Wednesday, June 15, 2011

Are all tornado warnings the same?

In the fallout from the large tornado death tolls this year, there are some who are finding fault with the warnings from the National Weather Service. Anyone who knows something about the weather must understand that weather forecasts, even those short-range forecasts we call "warnings" have uncertainty associated with them. The science of meteorology as of this moment simply doesn't permit forecasts without uncertainty. Nevertheless, either a warning is issued or it isn't -- technically, this means warnings are dichotomous (or binary): yes, a warning has been issued or no, a warning has not been issued.

At present, warnings come in two "flavors": severe thunderstorm warnings (coded "blue") and tornado warnings (coded "red"). By definition, a storm that produces one or more of the following -- large hail (the current threshold value is one inch diameter hail), strong "straight" winds (the current threshold is 50 knots or 58 miles per hour), or a tornado -- is a severe thunderstorm. Some storms for which blue warnings are issued go on to produce tornadoes, many storms for which tornado warnings are issued fail to produce tornadoes. Some storms for which no warnings are issued go on to produce severe weather (including tornadoes). The science of meteorology at its best offers no guarantees that these important distinctions can be made perfectly in advance. Generally speaking, a red warning is associated with a greater sense of threat than a blue warning -- tornadoes are far more likely to result in fatalities. But some severe thunderstorms can produce as much damage (or more) as most tornadoes (the vast majority of which fall well short of being classified as "violent" -- EF4 or EF5) and nontornadic severe thunderstorms can result in deaths, as well.

There are no tornado fatalities from false alarms! It's widely recognized by forecasters that the cost in lives for not issuing a warning can literally be infinitely greater than that associated with a false alarm. Forecasters aren't stupid -- they know that if they fail to issue a warning for a fatality-producing storm, there would be sharks of various sorts circling the waters around them. Nevertheless, our society demands from them that no storm-related fatality ever occur without warning. The problem is that the science simply doesn't permit that. If we were to issue a tornado warning for every storm that might go on to produce a tornado, the number of false alarms would become vastly larger than it now is. If we "raise the bar" for the threat level to reduce the number of false alarms, many more tornadoes will not be warned for in advance!

Assuming that the presence of a storm is necessary to issue a warning, some threshold threat level has to be achieved in order for the warning forecaster to decide to go with a warning, and presumably this must include the perceived threat of a tornado if the storm may need to be covered with a tornado warning. As it now stands, there is no uniformly-defined threat level associated with any of the decisions a warning forecaster has to make -- every warning forecaster is more or less on their own. Obviously, not every forecaster is a carbon copy of every other forecaster. Some make warning decisions better than others.

Recently, a new type of tornado warning has come into vogue -- the so-called "tornado emergency" (which has not been defined formally, either), presumably when a populated area is in the path of a very dangerous tornado. Implicitly, the perceived need for this is driven by the fact that the tornado threat varies from one situation to another. Yet another tough decision for a warning forecaster to make!

In my view of things, warnings should use probability to express different levels of threat for different phenomena: wind, hail, tornadoes. This permits the forecaster to use the science to estimate the likelihood of various outcomes in the path of a storm. It's an established fact that forecasters can become quite good at estimating uncertainty. I don't yet know how best to formulate probabilistic warnings -- I believe years of research are needed to inform us how probabilistic warnings could be made most effective. This is a complex problem that involves far more than the science of meteorology.

In the meantime, I would hope people who are not meteorologists would begin to understand that they bear the lion's share of the burden of responsibility for their personal safety. If they expect warning forecasters to be perfect -- they're actually putting themselves in considerable personal danger. One important factor in dealing with a tornado is having planned what to do in advance. Another is to be "situation aware" when thunderstorms are about rather than counting on someone else to do that for you.

25 comments:

Josh Johnson said...

Hey Chuck,

Long time, no talk. Hope you're well.

