Late warning for Springfield

[CT Rain]

What was the wording for the severe thunderstorm warning?...some of the severe warnings will say "there has been some weak rotation detected with this storm, while not immediately imminent, a tornado could still develop at any time" or something like that. Perhaps there was already enhanced warning in the severe storm warning. And who else knows what type of potential technical problem could arise too. But I would highly doubt they were unaware of the tightening couplet in that storm. Unless its a lot different than I remember, the radar will certainly let you know about it.

The SVR was issued at 4:18 and mentioned "weak rotation"

[Analog96]

Surely a program could look for more than just rotation such as the increasing gate-to-gate shear. What we are talking about here is fail safe/worst case scenarios where there is some error or human delay. There could also be a human overide where there is a built in delay so the warning isn't just issued and if a human cancels it obviously it doesnt go out but if its not cancelled its issued after a certain amount of time. That would cut down on false alarms and even if it wasn't as good as human analysis we are talking about lives and better safe than sorry. Just putting it out there.

As someone mentioned in this thread, when a rotating supercell pops up on the radar, there are audibles that can be heard very clearly throughout the NWSFO. I don't really think they need to have any more than what they already have in place now.

[Analog96]

The SVR was issued at 4:18 and mentioned "weak rotation"

At least they did that, but it sure had more than "weak rotation".

[Ginx snewx]

Someone needs to go to Congress and fight for funds to develop an auto text message system sent to all cell towers in a Tor warned area that all carriers would be mandated to send out.

[Analog96]

Someone needs to go to Congress and fight for funds to develop an auto text message system sent to all cell towers in a Tor warned area that all carriers would be mandated to send out.

Meh. People who pay per text message will complain.

[TheTrials]

As someone mentioned in this thread, when a rotating supercell pops up on the radar, there are audibles that can be heard very clearly throughout the NWSFO. I don't really think they need to have any more than what they already have in place now.

Assuming that is true, yes, that would seem like enough, but lets suppose in this situation for whatever reason whoever heard that warning couldn't get the text product out, would be nice to know that there was a fail safe. That's all.

[Analog96]

Assuming that is true, yes, that would seem like enough, but lets suppose in this situation for whatever reason whoever heard that warning couldn't get the text product out, would be nice to know that there was a fail safe. That's all.

I know it's true from interning at an NWSFO over the summer. I don't know how it would not work, but I assume something could always happen.

[TheTrials]

Meh. People who pay per text message will complain.

well, isn't a 911 call free? make it like that.

[Ginx snewx]

Meh. People who pay per text message will complain.

Opt out option, NWS system antiquated.

[TheTrials]

I know it's true from interning at an NWSFO over the summer. I don't know how it would not work, but I assume something could always happen.

Right, Im talkinh about last ditch fail safe. Again, not saying anythign about this situation or what happened or laying blame.

[radarman]

Because, for example, a lot of storms can spin up a low-level rotation signal in a non-convective event, such as a decent-sized Nor'easter. The storm may show "rotation" for a few minutes and not be anything at all. An "automated" warning service would probably put out a tornado warning for no reason in that case.

Respectfully disagree. There are a lot of ways to make your algorithm smart enough to maintain respectable POD & FAR. Much like a forecaster does, nothing precludes an algorithm from bringing in outside sources of information to get a sense of "situational awareness" and to warn accordingly based on Bayesian inferencing and a massive Nexrad historical archive to train on. Not saying humans should be kept out of the loop, but I am saying that with some investment an algorithm could be written that performs on par, or nearly on par, with a forecaster.

[Ginx snewx]

Respectfully disagree. There are a lot of ways to make your algorithm smart enough to maintain respectable POD & FAR. Much like a forecaster does, nothing precludes an algorithm from bringing in outside sources of information to get a sense of "situational awareness" and to warn accordingly based on Bayesian inferencing and a massive Nexrad historical archive to train on. Not saying humans should be kept out of the loop, but I am saying that with some investment an algorithm could be written that performs on par, or nearly on par, with a forecaster.

