Abstract

Cloud phase and relative humidity (RH) distributions at -67 degrees to 0 degrees C over the Southern Ocean during austral summer are compared between in situ airborne observations and global climate simulations. A scale-aware comparison is conducted using horizontally averaged observations from 0... Show moreCloud phase and relative humidity (RH) distributions at -67 degrees to 0 degrees C over the Southern Ocean during austral summer are compared between in situ airborne observations and global climate simulations. A scale-aware comparison is conducted using horizontally averaged observations from 0.1 to 50 km. Cloud phase frequencies, RH distributions, and liquid mass fraction are found to be less affected by horizontal resolutions than liquid and ice water content (LWC and IWC, respectively), liquid and ice number concentrations (Nc(liq) and Nc(ice), respectively), and ice supersaturation (ISS) frequency. At -10 degrees to 0 degrees C, observations show 27%-34% and 17%-37% of liquid and mixed phases, while simulations show 60%-70% and 3%-4%, respectively. Simulations overestimate (underestimate) LWC and Nc(liq) in liquid (mixed) phase, overestimate Nc(ice) in mixed phase, underestimate IWC in ice and mixed phases, and underestimate (overestimate) liquid mass fraction below (above) -5 degrees C, indicating that observational constraints are needed for different cloud phases. RH frequently occurs at liquid saturation in liquid and mixed phases for all datasets, yet the observed RH in ice phase can deviate from liquid saturation by up to 20%-40% at -20 degrees to 0 degrees C, indicating that the model assumption of liquid saturation for coexisting ice and liquid is inaccurate for low liquid mass fractions (<0.1). Simulations lack RH variability for partial cloud fractions (0.1-0.9) and underestimate (overestimate) ISS frequency for cloud fraction <0.1 (>= 0.6), implying that improving RH subgrid-scale parameterizations may be a viable path to account for small-scale processes that affect RH and cloud phase heterogeneities. Two sets of simulations (nudged and free-running) show very similar results (except for ISS frequency) regardless of sample sizes, corroborating the statistical robustness of the model-observation comparisons. Show less