Nice read - I hadn't stumbled across your blog until today. There's one line I don't agree with you: There are no tornado fatalities from false alarms!

Direct fatalities? Perhaps not. But, I have spoken with dozens of people in Alabama about the events in their life on 27 April. Many have told me that they have reached a point where they now ignore tornado warnings because "nothing ever happens." I make it a point to ask them if they once took warnings seriously - the majority say they once did and took action, but after doing that dozens of times with no sign of a tornado, they started ignoring the warnings.

I'd submit (with no scientific evidence on this - difficult/impossible to ever prove conclusively) that if the FAR stays high year after year, this apathy will grow, and the inaction of some will kill them. Yes, their own inaction is sometimes the direct cause of their deaths, but I think false alarms collectively play a role, too.

In my admittedly limited sample size, I found this trend was particularly noticeable in people under the age of 30.

Penny for your thoughts?

john said...

Good post. Just a note that the NWS directive has defined the use of "tornado emergency" in a tornado warning. The key element is confirmation of a tornado, and does not include population.

Here is some the text from the directive. "In exceedingly rare situations, when a severe threat to human life and catastrophic damage from a tornado is imminent or ongoing...", and "...this terminology should only be used when reliable sources confirm a tornado, or there is clear radar evidence of the existence of a damaging tornado such as the observation of debris."

Chuck Doswell said...

Josh ... my thoughts are already embodied in the blog. What your "many" people seem to be telling you is that they accept no responsibility for their own safety. They seem to be upset by the fact that their homes and lives were not affected by a tornado when they were in a warning!! Do they really think the NWS can tell them perfectly when a tornado is about to affect them, personally?

I'm not saying the NWS is completely without responsibility for the high FARs - what I am saying is that the existing system is structured on the counterfactual assumption that forecasters can distinguish perfectly between situations deserving a tornado warning and those not deserving such a warning. The existing system inevitably produces an overforecasting bias because of the asymmetric penalties for the different forecast errors.

Chuck Doswell said...

John ... thanks for the clarification. So a tornado "emergency" includes tornadoes out in rural areas, as well? Fascinating. That's probably the right thing to do from one standpoint (forecast the weather, not the impact!), but might not be right from all perspectives. How can a confirmed tornado create "catastrophic damage" in an open prairie?

It's evidently a decision that was made by someone in the NWS, likely without a lot of data to back it up one way or another. Please correct me if I'm wrong!

Josh Johnson said...

"Do they really think the NWS can tell them perfectly when a tornado is about to affect them, personally?"

I think they do. As a society, I think those who are completely detached from the weather/media/NWS world learned most of their "science" from junk television, Twister, CSI: Miami, hospital/doctor shows where everyone always lives because of a miracle cure, etc. They believe we can look at a tornado on radar and say "dude, it's an F-5" without seeing damage. Sadly, they are completely ignorant to the realities of science.

Sadly, we as TV meteorologists don't do much to counter it. At professional gunpoint by consultants, TV managers, and weather graphics vendors that only care about making a sale, we show minute-by-minute model output that WE know can't possibly be 100% accurate. They believe it is. We show "storm tracks" that give people an estimate on when a storm will arrive at their house - it can't possibly be accurate down to the minute...They think it is.

We must walk the line between honestly expressing variability and forecast confidence....and keeping the meteorologically ignorant consultants/TV managers happy. It's a tightrope.

Wendell Malone said...

Great read! I agree, people need to learn to be responsible for their own safety and become more proactive instead of reactive.

Scott said...

Hi Chuck, interesting blog entry as always.

"There are no tornado fatalities from false alarms!"

Don't the fatalities from false alarms come along later as a result of people ignoring what they perceive as warnings from an unreliable source?

In my opinion the present NWS system is adequate for my needs. In the case of the June 1 tornado about 90 miles from me in MA, I knew that June 1st bore watching 4 days earlier because of the convective outlook and New England being the target in the day 4 outlook. Day 3 was Michigan. Reminded me of the Flint-Worcester outbreak. That combined with the fact that other parts of the country had just got hit hard made me send an email to a bunch of friends advising them to watch the weather, especially on Wednesday. Unscientific and flimly evidence but in that case my hunch happened to be correct.

