Why isn't photogrammetry used in tornado windspeed measuring?

Joined
Jan 14, 2011
Messages
3,287
Location
St. Louis
If you have a known camera position and a known tornado location, why isn't photogrammetry capable of measuring windspeeds in tornadoes and used regularly? The main drawback is of course you can't see interior vortices inside of thick condensation or debris, but any visible debris or condensation features should be usable to prove the *minimum* wind velocity. Anyone know more abou this?
 
Dan,

I forwarded your inquiry to Randy Zipser and Dr. Joe Golden, both of whom were very active in photogrammetric studies of tornadoes during the heyday of tornado wind speed research at NSSL in the early to mid 1970's. Randy, whom also originally co-created the physical Stormtrack newsletter with Dave Hoadley in 1977, had this in-depth response to your question. With his permission, I've posted the entire message here:
If you have a known camera position and a known tornado location, why isn't photogrammetry capable of measuring windspeeds in tornadoes and used regularly? The main drawback is of course you can't see interior vortices inside of thick condensation or debris, but any visible debris or condensation features should be usable to prove the *minimum* wind velocity. Anyone know more abou this?

I think that the answer to the question [Dan] posed in the screenshot lies in the rapid advancement in technology.

Back in the 1970s, both radar, instrumentation, and satellite technology was much less sophisticated; thus the tools that were being used in tornado research back then were much more primitive (at least by today’s standards). Tornado chasers in the early days of organized storm chasing (including the NSSL chase program) had to use bulky mobile phone units (we called them ‘bricks’) to frequently “check in” to get guidance from the radar meteorologist at NSSL. By the mid-1990s, chase vehicles were fully equipped with not only continuous wireless communication, but also on-board real-time radar [e.g., Project VORTEX and the Doppler on Wheels (DOW)].

The same thing was happening with photogrammetry over this two-decade period. Back then, we were using 8mm and 16mm celluloid movies, whose images were often blurry and very shaky, and we employed very primitive manual tracing methods using the projection of these film images onto tracing paper attached to a darkroom wall as a screen! As you can imagine, errors in windspeeds were quite large, although difficult to quantify without statistical, objective-analysis routines (which also were very cumbersome to employ with photogrammetric analysis methodology). By the late 1990s, however, computer technology had advanced sufficiently so that reasonably-accurate, interactive algorithms had been developed to allow all these basic manual steps to compute tornado windspeeds on a personal computer, thus saving the researcher a tremendous amount of time and manual input. Just imagine what lies ahead using the evolving AI technology we now have!

The reason that we don’t hear much nowadays about photogrammetrically-derived windspeeds is because these can be determined to a high degree of accuracy directly from very-clear, gyroscopically-stabilized video imagery in a matter of seconds by a computer equipped with proper analytical software. Like all trends with advancing technology, tornado windspeeds are no longer the “mystery’ that they once were back in the nascency of tornado research. The “EF” tornado windspeed scale has also evolved with these same advancements in technology, combined with advancements in structural wind-engineering research over the same period (over a half-century now).

As Joe may still recall, during the defense of my master’s thesis, my thesis committee asked me to defend a statement I had made that “tornado windspeeds will likely not exceed 350 mph,” with little or no objective proof to justify that statement. I made that statement back in 1976, based solely upon verbal communications and largely-unpublished material from other researchers I had informally contacted (Jerome Blechman at the U of Wisconsin, and the Texas Tech Wind-Engineering group, for example). In those early days of tornado structural-damage research, the “gut feeling” at that time was that most tornado damage that was being observed could be explained with windspeeds at or below the 300 mph velocity range, as compared with a prevailing thought that tornado winds could possibly exceed 500 mph or even Mach 1. There was even some talk out of Fujita’s camp about adding an “F-6” category!

Luckily for me, my statement has successfully stood the test of time as tornado structural research has proven this observation to be accurate, so much so that Fujita’s original “F-scale” was replaced by the “Enhanced Fujita” (EF) scale in subsequent years, reflecting an overall reduction in tornado windspeeds based upon observed tornado structural damage. The EF scale is still in use today.

On May 3, 1999, Dr. Josh Wurman and the DOW research crew recorded a boundary-layer (near surface) windspeed of 318 mph. That is still the standard used 25-years later for “maximum” windspeed expected for an EF-5 tornado. Of course, there is no reason to expect that this “maximum windspeed” won’t be superseded in the future, but it has also stood the test of time.

All In all, we have a pretty good handle presently about tornado windspeeds, just as we know very well today why tornadoes form, how they form, and how and they are structured. The next frontier of research will be how to prevent them from forming in the first place, something that is likely a very long way off, due to the scale of forces at work. But with human ingenuity, we can never say “never”...

Randy Zipser
 
I'm with you Warren.... I think I read somewhere that the Harlan tornado was scanned by the DOW and got a reading of 224mph, but was rated EF3. I also fully understand the concept of the damage based scale, but I think we can do better. Maybe once most storms can be measured by radar or another accurate method, we'll see a change in policy. If photogrametry can be applied with a high degree of accuracy, hopefully we can make use of the technology and the abundance of chaser video to start getting some good data.
 
Blake, that is fantastic information, a big thanks to Randy for chiming in! With the multiple video angles in 4k we have now along with drone aerials, triangulation is now possible. Is there any talk on factoring this into the upcoming EF scale revisions?
 
