Assessment of the Cloud, Convection, Radiation and Energy budget for 20th Century IPCC AR 4th and 5th Simulations and their Satellite Sensor Simulator Outputs

Present-day shortcomings in the representation of cloud processes in general circulation models (GCMs) lead to errors in weather forecast and present climate simulations as well as account for a source of uncertainty in climate change projections. To improve these models, continued evaluation of the model cloud processes is needed against global satellite measurements. Remote sensing measurements of clouds and radiation are sensitive to multiple particle types such as cloud droplets and precipitation, whereas most GCMs, including all IPCC AR 4th and most AR 5th models, do not consider large precipitating particles. Thus, models not accounting for this are getting the top of atmosphere (TOA) balance incorrectly by tuning models' TOA radiative fluxes toward observations with possible compensating errors in radiative fluxes in their vertical distribution and at the surface as well as in their global energy budget estimation. To assess the fidelity of GCMs within CMIP3 and CMIP5 in simulating clouds, radiation, and the global energy budget, retrievals from several satellites are used. Cloud estimates are obtained by excluding precipitating cloud hydrometeors based on cloud classification, a precipitation surface flag, and profile information, a method determined by the estimated particle size distribution (PSD) from the CloudSat radar and CALIPSO lidar combined product. Surface sensible and latent heat fluxes from ECMWF and MERRA reanalyses, as well as CERES and SRB radiation data, are used for model radiation and global energy budget evaluations. Model evaluation also includes complimentary information from the CFMIP cloud and radiation satellite simulator output available in the IPCC 5th assessment. The magnitude and spatial distribution of CMIP3 and CMIP5 LWP somewhat resembles CloudSat total LWP without any filtering conducted for removing precipitation. Only a few zonal and annual mean vertical profiles from CMIP5 models are in pretty good agreement with CloudSat IWC estimates, while significant discrepancies are found in most of the models. Substantial differences are also found in the magnitudes of the radiation fields, cloud occurring frequency, cloud fraction and in their CFMIP output among the CMIP5 models, as well as in the spatial distributions and correlations against estimates from satellite observations. The largest bias in the global surface energy budget is identified with excessive net downward surface fluxes while the global energy budget is well in-balance at the TOA. In general, the comparison between CMIP3 and CMIP5 model fidelity using these measures shows no substantial improvement between the two successive CMIP archives in cloud hydrometeors, radiation simulations, and the global energy budget.