On the morning of the 1st I was monitoring the situation at work on the NWS website and kept their radar display on all day. If the heavy weather reached my area my plan was to stay at work until it was over. Sturdy building and I work below ground level so I that was a reasonably good plan.

The NWS tipped me off that June 1 was a day to watch and kept me up to date as to what was going on. I think they did a good job.

I do agree with the probabilistic warnings. I like the probabilistic displays on the convective outlooks.

If today's system was in place on 6/9/53 there is no doubt that the number of deaths in Worcester country would have been a fraction of the 94 that there were.

As for a high false alarm rate, if it can't be improved, the best thing to do is to go on an education campaign to educate the public on the reason for the FAR and why they should still pay attention when warnings are issued.

Have a nice weekend coming up.

Chuck Doswell said...

Scott ...

If one chooses to consider false alarms as leading to complacency, which in turn leads to fatalities ... yes. This point also was made by Josh earlier. However, this simply reflects the point I'm making that people are simply wrong in thinking that we can have absolute 100% confidence in every warning.

FARs can be improved, but not by leaps and bounds, using the present warning system, unless we are willing to accept many more tornadoes with no warnings. For red warnings, FARs are on the order of 70%. Everyone wants perfect warnings - like people in Hell want a glass of ice water. Not going to happen!!

BTW, FAR is the false alarm ratio, not the false alarm rate. There's a difference! The FAR is F/(F+H), where F is the number of times an event was forecast but not observed and H is the number of times the event was forecast to occur and was observed. The false alarm rate is often called the probability of false detection (POFD = F/(F+Z) where Z is the number of times the event was forecast not to occur and was not observed.

Going to probabilities allows warning forecasters to issue there estimated threat levels - something besides "Yes, there's a tornado warning in effect."

Scott Currens said...

I'd like to see the “Tornado Emergency” become an official warning type with strict criteria as to what qualifies as such. I think probabilistic forecast are great for media/private sector consumption but have concerns that many if not most of the GP would not understand them.

What should be done about tornadoes that occur in environments that only appear favorable for very week tornadoes? In the days following the Joplin tornado KC was blanketed by tornado warnings for very weak tornadoes occurring with low-topped storms in a cold core environment. I would argue that the environment was not capable of producing winds over 80 mph. Many businesses were shut down for hours because they were taking the warnings seriously given what happened in Joplin. I don't know what should be done, perhaps probabilistic warnings will help in these environments too.

Chuck Doswell said...

Scott C ...

I don't know how many times I've heard the objection that the "general public" doesn't understand probabilities. Sorry, but I'm not willing to concede this point. Please read my essay that's linked from the blog.

It seems obvious to me that the situation you describe about low-topped, "cold core" storms (like all of the myriad possible tornado situations) would be amenable to a more meaningful treatment via probabilistic threat assessment.

Anonymous said...

