The main objective of the present work is to use in-situ data to parameterize cloud fraction (Cs) as a function of both cloud condensed water content and its variability over scales of 10 km and 100 km. In-situ data were obtained during the Alliance Icing Research Study (AIRS) field project and represent mid-latitude winter stratiform clouds. The main observations used in the analysis were condensed total water content (qc) and liquid water content. It is shown that Cs, defined as the length of cloudy segments divided by the clear and cloudy segments along a flight leg, can change up to 5%–25% if the averaging scale is changed from 10 km to 100 km. A new parameterization of Cs versus the ratio of qc to its variance is suggested. The sub-grid scale variability is usually not known directly from the models and it needs to be obtained by prognostic or diagnostic equations. It is found that cloud sub-grid variability, represented by the standard deviation (σ) of qc, increases with increasing qc for both 10 and 100 km scales, and σqc for 10 km scales is larger by about 50% as compared to 100 km scales for a given qc value. These conclusions suggest that both averaging scales and sub-grid scale variability should be considered for cloud cover parameterizations.