NWS Central Region Impact Based Warning Experimental Product

[rdale]

You ARE using probabilistic warnings, you've admitted as such. (Correct me if I'm wrong.) The difference is that you are also acting as the filter - preferring "safe" than sorry with some clients, and preferring "very safe" to sorry for others. Nobody here is proposing ever telling the public that a 12.4% chance of a tornado is coming their way. That's what the dissemination system would filter out.

You're also passing judgement on a somewhat-controlled experiment that hasn't even begun based on uncontrolled situations over the past 13 years. I don't think that is a valid comparison.

 

[Mike Smith]

Nobody here is proposing ever telling the public that a 12.4% chance of a tornado is coming their way. That's what the dissemination system would filter out.

Rob, that is exactly what some in Norman are proposing. To quote Chuck Doswell, "I would have told Greensburg they have a 70% chance of a tornado fifteen minutes before the tornado arrived."

Chuck is not the only one. Look at Greg's proposed tornado warning graphics with 10%, 70%, etc. tornado probabilities here: www.norman.noaa.gov/nsww/talks/007_fri_stumpf.pdf

 

[Mike Smith]

You ARE using probabilistic warnings, you've admitted as such. (Correct me if I'm wrong.)

Consider yourself corrected.

Once the tornado warning threshold for a given client is reached, we warn. Those thresholds are fixed. That makes it binary at our end. Either the threshold is reached or it is not. It would not matter if we judged the tornado threat to be 40% above our threshold, our actions would be exactly the same. If we did something different when the threshold was exceed by that 40%, then they would be probabilistic warnings.

We do not give the customers probabilities in the warnings they receive nor to we put in words like "slight chance" or "likely."

Thus, we are not making "probabilistic" warnings.

 

[Mike Smith]

Rob,

I really want to understand your point of view but am having a hard time.

I wrote, "The current TE requires "reliable sources" and "clear radar evidence." I fail to see what is going to change between now and April 1 that will cause TE's on that date and thereafter to be more accurate than those of the last 13 years."

You replied:

You're also passing judgement on a somewhat-controlled experiment that hasn't even begun based on uncontrolled situations over the past 13 years. I don't think that is a valid comparison.

So, what is going to change scientifically between now and April 1 that is going to make a significant difference in tornado emergency accuracy? You keep saying I'm wrong about this. Look at what I wrote and tell me, specifically, how it is "invalid."

Thanks, Mike

 

[Mike Johnston]

When people hear a tornado warning, assuming they're even paying attention, it seems to me they want to hear the very basic facts of 1) what, 2) where, and 3) when. Complicating it with different headers, different threat levels, etc. will add very little value and maybe even cost precious time. Time in terms of composition of the message, reading/listening to more words in the header, and - finally - interpretation of the message. The most direct route seems pretty simple: tornado warning, tornado detected by [insert how], located at [insert where], moving at [insert speed and direction], headed torward [where].

And, no, it's completely unreasonable to expect the NWS to be all things to all people. If you're planning to take a 3-day outdoors hike, it's incumbent upon the individual to seek out and be attuned to the weather to a heightened extent, compared to a normal weekend for a householder. It is NOT incumbent upon the NWS to issue a personalized forecast based upon your hiking plans. That's where Greg's analysis is off the mark, IMO.

Now, if enterprises are particularly at risk to weather phenomenon, they may choose to pay a premium for customized coverage. That's where firms like Mike Smith's provide a valuable service. It's best left to them because they understand the specific needs of specific users - no way the NWS could even begin to fill that need.

 

[rdale]

Mike J - I saw you sneak in People will still hear exactly the same thing. I did a non-scientific survey of the ATL area that saw two hours of "debris on radar confirms tornado" when no tornado actually occurred. Nobody I asked knew a thing about any of that wording. They heard whatever the TV guys told them.

Chuck is not the only one. Look at Greg's proposed tornado warning graphics with 10%, 70%, etc. tornado probabilities here:www.norman.noaa.gov/nsww/talks/007_fri_stumpf.pdf

I didn't know Dr Doswell was on the group implementing probabilistic warnings - Greg, is that true?