Blake, that is fantastic information, a big thanks to Randy for chiming in! With the multiple video angles in 4k we have now along with drone aerials, triangulation is now possible. Is there any talk on factoring this into the upcoming EF scale revisions?
I've forwarded on to a few that are not on this forum that may have better insight. This would be an excellent inquiry or topic to discuss at the 31st AMS SLS conference in October [21-25], 2024.
 
If you have a known camera position and a known tornado location, why isn't photogrammetry capable of measuring windspeeds in tornadoes and used regularly? The main drawback is of course you can't see interior vortices inside of thick condensation or debris, but any visible debris or condensation features should be usable to prove the *minimum* wind velocity. Anyone know more abou this?
Some great info already in this thread.

I'll address the original wind speed question later, but first I can add that there are recent uses of photogrammatry that are not wind speed related that make high quality footage still of substantial value for certain uses.
  1. Dr. Leigh Orf has worked with specific chasers to obtain high quality footage for observatinal comparison of storm features with his high resolution supercell simulation features. As I recall, he can use the footage to judge if the simulation is behaving like real storms. Some of that same footage has been used to loosely calculate sub vortice and satellite transit ground speeds about a meso (Schyma and Talbot captured this footage and performed the computation, I believe).
  2. Several studies from academia/science and even chasers have used photogrammatry to obtain situation or timelime information from a weather event. For example El Reno has been analyzed extremely closely by people such as Skip Talbot using collected photos to produce one of the more detailed understanings of what occurred.
As for wind speed computations, Blake, and Randy via Blake, have already pointed out that we don't really need photogrammetry for wind speed. I will add several complications I can think of to trying to do so, even with 4K and up footage. I ran into these in the past when working an engineering project for an unnamed investor that wanted to attempt exactly what you mention, using wind speed estimates derived from video (I left that project for reasons, but here are some of the things considered as possible technical issues to make it work):
  1. For decent accuracy, fairly exact position would need to be known of the camera, as well as the objects moving in frame.
  2. Camera will need to be stationary to not make computation very hard (two moving things).
  3. The size of debris or other objects in frame might need to be known to help establish distance (are we seeing a piece of wood far out on circulation edge, or a roof right in the thick of it? If a center of mass pixel method is used, apprximations could be made especially with known references on the ground like power poles that can be measured, etc.
  4. Lens distortion (barrel, pincushion, mustache, complex) would need to be corrected and understood well on a fixed focal length lens during observation.
  5. Lens abberations can distort objects on edges or in corners of image
  6. Any focus issues or focus breathing can distort objects in frame
  7. Debris tumbling and feature morphing within a circulation makes it hard to track an transiting object in frame precisely
  8. Ground speed adds to rotational speed in damage that falls into our current speed estimates via damage indicators. Photo/video alone might have a hard time doing both at once; tracking a storm rotation and movement from a stationary location long enough for solid survey of a storm.
  9. Rain wrapping and multi vortex, and evolving appearance would likely complicate continuous tracking of features or debris for many tornadoes.
  10. Circular motion presented on a flat sensor (requiring polar coordinate translation probably). This would be much less problematic filmed far away with a long zoom than close and wide, as a tornadic circulation would appear much more like a flat surface to a zoomed lens that also will have low distortion compared to a wide lens. Still it may not be trivial to compute for high accuracy needs.
We never did resolve all those questions but had ideas, and also realized accuracy would often be limited and complexity high (expensive) when we already have radar. We were interested in the radar gap below 10m (for mobile radar) primarily. In the end, the things photogrammatery is good at are approximate/coarse tracking of large consistent features passing known visual references like a tornado ground speed from a stationary position with power poles in scene, or triangulation between photo video moments to learn about storms structures or event timelines.

Not sure if any of that was useful, but I thought I would add it in case it may be, since I had looked pretty far into this at one point.

I feel phased array radar is the holy grail of wind speed measurement. As for now, there are only a couple research units I know of through OU/Naval Research. Getting rapid beam scanned electronically steered radar near storms will really increase fidelity.
 
From a George Harrison song, Got My Mind Set On You...
"...it's gonna take money, a whole lotta spending money, it's going to take plenty of money...
...it's gonna take time, whole lotta precious time, it's gonna take patience and time..to do it right..." :)
But more seriously, here's a story from that era of research. I was walking through the hallway of the Meteorology Department in the Walker Building at The Pennsylvania State University on my way to speak with a new Ph.D. assistant professor named Dr. Gregory Forbes. My goal was to learn more about the April 3, 1974 Super Outbreak, unrivaled to this day for sheer atmospheric-force from a group of tornadoes.
I then noticed a "jumbo-sized" 😉 poster on a nearby wall. It featured photogrammetry of the Parker, Indiana F-4 tornado from that day, with varying-length lines pointing in many different directions labelled with differing wind speeds: Tangential velocities.
Basically, under the guidance of Ted Fujita, Forbes had earned his Ph.D. from the University of Chicago for that very dissertation topic. And that picture hanging in the hallway that I had seen earlier was the fruit of much of his labor. He had "problems with trees" while doing his research, but nonetheless determined in his work that the tornado & even numerous suction vortices featured faster wind speeds higher in altitude. You can read more about much of their work in both hard & soft research-literature from that time, journals and conferences, respectively.
After we spoke about the event and the aerial surveys, I asked him about the clutter of photographic slides that I noticed on the top of a file cabinet. He explained that people from seemingly everywhere send tornado images to him. I was envious; such items were less common back then. He held one up to the light as if searching inquisitively for a date encrusted on a rare gold coin painstakingly retrieved from a sunken, Spanish galleon many years ago.
 
Last edited:
Back
Top