Evaluating ice formation processes in tropical anvils using CRM-based simulations of MODIS and POLDER radiance and polarization measurements

Evaluating cloud resolving model (CRM) simulations of deep convection cloud top particle properties using satellite retrievals is advantageous owing to the extensive areal data coverage as compared with the limited spatiotemporal coverage and instrumental limitations of many aircraft in situ measurements. However, in the presence of ice crystals, uncertainties in the sensitivities of retrievals to assumptions regarding crystal habits, particle size distribution shape, and the vertical variability of such properties, often limits confident interpretation of comparisons to CRM outputs. For instance, it is generally not known how many optical depths below cloud top that retrieved effective radius represents, and we find that differences in ice effective radii in our tropical cumulonimbus simulations averaged over the first 2 or 7 optical depths commonly exceed 5 microns and can exceed 25 microns. Using MODIS and POLDER satellite measurements, here we present an evaluation of CRM simulations with a size-resolved microphysics scheme that aims to circumvent some of these problems by calculating multi-spectral, multi-directional total and polarized reflectance from the CRM results and comparing directly to the corresponding level-1 satellite measurements in a statistical manner. The forward model to simulate the measurements is consistent with the outcome of the CRM and the assumptions of the CRM microphysics, such as ice volume, projected area and aspect ratios as a function of ice type and size. Using a case study during the 2006 TWP-ICE campaign, we demonstrate an evaluation of the temperature level at which ice forms in the CRM and the modeled ice crystal sizes at cloud top. Furthermore, we discuss uncertainties related to the details of ice radiative properties that do not appear in the CRM microphysics, such as the assumed microscopic roughness of ice crystals.