LS Models Built from Partial Knowledge of the Flow

Since Thomson's (1987) provision of selection criteria for LS models, "known flow" has come to mean that the single-point pdf (PE) of the Eulerian velocity field is a mathematically-prescribed function of position. But for any real flow, one has available only partial information, usually in the form of a few low-order velocity moments: the LS model must be built from partial information.

Du et al. (1994a) took the view that from the available information, the most rational course of action is to form what is called the "maximum missing information" (mmi) pdf. If the information given is an ordered set of moments ( , j=1,2,...N), the mmi pdf is of form PE(w)=exp [ - ( a0 + a1 w + a2 w2 + ... + aN wN ) ] where the a's are determined by the given moments. Du et al. constructed an mmi pdf for vertical velocity in the convective (daytime) atmospheric boundary-layer (CBL), and derived the implied model for vertical dispersion, and elsewhere (Du et al., 1994b) demonstrated that forming the well-mixed model derived from an mmi pdf is the best solution yet provided to the problem of building an LS model from partial flow specification.

Back to John D. Wilson home page.

LS Models Built from Partial Knowledge of the Flow

Last Modified: 6 June 1995