Data Characterization for Code Comparison Results

The I in this refers to Rick Wagner


I am working on building a database for the runs, simulations and data sets in the CADAC that will be made publicly available. Part of this involves characterizing the results of a simulation. For the code comparison part of the KITP workshop, I am gathering advise on what quantities (derived scalar values) the participants think are appropriate.

These number are intended to encapsulate some of the actual results of the comparison, and to help with searching the results later. (In the Virtual Observatory realm, quantities like these are referred to as "Characterization Axes", like defining a parameter space.)


I think a list of about 10 of these numbers would be good start. From a short conversation with Mark Krumholz at the workshop, looking at the analysis tools Åke Nordlund has provided, and my own work with Alexei Kritsuk, here are some initial suggestions:

  • rms Mach number
  • maximum density
  • integrated power over a few large bins (e.g., 1 < k/kmin < 2, 2 < k/kmin < 4, 4 < k/kmin, etc.)
  • sigma, b from PDF of density
  • velocity power spectrum slope in the inertial range
  • structure function exponents for velocity
  • rms magnetic field strength for MHD simulations

It's very easy for me to imagine looking at table in a paper, showing these numbers for each code, after a single dynamical time, and getting a quick understanding of how they compare.

Two important aspects of whatever values are chosen, is that they should translate well between codes, and that it is realistic to expect all of the results to provide these numbers. (Where appropriate, maximum density has little meaning for incompressible turbulence, like magnetic field in a pure hydrodynamics simulation.) Another aspect, is that these numbers should be the results of analysis, not input parameters.

The goal of this is to build a set of common characteristics for data in the CADAC. We can always add more, and provided a means for users to define their own. But for now, I am only hoping for a few quantities that everyone can agree on to summarize the results of each simulation in the code comparison project.