Yep then the human generates one auto text warning. Simple yet effective, everyone has a cell phone including kids,elderly. KISS, Keep IT Simple Stupid

[Typhoon Tip]

The purpose of this thread and ramfications for impugning NWS is dubiously self-serving to me. I suspect that some of the issue here is really that some folks wanted the melodrama and excitement, and feel cheated because there wasn't blaring sirens and red flashing lights for people to get all manic in fear just that much more. They don't want their extreme weather party cut into - and perception is everything whether they know it or not.. and it is petty.

The same sort of vibe emanates from the chorus in the winter, when snow is handled even slightly off from what is or is not realized relative to a forecast; just waiting to pounce with drool! Wrong attitude to have though. Again, it's more a reflection of weather for entertainment/emotion rather than science, and needing some source to blame when those emotional expectations were not fulfilled. It's all stupid to begin with because no one should be investing that way.

I was on top of those cells yesterday and I thought it went over just fine. The reflectivity channels only hinted at a hooks until a tad later... I am aware of the gated this and that argument, but the velocity channels at KTAN its self did not show that quite as obviously as whatever these alterior product(s) origins are. Unless NWS uses some internal imagery that they don't show the publice - i don't know. Otherwise, from what was shown on KTAN's website's rad page, it was not abundantly clear there was a tornado on the ground - ground truth was key there; hense the appearance of a late dial up. I was watching both cells closely and thought the Springfield one - using KTAN's rad source - was somewhat suspicious looking when they issued a warning, and part of me even thought that might be the other way - trigger happy. I've seen countless "hooks" in the Plains with far better NWS radar velocity channel gates, and nothing more than a funnel cloud reported by storm chasers.

Give it a rest. There is no issue here with how it was handled. I promise, the next one you'll get ALL your party favors in, and not have any of its affects diminished.

[TheTrials]

The purpose of this thread and ramfications for impugning NWS is dubiously self-serving to me. I suspect that some of the issue here is really that some folks wanted the melodrama and excitement, and feel cheated because there wasn't blaring sirens and red flashing lights for people to get all manic in fear just that much more. They don't want their extreme weather party cut into - and perception is everything whether they know it or not.. and it is petty.

The same sort of vibe emanates from the chorus in the winter, when snow is handled even slightly off from what is or is not realized relative to a forecast; just waiting to pounce with drool! Wrong attitude to have though. Again, it's more a reflection of weather for entertainment/emotion rather than science, and needing some source to blame when those emotional expectations were not fulfilled. It's all stupid to begin with because no one should be investing that way.

I was on top of those cells yesterday and I thought it went over just fine. The reflectivity channels only hinted at a hooks until a tad later... I am aware of the gated this and that argument, but the velocity channels at KTAN its self did not show that quite as obviously as whatever these alterior product(s) origins are. Unless NWS uses some internal imagery that they don't show the publice - i don't know. Otherwise, from what was shown on KTAN's website's rad page, it was not abundantly clear there was a tornado on the ground - ground truth was key there; hense the appearance of a late dial up. I was watching both cells closely and thought the Springfield one - using KTAN's rad source - was somewhat suspicious looking when they issued a warning, and part of me even thought that might be the other way - trigger happy. I've seen countless "hooks" in the Plains with far better NWS radar velocity channel gates, and nothing more than a funnel cloud reported by storm chasers.

Give it a rest. There is no issue here with how it was handled. I promise, the next one you'll get ALL your party favors in, and not have any of its affects diminished.

I thought the public data was a different level than what the NWS offices see. Don't they have the same kind of data that these pay services offer which is better resolution and that show the shear etc?

[Typhoon Tip]

Also - I'm assumign that NWS is going to be doing surveys of their AF domain.... Any word on when those analysis will be put out.

[Typhoon Tip]

I thought the public data was a different level than what the NWS offices see. Don't they have the same kind of data that these pay services offer which is better resolution and that show the shear etc?

Maybe - If that is case, then there may be a better argument against - however, we also have to keep in mind this part of the country and how tough twister realization is to come by. We get them time to time, but that last thing NWS wants is apathy because of too much warning.

