HR: 15:20h
AN: NG12B-06    [PDF]
TI: Wavelet and Statistical Thresholding of convection plumes in High-Rayleigh number flows
AU: * Erlebacher, G
EM: erlebach@csit.fsu.edu
AF: Florida State University, Program in Computational Science and Information Technology 489 Dirac Science Library Tallahassee, FL, Tallahassee, FL 32306-4120 United States
AU: Yuen, D A
EM: davey@krissy.geo.umn.edu
AF: University of Minnesota, Minnesota Supercomputer Institute University of Minnesota, Minneapolis, MN 55455 United States
AU: Vasilyev, O V
EM: oleg.vasilyev@colorado.edu
AF: University of Colorado, Boulder, Department of Mechanical Engineering University of Colorado 427 UCB, Boulder, CO 80309 United States
AB: We present some novel approaches to the problem of analyzing the characteristic properties of temperature plumes in high Rayleigh number (Ra) convection flows. The datasets are computed at $Ra=10^6$ to $10^9$. As Ra increases, the characteristic width of the upwelling and downwelling patterns decreases as $Ra^{-1/2}$ and occupy a increasing small fraction of the volume. In the presence of ever increasing data set sizes, it is necessary to develop compression algorithms that are consistent with the physics of the flow (which is localized within these plumes).\\ We describe two approaches for plume identification. The first technique uses second generation wavelets combined with thresholding to only retain 1-5 percent of the dominant wavelet coefficients. The particular wavelets used permit a one to one identification between each wavelet and a point in physical space. By displaying those points corresponding to wavelets whose coefficients are above a user-specified threshold, the structure of the plumes becomes clear. \\ There is also a need to study the morphology of the plumes, including their surface, volume and other characteristics. We explore a computational approach that constructs probability density distributions (pdf/histogram) of variables integrated over the isosurface of temperature as a function of temperature. The novel approach lies in performing this computation using one pass through the data. By choosing appropriately the values integrated, one derives information on the distribution of functional values, gradients, and curvature. Maxima and inflection points become characteristic properties of the flow. We study their variation as a function of Ra. This technique is also applied to automatic color map selection; we illustrate the results on our datasets.
DE: 8121 Dynamics, convection currents and mantle plumes
SC: Nonlinear Geophysics [NG]
MN: 2003 Fall Meeting