep water. As shown in Part I of this paper (Wu et al. 1998), the long-term trend and diurnal variation of SST cannot concurrently be produced by a simple oceanic mixed-layer model, which omits the oceanic vertical mixing. The third factor is the horizontal advection of temperature and salinity. In a sensitivity test in which the observed horizontal advection of temperature (A T ) and salinity (A s ) was neglected, the model-produced SSTs using either the CRM-produced or the observed surface forcing are warmer than observed (not shown). 5. Concluding remarks We have demonstrated, through cloud-resolving modeling of tropical cloud systems in TOGA COARE, that numerically simulated 30-day mean top-of-atmosphere radiative fluxes and surface energy budgets using a conventional radiative transfer parameterization (i.e., without enhanced shortwave absorption) simultaneously agree with observations to within the measurement uncertainty. We have also demonstrated that the 30-day evolution and diurnal variation of the SST can be reasonably simulated by the 1D ocean model using the CRM-produced and observed fluxes which, in turn, verified the accuracy of CRM-produced surface fluxes. This is in marked contrast to atmospheric general circulation models and coupled atmosphere ocean models that have great difficulty in achieving an accurate energy balance. The comparison among the CRM, SCM, and observations suggests that important factors may include the accurate realization of cloud condensate and its vertical and horizontal distributions. We argue that this can be achieved only if cloud system regimes and transitions between regimes are correctly realized as the large-scale conditions evolve. CRMs explicitly treat cloud-scale and mesoscale dynamics and (prognostically) calculate condensate distribution with a horizontal grid interval of 3 km and vertical interval of about 200 m. Cloud-scale dynamics are explicitly coupled at small scales to the latent heating. In other words, the rates of change among the three phases of water in the microphysical parameterizations are coupled through cloud-scale dynamics that are, in turn, controlled by the environmental shear. These multiscale coupling and controlling properties generate strongly interacting dynamical, thermodynamical, and microphysical fields. Such interaction is difficult to achieve with parameterized convection and clouds at GCM resolution. The effects of cloud vertical distribution on the radiative heating profile and the atmospheric general circulation have been investigated by varying the vertical structures of clouds and modifying the cloud scheme in GCMs (e.g., Webster and Stephens 1984; Randall et al. 1989; Slingo and Slingo 1991; Wang and Rossow 1998). However, these modifications of cloud condensate fields require a physical or observational basis. Our successful evaluation of the bulk properties of convection and clouds against TOGA COARE observations shows that the CRM-produced results provide such a basis, albeit in a synthetic form. Observations alone cannot fully describe the cloud geometric associations, but CRMs driven by objectively analyzed large-scale fields are a

1MAY 2001 WU AND MONCRIEFF 1167 way to obtain cloud geometric characteristics for the mosaic treatment of cloud radiation interactions. Additional cloud-resolving simulations should be performed to determine sensitivity to different environments, microphysical schemes, radiation parameterizations, three spatial dimensions, and yet higher resolution (Wu and Moncrieff 1996; Grabowski et al. 1998). There is a dearth of datasets of sufficient dynamic range, especially a cloud condensate dataset, to properly evaluate long-term CRM simulations. While TOGA COARE provided comprehensive surface properties and convective properties, TOA measurements of comparable quality are not available. What is needed are simultaneous surface and TOA measurements accurate to a few watts per square meter. The suite of cloud-profiling instruments on TRMM (heavily precipitating convective systems) and in future on CloudSat (cloud condensate) and Picasso (thin cirrus), together with EOS PM/AM radiometer measurements and intensive observation periods in attendant field experiments are perhaps the only way the needed measurements can be obtained. Finally, the rather simple 1D upper-ocean model was found to be a sensitive discriminator of cloud system properties because deep convection (including freshwater fluxes and convective enhancement of surface fluxes), the cloud condensate field, and the treatment of cloud distribution in the cloud radiation interaction play important roles in realizing an accurate surface forcing. This supports the suggestion that an accurate representation of atmosphere ocean coupling in coupled GCMs may require accurate treatment of cloud-scale and mesoscale processes and parameterization