Irreducible 3D Radiative Transfer Effects in Multi-angle/Multi-spectral Radio-Polarimetric Signals (Not Noise!) from a Mixture of Clouds and Aerosol in a Single Large-Footprint Pixel

Although the Glory satellite mission failed at launch, the atmospheric observation strategy implemented in its Aerosol Polarization Sensor (APS) is alive and well since it is at least possible that another one will be built and launched. This strategy is based on APS's along-track scanning spectro-polarimetric measurement system that captures the three main Stokes vector elements (I,Q,U) at a large number (>200) viewing directions for 9 wavelengths emanating from a single pixel that is ~7 km in diameter at nadir and stretches into a ~7 x 20 km^2 ellipse at the most oblique views to be considered (~70 degrees). Two cloud cameras (CCs) were also onboard Glory to provide spatial context. If the relatively large APS footprint is cloud-free or fully-cloudy, then a 1D vector radiative transfer (RT) model is adequate for predicting the APS signals and, upon iteration over its input parameters, aerosol and cloud property retrievals are expected to be of high quality. And this level of accuracy is indeed required to make a real breakthrough in climate modeling where the radiative properties of aerosols and clouds remain one of the main sources of uncertainty. However, the CCs will often show that the APS's field-of-view is a spatially complex cloud scene, but where we are mostly interested in the ambient aerosols. Moreover, it is precisely these aerosols in contact with clouds that will influence their microphysical and optical properties, leading to the manifold indirect aerosol effects on the climate system that need to be far better understood in order to improve their representation in climate models. Therefore, the research presented here addresses the challenge of characterizing simultaneously aerosols and clouds in a single APS observation. Access to polarization can, at least in principle, be used to separate clouds and aerosols using the cloud-bow directions that will often be sampled by APS. In practice, however, we need to assess the extent of 3D polarized RT unfolding inside the APS pixel that cannot be estimated using a linear mixture of 1D vector RT (vRT) computations assuming either aerosol or cloud is present. Differences between the 1D vRT-based prediction and simulated APS data derived from a high-fidelity 3D vRT model is what we call "irreducible" 3D RT effects. To this end, we have used the vMYSTIC Monte Carlo 3D vRT model. Based on computations for a typical scene with a 3D cumulus cloud field embedded in a horizontally uniform aerosol, we find that the irreducible 3D vRT effects are in the APS's signal--not its noise--especially if the aerosol burden is significant. The cloud-bow region, which is key to any practical cloud-aerosol unmixing algorithm, is particularly vulnerable. Moreover, the adopted 1D vRT-based forward model is assumed to be very well informed about the actual aerosol/cloud properties, meaning that the predicted irreducible 3D vRT effects are a best-case scenario. In reality, the problem will be far more severe. We will nonetheless describe a promising path toward a mitigation scheme. We will also assess the impact of the 3D vRT damage on the joint aerosol-cloud property retrieval.