[Ginx snewx]

The purpose of this thread and ramfications for impugning NWS is dubiously self-serving to me. I suspect that some of the issue here is really that some folks wanted the melodrama and excitement, and feel cheated because there wasn't blaring sirens and red flashing lights for people to get all manic in fear just that much more. They don't want their extreme weather party cut into - and perception is everything whether they know it or not.. and it is petty.

The same sort of vibe emanates from the chorus in the winter, when snow is handled even slightly off from what is or is not realized relative to a forecast; just waiting to pounce with drool! Wrong attitude to have though. Again, it's more a reflection of weather for entertainment/emotion rather than science, and needing some source to blame when those emotional expectations were not fulfilled. It's all stupid to begin with because no one should be investing that way.

I was on top of those cells yesterday and I thought it went over just fine. The reflectivity channels only hinted at a hooks until a tad later... I am aware of the gated this and that argument, but the velocity channels at KTAN its self did not show that quite as obviously as whatever these alterior product(s) origins are. Unless NWS uses some internal imagery that they don't show the publice - i don't know. Otherwise, from what was shown on KTAN's website's rad page, it was not abundantly clear there was a tornado on the ground - ground truth was key there; hense the appearance of a late dial up. I was watching both cells closely and thought the Springfield one - using KTAN's rad source - was somewhat suspicious looking when they issued a warning, and part of me even thought that might be the other way - trigger happy. I've seen countless "hooks" in the plains with far better NWS radar velocity channel gates, and nothing more than a funnel cloud reported by storm chasers.

Give it a rest. There is no issue here with how it was handled. I promise, the next one you'll get ALL your party favors in, and not have any of its affects diminished.

Why would a fine discussion about the lack of advance warning be perceived as dubiously self serving? Improving performance of advanced warning and getting the message out without false alarms should be priority one,. I have a right to expect the best from a taxpayer point of view. I also can express my dissatisfaction, which I have not or have any. USA not CHINA

[radarman]

Yep then the human generates one auto text warning. Simple yet effective, everyone has a cell phone including kids,elderly. KISS, Keep IT Simple Stupid

We do this in Oklahoma... although not with SMS. Severe winds detected (algorithmically) trigger automated email, twitter, and NWS chat messages. These reference local GIS landmarks and give a distance, direction, and strength estimation, along with a motion estimator and a spatial extent estimation. We also threshold for frequency so as not to continuously blast these messages all the time when a severe event occurs.

[TheTrials]

Maybe - If that is case, then there may be a better argument against - however, we also have to keep in mind this part of the country and how tough twister realization is to come by. We get them time to time, but that last thing NWS wants is apathy because of too much warning.

One would think the proper program as a backup/last resort/fail safe would have little affect on too much warning and would have the opposite effect, especially in an area that doesn't see a lot of tornados and would probably benefit from enhanced computer interpretation just in case especially for a public that doesn't think they are going to see this.

[Ginx snewx]

We do this in Oklahoma... although not with SMS. Severe winds detected (algorithmically) trigger automated email, twitter, and NWS chat messages. These reference local GIS landmarks and give a distance, direction, and strength estimation, along with a motion estimator and a spatial extent estimation. We also threshold for frequency so as not to continuously blast these messages all the time when a severe event occurs.

Wow did not know this, should be made nationwide for Tor.

[Typhoon Tip]

Why would a fine discussion about the lack of advance warning be perceived as dubiously self serving? Improving performance of advanced warning and getting the message out without false alarms should be priority one,. I have a right to expect the best from a taxpayer point of view. I also can express my dissatisfaction, which I have not or have any. USA not CHINA

Because there was not lack of advanced warning... The area was in a Tornado Watch - and again, please read that more carefully: KTANS rad products were not as prodigiously indicative as to require warnings sooner.

You should expect the best - no argument on that. But I didn't see anyting from NWS position that could have expedited things.

