Abstract: In this study,a statistical cloud schemeis first introduced and coupled with a 1st-order turbulence scheme with 2nd-order turbulence moments parameterized by the timescale of the turbulent dissipation and the vertical turbulent diffusion coefficient.Then the ability of the scheme to simulate cloud fraction at different relative humidity,vertical temperature profile,and the timescale of the turbulent dissipation is examined by numerical simulation.It is found that the simulated cloud fraction is sensitive to the parameter used in the statistical cloud scheme and the timescale of the turbulent dissipation.Based on the analyses,the int roduced st at istical cloud scheme is modified.By combining the modified statistical cloud scheme with a boundary layer cumulus scheme,a new statisticaly-based low-level cloud scheme is given and tentatively applied in NCAR CCM 3.It is found that the simulation of low-level cloud fraction is markedly improved and the centers with maximum low-level cloud fractions are well simulated in the cold oceans off the western coasts with the new statistically-based low-level cloud scheme applied in CCM 3.It suggests that the new statistically-based low-level cloud scheme has a great potential in general circulation model for improving the low-level cloud parameterization.