I thought "tornado emergency" may have crept into the mix of products because the simple "tornado warning" no longer carried the impact it once did (when it was issued much less often). Once "tornado emergency" is issued often enough (and that day will probably come), something new will need to be invented to say "this is really serious". Regarding reducing FAR, I understand that to do so (in the present system) would require more "misses". I would think that such misses would mostly be for the "weaker" events with far less "impact" (although, impact is great to anyone who happens to be affected). Would not a more rare tornado warning that is more likely to be associated with an actual tornado result in a higher public response? I don't know. But it doesn't matter because the message (one can debate whether it's real or perceived) from on-high is that any miss in which casualties are involved is not acceptable. We tell users that a warning means the event is "imminent or occurring", but this is not true when it comes to TORs. My sense, without statistics, is that FARs are more common today than back in the pre-"no surprise NWS" days. I concur also with your point that people need to be responsible for themselves. The NWS does not save their lives. THEY do, using what we hope is reliable information from the NWS. And I also agree that a warning is really a very short range forecast of a certain type of weather. To my way of thinking, the ideal warning would be a short term forecast of a range of weather outcomes from a given storm, to include a tornado. Such short range forecasts would include NO call to actions. The forecaster need not spend any time deciding which CTAs to include because there are too many varied "actions" that may be best for a given person in a given circumstance. Some people may actually follow a particular CTA which isn't necessarily best for them. Short range severe weather forecast, no CTAs. The user decides what to do with the info in the probability table. Presumably this means that they would have some threshold for a given outcome that would cause them to implement their already-developed plan of action. But you are correct in that to develop such a system would take serious sustained effort, something not likely to be supported in today's climate. After all, lead time and POD are the only "metrics" that seem to matter, and achieving high values of each is not terribly difficult (or expensive) when FAR matters not.

Chuck Doswell said...

Anon ...

I agree that the perceived significance of a tornado warning has changed to the point where we have to have some new way of saying "No, we really mean it this time!!" And of course, it's like that "tornado emergency" will be (and has been) overused to the point where we would need to say something different to get people's attention.

The problem continues to be the fact that a red warning isn't allowed to convey a variable level of threat as indicated by probability. Not issuing anything for "weaker" events that still have some finite probability of a potentially serious tornado has at least two problems: (1) some of them wind up killing people, which clearly becomes an issue with the media and the "general public", and (2) it's not a accurate reflection of an intelligent warning forecaster's assessment of the situation.

I believe CTA's are pointless, and have a web essay about that here.

Rob Dale said...

I talked to a customer a few months ago about polygon warnings and they are sold on the system, but they also were much MORE interested in getting probabilities instead. Knowing the "controversy" I didn't lead them at all but just asked what they were looking for, and they said that some plants are large and need advanced lead time to shut down, while others could be sheltered in just 2-4 minutes. They felt that having probabilistic warnings would make those widely different decisions easier than the simple yes/no we have today.

Scott said...

How would you expect a probabilistic tornado or thunderstorm warning to be worded on NOAA WX Radio?

If there is rotation detected but no ground truth?

I assume that the probability would jump to 100 percent once you had a sighting?

The other issue is can NOAA WX radio handle warnings that may take longer to issue?

It seems to me that the legacy 1050 Hz tone warning tone then the SAME digital tones take quite a bit of time, then they have to have the computer generated voice announce the warning.

It will be interesting to see if a probabilistic warning works better. "You know, if they're not sure, maybe we had better keep an eye on the sky and watch for ourselves.

Chuck Doswell said...

Scott ...

At this point, going into the details of how probabilities should be worded is pointless. We need more information about how the public perceives such things and how best to address this.

The probability would not necessarily jump to 100 percent on a sighting. A warning is a forecast, and any forecast involves some uncertainty. Read my essay (linked from the article).

We have much to do if we want to introduce uncertainty into warnings in an intelligent way (unlike the way PoPs were introduced!).

Rob Dale said...

Scott... I would assume that many outlets to the public will "dumb them down" closer to a yes/no. Some customers would be able to take advantage of the probabilities, and some wouldn't.

And the last thing I'd want to let constrict the format is NOAA Weather Radio. More people get alerts on cell phones than NWR, so I'd keep that off to the side for the discussion.

Anonymous said...

Chuck,

You wrote:

Not issuing anything for "weaker" events that still have some finite probability of a potentially serious tornado has at least two problems: (1) some of them wind up killing people, which clearly becomes an issue with the media and the "general public", and (2) it's not a accurate reflection of an intelligent warning forecaster's assessment of the situation.