NOW, again, IF IF IF they are using some product they are not showing the public, I'd be happy to re think that positioin.

[TheTrials]

We do this in Oklahoma... although not with SMS. Severe winds detected (algorithmically) trigger automated email, twitter, and NWS chat messages. These reference local GIS landmarks and give a distance, direction, and strength estimation, along with a motion estimator and a spatial extent estimation. We also threshold for frequency so as not to continuously blast these messages all the time when a severe event occurs.

Do you think its possible for the algorithm to produce a reasonably accurate prediction about the towns that could be affected? So not just direction, but specific grid points about who will see the severe weather next?

[Analog96]

Because there was not lack of advanced warning... The area was in a Tornado Watch - and again, please read that more carefully: KTANS rad products were not as prodigiously indicative as to require warnings sooner.

You should expect the best - no argument on that. But I didn't see anyting from NWS position that could have expedited things.

NOW, again, IF IF IF they are using some product they are not showing the public, I'd be happy to re think that positioin.

Don't compare a tornado to a snowstorm. Most snowstorms, except for the HUGE blizzards, are incoveniences. Torandoes are much more than an inconvenience.

[The Dude]

The purpose of this thread and ramfications for impugning NWS is dubiously self-serving to me. I suspect that some of the issue here is really that some folks wanted the melodrama and excitement, and feel cheated because there wasn't blaring sirens and red flashing lights for people to get all manic in fear just that much more. They don't want their extreme weather party cut into - and perception is everything whether they know it or not.. and it is petty.

The same sort of vibe emanates from the chorus in the winter, when snow is handled even slightly off from what is or is not realized relative to a forecast; just waiting to pounce with drool! Wrong attitude to have though. Again, it's more a reflection of weather for entertainment/emotion rather than science, and needing some source to blame when those emotional expectations were not fulfilled. It's all stupid to begin with because no one should be investing that way.

I was on top of those cells yesterday and I thought it went over just fine. The reflectivity channels only hinted at a hooks until a tad later... I am aware of the gated this and that argument, but the velocity channels at KTAN its self did not show that quite as obviously as whatever these alterior product(s) origins are. Unless NWS uses some internal imagery that they don't show the publice - i don't know. Otherwise, from what was shown on KTAN's website's rad page, it was not abundantly clear there was a tornado on the ground - ground truth was key there; hense the appearance of a late dial up. I was watching both cells closely and thought the Springfield one - using KTAN's rad source - was somewhat suspicious looking when they issued a warning, and part of me even thought that might be the other way - trigger happy. I've seen countless "hooks" in the Plains with far better NWS radar velocity channel gates, and nothing more than a funnel cloud reported by storm chasers.

Give it a rest. There is no issue here with how it was handled. I promise, the next one you'll get ALL your party favors in, and not have any of its affects diminished.

So having a tornado warning in place while a tornado is on the ground constitutes needless "melodrama and excitement"? Interesting. I always thought their purpose was to save lives. I guess I learn something new every day...

Also, since when has there had to be a tornado on the ground before a warning was issued? Warnings are issued all the time based on radar-detected rotation.

[adamrivers]

Because there was not lack of advanced warning... The area was in a Tornado Watch - and again, please read that more carefully: KTANS rad products were not as prodigiously indicative as to require warnings sooner.

You should expect the best - no argument on that. But I didn't see anyting from NWS position that could have expedited things.

NOW, again, IF IF IF they are using some product they are not showing the public, I'd be happy to re think that positioin.

Maybe it didn't show up on BOX's radar....but it was pretty clear on ENX's as that is what radar I was watching from the Northampton storm as well.

[w1pf]

KBOXs 88D doesn't see the bottom few thousand feet of the atmosphere out here. I'm not a met, but I did spend some time at the Wichita WFO in tornado season while I was working on an ancient NWR program.. and those guys were very clear that upper air rotation is at best a so-so indicator of the probability of a tornado on the ground; you probably won't get one without upper air rotation, but you are pretty likely to not get one even with upper air rotation.

Unless you can see and measure that bottom slice of the atmosphere, you can't do automated warnings. IMHO.

