Thermal forcing of the Tibetan Plateau (TP) has large impacts on the Asian summer monsoon. In this study, we statistically analyzed the outputs from the global cloud-resolving model, NICAM (Nonhydrostatic ICosahedral Atmospheric Model). In order to investigate the convective activities, two brightness temperature datasets were compared, one derived from the satellite observation and another derived from the model. The model well simulated the spatio-temporal variations of convective clouds in April 2004. The diurnal cycle of clouds was better represented in NICAM, which shows only few-hour phase difference, than that in the reanalysis data. Three experiments changing horizontal resolution revealed that the higher resolution run conducts better representation of the diurnal cycle, especially on the nighttime disappearance of the high clouds. These results indicate that the global cloud-resolving model will improve the seasonal prediction of the Asian summer monsoon through better description on the thermal forcing of the Tibetan Plateau.
 The Tibetan Plateau (TP) has determinative influences on the earth's climate system as well as on the Asian monsoon climate [Hahn and Manabe, 1975; Kutzbach et al., 1989]. Dynamical role of TP forms the stationary ridge/trough structure of the planetary wave which significantly reduces the storm activities over North Asian arid region [Broccoli and Manabe, 1992]. The thermal forcing of TP has been known to modulate the seasonal evolution of the Asian summer monsoon [Yanai et al., 1992; Wu and Zhang, 1998; Hsu and Liu, 2003; Sato and Kimura, 2007]. Sato and Kimura  clarified the potential roles of TP on the monsoonal precipitation in South Asia using a simplified climate model with artificial TP heating. They found that the monsoon onset in North India is delayed owing to the TP heating during the pre-monsoon season. Taniguchi and Koike  suggested that, during April, the diabatic heating derived from the moist convection rises upper tropospheric temperature over TP. Although recent studies pointed out the importance of the spring thermal forcing, unfortunately, most of previous works have focused on the diabatic heating during the summer season (June–July–August). Thus, further investigations are mandatory for understanding the detail structure and diurnal features of the atmospheric heating over TP during spring season (March–April–May).
 As shown by the previous analytical studies using the reanalysis data [Yanai and Tomita, 1998], TP can be recognized as a weak heat source in the spring season. Recent progress in the satellite remote sensing provides information to examine cloud types using the reflectance from multi-channel spectrometer. Some studies using MODIS (Moderate Resolution Imaging Spectroradiometer aboard the Terra/Aqua) data suggested that the TP has prominently high frequency of high clouds during the spring season [Gao et al., 2003; Chen and Liu, 2005]. Toumi and Qie  showed the high flash rate per precipitation amount over TP during the spring. Fujinami and Yasunari  investigated the seasonal variation of the diurnal cycle of convective activity over TP. They found that, in April and May, convective activity is very pronounced over TP besides the well-known diurnal convective activity during the monsoon period [e.g., Kuwagata et al., 2001]. The diurnal features of the convective systems strongly affect the radiative balance by intercepting shortwave and emitting longwave radiation. Therefore, they should be properly simulated in the general circulation models (GCMs) in order to implement the huge impacts of TP on the climate system.
 A global cloud-resolving model, NICAM (Nonhydrostatic ICosahedral Atmospheric Model), has been developed at the Frontier Research Center for Global Change (FRCGC) [Tomita and Satoh, 2004; Tomita et al., 2005; Satoh et al., 2007]. With a few kilometers mesh interval over the globe, NICAM is intended to resolve mesoscale circulations associated with deep convection by calculating explicit cloud physics instead of cumulus parameterizations. Using the Earth Simulator, Miura et al.  performed the first 3.5 km grid experiment with realistic topography and land/sea distribution in the NICAM. The purpose of this study is to describe the convective activity over TP simulated in the Miura's experiments. We mainly focus on the diurnal cycle of convective activity in the NICAM which is evaluated by the satellite observations.
2. Experiments and Data
 This study mainly uses the data from NICAM experiments as shown by Miura et al. . The NICAM adopts the 40 vertical layers in the terrain following coordinate. Three kinds of experiments with different horizontal grid sizes (3.5, 7, and 14 km) are analyzed, in which the cumulus parameterization is absent in all runs. Experimental periods are 1–7th April, 1–10th April, and 1–31st April of 2004 in the 3.5, 7, and 14 km mesh runs, respectively. Initial conditions of atmospheric variables and soil moisture are given by National Centers for Environmental Prediction (NCEP) Global Tropospheric Analyses on 1.0 × 1.0 degree grids.
 Satellite observations are useful to evaluate the convective activities in the high resolution simulations. We use the infrared channel (10.5–11.5 μm) from the Geostationary Meteorological Satellite (GMS) launched by the Japan Meteorological Agency (JMA) to investigate the diurnal cycle of convective systems. Horizontal resolution of the GMS is 0.05 degree at nadir (140°E, 0°N), and the data is archived in every one hour. Since the satellite zenith angle for the area of Tibetan Plateau ranges from 55 to 80 degree, the resolution has range from 8 to 27 km, approximately. Brightness temperature (Tb) from GMS is analyzed in April during 1998–2002 because the GMS data are not available for April 2004 due to the problems in the satellite.
