Tuesday, April 17, 2012

A Rare Direct Measurement of a Tornado's Winds

As the tornado approached Wichita in the Saturday evening darkness, AccuWeather Enterprise Solutions' meteorologists in downtown Wichita and I (back home, I don't chase storms after dark) were closely watching the data from the Federal Aviation Administration's (FAA) Terminal Doppler Weather Radar (TDWR) located south of Wichita's Mid-Continent Airport. The 44 TDWR's were installed by the FAA to insure the safety of aviation in thunderstorm situations. These radars were a direct result of all of the airline crashes related to thunderstorms and wind shear from the 1960's to the 1980's.

These radars have much higher resolution wind velocity data than the National Weather Service's WSR-88D's plus they report more frequently (every minute versus, at best, 4.1 minutes from the -88D's). That resolution and timeliness is vital in tornado and severe thunderstorm situations.

Unfortunately, the FAA originally thought that extremely valuable data should be confined within the FAA. In Warnings, I describe the battles that had to be fought to free up that data for the wider meteorological community.  It was just last week the data at one-minute intervals began to flow.

With that background, I want to show you an amazing TDWR image from 9:47pm just before the tornado crossed from Sumner into Sedgwick County.

The purple oval in this wind display defines the edges of the mesocyclone which, typically, is a 5 mi. or so area of rotating air within a supercell thunderstorm. The -88D shows mesocyclones well. Unfortunately, from the point of view of a warning meteorologist, only about one in five mesocyclones produces tornadoes and it is often difficult to anticipate which is which.

But zoom deeply into the wind data.

It is one of the rare instances of a fixed (as opposed mobile research) radar directly observing a tornado. At the tip of the arrow are two light green pixels which is the relatively calm center of the tornado. Straight above, the bright green pixel represents winds of approximately 150 mph! We knew that a major tornado was headed toward our city.

Storm chaser Drew Pelz took this photo (below) of the tornado, illuminated by lightning, just north of U.S. 160 at the same time as the radar image. The rotating "wall cloud"(purple arrows) is the visual manifestation of the mesocyclone.

The radar depiction of the tornado, combined with the ground truth reports in the field, allowed plenty of time for people to get to shelter to save their lives as we focused in on the tornado's exact path.

Kudos to the National Weather Service and Federal Aviation Administration for making this happen.


  1. Amazingly enough, you could actually also make out the mesocyclone on the "normal" KICT radar (at least looking at the Level II velocity data) when the storm was at about the same position as I noted in this tweet: https://twitter.com/#!/wxdevin/status/191365164992434176

    Certainly not quite as high-resolution as the TDWRs (you certainly can't actually make out the tornado circulation) and the rapid-refresh is killer, but it's amazing that it was a powerful enough storm to notice something like that on radar!

  2. Yes, the mesocyclone showed up nicely on the KICT radar. Thanks for the comment.

  3. Mike, I don't believe you are able to resolve the tornado at that point you have shown. That is a very narrow tornado. The only way that could be resolved assuming it is around 15 miles from the RDA is if the tornado were a half mile wide.

    That is based off simple linear expansion of the beam. A one degree static beamwidth is around a mile wide at 60 miles. The TDWR has a static beamwidth of 0.55 degrees. Simple algebra then says that the beam at the point of intersection would be just over 1/8 mile wide. That means that to full resolve it, you would need four radials next to each other which totals around 0.5 miles in width. It would be much easier to see the depth of the tornado given the depth of each bin is around 150 meters.

    You reference relative velocity minimums. That could be from very low signal-to-noise ratios or from multiple folds in the velocity data given the relatively low Nyquist velocity that is inherent to a standard C-band radar. Without spectral plots this is impossible to decipher, and irresponsible to assume. In my opinion I'm sure that if you looked at the spectral data you would find that those bins are composed of what would statistically be considered white noise.

    Granted this simple spatial analysis would change drastically if the tornadic mesocylone moved closer to the RDA. Then it is possible you may be able to resolve it. Realistically the tornado would need to move within 2-9 miles to have a legitimate chance of being fully resolved.

    Chris M.

  4. The tornado was within "2-9 miles" of the radar, 4.2 to be exact.

    You might be confusing the location of the TDWR with the WSR-88D at Mid Continent. The TDWR is on Clearwater Road just east of Clearwater. You can see the radar site in the image displayed as a dark blue dot east of Clearwater.

    Given the very narrow beam spreading at that distance, I'm very confident it was sensing the tornado.

  5. Mike, I may have over-estimated the distance while you have under-estimated. I have downloaded the data and can confidently say you are most likely not resolving the actual tornado in the radar data. I have downloaded the data via NCDC. At 9:47 PM, the mesocyclone referenced is 9.2 statute miles (8.0 nm) from the RDA. I'll be generous and let you have 8 miles given some deviation from scan time to the time of the photo.

    Even then, you are looking at a static beamwidth of roughly 1/16th of a mile which equates to 330 ft. The bin depth remains the same.

