Radiative Closure Studies of How Cloud Property Retrieval Errors due to three-dimensional radiative effects Influence Our Understanding of Broadband Cloud Radiative Effects

Satellite retrievals of cloud properties, such as cloud optical thickness (τ) and cloud effective radius (re)based on the so-called bi­spectral method is widely used in climate studies and evaluation of Earth System Models (ESMs). These retrievals are usually based on the one­-dimensional (1D) radiative transfer (RT) theory that in many circumstances can result in significant biases due to the so-called 3D radiative effect. In this study, we investigate how cloud property retrieval errors due to the 3D radiative effects influence our understanding of broadband cloud radiative effects (CREs). To this end we developed a framework based on the combination of cloud fields from large eddy simulations (LES) and RT models able to simulate both 1D and 3D radiance and broadband fluxes, to understand the biases due to the 3D radiative effects in both radiance fields and cloud property retrievals, and the consequent influence on broadband CRE computations. Using this framework, we discovered that the broadband shortwave (SW) fluxes derived from the biased cloud property retrievals using 1D RT model (referred to as “1D-RT+retrieved clouds”) can provide reasonable broadband radiative energy estimates in comparison with those derived from the true cloud fields with 1D RT simulations (referred to as “1D-RT+true clouds”). The difference in the broadband SW flux between the 1D-RT+retrieved clouds simulations and the benchmark 3D RT simulations from the true cloud field (referred to as “3D-RT+true clouds”), depends primarily on the horizontal transport of photons in 3D RT, whose characteristics vary with the position of the Sun. In the high sun case with solar zenith angle (SZA) at 5°, the reflected fluxes at the top of the domain (TOD), the transmitted flux at surface and the absorbed flux, simulated based on 1D-RT+retrieved clouds are in excellent agreement with the benchmark results based on 3D-RT+true clouds, all within 7%. In the low sun case with SZA at 60°, the differences between the three sets of simulations are determined by how the cloud side illumination and shadowing effects compromise and cancel each other in the radiance, retrieval, and broadband fluxes. It is worth special note that the strong 3D effects in the low sun case can push a large fraction of pixels outside of the bi-spectral look-up-table (LUT), leading to retrieval failures (e.g., retrieved τ or re values larger than the maximum values in the LUT).

We found that it is important to include this population of failed retrievals in the estimate of broadband fluxes using the 1D-RT+retrieved clouds method to preserve their influences on the domain averaged results. Errors in CRE estimations at SZA 60° from the TOD reflected, surface transmitted, and column absorbed flux derived from retrieved cloud properties, could increase by as much as a factor of 6 if the failed retrievals were excluded in the LES scale domain­averaged analysis compared to estimates which include failed retrievals in the analysis. In conclusion, this study suggests that although the cloud property retrievals based on the 1D RT theory may be biased due to the 3D RT effects, they still provide an observational basis for the estimation of broadband fluxes.