Scaling properties of observed and simulated satellite visible radiances

Structure functions S q , which are related to power spectra and used to study turbulence, were computed for GOES‐13 visible radiances measured on 16 May 2015 over French Guiana and adjacent Atlantic Ocean. The nested Global Environmental Multiscale (GEM) numerical weather prediction (NWP) model was run for the same time and area. Cloud data generated by GEM over (300 km) 2 domains, with one‐way nesting ending at horizontal grid‐spacing of 0.25 km, were operated on by a 3‐D solar radiative transfer model with resulting radiances degraded to GOES‐13 resolution (~1 km) and S q computed for them, too. For GOES‐13 radiances, scaling exponents ζ (2) associated with S 2 , for separation distances between 5 km and 25 km, were typically >0.6 for deep convective and marine boundary layer clouds and <0.4 for shallow cumuli over land. ζ (2) for GEM agreed well with GOES‐13 for deep convective clouds. This suggests that the self‐organizing properties of deep convection in GEM exhibit realistic geometric features, a potentially important point given the link between cloud structure and precipitation, with the latter being much more difficult to measure and assess than visible radiances. Regarding radiances for GEM's marine boundary layer clouds, their S q differed markedly from GOES‐13's; better resembling fair‐weather cumulus. Likewise, GEM's shallow cumuli over land appear to have bypassed the “scattered” fair‐weather stage and went straight into more organized convection. Thus, it appears that comparing time series of S q for geostationary satellite data and corresponding modeled radiances has the potential to benefit assessment of cloud system‐resolving models.