    Given the photo you have shown, I cannot make accurate assertions about the dimensions other than the base of the tornado is possibly half of the height of the cloud base above ground and resembles the tube-type tornadoes. Now to sample the tornado fully with four radials, you would need the tornado to be 1320 ft wide which is 1/4 mile. Doubling that to figure out cloud height would yield a base height of 2,640 ft.

    Those numbers could be deemed as reasonable, but there is one final catch to this. Based off of the survey tracks here http://www.crh.noaa.gov/ict/?n=april14th_hpsutrack there is no way you are observing the tornado because at that point it has lifted. The track is the far northeastern one. Given the time you took the sample, the tornado has indeed lifted based off of that and the relative statistical certainty regarding the geo-location of the beam centerline.

    I have uploaded some images pulled from GRLevel3 Version 2.0 which brings in the same satellite data used in Google Earth plots. You can see the images with and without data at http://www.chrismaiers.com/MikeSmith/Map_Data.png & http://www.chrismaiers.com/MikeSmith/Map_ND.png . You can also verify my plotting by looking at the wider view image located at http://www.chrismaiers.com/MikeSmith/Map_Data_Wide.png . In this case the tornado’s ending point was figured by counting the land plots which are a quarter mile in size from the town of Conway Springs. The ending point listed in red is one plot north & about five east. The center of the mesocyclone shown in yellow is about two statute miles away as the crow flies. Again, that is far outside of statistical uncertainty of the geo-location of the beam centerline at this distance. The only way this would be possible is if the radar was mis-calibrated for due north & off-center from its true latitude and longitude coordinates. Both are unlikely.

    That being said there is no tornado at this point to sample. Therefore the mesocyclone is being sampled in the images you have posted.

    Also, those velocity minimums you are referencing at that time could also be contributed to a dealiasing failure which is easily noted by the discontinuity of the radial velocity data.

    Given those conditions it is nearly statistically and physically impossible that you are sampling a tornado with the TICH TDWR.

  6. At 8 miles, isn't the beam hitting the storm around 500 feet or so (8*5280*tan(0.5deg) + height of radar tower? Could that explain some of your disagreements?

  7. Hello, Chris,

    You are correct about the mileage, thank you. The scale of miles I was using was for 2 rather than the more traditional 1 mile and I didn't notice. Appreciate the correction.

    Unfortunately one of the links you provided does not work. I do not see any radar data, so I cannot comment on that.

    What I CAN comment on is that pattern (tornado within the mesocyclone) looks nearly identical to some of Bluestein's cases. So, the pattern looks reasonable.

    I think it is interesting that you contend there was no tornado when I posted a photo of the tornado taken within one minute of the time of the radar data's time stamp. Could it have partially lifted? Sure. I might technically be talking about a funnel cloud.

    Could the damage survey (conducted by ground, not air) have missed the tornado moving across otherwise empty wheat fields? Yes, to that also.

    Given the photographic evidence, combined with the typical radar pattern, and the fact that unfolding the red pixel WNW of the center yields 150 mph I believe it was a tornado or funnel cloud. That would equate to an F-3.

    Interestingly, F-3 was intensity level of the Wichita tornado -- which was spawned by this same supercell -- so that wind speed value seems reasonable.

  8. Mike, it is past my normal bed time, so I apologize if my spelling and grammar are off.

    My thought on the photograph is that the incorrect time & location may have been provided. Surveys and interviews have often revealed that chasers can be grossly unaware of their location and the time. Let's also not forget about the amount of fake photographs that are showing up for various events.

    I plotted one image without data and only the markers so you could see how the quarter-mile squares lay out without any data on top of them to establish point of reference. To eliminate the potential browser issue with PNG images I have uploaded three JPGs at:




    It is possible that a survey team may have missed damage in a field, but I would suspect that they would have noticed something like a power pole being knocked down or fences that were torn up when the suspected tornado crossed from field to field. Survey teams are told to cross back and forth along the suspected path to eliminate missed damage indicators.

    The contention of no tornado is based off of the plots that were generated by the local WFO’s survey. If you believe they are incorrect in their assessment, then you should take it up with them.

    Your contention about the pixel registering at 150 mph might be generous. We have no way to verify that assertion since spectral data is not provided in the base data stream for the TDWRs. However, based off of the more friendly S-band counterparts and what I’ve seen from klystron and TWT-based C-band datasets it is most likely a pixel that is all white noise, so that value cannot be trusted.

    As for the shape and character of the mesocyclone if I looked through enough cases I could easily come back with several mesocyclones that looked very similar but were also non-tornadic. If I recall, only around 10-20% of mesocyclones get to the point of possibly undergoing tornadogenesis.

    One pixel is also not resolving the tornado and gets back to an initial point in that most scenarios where the beam centerline is low enough to clip the vortex it is often entirely encompassed in one or two bins which does not yield a reliable measurement. If the Nyquist of the returned is exceeded but with a high level of uncertainty there is almost no way to recover that data thus making it unusable.

