Recent e-mail discussion on HSLC WFO Operations

A QLCS severe weather event occurred on April 11th and prompted an interesting discussion on the utility (futility?) of issuing tornado warnings for QLCS events. The discussion highlights some of the experiences and concerns of meteorologists participating in a CSTAR project examining HSLC convection. (FYI, I have done some minor editing to protect the innocent)

Andy Kula (HUN): “As some of you are aware,  we had another QLCS event in the southeast last night.  This time, with no initial tornado reports.  However, we are investigating a couple of meso-vortices that had damage paths that may need surveys in our CWA.  Other offices may be doing the same.

James Spann made a very interesting post on that may serve as additional motivation for our HSLC study.

I’m in agreement that we are creating some significant complanceny with the shear number of Tornado Warnings (and watches) in these situations. Perhaps this topic can serve as a discussion point on how we move forward with our research.”

Pat Moore (GSP)- “…it’s a moot point right now.  In my opinion, our hands are tied… Tornado probability of detection and warning lead time are part of our GPRA goals.  (Matt P…GPRA stands for Government Performance and Results Act)  In this part of the country, I speculate that at least 25% of our tornadoes are of the non-supercell variety.  In the GSP CWA, we average about 11 tornadoes a year, 2 of which are produced by a QLCS.  I haven’t pressed this issue because our tornado climatology only goes back to 1995 so it might not be statistically significant, and I also haven’t considered the last year or two (Justin would have better statistics).  Anyway, if we just give up on 2 out of 11 tornadoes a year, that’s 18% of our possible warning hits.  That means we have to hit every single other tornado event in order to reach the GPRA goal.  We can’t afford to miss even a single other tornado.  8 out of 11 is a POD of only 72%.  Failure!  …WE HAVE TO TRY to warn for these weak tornadoes, or else we probably will not meet our goal.  (Aside…how nice it must be for Srn Plains offices to get so many supercell tornadoes that they might ignore the weak spin-ups (he says, sarcastically).)

Either way, NWS offices need to be consistent with their strategy.  We can’t have one office putting out lots of tornado warnings trying to detect all the weak spin-ups, while a neighboring office not putting out any warnings because they have decided they are not going to go after the spin-ups.”

Justin Lane (GSP)- “I agree with Andy completely. I always use the following example…

A 1-mile track EF0 associated with a low-topped QLCS knocks down a few trees and blows the roof off a barn in a rural area 20 miles outside of Charlotte. You issued a successful TOR with 12 minutes of lead time. Your POD is 1.0, FAR is 0, and average lead time is 12.5 minutes. According to the stats, you did a great job, all within national GPRA goals. However, what if that warning included much of the greater Charlotte metro area? You just sounded the alarm for more than a million people for a minor event that basically affected one person. Furthermore, you most likely issued the TOR with relatively low confidence. However, confidence was very high that IF a tornado developed, it would be weak and short-lived. Have you really performed good public service? What if you issued an SVR with a call-to-action stating “A BRIEF WEAK TORNADO MAY ALSO DEVELOP WITHIN THE WARNING AREA.”

On the other hand, Pat is also correct. We are beholden to the rule that we are expected to issue successful TORs for all tornadoes, weak or no. As Pat said, GSP’s TOR stats fluctuate from year-to-year based upon how many supercell days we get. Some years we get a couple of days with tornadic supercells, and our stats are great. Some years we get none, our stats are awful…

I look at this as a verification issue, not a warning decision making issue. If we feel a tornado is likely, we should issue a TOR. However, if we think a tornado is possible, but not likely, and that IF one develops it will be weak, we should be able to issue an SVR without being penalized if that EF0 or EF1 develops. I would really like to see a grassroots effort from the field to push for this verification change. In the interim, we should continue to search for a scientific solution to this problem.”

Moore- “…if only we could have the option of issuing a Severe Thunderstorm Warning with possible brief tornado wording and then not take the verification hit IF the weak tornado actually happens.  Making that change would probably make our work easier and less stressful.