 In order to compare NICAM results with GMS derived Tb (GMS-Tb), radiative transfer model developed by JMA [Owada, 2006] is used. The model computes the Tb value using the three dimensional distributions of microphysical and surface variables from NICAM, such as cloud liquid water and cloud ice contents. Since the three-dimensional variables are available only in 14 km grid run due to the limit of storage, the comparison of NICAM-Tb and GMS-Tb are shown for 14 km grid run. Hereafter, local time (LT) is used (LT = UTC + 6).
 Hovmöller diagram of NICAM-Tb from 14 km grid run is shown in Figure 1 as averaged over 30–35°N. Clear diurnal cycle of Tb is repeated over TP. This result is consistent to the previous work by Fujinami and Yasunari  in respect that the TP experiences the dominant diurnal cycle of convection in April. In many cases, active convection systems that generated over TP decease over TP. Some developed convective systems travel further east, and move out of TP (cf. arrows in Figure 1). The propagating convective systems tend to appear in association with the synoptic-scale forcing like troughs, in which their characteristics are very similar to that shown by Wang et al. [2004, 2005] although their interests are further east and in summer season. Asai et al.  studied the diurnal variability of cloudiness over East Asia, and indicated that the eastward propagation of the cloud clusters originated over TP are important process governing the diurnal variation of cloudiness over East China. A large portion of the moisture might be supplied from outside of the plateau in the propagating cases as it was speculated for the monsoon season [Tian et al., 2001].
Figure 2 compares the diurnal cycles of brightness temperatures derived from GMS and NICAM. GMS observation (Figures 2a–2d) indicates clear diurnal variation with minimum Tb in 18 LT over the entire area of TP. The lowest Tb appears over the Kunlun Mountains and northern TP indicating the active convective systems around the sunset. The Tb tends to increase after 18 LT; thus, the majority of convective systems weakens or disappears until the midnight. The NICAM-Tb (Figures 2e–2h) behaves a similar pattern to that in GMS showing active convection over northern TP in 18 LT. Additionally, the Tb tends to increase in 00 LT indicating disappearance of the convective systems as that observed by GMS. On the other hand, in the NCEP/NCAR (National Center for Atmospheric Research) reanalysis data [Kalnay et al., 1996], the outgoing longwave radiation (OLR) remains decreasing throughout the night, and having the minimum value during 00–06 LT (Figures 2i–2l). Thus, the diurnal cycle of the cloud characteristics are better represented in the NICAM than those in the reanalysis data despite the reanalysis data is often used to evaluate the diabatic heating rate. However, NICAM overestimates the convective activities over the Himalayas although they do appear in the GMS observation as well. The NICAM tends to show lower Tb than that in GMS even with or without cloud cover over the analyzed domain (70–105°E, 25–45°N).
 The statistical comparison of diurnal cycles is carried out between NICAM-OLR and GMS-Tb. We analyzed the grid pixels corresponding to TP which is defined as the altitude exceeding 3000 m in the area of 70–105°E, 25–45°N. Figure 3 shows the diurnal cycle of the probability density distributions of GMS-Tb and NICAM-OLR. During night, from 18 LT through 06 LT, probability density (PD) of GMS-Tb is very high around 260–270 K indicating that the land surface is exposed due to very low cloudiness in most area of TP. Kurosaki and Kimura  suggested that the cloud frequency in pre-monsoon season (May) is very low over TP in the morning, which agrees to our statistical analysis. The high PD area is gradually shifting to 300 K before 12 LT indicating the surface warming due to the solar radiation. The high-PD track becomes ambiguous after 12 LT, and the high-PD area suddenly reappears around 230 K in 15 LT. The high-PD around 230 K persists until 18 LT, which reconfirms the existence of the high-clouds due to developed convection in most area of TP as indicated in Figure 2. These convective systems disappear immediately after 18 LT as shown by the high-PD around 270 K at 21 LT. The right panel of Figure 3a shows the probability density distributions from GMS observation in daily mean and in 18 LT. Prominent maximum exists around 265 K in the daily mean. On the other hand, two peaks appear in 18 LT around 230 K, corresponding to the high-clouds, and around 270 K, corresponding to the land surface.
