Kelly, Joe, authorChiu, Christine, advisorMiller, Steven D., committee memberVenkatachalam, Chandra, committee member2023-06-012024-05-262023https://hdl.handle.net/10217/236620Marine stratocumulus clouds are a critical component of Earth's radiation budget and remain a key source of uncertainty in climate projections. Better representing these clouds and their interactions with radiation, precipitation and aerosols in models necessitates observations of three-dimensional (3D) cloud fields. While passive satellite observations provide critical information on cloud properties globally, their retrievals lack information on vertical structure. Most retrieval methods also assume one-dimensional (1D), plane parallel clouds, leading to significant retrieval errors for both stratocumulus and cumulus regimes. In contrast, observations from active sensors allow for the probing of cloud vertical structure. However, active sensor data are limited in coverage. Combining active and passive satellite observations provides an excellent opportunity to reconstruct the 3D cloud fields. To provide 3D cloud property fields that do not suffer from errors introduced by the plane-parallel assumption, 3D radiative effects must be incorporated during the retrieval process. In this thesis, the impact of 3D radiative effects on 1D retrievals of cloud optical and microphysical properties is quantified, focusing on contrasting illuminated and shadowed pixels. When evaluating 1D retrieval on a synthetic cloud field, it is found that shadowed pixels had a larger magnitude of mean optical depth bias (–12) than illuminated pixels (3) at small solar zenith angles, while shadowed pixels had a lower magnitude of mean optical depth bias (–5) than illuminated pixels (12) at large solar zenith angles. For effective radius, the mean biases in shadowed and illuminated pixels are respectively 3.9 μm and –4.9 μm at large solar zenith angles. At small solar zenith angles, shadowed pixels had a smaller mean effective radius bias (0.8 μm) than illuminated pixels (–3.8 μm). By incorporating 3D radiative effects into the retrieval of the synthetic cloud field, the range of retrieved optical depth errors is greatly reduced from [–50, 100] to [–30, 40]. In addition to the synthetic dataset, we highlight a real-world case from the Variability of the American Monsoon System (VAMOS) Ocean-Cloud-Atmosphere-Land Study Regional Experiment (VOCALS-REx), serving as a potential dataset for evaluating 1D and 3D retrievals. Cloud microphysical properties were derived from in-situ cloud probe measurements collected from a profiling flight and three longer and horizontal transects that were within 1 hour of the A-Train overpass. In this particular cloud profile, the cloud droplet number concentrations ranged between 100–150 cm–3 and were relatively constant with height; cloud liquid water content increased approximately linearly with height, following a sub-adiabatic growth rate of 1.4 g m–3 km–1. We have found that properties from three horizontal transects have similar cloud statistics and structures. Applying the retrieval method to real-world data proved challenging due to the limited vertical information available from satellites about clouds near the surface and due to the inherent uncertainties of comparing cloud fields at different times. Lastly, to incorporate 3D radiative effects in the retrieval process, we have developed 3D shortwave radiative transfer emulators for stratocumulus and cumulus cloud fields using a convolutional neural network. The emulators were trained on cloud fields generated from the Large Eddy Simulation (LES) and specific sets of solar and viewing geometry and aerosol conditions. The performance of emulators was evaluated against a testing dataset in which the truth reflectance was computed by a 3D radiative transfer model with a subset of LES output as the input cloud fields. Overall, the predicted reflectance at the top of the atmosphere in the visible and near-infrared spectral regions has mean relative errors smaller than 2%, and the 15th and 85th percentile errors are generally less than ±10% for all setups. This type of emulator can be integrated into remote sensing applications and allow 3D radiative effects to be integrated effectively into advanced retrieval methods.born digitalmasters thesesengCopyright and other restrictions may apply. User is responsible for compliance with all applicable laws. For information about copyright law, please see https://libguides.colostate.edu/copyright.cloud remote sensingmachine learningVOCALSCloudSat3D radiative transfer emulatorMODISText