Knowledge of cloud vertical structure is important for meteorological and climate studies due to the impact of clouds on both the Earth’s radiation budget and atmospheric adiabatic heating. Yet it is among the most difficult quantities to observe. In this study, we develop a long-term (10 years) radiosonde-based cloud profile product over the Southern Great Plains and along with groundbased and space-borne remote sensing products, use it to evaluate cloud layer distributions simulated by the National Centers for Environmental Prediction global forecast system (GFS) model. The primary objective of this study is to identify advantages and limitations associated with different cloud layer detection methods and model simulations. Cloud occurrence frequencies are evaluated on monthly, annual, and seasonal scales. Cloud vertical distributions from all datasets are bimodal with a lower peak located in the boundary layer and an upper peak located in the high troposphere. In general, radiosonde low-level cloud retrievals bear close resemblance to the ground-based remote sensing product in terms of their variability and gross spatial patterns. The ground-based remote sensing approach tends to underestimate high clouds relative to the radiosonde-based estimation and satellite products which tend to underestimate low clouds. As such, caution must be exercised to use any single product. Overall, the GFS model simulates less low-level and more high-level clouds than observations. In terms of total cloud cover, GFS model simulations agree fairly well with the ground-based remote sensing product. A large wet bias is revealed in GFS-simulated relative humidity fields at high levels in the atmosphere.