Warnings based on Rotating Storms vs. Actual Tornado Touchdown

[Todd Renshaw]

Satellites are not able to determine tornadoes by any stretch. There are some signals that can be found in storms which imply a tornado possibly, but nothing that can be relied upon.

Of course. But I was (mostly) speaking of future radar technology. Or at least cutting-edge technology that has not been implemented in this way before.

 

[Stan Rose]

* These are my personal opinions and in no way reflect the position of the National Weather Service or NOAA *

I understand this problem very well. One change which will help, and one in which the NWS is heading though not quite there (as evidenced by the warning sirens blowing in Chadron NE twice the other day when the outflow dominant sup was miles and miles away...) is to tighten up the warnings. More frequent updates with a smaller geographical area warned. that will lesson the sense of security for those in the 'box'. Another, which i see as the near future, is that the curse of too many chasers has at least led to the good of nearly ubiquitous live feeds and reports. For the 'big' days, its almost a gaurantee that there will be multiple chasers on a storm (and for the low risk or outside-the-box days, the warning forecaster can be more cautious, but still be conservative). I issued warnings for the Campo May 31 storm last year while i watched the storm develop live in real time. I could see on the internet what could not be seen by radar. No need to guess or determine the trustworthiness of being there. Maybe not as good as being there in person--but its getting close!

 

[Robert S. Thompson]

I believe the NWS was using "tornado emergencies" when the long track tornadoes were on the ground in AL. So maybe "tornado warning" when Doppler indicated or suggested by the public, and "tornado emergency" when confirmed by trained spotters.

Also, at least in my area, smarter sirens would help. We've had a bunch of polygon tornado warnings this season or rural parts of my county (where they have few/no sirens) yet they were sounding for the entire county including the main cities. But that takes money, and the county doesn't have any.

RST

 

[Dave Kaplow]

It seems to me there is one thing well within the realm of possibility that we could attempt to do to improve the situation, and that is to make the TVS algorithm more discriminating. I realize this has almost certainly been looked at already by greater minds than mine, but surely there must be something out there that could be added to the algorithm to help cull the false alarms. Supposing someone did want to take a crack at this, they would need a database of historic WSR-88D data containing many examples of both tornadic and non-tornadic TVS-triggering storms, along with detailed records of the actual mesoscale environment of each cell at the time of the TVS. What I’m imagining here is an effort similar to the Netflix Prize, where the database would be made available to anyone who wanted to take a shot at it. There could be a prize or series of prizes for those who manage to make genuine improvements to the TVS algorithm. Granted, the meteorological community probably couldn‘t come up with a million dollars like Netflix did, but still, I think the crowd-sourcing concept is worth a look. The TVS could definitely stand some improvement, I can’t tell you how many times I’ve seen obviously non-tornadic cells get T-warned by OKX on the basis of (presumably) nothing more than a transient TVS popping up. A little more discrimination in the TVS algorithm would go a long way toward reducing the false alarm rate.

 

[rdale]

The database (L2) and the software (CODE) have been freely available for a long time.

http://www.weather.gov/code88d/

I would be shocked if any professional meteorologist issued a warning or interrupted TV just because of a TVS. The algorithms are 'look at me' flags, not 'issue a warning for me' flag. Anyone using radar alone to issue a warning needs to be taken off the front lines and retrained in the first place...

 

[Dave Kaplow]

The database (L2) and the software (CODE) have been freely available for a long time.

http://www.weather.gov/code88d/

I would be shocked if any professional meteorologist issued a warning or interrupted TV just because of a TVS. The algorithms are 'look at me' flags, not 'issue a warning for me' flag. Anyone using radar alone to issue a warning needs to be taken off the front lines and retrained in the first place...

Yeah, I know that’s how it’s supposed to work in theory… I think that in practice, however, especially in areas that are outside the traditional tornado alley(s) and that have very high population density, it’s possible that the TVS sometimes gets a little more respect than it perhaps deserves. But of course I don’t know what’s actually behind any given warning, I guess I’m just assuming the TVS sometimes plays a big role. That’s interesting that the database is already available, but I guess I should have anticipated that. Now all we need is for someone to put up a million dollars!

 

[Jake Orosi]

I find it humorous that people are complaining when tornado warnings are issued for storms that are "simply rotating"; considering that a rotating storm is the best indicator that the initial stages of tornado formation are in progress.

Perhaps it's just as simple as keeping the tornado warnings for storms where radar actually shows a debris ball, or trained spotters have reported a tornado, and just t-storm warning a storm that's just rotating or shows a TVS. Let's face it, a rotating storm will probably be creating some kind of severe weather. The only problem is, as lackadaisical as people are becoming when it comes to reacting to a tornado warning, the apathy in response to a "severe thunderstorm warning" is already pretty much complete. I don't think I know a single non-meteorologically-inclined person who thinks a severe thunderstorm warning indicates or even suggests any kind of danger. The NWS may as well call it a "sorta scary storm alert".

