Accurate ground-based remotely sensed microphysical and optical properties of liquid water clouds are essential references to validate satellite-observed cloud properties and to improve cloud parameterizations in weather and climate models. This requires the evaluation of algorithms for retrieval of cloud microphysical and optical properties using ground-based remote sensing observations, because there are large differences between the cloud property retrievals of various algorithms due to the differences in the applied retrieval theories, assumptions, retrieval inputs and constraints. This thesis focuses on three commonly used vertical cloud models for the parameterization of the in-cloud vertical structure in cloud property retrieval schemes. The objective is to explore the impact of the vertical cloud models on the computations of microphysical and optical properties of liquid water clouds and to evaluate their uncertainties. This information can help to improve current liquid water cloud property retrieval schemes and to increase the accuracy of the obtained cloud physical properties. A comparison of three algorithms with different vertical cloud models for the retrieval of liquid water cloud microphysical and optical properties is performed. In the first algorithm, the vertical structure of the cloud is parameterized as being vertically homogeneous (Vertically Uniform, VU). In the second algorithm, the used vertical cloud model originates from an adiabatic model (Scaled Adiabatic Stratified, SAS) and the third algorithm relies on a vertical model, which considers the impact of cloud top entrainment mixing processes on the cloud microphysical properties (Homogenous-Mixing, HM). All three algorithms use observations of the cloud radar reflectivity, the microwave radiometer obtained liquid water path (LWP) and the cloud geometrical thickness from lidar and cloud radar. They require a priori assumptions on the cloud droplet size distribution (DSD). Hence, the gamma function is used to parametrize the DSDs and possible values for the gamma DSD shape parameter are evaluated from reanalyzed in-situ observations. All three algorithms investigated here retrieve vertical profiles of the liquid water content (LWC), the droplet concentration, the effective radius, the visible optical extinction and the visible optical depth. The differences between the cloud property retrievals of each algorithm are explained on the basis of remote sensing observations that appear to be typical for low-level water clouds. The results of the VU cloud model per se lack detailed information on the vertical distribution of the cloud property retrievals. Under adiabatic conditions, the retrievals of the SAS and the HM models are equivalent, while the vertical distributions of the LWC, the effective radius and the optical extinction differ substantially under non-adiabatic conditions, especially at the cloud boundaries. The droplet concentrations of the SAS and HM models are very close to each other for both conditions. The model of uniform cloud properties yields values of the droplet concentration that are 25% lower than those from the models of non-uniform cloud properties. Interestingly, the differences between the cloud microphysical properties lead to very similar values of the retrieved visible optical depths. Sensitivity and error analyses suggest that the droplet concentration retrieval is generally most strongly affected by errors in the radar reflectivity and the LWP, while the retrievals of the effective radius are most robust in all three models. The retrievals of the optical depth and the effective radius are less affected by the variations in the DSD shape parameter as compared to the impact of the errors in observations. In contrast, the droplet concentration is more sensitive to changes in the gamma DSD shape parameter. Consequently the DSD shape parameter should be known a priori with reasonable precision. In order to evaluate the validity of the cloud property retrievals, the three algorithms are applied to synthetic surface remote sensing observations of a modeled liquid water cloud layer. The retrievals are compared with the physical properties of the modeled cloud layer as a function of the cloud height. Applying the algorithms to the best estimate “observations” and the assumed value for the DSD shape parameter leads to consistent HM model cloud property retrievals. In turn, significant overestimations of the SAS model LWC (50%) and the effective radius (10%) occur at cloud top where the SAS model retrieves the maximum values in the profiles. In all layers below the cloud top, the SAS cloud model retrievals of the LWC and the effective radius are very close to the modeled ones, because the true properties are increasing nearly adiabatically. As expected, the differences in the LWC and the effective radius profiles are largest upon the application of the VU model, which significantly overestimates both properties in the lower levels and underestimates them in the upper height levels. The very simple assumption that all cloud properties are uniformly distributed leads to a significant underestimation of the droplet concentration by about 20%. The SAS model droplet concentration is only slightly overestimated by 7%. Nevertheless, all cloud model retrievals of the optical depth agree well with those of the modeled cloud layer. To evaluate the performance of the cloud property retrievals obtained from real remote sensing observations, a broadband shortwave (SW) radiation closure analysis is performed for a selected water cloud case study. The SW fluxes at the surface calculated on the basis of the cloud properties of VU, SAS and HM models agree well with the surface radiation observations. The mean difference between the simulated and the measured SW fluxes is 2 W/m2 to 5 W/m2 with a standard deviation of 13W/m2. The uncertainty in the simulated fluxes can be explained by the uncertainty in the LWC and the effective radius due to errors in the LWP, the reflectivity and the assumption on the gamma DSD shape parameter. The three presented retrieval methods provide reliable cloud optical depth values for the selected water cloud case study. The different vertical distributions of the LWC and the effective radius, as well as differences in the droplet concentration, have a minor effect on simulating SW fluxes, because they lead to similar values of the optical depth. The present work shows that the liquid water cloud property retrievals obtained from the remote sensing observations depend on the model that is used to describe the vertical cloud structure. It shows that systematic deviations between the microphysical cloud properties of the VU, SAS and HM cloud models exist, especially regarding the droplet concentration, the LWC, the effective radius and the optical extinction at the cloud boundaries. The cloud microphysical properties estimated using the HM model parametrization show the best performance. The SAS cloud model can represent the vertically resolved microphysical properties well if they are very close to being adiabatic. Clearly, there are significant deviations in the cloud microphysics from the clouds that are parameterized as being vertically homogeneous (VU model). The different combinations of the microphysical properties in the three models lead to almost equivalent VU, SAS and HM optical depth retrievals, which agree well with the values of the modeled liquid water cloud. They are all able to reproduce the surface shortwave broadband radiative flux. However, by modeling clouds as being vertically homogeneous, sufficient accuracy in both the microphysical and the optical property retrievals cannot be achieved.