The work that we are doing could be part of the foundation of a grass-roots effort to change the way we verify these events.  We always say these brief tornadoes are difficult to warn for, but has anyone ever tried to quantify the limit of predictability for them?  Perhaps one of the outcomes of this project (from the NWS side) could be something that could be added to the body of work that shows how our customers might be better served by making the change.”

Trisha Palmer (FFC)- “In theory I like the possibility of SVR with “a weak tornado possible” wording – we have SWA (SPS) text here that we’re testing out for that very case:




However, this all assumes that the tornado would, indeed, be “weak”.  In my spreadsheet of 40 potential HSLC cases spanning just over 5 years, I have 12 EF3s and 24 EF2s (if I counted right) – a lot of tornadoes that are definitely not “weak”.  With these tornadoes, around 5 of each may be more supercellular than QLCS in nature, with slightly higher CAPE, but that still leaves 7 EF3s and 19 EF2s that likely came from QLCSs.

As a warning forecaster, I haven’t yet found a way to distinguish an EF3 from an EF0 on radar – I’m sure we’ve all seen scary velocities that ended up being nothing, and then a weak meso that ended up putting down a tornado.  Unless I can be absolutely sure that it’ll be a weak tornado, I can’t justify NOT issuing the TOR.”

Steve Keighton (RNK)- “I tend to agree with Trisha. The science and radar resolution just does not give us the confidence in almost all situations to distinguish between a “weak” tornado or otherwise.  While for us EF2+ are quite rare, I know at least several of the EF2s have come with subtle signatures, partly due to range from the radar. Most of our EF2+ tornadoes have been with fairly obvious supercells, but not all.”

Lane-   “Absolutely. As I said…if we believe a tornado is likely, we should issue a TOR, case closed. However, I don’t believe we should be penalized for not issuing a TOR for the weak tornado that develops in association with subtle, rapidly developing signatures (i.e., our “broken-S” events). I believe that there’s way too much guess-work that goes into making TOR decisions for these events. In light of the fact that most of the radar signatures in these events develop as the tornado is developing, it’s basically as if we are expected to “warn on forecast.” Again, I believe it’s a verification issue (allowing a SVR to verify EF0/1 tornadoes), not a WDM issue.”

Kula-The response of the public (and the media) to Severe Thunderstorm Warnings appears to be waning the past several years.  I don’t have scientific evidence of this other thant speaking with folks, and watching how they are broadcast.   This would be a question for the WAS*IS folks. 

Have you noticed the term “Severe” is usually not on the TV scroll anyway?   How can people take them seriously.  I’d rather see some additional threat specific warnings as opposed to the generalized severe thunderstorm.  I.E.  extreme winds, hail, weak tornadoes.  Then leave the Tornado Warning and/or Emergency for EF2 or higher?  Just throwing that out for consumption.

With the advent of social media, hopefully we can improve severe thunderstorm response.  The hail diameter change, albeit a good thing to reduce numbers of warnings in some cases, doesn’t help with the response.  So, if we were to emphasize weak tornadoes in SVR’s, I think we have to rethink all of our products, naming, size, CTAs.  I’d argue that many storms that knock trees down are not “severe” by the wind speed definition.  We issue SWA’s for sub-severe activity, but not sure of their value either.”

