Representing cloud overlap with an effective decorrelation length: An assessment using CloudSat and CALIPSO data

This study commenced testing the hypothesis that a vertically constant, effective, decorrelation length *cf can be used to represent overlap of cloud for the purpose of performing radiative transfer calculations in global climate models. It was assumed that total cloud fraction resulting from multiple layers of overlapping fractional cloud can be described as a linear combination of maximum and random overlap, with the weight defined by exp(−Δz/*cf) where Δz is distance between layers. Cloud masks and water contents (CWC) obtained from CloudSat and CALIPSO satellite data, for January 2007, were used to solve for *cf. Benchmark shortwave (SW) and longwave (LW) broadband flux profiles were computed for 500‐km‐long retrieved cross‐sections via the independent column approximation (ICA). Then *cf was found and used, along with corresponding profiles of cloud fraction and CWCs, in a stochastic cloud generator suitable for use in global climate models (GCMs), and ICA fluxes computed for the generated fields. When clouds were homogenized horizontally, zonal mean bias errors for SW cloud radiative effect (CRE) at the top of atmosphere (TOA) were generally <±3 W m−2, with LW counterparts <2 W m−2, similar to changing cloud particle size by ∼15%. For inhomogeneous clouds, SW CRE biases jumped to typically −5 W m−2 partly because of limitations with the generator. When *cf = 2 km (near global median) was used ubiquitously in the generator, was overestimated slightly, mostly by clouds above ∼10 km, and CRE errors grew by just ∼10% to 20%. Exposing too much high cloud to space produced local SW heating rate biases of ∼15%. While optimal effective decorrelation lengths differ for SW and LW radiation, which in turn generally differ from *cf, it appears that use of *cf will suffice for both bands. The impact of using *cf in GCMs remains to be seen.