On the other hand, and he can answer, did Greg ever say that the only product going to the public is a map like that? That goes against everything I've heard... Even goes against the example on slide 34.

Once the tornado warning threshold for a given client is reached, we warn. Those thresholds are fixed.

Exactly - that's a probabilistic warning. I don't see the difference. If I as an EM only want to be alerted when the NWS says there is a 10% probability of a tornado in my county, I never would have to see a "there is a 10% chance" bulletin. The system I use waits until the probability is > 10% and then sends "tornado threat in your county." Exactly what you are doing. Just because the numbers aren't going out doesn't make a difference, you have a threshold and you don't warn until it exceeds that threshold. The thresholds are different based on your users' needs. Again tell me where I'm wrong, I never worked at AW.

Either the threshold is reached or it is not.

Exactly. If I am Jones Manufacturing and it takes me 30 seconds to evacuate the plant, I don't want to be alerted until the 80% threshold is passed. The alert to me would not be "80% chance" but "Tornado imminent." You don't need to have a number to be a probabilistic warning.

If we did something different when the threshold was exceed by that 40%, then they would be probabilistic warnings.

Somehow your definition is different than the people in this thread discussing it.

We do not give the customers probabilities in the warnings they receive nor to we put in words like "slight chance" or "likely." Thus, we are not making "probabilistic" warnings.

Honestly I don't know who told you that. You are. At least the way it's being discussed here.

So, what is going to change scientifically between now and April 1 that is going to make a significant difference in tornado emergency accuracy?

Mike - the TE has been an informal product for years, and was not verified formally except by people like Patrick. There was no rigid ruleset nor tracking mechanism. Now there is. I'm not saying forecasters are more accurate, I'm saying they are now being formally notified of the criteria and will be tracked.

Let me rephrase this... You often use your blog for forecasts or nowcasts. Do you verify each and every one of those forecasts? Track them all and watch your accuracy? Or is it just something you put out your thoughts about and let fly. I assume the latter. In my experience, I am much looser with a forecast I make for my personal enjoyment than a forecast I make for my customers. Think of it that way. TE's were just a neat phrase in the SVS over the past decade. Now they are the real thing. I don't get how you can assume nothing will change in the way they are issued because of it.

 

[Mike Smith]

You often use your blog for forecasts or nowcasts. Do you verify each and every one of those forecasts? Track them all and watch your accuracy? Or is it just something you put out your thoughts about and let fly. I assume the latter.

I verify all of my major forecasts. Two recent validations below. The more important the forecast on my blog, the more likely I am to validate it:

www.mikesmithenterprises.com/2011/11/so-how-did-that-drought-forecast-work-out/

www.mikesmithenterprises.com/2011/12/how-good-or-bad-were-the-blizzard-forecasts-the-response-part-2/

In my experience, I am much looser with a forecast I make for my personal enjoyment than a forecast I make for my customers. Think of it that way. TE's were just a neat phrase in the SVS over the past decade. Now they are the real thing. I don't get how you can assume nothing will change in the way they are issued because of it.

Rob, you keep turning this around. I have - twice, went back and checked -- stated my reasons for my contention. You keep telling me you "don't get" my position but have failed to answer my question about what is going to change between now and April 1.

If you are telling me that the NWS offices have been going against their own rules (post #39 in this thread) when issuing tornado emergencies, you are claiming something that is rather alarming: Cavalier treatment of the public in a situation of extreme danger. I do not believe that is the case.

What I do believe is that NWS offices issue TE's in good faith but they bust so often because we don't have the scientific knowledge to do it. I've explained this several times. Please tell me why you believe the science is there to issue a level 2 or 3 warning accurately and how that science has improved since Patrick's verification. Otherwise, lets get off this point and agree to disagree.

