Listed below are the ideas which have been proposed, along with images representing prototype implementations of each technique. Unless otherwise stated, the dataset being used for these examples is synthetic, and the grid size is 30 by 90 by 12 grid cells. Data Reduction Techniques Depth Averaging Combined Flux Proposed Visualization Techniques Cones for Flux 2D Icons for Flux Streamlines Backwards Streamlines Depth-Colorized Streamlines Time Markers Scaled Flux Blobs Small Data Examples Small Data Sanity Check on Reservoir G Reservoir G
The technique which was suggested is something like this: at each grid location, create an icon who's shape is a function of two parameters. One of the parameters is SO or SOM (or some other variable to be determined). The other is the flux (for oil, I'd guess). The idea is to have what is basically a sphere (that's large enough to see) in areas where SOM or SO is HIGH and the flux value is LOW. And to have footballs or ellipsoids in areas where SOM or SO is high and flux is HIGH. And voila!
You'll see that in that same area of high SOM and low flux a bunch of spheres. Elsewhere, where flux is significant in magnitude, the icons get more oblongated. The sizes of the icons don't really mean much because of a simple trick I put in to get the spheres to stand out: the major and minor axes of the footballs are scaled DOWN by a value which is the ratio of the major and minor axes. Which means that spheres stay big and ellipsoids get smaller. There are a couple of other issues which I'll leave out for now...
Here's a snapshot from a few time steps later in the simulation..
Use of color to highlight a change in a scalar variable.
The first example shows using the location of well "W12" (presumably a water injector). Rather than beginning the streamlines computation at the location of the well, which would result in flow directly out of the grid to the bottom because of a strong downward flow, we seed the points at some distance away from the well. This distance is under user control.
After positioning the location of a virtual well into the flow field, and decreasing the density of sample points around the well, streamlines are produced which eminate from a virtual cylinder, of a user-specifiable radius, around the virtual well. Note that these streamlines were computed using the LBL-written streamlines code. The numerical method being used is the second order Runge-Kutta form of time-velocity integration.
Backwards streamlines computed from a virtual well placed at what looks like an oil producing area. A radius of about "1200" (in the local coordinate system) spreads the seed points out around the well. The streamlines are depth-colorized, and some effort in tweaking renderer parameters was needed (depth cueing and adjusting of the front/rear clipping planes) to effectively "hide" data which is "far away". Click here for super hi-res version (1800x1600) of this image. This image computed from "live" data from the "C" reservoir.
Another example using two layers of cones. One layer represents a depth- averaging of the first 3 simulation layers, the other represents a depth-averaging of the bottom four simulation layers. The seed points for the streamlines are placed at some distance radially from the virtual well to achieve better dispersion.
Same as previous picture, but added time-markers along the streamlines.
Some issues to consider (for the above image) are:
A first cut at putting flow BARBS on the grid blocks. The barbs are oil flux, and the boxes are colored according to the pressure in the init map file. I only put barbs on one of the 8 block faces.
However, when "similar" simulation layers are combined, the amount of data to visualize and interpret is reduced by an order of magnitude. In this image, the top few simulation layers are combined into a single layer of flux; similarly with the bottom few simulation layers. The general trend in each of the top few and bottom few layers is much more obvious in this image than in the image containing a geometric icon at every node.
These images make the same point, but were computed using "live" data from the "C" reservoir.
Check out a couple of short, 13-frame MPEG movies: Per-Cell Cones Movie and Depth-Averaged Cones Movie. (Note: you may want to tweak your Mailcap file so that mpeg_play will loop explicitly, since these movie loops are so short.
This image is from a "live" reservoir. Cones are placed at grid coordinates in which the flux data is non-zero. Again, green represents oil flux, and red represents gas flux.
Same as above, but with color-coded transmissiblity magnitude at the grid boundaries.
Mpeg movie containing 50 time steps.