Lane (inserted comment)- Again, I’m in agreement with Andy. As someone who does a lot of storm surveys and peruses a lot of news articles for verification purposes, I’m to the point that my head is going to explode the next time I hear, “there was only a severe thunderstorm warning in effect (so I took no action),” or “I don’t care what anyone says. It was obviously a tornado” every time someone has some trees uprooted in their yard. (Heard this A LOT with the April 4-5 event). It’s difficult to say if this is a communication failure on our part, or if the public just isn’t paying attention (probably a little of both). One thing I am sure of: our wordy warnings can’t be helping matters. Technological gadgets have transformed our society into one of instant gratification. If you can’t give most people important info in a 3-5 second sound byte, they don’t think it’s worth hearing. (I recently had media training at the IMET workshop. So, I’m not completely making that up). In addition, more and more people are going to be receiving warning messages on their mobile devices as a preference, rendering wordy warnings impractical. As an example, why do we bother putting wind speeds in our warnings? That means absolutely nothing to most people in the public. They want to know the where, when, and the what (meaning the IMPACT, not the numbers). In my opinion, a warning for the April 4-5 event worded in the following way… SEVERE THUNDERSTORM WARNING FOR…GREENVILLE COUNTY…INCLUDING THE CITIES OF GREENVILLE…GREER…TAYLORS… FOR HURRICANE FORCE WINDS CAUSING WIDESPREAD TREE DAMAGE AND  STRUCTURAL DAMAGE TO OUTBUILDINGS AND MOBILE HOMES…UNTIL 2 AM. (End!) would probably illicit a much better response from the public than the eloquently written but very wordy products we currently employ. Now that I’ve completely hijacked the discussion…

Steve Nelson (FFC)- Am interested in other WFOs’ list of cases.  Are you all getting relative frequencies (for the obvious QLCS tornado cases) of the EF ratings as we are?  Trapp et al 2005 found relative frequencies of EF ratings from all 1998-2000 tornadoes as shown…

Notice there is a slight higher relative frequency of QLCS tornadoes compared with supercell at EF-1 and does not go below supercells until EF-3 and 4.

Also, to add to Trisha’s last point about radar estimation of tornado intensity, see the slide comparing the two tornadoes from 30 November 2010 that I presented at our media workshop a few weeks ago.  The high-probability tornado signature on the left velocity image only yielded an EF-0 tornado, the worst of the damage is in the right photo. The lower-probability tornado signature on the right velocity image produced the EF-2 Buford tornado (left photo) that was as close as you could come to a major disaster–3pm on a weekday in a high-density county, barely missed a high school, two elementary schools, and a school bus full of kids on their way home, no warning issued (SVR or TOR), yet no injuries or fatalities.

Kula- Glad you brought that slide to our attention again Steve.  I’ve seen very few QLCS tornadoes exhibit classic “gate to gate” velocity couplets.  Usually it is very close to the radar, and even then, not always.  We had a case from December ’09 within about 15 miles of KHTX (Hytop AL) produce an EF2 without gate to gate.  The circulations were pretty obvious, but still…

Hunter Coleman (CAE)- My opinion is similar to others that have spoken in that for the weak tornado spin-ups possible within a QLCS line that are usually short lived and low confidence, I like the idea of going with a severe and including wording about possible weak tornado in a CTA or in the text and when you have high confidence in a tornado then issue the TOR regardless if you expect it to be weak or not.  Also, as it has been mentioned, some of these spin ups occur on such a short time frame and last for only a few minutes it is possible that it may occur between volume scans since our fastest VCP is 4 minutes, so we may never see it in the data (I understand that this circumstance  is not all that likely).  I think if the verification measures could be adjusted as suggested in this thread, that may help the warning forecasters with these types of events.  Wouldn’t it be nice to have some CASA radar data (1 minute resolution) for some of these QLCS events!

Steve Zubrick (LWX)- Radar detection of subtle tornado signatures associated with QLCS varies.

Here at LWX, in addition to the 88D, we have access in AWIPS in real-time to >>4<< FAA TDWRs. Each TDWR offers 1-minute resolution at the lowest scan (and 3 min at some of the upper scans).

Having 5 radars to examine keeps one on their toes during QLCS environments!

Last week, we issued 5 TORs during a pre-dawn QLCS (and HSLC) event on April 5, 2011. Initially, only one of the 5 TORs verified (with a tenth of a mile long!). Damage was very spotty and mostly minor.