 Diurnal cycles of PD in NICAM-OLR are illustrated in Figures 3b–3d from three different resolution runs. In three cases, OLR increases from 240 Wm−2 in the morning until 12 LT as the surface temperature increases, which agree well to the GMS observations. The prominent peaks of PD appear in all cases around low-OLR area (160–180 Wm−2) after 12 LT, especially during 12–18 LT indicating the development of convective systems. In the 14 km grid run, convective systems are hard to decay or high clouds still exist in 18 LT because the PD remains high in the lower OLR zone (Figure 3b). Additionally, the high-PD in the high OLR zone, corresponding to the land surface, appears only later than 00 LT. The failure in simulating nighttime disappearance of the high clouds tends to be improved as the horizontal resolution becomes higher. The high-PD around 240 Wm−2 is more prominent after 21 LT in the 3.5 km grid run, which indicates that the daytime convective systems correctly diminish after sunset. The daily mean PD distributions also support this tendency since the PD is more concentrated around 240 Wm−2 as the resolution becomes higher. Therefore, the higher resolution runs will improve the description of the diurnal cycle of clouds, especially the duration without cloud cover.
4. Discussion and Summary
 This study compares the diurnal cycle of convective systems over TP between that observed by the satellite and that obtained by the global cloud-resolving simulations. Since the TP has huge impacts on the earth's climate systems as an elevated heat source [Yanai et al., 1992; Hsu and Liu, 2003; Sato and Kimura, 2005, 2007], diabatic heating in the atmosphere should be properly simulated in the GCMs. Three-dimensional cloud distribution is crucially important due to its strong diurnal dependency, the elevation and duration of which alter the radiative balance.
 One month simulation by NICAM well reproduces the diurnal cycle of convective activities over TP. The diurnal phase of convective activity is better reproduced in NICAM than that in reanalysis data as diagnosed by OLR. However, more improvements may be necessary in the following parts. The initiation time of the convective systems is earlier in all resolutions of NICAM runs than that observed by GMS, which consequently reduces the downward shortwave radiation absorbed at the TP surface. The land surface condition in the NICAM is simply given by the bucket model, which may induce time lag of the convection development. The MODIS land surface products could be one of the possibilities to improve of the model. At the same time, as revealed by the Tb comparison in Figure 2, the developed convective systems in NICAM tend to exhibit higher cloud top height than that evaluated by GMS observation. Too strong convection over the Himalaya may be attributed to the strong upslope flow owing to the smoothed topography. Further analysis on precipitation along the Himalaya is essential with the aid of satellite observations like TRMM (Tropical Rainfall Measuring Mission).
Figures 2 and 3 show that the pronounced high-cloud frequency, which was revealed by the recent remote sensing observations, is strongly linked with the diurnal cycle of convective systems rather than the time-independent cirrus clouds. Cloud detection by MODIS sensor is very beneficial; however, they may be strongly dependent on the sounding time due to the orbital reasons. In order to diagnose the climatological state of the clouds, the high-resolution cloud-resolving model is useful to know the diurnal variations. The NICAM is expected to improve the representation of the diurnal cycles of convection and precipitation since it does not use cumulus parameterization which sometimes causes the time lag in precipitation diurnal cycle in the climate models [Collier and Bowman, 2004; Neale and Slingo, 2003; Dai et al., 1999]. Additionally, once the forward models, which can compute the observable variables like Tb, are implemented in the high resolution GCM, the simulated variables are able to compare directly with the satellite observations. This process should be useful to enhance the model performance and, in turn, to improve the retrieval algorithms of the satellite observations.
 Resolution dependency on the diurnal cycle of cloud activities is investigated using NICAM runs with three different horizontal resolutions. In the 14 km grid run, daytime convective systems are likely to remain until midnight. The 3.5 km grid run shows better description of the clouds, in which the high-clouds are correctly deceased after sunset, and cloud free grids are properly simulated. Similar resolution dependencies were reported in some of the cloud-resolving models. The coarser resolution induces the delay of convection initiation and mature state [Petch et al., 2002; Weisman et al., 1997]. In the 3.5 km grid run, disappearance time of the high clouds is still three hours later than that observed by GMS albeit even better than the reanalysis data; therefore, further improvements or higher resolution are needed. In any case, our results indicate that the global cloud-resolving model is expected to improve the thermal forcing of TP which consequently leads to enhance the seasonal prediction of the Asian summer monsoon.
 We acknowledge the NICAM group members in FRCGC (H. Tomita, S. Iga, T. Nasuno, A. T. Noda, K. Oouchi) and CCSR (W. Yanase, T. Mitsui, Y. Niwa, K. Suzuki) for developing the model. Discussions with H. Fujinami T. Yoshikane and M. Hara greatly helped to summarize the results. Radiative transfer code to evaluate NICAM-Tb was provided by H. Owada in JMA. One of us (T.S.) is supported by Japan Society for the Promotion of Science (JSPS) as a research fellow. GMS data was obtained by Kochi University. This study was financially supported by CREST (Core Research for Evolutional Science and Technology) program of JST (Japan Science and Technology Agency). The NICAM simulations were carried out on the Earth Simulator at the Earth Simulator Center of the Japan Agency for Marine-Earth Science and Technology.