Analysis of Patterns of Atmospheric Motions at Different Scales by Use of Multiresolution Feature Analysis and Wavelet Decomposition

摘要

This report describes the work accomplished during ONR Grant No. N00014-92-J-1223, R&T Project Code 3226028-10. We have been successful in developing an entirely new technique for analyzing the form and spatial distribution of eddies in observed and simulated atmospheric flows. This technique, which we call multiresolution feature analysis (MFA) is applicable to other types of flow as well as atmospheric. To date, MFA has been used to determine the preferred patterns of motion for atmospheric flows observed over the high plains, and for a comparable large eddy simulation (LES). The technique was also used to obtain estimates of the support dimension of the turbulence for the same observed and LES flows. The results showed that the subgrid parameterization used in the LES did not capture all the intermittency of the observed flow. Fast wavelet transform techniques have also been implemented and applied during the course of this grant. The decay of turbulence in a stratified laboratory flow was analyzed to show that mixing and restratification tend to be governed by a few major mixing events, rather than many small ones. It also showed that the major eddies at one scale do not coexist with those at other scales. This suggests that the smaller eddies in a turbulent cascade are not embedded in the larger ones, but may be attached to them. The methods developed during the grant have been demonstrated to be workable, and have already provided some new physical insights into the nature of small scale motions. We expect to apply the analysis tools that we have developed to other flows and to modify them to provide more information about the nature of the spatial attachment of eddies at one scale to those at another. To make future applications easier, we have begun implementing and testing interpolation techniques that can be used to provide estimates of missing observations, and more regularly spaced values for the numerica.