Characteristics of Cloud Size of Deep Convection Simulated by a Global Cloud Resolving Model over the Western Tropical Pacific

Cloud horizontal size distributions of cloud clusters were analyzed for global cloud resolving simulations with the global nonhydrostatic model NICAM whose mesh interval is about 3.5 km and 7 km. The 3.5 km-mesh simulation was performed for 7 days starting at 00 UTC 25 Dec 2006 by giving an initial condition of reanalysis data, while the 7 km-mesh simulation was run for 32 days from 00 UTC 15 Dec 2006. We used outgoing long-wave radiation (OLR) simulated by NICAM to calculate size distributions of deep convection, and compared them with those analyzed using equivalent blackbody temperature (T-BB) of the infrared channel of 11 mu m of the Japanese geostationary meteorological satellite (MTSAT-1R). We selected two threshold temperatures, 208 K and 253 K, to identify deep convective areas including anvil cloud. Specifically, we call clouds defined by the 208 K-threshold deeper convective clouds. Over the tropical region covering the maritime continent and the western tropical Pacific (10S-10N, 90E-160W), we examined the size of cloud areas defined by the two TB, threshold values and corresponding threshold values of OLR of 90 W m(-2) and 210 W m(-2), which were chosen by comparing cumulative histograms of T-BB and OLR in this region. Resolution dependency by NICAM shows that the overall cloud size distribution of the 3.5 km-mesh simulation is much closer to that of the MTSAT-1R observation than that of the 7 km-mesh simulation. Size distributions of deep convection in both simulations indicate nearly lognormal as is seen in the MTSAT-1R observations. The 3.5 km-mesh simulation shows slightly less frequency than the MTSAT-1R observation for smaller size of deeper convection, and it does not reproduce very large clouds. When comparing cloud characteristics over land and ocean, simulated cloud size statistics are slightly closer to the MT-SAT-1R observation in the maritime continent region (westward of 160E) than in the open ocean region (eastward of 160E). A comparison of temporal variation of cloud area shows that the 3.5 km-mesh simulation captures clear signals of diurnal cycles over the maritime continent, together with amplification associated with the Madden-Julian Oscillation (MJO) event. Morning and afternoon difference of convective activity over a large island within the maritime continent is also simulated by 3.5 km-mesh simulation. When one uses a global cloud resolving model for climate studies, the analysis of cloud size distributions gives another dimension to improve the cloud properties of simulations. It is not only relevant to realistic representations of deep convection, but is also useful for improving the radiation budget of global cloud resolving simulations.