Examination of available TDWR data showed small “spins”; i.e., meso-vorticies, along the QLCS line. Here’s one example of a base vel/ref image showing meso-vorticies (the one just west of Waldorf (lower center) produced the one-tenth of a mi EF-0 shortly after this image was taken). Also note a meso spin just west of Clinton (near the word “Prince Georges” upper center; this spin did not produce any evidence of tornado-like damage…but did blow off shingles and siding on 3 residences):

I spent all day on Monday (Apr 11, 2011) driving around searching for damage in areas where no damage was reported by EMs/Spotters. I surveyed along the tracks of 3 meso-vortexes plotted from several FAA Terminal Doppler radars…Based upon my survey, I found minor damage in spots along each meso vortex track.

Matt Parker (NCSU)- This very topic came up (spurred by the same event) on another severe storms discussion list to which I belong.  Some themes that emerged from the other list include:

-Some people strongly suspect that the number of *non*-tornadic QLCS circulations that occur in a given year is considerably higher than we have previously realized. We are detecting these circulations much more readily now that TDWR and super-resolution WSR-88D are available in real-time operations. One participant who monitors warning performance on a national scale asserts that it’s becoming quite predictable that when a QLCS approaches an 88D or TDWR site, the WFO will usually begin blanket TORs as the circulations are detected, and then the warnings will fade back to SVR as the QLCS recedes from the radar (simply due to the lack of resolution of the small-scale vortices).

-To follow up this thought, one participant suggested that, at the heart of the matter, we have a conceptual model problem. Such a problem arises because we had ~12 years of non-super-resolution 4-bit 88D field experience, so that if a QLCS circulation was detected, it necessarily had to have grown big and strong enough to be sampled. Now with 8-bit data resolution, super-resolution and TDWR we are much more efficient at detecting much smaller and weaker circulations along lines, yet we are still applying the conceptual model that these are just as likely to produce EF2 or EF3 tornadoes as their bigger and stronger counterparts we had earlier experience detecting.

-Some people are skeptical about whether or not what many of us would conventionally call “QLCS tornadoes” are not actually predominantly associated with embedded supercells.  It seems to me that we have some pretty clear-cut evidence of non-supercellular QLCS tornadoes, but it may be a question of the proportion of embedded supercells that are tornadic vs. the proportion of “other” mesovortices that are tornadic.  At any rate, it highlights the need to know more about the distribution of tornadic and non-tornadic vortices (both supercellular and other).

-As I understand it, false alarm rate is also a GPRA metric, so clearly we’re not just talking about the issue of missing the GPRA POD goal.  It’s maybe more pointed to say it’s the perceived asymmetric penalty function of missing tornadoes vs. issuing false alarms.  As I’ve said before, as a NOAA outsider this is an issue that I personally don’t want to touch with a 10 foot pole.

I am largely relaying what others have said here.  But, it’s been healthy trying to synthesize what I am hearing/learning from this group with what the other group is talking about.  Clearly, this is a problem whose time has come.  *Both* groups seem to be iterating toward the idea of issuing some kind of SVR with enhanced wording, so perhaps there is a chance that something will actually happen here.  It would be great if our regional HSLC study can help inform such decisions.

Jonathan Blaes (RAH)- I think Dr Parker is right on with attributing some of the issue to a conceptual model problem and applying these somewhat dated approaches to new data sets with increased spatial and temporal resolution. In addition, our regional goals (ER) include a TOR POD of 70 or greater and a FAR of 72 or less. The message to forecasters is to cover the TOR and that a large FAR is acceptable.

While not a QLCS event, I have a TDWR loop from July of 2008 of unorganized pulse convection in which the TDWR one minute resolution shows numerous down burst signatures in which a WSR-88d would miss. We are seeing some much more than we used to and some of this is a little scary – there is a lot going on outside our windows that we may not have realized or appreciated before.

This entry was posted in General Information, High Shear Low Cape Severe Wx. Bookmark the permalink.

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s