Wavelet-based lossy compression of turbulence data

Abstract

Summary form only given. Modelling of turbulence is among a class of Grand Challenge applications that severely strain data storage capabilities. The rate at which data is generated can actually limit the performance of the application. One tool that can assist is data compression. Unfortunately, lossless compression performs poorly on such data. Lossy compression is the only choice for turbulence data. Lossy compression preserves the absolute characteristics of turbulence data well. However the statistics of the data are poorly preserved. It is believed that various structures in turbulence determine the behaviour of the various statistics. Designating these structures as regions of interest and using a region-of-interest (ROI) coder is one potential approach to preserving these statistics better. In this research, data from a quasi-geostrophic model of incompressible turbulence is used. The output data of this model is vorticity.