    Of course let’s not forget that there may be other non-meteorological scatterers around that can make returns read odd values. I’ve seen flocks of water fowl clock in at 70 mph. There is also the potential for super-refraction and ground-based echoes.

    If you want to fully resolve that funnel cloud the way you are referencing that would be one gigantic funnel—1,000+ ft wide? More likely you are registering some of the winds associated with the wall cloud or parent mesocyclone.

    You are also referencing the damage level. In published surveys they only reference the peak intensity. That is not representative of the entire path. At the last damage indicator it was only EF-1. Your assumption of a constant 150 mph wind in the tornado and mesocyclone constitutes a fallacy in this matter.

    Also keep in mind that there are two damage paths. The one with EF-3 damage is on the opposite side of SR-49. This would suggest a completely different parent supercell.


  9. Chris, I know the chaser involved and am certain the photo was not a fake. I have no way knowing the accuracy of the time he provided.

    I was able to look at the radar data, thank you. However, are we looking at the same data? Mine was +1.0° tilt. Is that what you are showing? I ask because at first glance it does not appear to be the same.

    It has been a VERY long few days and will take this up again tomorrow.

  10. Mike, I'm looking at the 0.5 degree tilt which would show a vortex on the ground if there was one to see.

  11. Hi Chris, your 5:04am comment clears up our disconnect, I think. I was not and am not contending that the TDWR was showing ground-level winds. I realize that would have been impossible.

    I am contending that it showed the tornado's (or funnel cloud) winds within the mesocyclone. In other words it is showing the winds of the "vertical column of rotating air in contact with the ground" (definition of tornado) several hundred feet above the ground.

    Does that clear it up? Thanks for all of your comments.

  12. Mike, it really doesn't clear up anything. I'll still contend that there was no tornado to sample at this time given the survey results and the given time of the data. To claim otherwise is nothing more than speculation as opposed to a verifiable fact or theory.

    If the photo of the tornado is examined closely, one can make out what might be considered to be a debris cloud. This would suggest the photograph was taken earlier than the stated time. If you have the original, the metadata embedded in the JPG might give the time listed on the camera. However there is the real possibility of the time not being properly set in the camera as well.

    Another reason I think the time is incorrectly listed is the amount of light in the lens. Sunset on April 14th in Wichita was 8:05 PM local time. If this was a full hour and 42 minutes after that, I would suspect it would be pretty much pitch black. You can find the sunset times in the daily climate summaries on this page: http://www.nws.noaa.gov/climate/index.php?wfo=ict

    An additional point I'll make on that data is that you have the core of the supercell right over the radar. It's likely that the signal is being highly attenuated on both send and receive. That would potentially explain errant pixels resulting from poor signal-to-noise ratios and possibly why you think that the radar may be sampling a funnel/tornado--I'm unclear on what you're referencing at this point since you seem to be going back and forth on this.

    I've maintained from the start that the radar was sampling the mesocyclone. At this point in time we're debating I think that it's been established that there was no tornado on the ground to sample. If you are referencing smaller vortices within the parent meso, I believe those have only been fully resolved with X, K or W-band radar within the actual wall clouds or storm base. C-band would likely be unable to fully resolve those given the smallest resolution I've seen accurately produced is 60 meter bins.

    One final point I’ll make is that your title is incorrect as well. Radars do not directly sample anything. They are covered under the umbrella of remote sensing.

    At this point I think I’ve proven facts regarding the sampling of anything other than a mesocylone being a pretty big stretch. I’m going to leave the conversation as is since I feel the debate has been exhausted.


  13. 88D's can have volume scan times less than 4.1 minutes, if AVSET is active, thus the upper scans are not needed (returns not detected). Nothing major, and still not as fast as TDWR's

  14. Chris,

    The photo is a still frame shot from our video of the storm illuminated by lightning. It was pitch black during the storm. We followed this storm on Anson Rd from 160 and witnessed this tornado form. NW of 160. We also witnessed the larger bigger tornado to the west of this tornado's location. We did not have a live stream of this tornado due to using band width on our radar feed. I believe Mike's radar shot was taken right as the tornado was lifting just north of Conway Springs Road. The timing of photo and the radar may not be sync'd at the same exact moment. I think number one point remains that a significant event was captured on radar as well as in the field and reported live on Wichita Radio to warn folks in the Clearwater; Haysville, and Wichita areas 10-30 minutes before striking.

    Our number 1 goal is to warn the public and save lives.

    Drew Pelz

  15. A close-up of a tornado from a mobile Doppler radar is here: /www.easternuswx.com/bb/index.php?/topic/170300-mobile-doppler-radars/ Unfortunately, I don't believe images can be posted in comments.

    The left image is velocity and it is virtually the same pattern (wider mesocyclone, tornado structure in the middle, weak echo in the center of the tornado) as depicted in the TDWR above.


Note: Only a member of this blog may post a comment.