 

[Mike Johnston]

Yeah, we'll see how the verification of TE's, and levels 1, 2 and 3 actually go. I predict the verification stats won't exactly be transparent to the public, nor will anything in this actually advance the science itself. Let's just put the onus on the forecasters: keep up a good POD while steadily reducing the FAR; that's what it really should be about. All of these tiers, warn on forecast philosophies, etc. are just obfuscations of what the basic mission should be about.

 

[rdale]

I didn't say your major forecasts - I asked if you verify all of them. The answer is no. TE's were not verified by the NWS at all (as far as I can tell, feel free to correct me) and IMO you can't jump to conclusions about an experiment when they will be formally verified.

If you are telling me that the NWS offices have been going against their own rules (post #39 in this thread) when issuing tornado emergencies

I am saying that different offices have had different philosophies when it comes to TEs. For years there were no rules at all, which you are certainly aware of, so they couldn't have "gone against them" anyways since they didn't exist.

What I do believe is that NWS offices issue TE's in good faith but they bust so often because we don't have the scientific knowledge to do it.

That is certainly a possibility. The possibility exists that you are wrong though, and having set criteria along with valid verification will help us answer that question. To declare that as a closed case is certainly not warranted by evidence at this time.

And Mike J - these aren't obfuscations. They are helping to differentiate between a gustnado spinup and an EF5 wedge, both of which are currently the exact same product with the exact same header and the exact same tags. Now they won't be.

 

[Mike Johnston]

Rob,

You keep saying the proposed new warning structure will help differentiate, but at the same time saying that the public will receive the exact same messages as they have been before. So, which is it? I'm just a layman when it comes to the meteorlogical world. You're a professional and sound like you're up on every technical bulletin. If there is a large or damaging tornado with ground truth verified, why not just state this in the text of the warning? Like I said - what, where and when are the questions the public wants to know, why not just be as concise as possible without getting caught up in definitions of headers, tags, etc.?

 

[rdale]

Because nobody in the public gets their warning by reading the text of a TOR. They get from TV and text. Now Tv weather casters who don't know radar and aren't in NWSChat can specifically see which storms to focus on if they can't figure out on their own.

And for automated services, this is a dream. If someone only wants an alert for actual tornado instead of radar based, the tags let that be a filter setting. If a normal SVR is no concern for a hospital EM, but he wants to know when 2" hail is possible, the tags do that too.

So even if the tier experiment doesn't work, the tags are an incredible and overdue advance for automated alert services.

 

[Mike Johnston]

What do you mean "nobody in the public gets their warning by reading the text"? The red scroll at the bottom of TWC's broadcast is the text of the warning. Driving in a car? If the station you're listening to breaks in for warnings, you often hear the text verbatim. Use a NOAA weather radio that alerts you at night? You hear the text. Even TV mets that break in for live coverage begin by at least summarizing the essentials of the NWS warning text before they proceed to their own graphics, radar analysis, etc. Every tornado situation is at least somewhat unique. Why not describe the situation with specific words rather than trying to come up with a multitude of codes and classifications that could never replace the English language no matter how comprehensive?

 

[rdale]

That adds up to about 10% given the research over the past few years. Which leaves 90% to other sources. Sorry if my "nobody" was too general, but I'd even surmise that most of the 10% don't hear anything past "tornado warning" and "my county."

Why not describe the situation with specific words rather than trying to come up with a multitude of codes and classifications that could never replace the English language no matter how comprehensive?

They are - I think you are misunderstanding as there are no "codes and classifications." Unless I'm missing something in the PDD - but feel free to post it because I'm not seeing it. The "tier structure" is a way to describe these - that phrase will not be used in the warnings.

 

[Mike Smith]

The codes, a/k/a tags, are at the end of the warning and they are changing. It's in the PDD.

 

[Mike Johnston]

So, the warning sources of: local TV, The Weather Channel, commercial radio, and NOAA weather radio combined represent only 10% of the total communications of how people receive weather warnings? Wow...that is an eye opener. I wonder how the remaining 90% receive their warnings? Text messages, twitter, perhaps the Farmer's Almanac?