Comparison of high‐level clouds represented in a global cloud system–resolving model with CALIPSO/CloudSat and geostationary satellite observations

Vertical and horizontal distributions of high‐level clouds (ice and snow) simulated in high‐resolution global cloud system–resolving simulations by the Nonhydrostatic Icosahedral Atmospheric Model (NICAM) are compared with satellite observations. Ice and snow data in a 1 week experiment by the NICAM 3.5 km grid mesh global simulation initiated at 0000 UTC 25 December 2006 are used in this study. The vertical structure of ice and snow represented by NICAM was compared with Cloud‐Aerosol Lidar and Infrared Pathfinder Satellite Observation (CALIPSO) and CloudSat observations. High‐level clouds (cumulonimbus and cirrus type clouds) classified by the split window (11 and 12 mm) data on board geostationary meteorological satellites (GMSs) were used for comparison of the horizontal distributions of ice and snow in NICAM. The vertical distributions of ice and snow simulated by NICAM qualitatively agree well with those of cloud signals observed by CALIPSO and CloudSat. We computed corresponding cloud lidar backscatter coefficients and cloud radar reflectivity signals from ice and snow data of NICAM using Cloud Feedback Model Intercomparison Project (CFMIP) observational simulator packages. The contoured frequency by altitude diagram for the cloud lidar backscatter coefficients shows lower frequency at higher altitude of 8–14 km by NICAM than CALIOP observations. This suggests that the amount of ice is not well represented in NICAM. The simulated cloud radar reflectivity signals by NICAM indicated higher frequency at 8–10 km altitude than CloudSat observations, although there were some differences between over oceans and continents. This implies that the amount of snow is larger in NICAM simulations. The horizontal pattern of ice clouds (column‐integrated ice and snow of greater than 0.01 kg/m2) in NICAM shows good agreement with that of high‐level clouds identified by the split window analysis. During this 1 week simulation, 48–59% of ice clouds in NICAM matches with observed high‐level clouds. The cross correlation between the spatial distributions of simulated ice clouds and satellite‐observed high‐level clouds is 0.40–0.51, and the equitable threat score is 0.31–0.45. Furthermore, temporal variations of column‐integrated ice clouds in NICAM are compared with high‐level clouds classified by the split window at the decaying stage of deep convection over the tropics. The results indicate that the mean decaying speed of ice clouds of NICAM and high‐level clouds by satellite observations agrees well for this analysis area and period, although the variances are larger in NICAM. This implies that the fall speed of snow in this NICAM experiment is appropriate to depict the decay of anvil clouds by compensating for the excess of snow in NICAM simulations, when we assume that the decay of anvil clouds is largely controlled by the evaporation of ice and snow.