Large eddy simulations (LESs) of continental boundary layer clouds (BLCs) observed at the southern Great Plains (SGP) are used to study issues associated with the parameterization of sub-grid BLCs in large scale models. It is found that liquid water potential temperature _l and total specific humidity q _t , which are often used as parameterization predictors in statistical cloud schemes, do not share the same probability distribution in the cloud layer with _l skewed to the left (negatively skewed) and q _t skewed to the right (positively skewed). The skewness and kurtosis change substantially in time and space when the development of continental BLCs undergoes a distinct diurnal variation. The wide range of skewness and kurtosis of _l and q _t can hardly be described by a single probability distribution function. To extend the application of the statistical cloud parameterization approach, this paper proposes an innovative cloud parameterization scheme that uses the boundary layer height and the lifting condensation level as the primary parameterization predictors. The LES results indicate that the probability distribution of these two quantities is relatively stable compared with that of _l and q _t during the diurnal variation and nearly follows a Gaussian function. Verifications using LES output and the observations collected at the Atmospheric Radiation Measurement (ARM) Climate Research Facility (ARCF) SGP site indicate that the proposed scheme works well to represent continental BLCs.