Mark Rast is an associate professor in the Department of Astrophysical and Planetary Sciences at the University of Colorado, Boulder. His primary research interests are astrophysical fluid dynamics with an emphasis on stellar convective dynamics and scale selection, turbulence, the excitation of the solar p-modes, and the origin of solar/stellar irradiance variations. His interest in large scale numerical simulations includes the development of visualization and analysis techniques essential to their scientific understanding. In addition to these theoretical and computational efforts, Professor Rast serves as the principal scientist of the Precision Solar Photometric Telescope (PSPT) at Mauna Loa Solar Observatory, which obtains full disk images of the Sun at five wavelengths with 0.1% photometric precision.

Sun-like stars are convectively unstable in their outer layers. Turbulent motions are driven by heat production in the interior and radiative loss from the photosphere. The radiative losses are nonuniform over the solar surface, and vigorous new downflow plumes form at those sites where advective heat supply from below fails to support the radiative energy losses from above. These plumes play a crucial role in the dynamics of the flow [1], interacting to form larger convective scales in the surface layers, and possibly descending through the entire highly stratified convective layer to play a key role in the transport of heat, momentum, and magnetic field into the overshoot region below. The dynamics, stability, and transport properties of these flows are just beginning to be understood. That understanding is being achieved through very high resolution hydrodynamic and magnetohydrodynamic simulations.

While resources exist for such simulations of sizes 2048³ and larger, the essential scientific analysis and visualization of the results proves daunting. Post-simulation analysis is essential because these studies typically focus on physical process within the domain rather than statistical properties of the solution. The locations of interest, the nature of the analysis to be undertaken, and the relevant secondary quantities to be derived, are often not known or even knowable before the simulation is computed. Visualization to locate sites of interest and illuminate spatial relations between variables and analysis to quantitatively test scientific hypotheses on interactive time scales are both essential in order to interrogate the data and learn from the solution. Typically, many hundreds or thousands, and soon tens of thousands, of central processing units are available for batch mode simulations, but only a few to postprocessing scientific analysis. This mismatch means that the scientific return realized from the computational investment is limited by the availability, or lack there-of, of appropriate post-processing resources, both software and hardware.

The volume visualization in Figure 1 shows the enstrophy of a fully-developed compressible downflow plume. The plume was initiated and maintained by a fixed temperature perturbation imposed on the upper boundary (three-dimensional version of Case E in [2]). It is subject to vigorous secondary instability that leads to the successive penetration and disruption of the leading vortex torus by the stem flow from behind. This process generates the tangled mass of vortex filaments at the plume head. The computational grid employed was horizontally periodic and highly nonuniform. Only the central one-third and bottom one-third of the domain is shown after uniform resampling. The image was produced using an open source analysis/visualization package developed at NCAR named VAPOR [3].

[Download]

If you publish your work using this data set, please make acknowledgment of VisFiles: The data set is made available by Dr. Mark Rast at the University of Colorado-Boulder through US Department of Energy's SciDAC Institute for Ultrascale Visuaization.

[Pellet Data]

 
Return to Meet the Scientists