Direct Comparisons of Ice Cloud Macro- and Microphysical Properties Simulated by the Community Atmosphere Model CAM5 with HIPPO Aircraft Observations

Abstract:

A dominant source of uncertainty within climate system modeling lies in the representation of cloud processes. This is not only because of the great complexity in cloud microphysics, but also because of the large variations of cloud amount and macroscopic properties in time and space. In this study, the cloud properties simulated by the Community Atmosphere Model version 5.4 (CAM5.4) are evaluated using the HIAPER Pole-to-Pole Observations (HIPPO, 2009-2011). CAM5.4 is driven by the meteorology (U, V, and T) from GEOS5 analysis, while water vapor, hydrometeors and aerosols are calculated by the model itself. For direct comparison of CAM5.4 and HIPPO observations, model output is collocated with HIPPO flights.

Generally, the model has an ability to capture specific cloud systems of meso- to large-scales. In total, the model can reproduce 80% of observed cloud occurrences inside model grid boxes, and even higher (93%) for ice clouds (T≤-40°C). However, the model produces plenty of clouds that are not presented in the observation. The model also simulates significantly larger cloud fraction including for ice clouds compared to the observation. Further analysis shows that the overestimation is a result of bias in relative humidity (RH) in the model. The bias of RH can be mostly attributed to the discrepancies of water vapor, and to a lesser extent to those of temperature. Down to the micro-scale level of ice clouds, the model can simulate reasonably well the magnitude of ice and snow number concentration (N_i, with diameter larger than 75 μm). However, the model simulates fewer occurrences of N_i>50 L^-1. This can be partially ascribed to the low bias of aerosol number concentration (N_aer, with diameter between 0.1-1 μm) simulated by the model. Moreover, the model significantly underestimates both the number mean diameter (D_i,n) and the volume mean diameter (D_i,v) of ice/snow. The result shows that the underestimation may be related to a weaker positive relationship between D_i,n and N_aer and/or the underestimation of N_aer. Finally, it is suggested that better representation of sub-grid variability of meteorology (e.g., water vapor) is needed to improve the formation and evolution of ice clouds in the model.