Earthquake simulation visualization

by Yuan Hong, Thomas Kerwin, and Boonthanome Nouanesengsy

• [Home]
• [Contact]

Task

For the final project in CSE 788.14K (Fundamentals of Visualization), groups are asked to visualize the IEEE Visualization 2006 Contest dataset.

Directions for the project

One of the most important things in this visualization is showing the geographical context of the data. This includes:

• Add the borders of california
• Mark the ocean and the land
• Mark the location of cities
• Mark the basin locations
• Show legend information such as scale, compass and time

Our target will be for high quality two-dimensional display rather than real-time interactive images. We will use the small multiple technique to give more information in a smaller space. We want to have high data density and integrate time series information into static images.

We want to focus on the subset of the data given more importance by the study. However, we will include some displays of the full data to increase contextural information.

We plan to write some custom software using OpenGL to integrate the context information into the data, but we are going to use a pre-written volume rendering library to handle the volumetric data.

Data processing

The original datasize was over 60GB. To make the data more managable, We took every tenth timestep, lowering the data we had to deal with to 6GB. We also wrote scripts to take the mean velocity magnitude (speed) of the 22 vector fields that were left.

Using Volsuite, we extracted an isosurface from the stiffness dataset that represented the edges of the basins talked about on the data explaination page. We rendered this mesh with the mean speed data, converted from floating point to 1 byte data. We applied a transfer function so the data would stand out against the basin.

Histogram

We made a two-dimensional histogram of rock stiffness vs. mean speed. On the vertical axis is stiffness, with lower numbers at the top. The horizontal axis is velocity, with lower numbers on the left. A program went through all of the voxels in the two data volumes and incremented a count for each bin in a 1024x1024 grid.

This histogram has two interesting features. One is the larger amount of higher velocity datapoints in lower stiffness rock. This seems to correspond with our common-sense ideas about stiffness. The second feature is the marked sections that the data is divided into. Whether these show some physical characteristic of rock or are artifacts from assumptions made when generating the simulation, we do not know.

Gradient magnitude

Applying a central differences filter to the mean speed gives us a gradient. The magnitude of this is an acceleration value. By plotting this with the basin mesh, we can see higher activity in the main basins near the center of the disturbance. The values go from dark red to bright orange.

Vector field visualization

By treating the velocity data as a flow field, we can generate movies based on this field. This short movie shows the pressure wave coming out of the epicenter of the event and moving outwards.

The waves are computed through 2D streamlines (XY plane). Seeds are dropped uniformly along the X and Y axes. The depth Z (42) is selected around the basin. In each time step stream particles are calculated and rendered in a blue color. From the animation we can find out that the earthquake wave occurs at one spot (southwest) and spreads out to its surroundings and finally occupies all the space.

Composite image

In this image a map of southern California is overlayed with an image of the mean of the velocity magnitude over all 22 sampled timesteps. The colors for velocity range from white, the weakest velocity magnitude, to red to yellow, the strongest. The averages were obtained from sampling every tenth timestep. The blue areas are basins, and were extracted by displaying an isosurface from the scalar stiffness field. Also pictured are four cities, designated by the black open circles. At the bottom of the image are four line graphs displaying the velocity magnitude at the cities. Values were recorded every tenth timestep. The x-axis for the line graph represents the timestep, and the y-axis represents the magnitude of the velociy, in meters per second.