The sheer size of modern scientific data presents significant bottlenecks that limit the value of the data. Fortunately, a scientist often only needs access to small subsets of the data at its full resolution. The identification of these "interesting" subsets is part of the data exploration process. An effective data exploration environment for very large data sets requires a hierarchy of data representations at increasingly finer resolution, along with an accompanying "error" measure that helps to identify the authenticity of each representation level. One example of research that could benefit significantly from such a representation is the study of clear air turbulence in the atmosphere. Using techniques of computational fluid dynamics (cfd), direct numerical simulations resolve both the smooth wave motion over most of the domain, and the turbulence, which occurs in small patches. These turbulent patches are important for airplane travel, satellite communications, and environmental studies. Simulations such as require hours of supercomputer time and produce far more output data than can be "downloaded" and visualized without time-consuming trial-and-error. This project aims at developing tools to transform such data into a multiresolution hierarchy with error measures and to provide associated interactive access and rendering functionality. Such an environment will provide researchers with interactive access to an order of magnitude more data than is otherwise possible. Such access will allow researchers to interpret their data in far less time than is now possible, and as a result, they will be able to run their simulations at ever more accurate resolutions. www.cs.unh.edu/mdr/
