Radar tornado detection in tropical systems

[Matthew E. Engelbrecht]

In watching the plethora of tornado warnings coming out of Louisiana, Mississippi and Alabama this morning with landfalling hurricane Gustav, I have been attempting to correlate what radar signatures are triggering the warnings where only radar indicating a tornado is mentioned in the warning text. Using GRLevel3 I have yet to find any velocity couplets either in BV or SRV in the 2 or 3 volume scans immediately preceding the issuance of the warning. On the reflectivity side, it seems that cores >=50 dBz that extend up to about 5K ft. and persistent supercell-like structures like those I show below are present immediately prior to and during the warnings.

Can anyone confirm/deny these observations and possibly shed some light on the guidance used by operational forecasters for issuing tornado warnings in tropical systems?

 

[GPhillips]

Level 3 data is going to have problems depicting the rotation in many of these situations. Besides the resolution issue, the level 3 SRM uses the average storm motion for all identifiable cells from the previous volume scan. This can be quite far off when you have storms moving in many directions, such as near a tropical cyclone.

Below are comparison images for Level 2 and Level 3 from 1706Z from the New Orleans 88D. The average storm motion used for level 3 was 153 at 49 kts. Tracking the tornado warned cells in level 2, I got 137 at 49. Only 16 degrees of difference, but along with the resolution difference, you can see the outbounds in the couplets in the level 2 image west of Woolmarket and near Biloxi, whereas you can't in the level 3 image. Even level 2 doesn't look impressive, but then it usually won't at 40 to 50 nm in these tropical situations.

 

[Matthew E. Engelbrecht]

I see. I had a feeling that level 3 vs. level 2 might end up being the difference and your example illustrates it quite well. As you pointed out, the couplets don't seem impressive but at least they exist in the data, and that helps me understand why I am not seeing them

Thanks!!!

 

[Patrick Marsh]

You can also use spectrum width. In a hurricane the winds should (key word is should) have relatively low spectrum width values. If you have an area of high SW then you would theoretically have an increase in the variation of wind speed over the size of the bin...which would be the case if there was rotation.

The question then becomes identifying regions where high SW is associated with possible tornadoes and where it is from something else. There is a paper or conference preprint on using SW in hurricanes for tornado identification, but I can't remember the title...

 

[Glen Romine]

There is a paper or conference preprint on using SW in hurricanes for tornado identification, but I can't remember the title...

One I'm familiar with is Spratt et al. (1997).