STORM TRACK: September 30, 1983 (Volume 6 Issue 6)

Organized storm chase efforts have traditionally placed a high priority on obtaining movies of tornadoes. It has long been felt that much could be learned by analyzing these movies to determine airflow, using tornado photogrammetry techniques. Photogrammetry is simply the science of making measurements from photographs.

Various chasers have pioneered in tornado photogrammetry; among them Joe Golden, Dan Purcell, Bob Davies-Jones, Randy Zipser and myself. Early efforts involved measuring the movement of tags (dust, debris, or cloud) in two dimensions in the plane, oriented parallel to the film, at the tornado. At the University of Chicago in the mid-70's, Greg Forbes improved photogrammetry techniques by using more precise mathematical formulations and assessing uncertainties in the velocity estimates.

While working at Texas Tech University (on the magic High Plains at Lubbock), I further refined techniques and performed detailed analysis on footage of two tornadoes from the May 28 Tulia, Texas storm. Basically, tornado photogrammetry involves measuring the movement of a tag on the projection screen, and then, by estimating how far the tag is fram the camera and knowing what length of time the movie covers, the velocity can be determined.

But one camera can only provide information in two dimensions, so one or more pieces of information is needed: an estimate of what "orbiting radius" the tag is following around the tornado center. Finally, "conservation of mass" is assumed so that how much air is flowing into or out of the tornado can quickly be computed from knowledge of how much air is flowing upward through the tornado.

The geometry of the problem is really fairly complicated. Uncertainties can be greatly reduced by doing a careful survey after the tornado. Several types of information must be obtained. The camera site must be located exactly, and the distance from the camera to the tornado ascertained. Also the angular separation and elevation angles of as many distant landmarks as possible must be measured with a transit.

In the past two years, several tornadoes have been filmed from two or more camera sites at the same time. This opens the possibility of doing highly accurate photogrammetry, using fewer assumptions and substantially simpler geometry. I will be analyzing footage of these tornadoes and bringing photogrammetry into the high-technology age during the next year at the University of Illinois. (Editor's note: This article was submitted to ST during the winter of 1982-1983)

This is how I envision doing photogrammetry on the backlog of tornado footage now available. Each movie frame will be digitized; that is, brightness of the picture will be measured at perhaps several million points by a fancy piece of equipment. Then, using an image processing and analysis computer system, I'll remove scratches, bounciness and zooming from the movie segments.

The computer will allow me to look at all of those features that have one of 256 levels of brightness or certain combinations of brightness. So, individual tags will be highlighted and tracked with a cursor on a television monitor. The location of the camera and the position of a tag marked by the cursor form a line in space, and the tag is located some distance along this line. The exact position along the line can be determined by looking at a simultaneous image from a camera at another site. The computer will automatically display a line in the second-camera image, which represents the camera-to-cursor line in the first image. It is then up to the analyst to decide, by looking at the second-camera image, where along the line the tag actually is. Once that position is determined, the computer "knows" exactly where in space the tag is located. Its movement from one frame to the next gives the velocity.

The technique just outlined will produce the first fully three dimensional pictures of tornado structure and flow ever produced. Also, new movies can be produced which are perfectly stable and will have enhanced contrast and image quality. These can also have graphical presentations superimposed, such as arrows showing how the air is flowing.

Hopefully, the procedures can be so automated that large amounts of footage can be quickly analyzed. Perhaps by spring, I will have made a dent in the two-year backlog of movies, and we can obtain more simultaneous films for future analysis.