Simulating clouds is an intricate challenge for global climate models (GCMs). In pursuit of improving the fidelity of future climate prediction, a better understanding of cloud radiative effects and the environmental conditions for cloud formation is needed. To estimate cloud radiative effects, models must accurately represent physical processes occurring during cloud formation and evolution, in order to capture cloud microphysical properties (e.g., mass and number concentrations of cloud hydrometeors) and macrophysical properties (e.g., vertical and horizontal extent, cloud fraction) (Liou, 1992;Liou & Wittman, 1979). Therefore, systematic identification of these physical processes is one Abstract A comparative analysis between observational data from McMurdo Station, Antarctica and the Community Atmosphere Model version 6 (CAM6) simulation is performed focusing on cloud characteristics and their thermodynamic conditions. Ka-band Zenith Radar (KAZR) and High Spectral Resolution Lidar (HSRL) retrievals are used as the basis of cloud fraction and cloud phase identifications. Radiosondes released at 12-h increments provide atmospheric profiles for evaluating the simulated thermodynamic conditions. Our findings show that the CAM6 simulation consistently overestimates (underestimates) cloud fraction above (below) 3 km in four seasons of a year. Normalized by total in-cloud samples, ice and mixed phase occurrence frequencies are underestimated and liquid phase frequency is overestimated by the model at cloud fractions above 0.6, while at cloud fractions below 0.6 ice phase frequency is overestimated and liquid-containing phase frequency is underestimated by the model. The cloud fraction biases are closely associated with concurrent biases in relative humidity (RH), that is, high (low) RH biases above (below) 2 km. Frequencies of correctly simulating ice and liquid-containing phase increase when the absolute biases of RH decrease. Cloud fraction biases also show a positive correlation with RH biases. Water vapor mixing ratio biases are the primary contributor to RH biases, and hence, likely a key factor controlling the cloud biases. This diagnosis of the evident shortfalls of representations of cloud characteristics in CAM6 simulation at McMurdo Station brings new insight in improving the governing model physics therein.Plain Language Summary Global climate models (GCMs) historically struggle to accurately estimate the amounts and types of clouds over the polar regions. Cloud cover and thermodynamic phase directly influence Earth's radiation budget and the accuracy of future climate prediction. Particularly, Antarctic ice sheet is vulnerable to a changing climate through interactions with atmosphere and ocean, and the impacts of clouds are still not well understood. In this study, shortcomings of cloud representations in the CAM6 model were diagnosed by comparing with observational data (ground-based remote sensing and radiosondes), which encompassed a year of measurements at McMurdo Station, Antarctica. Cloud fraction...