The ash cloud visualization is based on the output of a model developed by Hiroshi Tanaka of the Geophysical Institute of the University of Alaska and Tsukuba University, Japan. Using meteorological data and eruption parameters for input, the model predicts the density of volcanic ash particles in the atmosphere as a function of time. The three dimensional Lagrangian form of the diffusion equation is employed to model particle diffusion, taking into account the size distribution of the ash particles and gravitational settling described by Stokes' law. Details of the model are given in [2, 3].
The meteorological data required are winds in the upper atmosphere. These are obtained from UCAR Unidata in NetCDF format. Unidata winds are interpolated to observed conditions on 12 hour intervals. Global circulation models are used to provide up to 72 hour predictions at 6 hour intervals.
The eruption parameters for the model include the geographical location of the volcano, the time and duration of the event, altitude of the plume, particle density, and particle density distribution. The time and duration of the eruption are generally determined from seismic instruments in the vicinity of the volcano. Several of the most active volcanoes in Alaska are also monitored by slowscan video cameras, which can provide a more accurate indication of the time and duration of the episodes when ash is actually being produced during an eruption. The particle density distribution is set to match observed distributions from other measured eruptions. The particle density itself is entirely relative, as there is no available method to estimate the amount of material injected into the atmosphere at the time of the eruption. After the fact, it is theoretically possible to compare ashfall on the ground with model predictions, but this involves extensive field work in what are usually wilderness areas.
The raw output from the model for each time step consists of a list of particles with an (x,y,z) coordinate for each particle. An AVS module reads the particle data and increments the particle counts for the cells formed by an array indexed over (x,y,z). We chose a resolution of 150x150x50 for the particle density array, which equals 1.1 million data points at each solution point in time. For the video animation, we ran the model with a time step of 5 minutes. For 13 hours of simulated time, the model produced 162 plumes, amounting to approximately 730 MB of integer valued volume data.