A Comparison of Wintertime Cloud Microphysical Properties Derived from MODIS and CALIPSO with In-situ Measurements over the Mid-latitude Southern Ocean

The best constructions of the Earth's climate continued to be challenged by a poor representation of the unique clouds over the Southern Ocean (SO). Due to the remote nature of the SO, satellite-derived products are crucial for our current understanding of the climate system in this region, and for the evaluation and improvement of global models. However, the SO features some of the most extreme conditions in the world (e.g. pristine air mass, strong winds, large fraction of sea spray, etc.), making the accuracy of many retrieval products questionable. Moreover, the use of empirical relationships in retrieval algorithms are derived almost entirely on Northern Hemisphere data sets, themselves often limited.

In this study, wintertime in-situ observations of cloud effective radius (r_eff), cloud droplet number concentration (N_d) and cloud thermodynamic phase from eleven flights over the mid-latitude Southern Ocean (SO) (43–45^oS, 145–148^oE) are compared against retrieval products derived from the MODerate-resolution Imaging Spectroradiometer (MODIS) and Cloud-Aerosol Lidar with Orthogonal Polarization (CALIOP) observations. The eleven missions were designed to make observations underneath the A-Train constellation overpasses with the collocated in-situ observations being constrained to a 30-minute window. During these eleven flights, open mesoscale cellular convection (MCC) was observed to be predominant, although closed MCC was observed during two flights. For open MCC, clouds were commonly observed to be precipitating intermittently, to be patchy in spatial extent, and to be intermittently mixed-phase. As such, these clouds were rarely ideal for satellite retrievals.

Compared to the in-situ observations of the cloud thermodynamic phase, the CALIOP cloud-top phase and MODIS cloud phase optical property (CPOP, collection 6) products are found to consistently underestimate the frequency of occurrence of mixed-phase clouds, whereas the MODIS cloud-top phase (infrared based) product shows a better qualitative agreement despite the relative large fraction of uncertain class.

Focusing on the liquid phase clouds only, the MODIS retrieved r_{eff_2.1} (r_eff from the 2.1-μm channel) overestimates the in-situ r_eff for non-drizzling cloud samples (by ~13 μm, on average) and, to a lesser extent, for lightly drizzling samples. Conversely, r_{eff_2.1} underestimates the in-situ observations for heavily drizzling samples by ~10 μm, on average. Overall, the MODIS r_eff shows much weaker variability than the in-situ observations. The overestimation of r_eff for non-drizzling cloud samples is much greater than that reported for stratocumulus clouds over South East Pacific from the VAMOS Ocean-Cloud-Atmosphere-Land Study Regional Experiment (VOCALS-REx). The retrieved cloudy pixel population for the 2.1-μm channel is higher than those with 3.7-μm channel and this is outstanding in most drizzling cases. Overall, r_{eff_3.7} was in a better agreement with the in-situ observations than r_{eff_2.1} for non-drizzling and light drizzling conditions, but r_{eff_2.1} was in a better agreement than r_{eff_3.7} for heavily drizzling conditions. No specific bias was evident for MODIS retrievals made at large solar zenith angle (> 65^o), although the sample size is quite limited. In the broken or patchy cloud field with a mixed presentation of ice, liquid and large drizzling particles, the large inhomogeneity and 3-D effect are expected to be the primary sources of errors for the r_eff retrievals, although the differences between the effective variance prescribed in the precomputed look-up table and the actual range of variability are not negligible. Ice contamination and sub-pixel variability may also be important factors for the retrievals bias.

The computed N_d based on MODIS retrievals, however, is largely consistent with the in-situ observations, regardless of the presence of drizzle. However, the N_d of the outstanding two high N_d cases featuring with closed MCC observed during the flights is highly underestimated. These biases may be associated with the overestimated r_eff combined with the errors in the retrieved cloud optical thickness, as well as uncertainties in the algorithms used for computing N_d.