Publication acceptée, avril 2013

Level1 measurements, including cross-polarized backscatter, from the Cloud-Aerosol Lidar with Orthogonal Polarization (CALIOP) lidar have been used to document the vertical structure of the cloud thermodynamic phase at global scale. We built a cloud phase identification (liquid, ice, undefined) in the GCM-Oriented Cloud-Aerosol Lidar and Infrared Pathfinder Satellite Observation (CALIPSO) Cloud Product (GOCCP) and analysed the spatial distribution of liquid and ice clouds in 5 JFM seasons of global scale observations (2007-2011). We developed a cloud phase diagnosis in the Cloud Feedback Model Intercomparison Program Observation Simulator Package (COSP) to evaluate the cloud phase description in the LMDZ5B climate model. The diagnosis in the simulator is fully consistent with the CALIPSO-GOCCP observations to ensure differences between the observations and the “model + simulator” ensemble outputs can be attributed to model biases. We compared the liquid and ice cloud vertical distributions simulated by the model with and without the simulator.to quantify the impact of the simulator. The model does not produce liquid clouds above 3 km, and produces ice instead of liquid at low- and mid-altitudes in polar regions, as well as along the Inter-Tropical Convergence Zone. The model is unable to replicate the observed co-existence of liquid and ice cloud between 0 and -40 °C. Liquid clouds dominate for T > -21 °C in the observations, T > -12 °C in the model + simulator, and T > -7.5 °C in the model parameterization. Even if the simulator shifts the model cloud phase parameterization to colder temperature because of the lidar instrument peculiarities, the cloud phase transition remains too warm compared to observations.