 

[rdale]

keeping the tornado warnings for storms where radar actually shows a debris ball, or trained spotters have reported a tornado, and just t-storm warning a storm that's just rotating or shows a TVS.

First off - "debris ball" is WAY overused. Until you are looking with dual-pole, you can't be sure that it's debris.

But the problem with your theory is that 1) areas with low spotter coverage would never get tornado warnings 2) many tornadoes are short lived, as in 10 minutes or less. Considering it would take a minimum of 3-5 minutes to get a report in and the warning sent back out, by the time you're sending it the damage is pretty much done.

 

[twarner]

2) many tornadoes are short lived, as in 10 minutes or less. Considering it would take a minimum of 3-5 minutes to get a report in and the warning sent back out, by the time you're sending it the damage is pretty much done.

The timing here illistrates a limitation of the NWS radar. Potentially by the time they see the possible "tornado signature" and issue a warning the event may be over due to the scan times. Why doesn't the NWS change the software code to scan more frequently and get quicker updates? Or at least have a mode to switch to when you have a situation where more frequent updates would be useful? Is it a data infrastructure issue?

 

[rdale]

Why doesn't the NWS change the software code to scan more frequently and get quicker updates?

Because then you lose the volume scanning ability. The dome can only swing so fast. PAR will help but that's at least a decade away.

Or at least have a mode to switch to when you have a situation where more frequent updates would be useful?

AVSET will help in some of those cases, but again you can't do fast and volume. Either slow with a full volume scan, or fast with just the first tilt. Which is why many TV stations put their live radar online so the NWS can use them in conjunction.

 

[Jake Orosi]

First off - "debris ball" is WAY overused. Until you are looking with dual-pole, you can't be sure that it's debris.

But the problem with your theory is that 1) areas with low spotter coverage would never get tornado warnings 2) many tornadoes are short lived, as in 10 minutes or less. Considering it would take a minimum of 3-5 minutes to get a report in and the warning sent back out, by the time you're sending it the damage is pretty much done.

Oh don't get me wrong, I think the current system is fine and that the reasons people give for choosing to ignore tornado warnings that they receive are ridiculous, and that the "problem" isn't warning procedures or radar detection, but peoples' cavalier attitudes toward weather and refusal to be proactive about it. I guess I was just suggesting that changing the system in a way that will address those peoples' "complaints" can only result in a less-efficient system and more deaths and injuries.

 

[Greg Stumpf]

Why doesn't the NWS change the software code to scan more frequently and get quicker updates?

The faster you rotate the antenna, the more beam smearing will occur. This can be compensated by accepting fewer pulses per sample volume, but then you will increase the noise in the data. There have been attempts to overcome this via oversampling: http://bit.ly/mgkFhk

Or at least have a mode to switch to when you have a situation where more frequent updates would be useful? Is it a data infrastructure issue?

The WSR-88Ds can operate in different modes that can be machine-selectable or manually overridden.

One other way to speed up volume scans is to reduce the number of vertical tilts. Since the data nearest the ground has more value in warning decision making, interlacing additional 0.5 degree elevation scans in the volume scan also helps - this is what TDWR radars do now.

Or, have an electronically-steered beam like a phased-array radar (PAR).

 

[KrissyHeishman]

Just chiming in here with a few ideas..

1) NWS has a Text alert system. (http://inws.wrh.noaa.gov/) To date, I don't know anyone that doesn't have a cell or is not with someone with a cell at all times. So that should help solve the problem of getting the message out there. Similarly, there could easily be a system where social network users are notified via those systems by the NWS, in case they do not want to receive texts. Information for this system could really be getting out more than it is already. Public service announcements are FREE to put on air, should they be accepted by the stations/networks. It's actually one way indy/novice videographers get their start.

Problems I see with this.. Cell/data tower overload, loss of power, etc. Sadly, this system sounds great, in theory, but I'm just not so sure about how reliable it would be in a severe weather situation. Still, I think it's worth looking into because, with today's rapidly developing technological society, I'm sure we can come up with some sort of fix/workaround. Maybe something satellite based? I dunno.

2) I'm not too familiar with how the weather sirens work, and if they all work the same or they are region/county/city specific, but it would be greatly beneficial to have shorter warning bursts to indicate the need to be aware of potentially incoming severe weather. I know in Dallas, where I grew up, they would do something like this from time to time, but the so-called "short" bursts seemed to last almost a minute. Would it be possible to have a series of say, 3, really shorty bursts?? Almost like Morse code, or the NWS Alerts for the radio. If this is not possible, perhaps installing a new siren-type system on the same siren-poles that would do the trick?

So, maybe you can have 3-short burts for "hey, turn on the tv, radio or get online/noaa", a series of 30-second or 1-min bursts for "tornado warning in your area, possible tornado, prepare shelters "& a long, consistent sirent for "verified touchdown. seek shelter immediately. meteorologists have predicted you are in the path of a tornado".