-----

Your #2 is right since we now have a binary system for telling people about tornadoes, a system skewed toward false alarms because of #1. I've thought that perhaps, within our current system, improved response in the "big" event could be achieved by reducing FAR via letting "lesser" events go, even though those events certainly have potential to adversely affect "some" people. But I guess there is no way to know whether X more people heeded warnings of the F5 because some weaker events were let go (even though intuitively this would be the case). It sounds callous to go down this path of reasoning because "all" people deserve our best assessment of a severe weather threat. But we cannot do that in our present system because to ensure that the greatest number of people are "warned" for an actual tornado, we must go ahead and "warn" over almost the entire range of uncertainty, which leads to many people being "warned" about something which doesn't actually occur. This all leads me to concur that the best "service" we can offer is to convey our level of uncertainty of a given outcome, and let the user decide what (if anything) needs to be done with that information.

Amen on CTAs. I worked svr wx events for years and never thought CTAs were useful. The warning should simply serve as a notice to implement "your" plan of action, not to try and help you devise a plan on the fly.

I once talked to the former official in charge of the old Abilene TX weather office. He relayed an account of a tornado (around 1960 I think) out west of town. They at the office simply went out into the parking lot to watch it. He said they didn't bother with a warning because it wasn't going to bother anyone! Simpler times.

Scott said...

I was watching, courtesy of Youtube a television broacasting warnings to the residents of Greensboro Kansas. They repeated the tornado emergency issued by the NWS and showed on radar where the circulation was.

In the case of a confirmed tornado on the ground heading straight for a populated area, it seemed appropriate to use a term that is stronger than the usual tornado warning. Tornado emergency is a strong term and is a little out of character with the normally understated NWS language but it certainly makes sense.

One of the problems with a binary warning system is that the technology doesn't exist to be able to dumb it down to that level.

The POD is high but I could have a high POD too by issuing a warning for every single thunderstorm.

As a non scientist and non meteorologist it would be helpful to me to know if the report is based only on radar or if there are any spotter reports, and if the spotter reports are for funnel clouds or for tornadoes.

As far as CATS, it's unlikely that they're going to stop advising people what to do.

If it's indicated on radar then it makes sense to recommend that people either keep an eye on the sky if that's possible, or go to their shelter.

If spotters have confirmed funnel clouds then that adds to the urgency of the situation and that fact needs to be conveyed.

If spotters have confirmed a tornado then that is the most serious situation and people need to move fast.

With the binary warning the false alarm rate is going to be high.

If you tell people exactly what information you have then the false alarm rate drops.

We're seeing rotation on radar, recommend you keep an eye on this. That's not a false alarm, that's reporting the truth regardless of what happens.

Spotters reporting a funnel on the ground, moving in X direction, get the appropriate warning out, no false alarm there.

The binary warning system also makes it seem like people do not need to take any responsibility for their own safety. "Don't worry, listen to us, we'll tell you when to run for cover."

If you run for cover for every tornado warning, most likely you will be safe, of course, the false alarms are likely to get you out of that habit and fast.

It seems to me that the NWS has gone down the wrong road by dummying down the system to the point of yes/no.

This is one problem that does not have a boolean answer.

My question for you Chuck is via radar alone, how much confidence can they get that the circulation is indeed reaching the ground or is going to reach the ground?

When dealing with RF links, the distance to the horizon depends on your HAAT or height about average terrain. If you want your radar to see down near the ground, you put it up higher. 1000 feet would be good for about 40 miles. It would seem to me that putting your radar on top of the highest building in a metro area you could see down to ground level for as far as you really need to, in many directions. Is this being done anywhere? The radars that I have seen were not mounted very high. 100 feet, 200 feet, fairly low.

Tower mounting at 1000 feet a rotating dish in a radome is most likely out of the question but it may be possible to tower mount say 3 radars each with 120 degree views using electronically steered antennas.

Since one tall building may only be good for certain directions, several radars could be used to cover a metro area with coverage down to the ground.

With the present Nexrad WSR88D's how much confidence can you realistically get that you've got rotation that is reaching the ground or going to reach the ground?

