Abstract We present a method for retrieving cloud optical depth that applies 3D radiative transfer to utilize the combination of polarimetry and multi‐spectral imagery that is newly available from satellite missions such as the Plankton, Aerosol, Cloud, Ocean, Ecosystem (PACE) Mission. Due to the approximate spectral invariance of scattering by clouds, this combination of measurements is sensitive to the mean number of scattering events experienced by visible radiation. Using a hierarchy of synthetic cloud fields ranging from idealized cloud geometries, stochastically generated cloud fields with idealized microphysics, and those produced by Large Eddy Simulations (LES), we demonstrate that the combination of visible reflectance and the mean number of scattering events skillfully predicts the in‐cloud mean of the optical depth at both 8 and 1 km resolution, with coefficients of determination exceeding 0.92 and 0.86, respectively. Further out‐of‐sample testing on LES cloud fields show that a multi‐linear regression trained on stochastically generated cloud fields reduces the relative root‐mean‐square error from 29% under the plane‐parallel homogeneous assumption to less than 14% at 6 km resolution. Biases in 1 km resolution retrievals of trade cumulus are reduced from −74% to −40%. Uncertainties from instrumentation and atmospheric correction add up to 20% additional uncertainty in cloud optical depth for the LES cloud fields. With this method, the PACE mission can provide the first global estimate of cloud optical depth that accounts for cloud heterogeneity.
Loveridge et al. (Sat,) studied this question.