We didn't have any CASA radars to give us a peek at what was going on down low; they aren't tied in operationally, and they are all out in OK (except for the one in Amherst, which can't see over the holyoke range)

[radarman]

Do you think its possible for the algorithm to produce a reasonably accurate prediction about the towns that could be affected? So not just direction, but specific grid points about who will see the severe weather next?

Certainly, within a reasonable distance of course. At the moment every minute we generate a 20 minute nowcast on a standard but fairly high-end desktop machine. The nowcast takes approximately 1 minute to run. It's essentially an advection model though, and does not predict tornadogenesis, nor handle complex storm interactions or lifecycle changes.

For those we use an OU forecasting algorithm, however the run time is much longer and it requires a supercomputer at the moment.

With a conical error model it is fairly easy to name towns, even sub-regions of towns (depending on size), however communicating location information in coordinates we have found to be non-ideal. People have trouble relating. It's better to be a little more general and use town names, villages, or major infrastructure. (highways, etc.)

[adamrivers]

KBOXs 88D doesn't see the bottom few thousand feet of the atmosphere out here. I'm not a met, but I did spend some time at the Wichita WFO in tornado season while I was working on an ancient NWR program.. and those guys were very clear that upper air rotation is at best a so-so indicator of the probability of a tornado on the ground; you probably won't get one without upper air rotation, but you are pretty likely to not get one even with upper air rotation.

Unless you can see and measure that bottom slice of the atmosphere, you can't do automated warnings. IMHO.

We didn't have any CASA radars to give us a peek at what was going on down low; they aren't tied in operationally, and they are all out in OK (except for the one in Amherst, which can't see over the holyoke range)

Anyone know what the ball (radar??) on top of Soapstone Mtn. in Somers is?? Is that a radar?

[radarman]

KBOXs 88D doesn't see the bottom few thousand feet of the atmosphere out here. I'm not a met, but I did spend some time at the Wichita WFO in tornado season while I was working on an ancient NWR program.. and those guys were very clear that upper air rotation is at best a so-so indicator of the probability of a tornado on the ground; you probably won't get one without upper air rotation, but you are pretty likely to not get one even with upper air rotation.

Unless you can see and measure that bottom slice of the atmosphere, you can't do automated warnings. IMHO.

We didn't have any CASA radars to give us a peek at what was going on down low; they aren't tied in operationally, and they are all out in OK (except for the one in Amherst, which can't see over the holyoke range)

That's actually not true...

I do totally agree that the height off the ground does preclude saying accurately whether you have a TOG or not, however, a warning on a funnel cloud should not be considered a false alarm. And probably nor should a warning on a very strong meso, although neither of these generally will have any ground truth on which to judge your results, except possibly spotter reports.

One other thing to remember is the human forecaster has the same problems as the algorithm with regards to off the ground radar data. So I'm not suggesting an automated warning could be much better than the forecaster's warning, but as I said, on par or nearly so.

[The Great Zo]

I do totally agree that the height off the ground does preclude saying accurately whether you have a TOG or not, however, a warning on a funnel cloud should not be considered a false alarm. And probably nor should a warning on a very strong meso, although neither of these generally will have any ground truth on which to judge your results, except possibly spotter reports.

From a classic mesocyclone standpoint, one would expect to see moderate to strong circulation in the mid levels of a strengthening supercell before a "surface" reflection appears and a tornado occurs. That seemed to be the case west of Springfield, with signs of that appearing on both KBOX and KENX between 4:10 and 4:20 (50+ knots each way, gate-to-gate across multiple gates). Of course, in the Springfield area, there is no radar close enough to determine conditions at the surface -- the lowest slices of KBOX and KENX are both over a mile high.

Far-from-radar situations like this present complications to the warning process, and sometimes put more emphasis on ground truth reports. Still, combining certain radar signatures with a storm-scale conceptual model, and the storm's expected evolution based on a detailed mesoscale analysis, can lead to early warnings and lead time before a tornado actually touches down.