Problems I see with this, however are.... Liability, potential for inaccurate forecasts/warnings & desensitization People are desensitized to sirens these days in the plains, valley, etc. but creating such a detailed system for warnings may have a similar effect. People may not heed a "possible tornado" warning when, in fact, there was a tornado, just not spotted. or, shortly after this warning was made, a tornado winds up and touches down on someones house immediately after, yet before the sustained siren went off & thus creating a minimal, yet potentially dangerous window where there is simply no way to know for sure if you are in danger or not in between the warning "steps" or w/e..

Additionally when the sirens are going off, you are hearing more than one siren in your area. If there are sirens signalling with intermittent series, it would prolly sound like a jumbled mess. I'm sure, however, that with today's technology, audio engineers can figure out a way around this. Still, a roadbump, nonetheless.

3) Well, I had a 3rd idear.. but it's totally left me now.. hehe.. I'll post it up tho when I can remember what it was. @_@



And touching the subject with the reporting/spotting accuracy..

I'm not really sure about a fix for this yet, but I do see what could be causing a lot of problems. First, I'd like to say that the twitter/social network/texting ideas sound great but would be potentially undependable based on the same reason for "#1" - cell/data towers & power. Therefore, a threshold-type algorithm would be impractical but, possibly, having forecasters or their associates monitor them manually and make determinations themselves might be more practical.

Sadly, the thing that's best for forecasting has also become the worst, imo - the advent of modern mass-communication technology. It's given us so many routes to receive reports, however, that is also the problem in itself. Just watch any news station during a recent outbreak and you will see them mention all the various sources that they are receiving information from. I think this IS great, don't get me wrong, but it also seems to cause confusion and clutter within the studios and potentially slow-down forecasts.

Not a solution, but just a thought.. If there could be more organization on how spotters & civies send in information, then this might help streamline things, logistically. I know in Dallas, SKYWARN communicates with the meteorologists via Ham Radio, which is (imo) highly effective. However, Hams are few and far between these days. =\

All-In-All, I think there is one thing that just about anyone can do or request/nag/demand be done with little monetary or physical effort, considering the alternatives that will help solve a lot of these issues is and that is... more PSA's, more iniative and proactiveness, and less reactiveness.


OK IM DONE.. I swear.. and sorry for the epically long post. =[

 

[rdale]

1) The NWS alert system is for EM/media/related. But I agree, there are plenty of free systems to do that. Text messaging is MUCH more valuable than NOAA Weather Radio for one big reason -- market penetration. Less than 5% of the public has a NWR. Nobody takes that radio around with them while mobile. Cell phones are a little better at that

2) Sirens are simply made to let people know "bad things could be coming, turn on your TV or check the web and find out why."

I'm not sure what you mean about a lack of organization. There is a very good system for reporting that (SpotterNetwork) and a great system for coordinating with all partners (NWSChat.)

 

[Patrick Marsh]

there is no algorithm that we know of YET. everything in this world is made up of math, and numbers, there is a code, it just hasn't been discovered yet and it may not make sense to us yet. and we may never find it, but I believe that you thinking there is no code, is just your opinion, as my belief that there is a code is my opinion

It's infinitely more complex than just finding the correct equations. Comparatively speaking, that's actually the easy part -- and it is far from "easy".

Just because we can find the equations, doesn't mean we know how to solve them. The atmosphere is governed by the partial-differential equation(s) known as the Navier-Stokes equation along with the other fundamental laws of the universe. The problem here is that we don't know how to solve the Navier-Stokes equation except in very idealized situations. Everything else is an approximate solution. Even though we can get the approximate solution correct out to several decimal places, Lorenz showed that this isn't good enough. What most people would consider "noise" (the part of the solution well right of the decimal point), turns out to have potentially dramatic impacts on the final solution.

Assuming we could solve the Navier-Stokes equation(s), there is still the issue of modeling the continuous world we live in by a series of discrete grid boxes. The whole notion of the "butterfly effect" can be explained as details that occur sub-grid scale (beneath the resolution of the model) can potentially have dramatic effects on the final solution. We already saw this with our global models that cannot resolve convection. We have to parameterize convection and hope that we get the impacts correct. Moving to convection-allowing models has improved the impact of thunderstorms on numerical forecasts, but we still aren't modeling all of the thunderstorm, because most of the thunderstorm still occurs sub-grid scale. As a general rule of thumb, you need a minimum of 4 grid-boxes, and really at least 8 grid-boxes, to be able to resolve a feature in a numerical model. This means to resolve a 2-km thunderstorm, you need a minimum resolution of 500 meters and really a resolution of at least 250 meters.

Just having the computing power doesn't mean that we can correctly model the atmosphere. We then have to have correct initial conditions at the resolution of the model. This means that to model the 2-km thunderstorm, we'll need surface (and upper air!) instrumentation every 250 meters. Can you imagine how dense of observational network would have to be to correctly model even the most intense tornadoes? What about the weaker, and typically smaller, ones?

As challenging as it will be to build the computing resources to improve the precision of the solutions to the equations we already know, getting observations on higher resolution scales is the more difficult to achieve.