Doppler measurement of tornado wind speeds

[JP Santiago]

This thread discussing the tornado damage survey of the recent storms in the DFW area (http://www.stormtrack.org/forum/index.php?showtopic=11278&st=0) has a very interesting discussion about the Fujita Scale and the evolution of its use to today. It got me wondering if it was possible for the current NEXRAD radars to estimate tornado wind speeds? I know this can be done by the mobile units, but is such a capability possible with the main NEXRAD units or is their resolution too coarse to make that measurement?

 

[Jeff Snyder]

The resolution is the main problem with using 88D velocity data to estimate (note that all radar velocity data are estimates given assumptions and averaging that occurs during the signal processing phase), though the fact that the 0.5degree elevation scan will be quite a way above ground is important too. Otherwise, with 250m radial and 1 degree azimuthal resolution, you'd need any potential tornado to be either very large or very near the radar to get any actual velocity data from the tornado itself (as compared to the mesocyclone). Even if you do have a large tornado, the beam will be at least a hundred feet above ground, and that's if the tornado within 2 miles of the radar.

The better bet is mobile radar data. The DOW radars are all X-band radars and have the capability to directly capture the tornad. The UMass radars are X-band and W-band, so they too are used to directly sense tornadoes and other small-scale phenomena.

 

[Aaron Kennedy]

Of course then the issue is are you measuring the tornado itself or the debris?

 

[Jeff Snyder]

Of course then the issue is are you measuring the tornado itself or the debris?
[/b]

Ah yes, forgot about centrifuging. I suppose that'd be third in line, behind resolution issues and beam height. Regardless, that is an important thing to note... I'd think that a tornado with 300mph winds wouldn't necessarily loft debris that moves at 300mph. Well, perhaps the leaves and papers do, but large objects (those which return large amounts of energy) would likely move much slower than the mean winds.

 

[JP Santiago]

Thanks for the reply, Jeff. Has there ever been any study to try and correlate 88D radar features/data with tornado wind speeds? Basically has anyone looked at seeing if particular features or data easily meaured with the 88D correlate with a given level of tornadic wind speed?

I don't know if that sounded clear, but here's an analogy from my day job, medicine- for diabetics we have a blood test called hemoglobin A1C that gives us the general trend of a person's blood sugar levels for approximately the past 3 months. While it can't give us precise information on what a person's blood sugar does on a day to day basis, there is a linear relationship with the levels of HgbA1C and a person's average daily blood sugar.

 

[rdale]

"Basically has anyone looked at seeing if particular features or data easily meaured with the 88D correlate with a given level of tornadic wind speed?"

Yes. But due to all the problems Jeff laid out - it cannot be done.

 

[Jeff Snyder]

It should be noted that there are developing techniques to try to extract more information from doppler radar data (IIRC, much of it involves the use of Level-I NEXRAD data). For example, the Tornado Spectral Signature is currently being investigated by some researchers as OU and NSSL. We talked about it last semester in my radar class, but again IIRC, it essentially involves looking at spectrum power density for the valid +/- Nyquist range... If the returned power spectrum is high and is nearly equal across the Nyquist velocity range, that signifies that there is multiple velocity folding occurring with that range bin. This is just one technique being investigated, but there are also techniques that attempt to increase the resolution of the radar data (decreasing dwell time allows one to oversample azimuthally, etc).

As for correlating 88D data with actual tornado wind speeds... Not really that I know of. We have a hard enough time determining tornado wind speeds to begin with, so any comparative study of tornado winds and 88D would be difficult I would think. The know Dr. Wurman has worked with the DOW data in regards to tornado wind speed (and F-scale rating and damage production, which I think he did on the Spencer SD tornado). Again, however mobile doppler radars have a massive benefit in this regard as compared to the 88Ds.

 

[Robert Edmonds]

It should be noted that there are developing techniques to try to extract more information from doppler radar data (IIRC, much of it involves the use of Level-I NEXRAD data). For example, the Tornado Spectral Signature is currently being investigated by some researchers as OU and NSSL. We talked about it last semester in my radar class, but again IIRC, it essentially involves looking at spectrum power density for the valid +/- Nyquist range... If the returned power spectrum is high and is nearly equal across the Nyquist velocity range, that signifies that there is multiple velocity folding occurring with that range bin. This is just one technique being investigated, but there are also techniques that attempt to increase the resolution of the radar data (decreasing dwell time allows one to oversample azimuthally, etc).

As for correlating 88D data with actual tornado wind speeds... Not really that I know of. We have a hard enough time determining tornado wind speeds to begin with, so any comparative study of tornado winds and 88D would be difficult I would think. The know Dr. Wurman has worked with the DOW data in regards to tornado wind speed (and F-scale rating and damage production, which I think he did on the Spencer SD tornado). Again, however mobile doppler radars have a massive benefit in this regard as compared to the 88Ds.
[/b]

I had been thinking a little bit about what has been mentioned above before. I'm sure there is probably more information that we could be getting if there was more spectral analyzing. When at the storm chasers convention there was a mention about how generally the range of water droplet size is unknown in a storm. I was thinking though, there should be more or less interference with certain frequencies depending on the water droplet size range. I'm not sure though because it depends on the amount of the internal reflection. Either way interesting ideas that should be looked into. Perhaps they should talk to an astronomer, they get tons of info off spectral information.

 

[Jeff Snyder]

There is quite a bit of ongoing research in the area of power spectra and time-series data (i.e. Level-I data). The ORDA upgrade (including the Sigmet RVP8 signal processor) opens the door for more research and improvement, from enhanced clutter suppression (better filtering, etc) to "super-resolution".

Here's a little presentation on spectra processing (use of Level-I / I(t) and Q(t) data) --> Tornado spectral signature observed by WSR-88D (Yu et al.)

There's quite a bit of info out there associated with 88D super-resolution, TSS, and other relatively new research...

Tornado spectral signature observed by WSR-88D (Yu et al.)
RADAR OPERATIONS CENTER (ROC) EVALUATION OF PROPOSED SUPER RESOLUTION TECHNIQUES FOR THE WSR-88D (Warde et al.)
Technique for Improving Detection of WSR-88D Mesocyclone Signatures by Increasing Angular Sampling (Wood et al.)
Adaptive Signal Processing for WTC Mitigation (Concerning the wind turbine clutter near the KDDC radar)

Yu, T.-Y., G. Zhang, A. Chalamalasetti, R. J. Doviak, and D. Zrnic, 2006: Resolution Enhancement Technique Using Range Oversampling, J. Atmos. Oceanic Technol, in press.
Brown, R. A., B. Flickinger1, D. Schultz, P. Spencer, V. Wood, and C. Ziegler, 2003: Experimental High-Resolution WSR-88D Measurements In Severe Storms. NSSL Final Report FY04