How reliable are the spotter networks in your area Chuck. You've said good things about spotters. Can they be relied upon? Are there enough of them?

Chuck Doswell said...

Scott,

That was a very long post! It's not evident to me that such a lengthy set of comments was needed. OK - let me deal with your remarks ...

Why would it be helpful, as a NON-meteorologist, to know whether a warning was based on radar versus a spotter report? Presumably, you're asking this on behalf of those who wish to do their own personal interpretation of the threat level. If a properly trained meteorologist does that assessment, and is allowed to expressed that threat level in some terms (probabilistic or otherwise), can non-meteorologists presume to know better than the meteorologist how to interpret the data s/he is using to assign that threat level?

Whether or not my opinion can convince the NWS to drop CTAs is irrelevant for purposes of this blog.

In what follows that comment, you have a laundry list of indicators currently available to warning forecasters, who presumably know how to incorporate them into a graded threat level, should they eventually be allowed to issue products that convey that.

Your question concerning radar ... radar is a powerful tool for issuing warnings but it has its limitations, and should never be depended upon as the sole tool for issuing warnings. In good offices, radar isn't used as the only tool. Radar will never be capable of making 100 percent accurate determinations of the likelihood of a tornado during the lifetime of a tornado warning.

Spotters are another important tool that's underutilized in some places - perhaps owing to a lack of good spotters in some CWAs, or perhaps because some WFOs just don't use them as they should be used. Like radar, spotters also have limitations, especially in some regions of the US where visibility is limited by various factors -- everywhere after dark.

Scott said...

Why would it be helpful, as a NON-meteorologist, to know whether a warning was based on radar versus a spotter report?

--------------------
In keeping with the idea and fact that we are responsible for our own safety, and taking into account the high number of false alarms when it comes to tornado warnings, a spotter report adds credence to the warning as well as adds urgency to the situation.

In short, adding in spotter reports I think benefits everyone.

Even if the NWS implemented your system, and I can't understand why they haven't implemented it decades ago, I would still want to know if there was ground truth.

Hope you had a good forth of July weekend.

Take Care

By the way, just finished with a really good book on the 1966 Topeka KS tornado. "And Hell Followed With IT' by Bonar Menninger. 325 or so pages. He does a great job of telling the story from the point of view of the survivors.

It's interesting that he also took the time to issue a warning on it's own page indicating that some of the stories in the book involve taking shelter under underpasses and points out that it's a bad idea and why.

I'm also reading through an interesting book called "The Unthinkable" Who Survives When Disaster Strikes and Why. By Amanda Ripley.

Very interesting book.

Chuck Doswell said...

Scott,

Do you believe in the credibility of those issuing NWS warnings? If so, you should be able to accept their interpretation of the data used to formulate their warning. If not, and we include ANY of those data, where does that stop? A LOT of information goes into that warning decision. Many people think the warnings are too long already!

If you DON'T have confidence in the accuracy of the NWS warning forecasters, do you believe 100 percent in the accuracy of spotter reports? Might you now want to know WHO that spotter was? The warning forecasters take that into account, of course.

Rob Dale said...

And Scott - the text of the warning DOES indicate whether it is a radar only warning or whether a spotter report helped push the button. What is needed is a computer-readable line in the bulletin to make that easier to interpret automatically.

Scott said...

The NWS has shown that it is very good at detecting the major events and most of the minor ones. If you seek shelter when they recommend that you do so, you're unlikely to have any problems. You do have to be tolerant of false alarms. You have done a fine job already of explaining why the false alarm rate is higher than anyone would like to see it.

As for the spotters, I have looked at the NWS presentation for spotters, and can see that it easy to make a mistake and issue a false positive as well as not see circulation that is rain wrapped.
So, 100 percent confidence in the spotters is not warranted.

Reporting that a confirmed tornado is in the area seems to me to be a crucial piece of information. A unique piece of information as well.