of subgrid-scale air sea interaction (Webster 1994; Webster and Lukas 1992). Dynamically consistent surface radiative, heat, and momentum fluxes derived from the CRM provide a unique synthetic forcing to examine the response of upper-ocean structure to cloud-scale processes and improve the parameterization of subgrid-scale air sea interaction in coupled GCMs. Acknowledgments. The authors benefited from discussions with many individuals, in particular, Bill Collins, Jeff Kiehl, Xin-Zhong Liang, and V. Ramanathan. We would like to thank Jim Hack and John Pedretti for providing the SCM of CCM3 and Sherrie Fredrick for installing the SCM on the workstation. We are very grateful to Ralph Milliff and Bill Large for providing the 1D ocean model and for their generous help. Thanks also go to Bill Collins for providing GMS1 and GMS2 data; Yuan-Chong Zhang for processed ship data; and Ming-Dah Chou and Shu-Haien Chou for their surface heat and radiative fluxes. Comments on the manuscript by Steve Krueger, Wojciech Grabowski, Ralph Milliff, C.-H. Sui, and Joe Tribbia are gratefully acknowledged. This work is part of the NCAR Clouds in Climate Program. REFERENCES Anderson, S. P., R. A. Weller, and R. Lukas, 1996: Surface buoyancy forcing and the mixed layer of the western Pacific warm pool: Observations and 1-D model results. J. Climate, 9, 3056 3085. Antonissen, E., 2000: On the evolution of the western equatorial Pacific warm pool during the TOGA COARE IOP. Ph.D. dissertation, College of Oceanic and Atmospheric Sciences, Oregon State University, Corvallis, OR, 165 pp. Chou, M.-D., W. Zhao, and S.-H. Chou, 1998: Radiation budgets and cloud radiative forcing in the Pacific warm pool during TOGA COARE. J. Geophys. Res., 103, 16 967 16 977. Chou, S.-H., C.-L. Shie, R. M. Atlas, and J. Ardizzone, 1997: Air sea fluxes retrieved from special sensor microwave imager data. J. Geophys. Res., 102, 12 705 12 726. Clark, T. L., W. D. Hall, and J. L. Coen, 1996: Source code documentation for the Clark Hall cloud-scale model: Code version G3CH01. NCAR Tech. Note, NCAR/TN-426 STR, 137 pp. [Available from NCAR Information Service, P.O. Box 3000, Boulder, CO 80307.] Collins, W. D., A. Bucholtz, and F. P. J. Valero, 1997: Derivation of top-of-atmosphere fluxes from geostationary satellites using high-altitude aircraft measurements: Results from COARE and CEPEX. Preprints, Ninth Conf. on Atmospheric Radiation, Long Beach, CA, Amer. Meteor. Soc., 198 202. Cronin, M. F., and M. J. McPhaden, 1997: The upper ocean heat balance in the western equatorial Pacific during September December 1992. J. Geophys. Res., 102, 8533 8554. Doelling, D. R., D. F. Young, R. F. Arduini, P. Minnis, E. F. Harrison, and J. T. Suttles, 1990: On the role of satellite-measured narrowband radiances for computing the Earth s radiation balance. Preprints Seventh Conf. on Atmospheric Radiation, San Francisco, CA, Amer. Meteor. Soc., 155 160. Fairall, C. W., E. F. Bradley, D. P. Rogers, J. B. Edson, and G. S. Young, 1996: Bulk parameterization of air sea fluxes for Tropical Ocean Global Atmosphere Coupled-Ocean Atmosphere Response Experiment. J. Geophys. Res., 101, 3747 3764. Feng, M., P. Hacker, and R. Lukas, 1998: Upper ocean heat and salt balances in response to a westerly wind burst in the western equatorial Pacific during TOGA COARE. J. Geophys. Res., 103, 10 289 10 311. Grabowski, W. W., X. Wu, and M. W. Moncrieff, 1996: Cloud resolving modeling of tropical cloud systems during Phase III of GATE. Part I: Two-dimensional experiments. J. Atmos. Sci., 53, 3684 3709.,,, and W. D. Hall, 1998: Cloud resolving modeling of tropical cloud systems during Phase III of GATE. Part II: Effects of resolution and the third spatial dimension. J. Atmos. Sci., 55, 3264 3282. Hack, J. J., 1994: Parameterization of moist convection in the National Center for Atmospheric Research Community Climate Model. J. Geophys. Res., 99, 5551 5568., and J. A. Pedretti, 2000: Assessment of solution uncertainties in single-column modeling frameworks. J. Climate, 13, 352 365. Holtslag, A. A. M., C.-H. Moeng, 1991: Eddy diffusivity and countergradient transport in the convective atmospheric boundary layer. J. Atmos. Sci., 48, 1690 1698. Hong, S.-Y., and H.-L. Pan, 1996: Nonlocal boundary layer vertical diffusion in a medium-range forecast model. Mon. Wea. Rev., 124, 2322 2339. Kessler, E., 1969: On the distribution and continuity of water substance in atmospheric circulations. Meteor. Monogr., No. 32, Amer. Meteor. Soc., 84 pp. Kiehl, J. T., 1998: Simulation of the tropical Pacific warm pool with the NCAR climate system model. J. Climate, 11, 1342 1355., J. J. Hack, and B. P. Briegleb, 1994: The simulated Earth radiation budget of the National Center for Atmospheric Research Community Climate Model CCM2 and comparisons with

