Abstract: Satellite cloud remote sensing provides us the opportunity to study the spatial and temporal distributions of marine boundary layer clouds, as well as their connections with environments on a global scale. However, cloud remote sensing is not without difficulties; retrievals require numerous simplifying assumptions, placing limits on our understanding of cloud processes. Passive remote sensing retrievals often assume that clouds are homogeneous slabs, when in reality, these clouds often have complex inhomogeneous vertical and horizontal structures. Enhancing our understanding of how cloud inhomogeneity influences passive cloud remote sensing requires comparison between cloud retrievals and the underlying cloud properties. In observational data-sets this can become problematic, as it is difficult to compare satellite and airborne measurements because they have both different observed spatial scales and sensitivities to cloud properties. To avoid these complications, this work is based on a satellite retrieval simulator – a Large-Eddy Simulation (LES) cloud model coupled to radiative transfer and retrieval algorithms. The LES-satellite simulator can be used to study the source of retrieval biases. It provides the underlying realistic cloud structure as a reference, informing conclusions about its impact on various cloud retrieval methods. In the first step we focus on cloud vertical profile, finding that the selection of appropriate vertical profile assumptions for the retrieval of cloud liquid water path. Confirming previous studies, drizzle and cloud top entrainment of dry air are identified as physical features that bias liquid water path retrievals away from adiabatic and toward homogeneous profile assumptions. The mean bias induced by drizzle-influenced profiles was shown to be on the order of 5–10 grams per meter squared. In contrast, the influence of cloud top entrainment was found to be smaller by about a factor of 2. A theoretical framework is also developed to explain variability in LWP retrievals by introducing modifications to the adiabatic effective radius profile. The second step focuses on horizontal inhomogeneity and exploring a comparison of both the bispectral and polarimetric cloud retrieval techniques. Using the satellite retrieval simulator we are able to verify that at high spatial resolution (50 meters) the bispectral and polarimetric retrievals are indeed highly correlated with one another. The small differences at high spatial resolution can be attributed to different sensitivity limitations of the two retrievals. In contrast, a systematic difference between the two effective radius retrievals emerges at coarser resolution. This bias largely stems from differences related to sensitivity of the two retrievals to unresolved inhomogeneities in effective variance and optical thickness. The influence of coarse angular resolution is found to increase uncertainty in the polarimetric effective radius retrieval, but generally maintains a constant mean value. The third study focuses on 3-D radiative effects influencing both total and polarized reflectances and retrievals. Comparisons between the 1-D and 3-D reflectances are made in order to study horizontal photon transfer and radiative smoothing. We find noticeable differences between the total and polarized reflectance 3-D effects, with radiative smoothing and roughening occurring at different scales as well as viewing geometry dependence. Despite these apparently strong 3-D effects on polarized reflectances, the polarimetric retrieval is robust to the influence of 3-D effects – with only sub-micron biases in the retrieval of effective radius.