We have finally gone past the stage of being in a watch and are truly in the warning stage. Why, we now have something on the ground to warn about.

As for the amount of information given about a warning, I think that should depend upon the medium being used and how much information that it can handle.

On the web I would like a short warning with the essentials with the option of getting as much information as possible. This information would be mainly for personal interest and for study. It doesn't necessary have to be available right away but after the event and things calm down it would be nice to have access to that information.

When I heard on NOAA WX Radio on June 1 that there was a tornado, on the ground as they say, that was an attention getter. It says that this storm is serious.

Now, is that necessary? It depends on your audience and what you are trying to accomplish.

If your goal is simply to issue a technically accurate forecast and whether anyone pays attention is not your concern then issue your warning however you see fit.

If the goal goes beyond that then I believe that you need to spell out the risk, and recommend stops to take. The two need to match. You don't send people to shelters for a small risk, you tell them to keep an eye on the sky and be prepared to take shelter immediately. If they cannot keep an eye on the sky then it is best to take shelter.

If your audience is people who are weather savvy then you can skip the CTA.

Your other choice is to publish what everything means and what people should do on the NWs website, and issue warnings with no CTAs.

Do you have any idea of who relies on NWS instead of local media?

Thanks for the reply.

Scott said...

Thank you Rdale. I have heard that on NOAA but didn't realize that it was standard policy. Makes sense.

I agree, but I think they will need to do a parallel system for data warnings. The current NWR is too busy when the warnings start going out to fit data packets in, in my opinion, unless the NWS becomes more economical with words and speeds the text up a bit.

One data channel, shared by all base stations, either timed transmissions or carrier sense multiple access and you could have a very good warning system.

Use 4800 bps, make it easy to interface a scanner to a computer with soundboard and decode the data as its coming in.

They probably should plan on phasing out the voice service in 20 years. They're using TTS now. Rather than broadcast the TTS using FM, it makes more sense to send the information as data and either have a readout on the device or do the text to speech on the device.

If they make the data rate 4800 then inexpensive devices that connect to the speaker jack of a scanner would be able to decode the data.

The nice thing about them using FM is that you can listen to NWR on any scanner or communications receiver.

The other advantage is that it's a relatively simple system. Theoretically the system requires a transmitter, 300 to 1kw of transmitter power, a generator, and a satellite receiver. Preferably C band so there is no problem with link budget due to rain. The dish should be protected from movement with a radome. Backup power via battery then generator.

Absent a direct hit, the system should stay on the air.

The satellite link is preferable due to reliability issues.

Internet warnings to phones are nice but the problem is that there can be significant delay due to system usage. There are also many points of failure.

Thanks again for your comments Rdale. Take care and have a nice weekend.

Matt said...

I concluded that the Tornado Emergency was not a good road to go down when I witnessed in "VORTEX offspring" in Panhandle, TX an F1 tornado was given a "Tornado Emergency" on NOAA Weather Radio. I was looking AT the tornado. It was daylight and everyone in Panhandle was looking at it SW of town. It petered-out before in struck Panhandle. I wrote about this in my dissertation and you are correct in what you say in the Blog. The 2000 population in Panhandle is ~2400 on in Carson County, TX it was ~6500. One, you cannot say a Tornado Emergency is going on for Panhandle because it is too rural. And F5 going for Panhandle still would still not be pertinent (and it was for a small town in KS too - Greenburg!). It you have to do it, it should be in urban areas for long-track tornadoes like the first one (OKC) which are few and far between! The "Tornado Emergency" took off And NWS did a "bunch of them" including Panhandle, Greenburg, Indianapolis, OKC not counting the 1999 tornado, etc. It has maybe did a slow death, although with as few tornadoes as we had this year, I do not know? We should stick to the Watch-Warning Program and get the FAR down and areal areas of the warning smaller.

Dr. Matthew Biddle is you do not know who I am!