1168 JOURNAL OF THE ATMOSPHERIC SCIENCES VOLUME 58 the Earth Radiation Budget Experiment (ERBE). J. Geophys. Res., 99, 20 815 20 827.,, G. Bonan, B. A. Boville, D. Williamson, and P. J. Rasch, 1998: The National Center for Atmospheric Research Community Climate Model: CCM3. J. Climate, 11, 1131 1149. Koenig, L. R., and F. W. Murray, 1976: Ice-bearing cumulus clouds evolution: Numerical simulation and general comparison against observations. J. Appl. Meteor., 15, 747 762. Large, W. G., J. C. McWilliams, and S. C. Doney, 1994: Oceanic vertical mixing: A review and a model with a nonlocal boundary layer parameterization. Rev. Geophys., 32, 363 403. Lin, X., and R. H. Johnson, 1996: Kinematic and thermodynamic characteristics of the flow over the western pacific warm pool during TOGA COARE. J. Atmos. Sci., 53, 695 715. Ma, C.-C., C. R. Mechoso, A. Arakawa, and J. D. Farrara, 1994: Sensitivity of a coupled ocean atmosphere model to physical parameterizations. J. Climate, 7, 1883 1896. Minnis, P., D. F. Young, and E. F. Harrison, 1991: Examination of the relationship between outgoing infrared window and total longwave fluxes using satellite data. J. Climate, 4, 1114 1133. Moncrieff, M. W., 1981: A theory of organized steady convection and its transport properties. uart. J. Roy. Meteor. Soc., 107, 29 50., 1995: Mesoscale convection from a large-scale perspective. Atmos. Res., 35, 87 112. Randall, D. A., Harshvardhan, D. A. Dazlich, and T. G. Corsetti, 1989: Interactions among radiation, convection, and large-scale dynamics in a general circulation model. J. Atmos. Sci., 46, 1943 1970. Slingo, A., and J. M. Slingo, 1991: The response of a general circulation model to cloud longwave radiative forcing. Part II: Further studies. uart. J. Roy. Meteor. Soc., 117, 333 364. Slingo, J. M., 1987: The development and verification of a cloud prediction scheme for the ECMWF model. uart. J. Roy. Meteor. Soc., 113, 899 927. Smagorinsky, J., 1963: General circulation experiments with the primitive equations. I. The basic experiment. Mon. Wea. Rev., 91, 99 164. Stephens, G. L., and S.-C. Tsay, 1990: On the cloud absorption anomaly. uart. J. Roy. Meteor. Soc., 116, 671 704. Stockdale, T., M. Latif, G. Burgers, and J.-O. Wolff, 1994: Some sensitivities of a coupled ocean-atmosphere GCM. Tellus, 46A, 367 380. Sui, C.-H., X. Li, K.-M. Lau, and D. Adamec, 1997: Multi-scale air sea interactions during TOGA COARE. Mon. Wea. Rev., 125, 448 462. Tian, L., and J. A. Curry, 1989: Cloud overlap statistics. J. Geophys. Res., 94, 9925 9935. Troen, I., and L. Mahrt, 1986: A simple model of the atmospheric boundary layer: Sensitivity to surface evaporation. Bound.-Layer Meteor., 37, 129 148. Wang, J., and W. B. Rossow, 1998: Effects of cloud vertical structure on atmospheric circulation in the GISS GCM. J. Climate, 11, 3010 3029. Webster, P. J., 1994: The role of hydrological processes in oceanatmosphere interactions. Rev. Geophys., 32, 427 476., and G. L. Stephens, 1984: Cloud-radiation interaction and the climate problem. The Global Climate, J. T. Houghton, Ed., Cambridge University Press, 63 78., and R. Lukas, 1992: TOGA COARE: The Coupled Ocean Atmosphere Response Experiment. Bull. Amer. Meteor. Sci., 73, 1377 1416. Weller, R. A., and S. P. Anderson, 1996: Surface meteorology and air sea fluxes in the western equatorial Pacific warm pool during the TOGA Coupled Ocean Atmosphere Response Experiment. J. Climate, 9, 1959 1990. Wu, X., and M. W. Moncrieff, 1996: Recent progress on cloud-resolving modeling of TOGA COARE and GATE cloud systems. Workshop Proc. New Insights and Approaches to Convective Parameterization, Reading, United Kingdom, ECMWF, 128 156., and M. A. LeMone, 1999: Fine structure of cloud patterns within the intraseasonal oscillation during TOGA COARE. Mon. Wea. Rev., 127, 2503 2513., W. W. Grabowski, and M. W. Moncrieff, 1998: Long-term behavior of cloud systems in TOGA COARE and their interactions with radiative and surface processes. Part I: Two-dimensional Modeling Study. J. Atmos. Sci., 55, 2693 2714., W. D. Hall, W. W. Grabowski, M. W. Moncrieff, W. D. Collins, and J. T. Kiehl, 1999: Long-term behavior of cloud systems in TOGA COARE and their interactions with radiative and surface processes. Part II: Effects of ice microphysics on cloud radiation interaction. J. Atmos. Sci., 56, 3177 3195. Yu, J.-Y., and C. R. Mechoso, 1999: A discussion on the errors in the surface heat fluxes simulated by a coupled GCM. J. Climate, 12, 416 426. Zhang, G. J., and N. A. McFarlane, 1995: Sensitivity of climate simulations to the parameterization of cumulus convection in the Canadian Climate Centre general circulation model. Atmos. Ocean, 33, 407 446. Zhang, Y.-C., W. B. Rossow, and A. A. Lacis, 1995: Calculation of surface and top of atmosphere radiative fluxes from physical quantities based on ISCCP datasets. 1: Method and sensitivity to input data uncertainties. J. Geophys. Res., 100, 1149 1165.