Spurious Oscillation in Simulating Boundary-Layer Cumulus Clouds with Third-Order Turbulence Closure Models
Cheng, A.(a) and Xu, K.-M.(b), Center for Atmospheric Sciences, Hampton University, Hampton, VA (a), Atmospheric Sciences, NASA Langley Research Center (b)
Thirteenth Atmospheric Radiation Measurement (ARM) Science Team Meeting
A hierarchy of third-order turbulence closure models are used to simulate boundary-layer cumulus clouds from the Atmospheric Radiation Measurement in this study. A liquid-water spurious oscillation is found in the Level-3 model, which predicts all third moments. The period of the oscillation is about 1000 s, which is resulted from the interaction of the mean liquid water gradient and the liquid water buoyancy terms in the third-moment equations. A reasonably large diffusion coefficient and a large dissipation at its originating level are needed in order to eliminate the moist spurious oscillation. However, this approach also diminishes the real signals for shallow cumulus clouds so that the vertical distributions of cloud fraction and liquid water mixing ratio are distorted. A better approach is to parameterize some third moments in the third-order closure models (Level-1 and Level-2 models), especially those related to liquid water buoyancy. Level-1 model, which only predicts the skewness of vertical velocity, is shown to produce reasonably good simulations and prohibit the moist spurious oscillation efficiently. A drawback of this model is that some information of the third moments may be lost. A Level-2 model would be more desired, which predicts more third moments than a Level-1 model. Another difficulty with the third-order closure models is the closure assumption on the fourth moments. The vertical distributions of cloud fraction and liquid water mixing ratio simulated by third-order closure models are very sensitive to the closure assumptions. It is found that the closure assumptions based upon the quasi-Gaussian and double-delta distributions are not suitable for shallow cumulus clouds. The closure assumption based upon the double Gaussian distribution produces the qualitatively right cloud distribution.It is also found that clipping approximation, which is usually associated with the quasi-Gaussian closure, can hide the instability problem that arises from the quasi-Gaussian closure assumption in the Level-3 model and also produces incorrect cloud distributions.
Note: This is the poster abstract presented at the meeting; an extended version was not provided by the author(s).
