Worst Science Story of the Week

Worst Headline: “Tornado Forecasting Eludes Weather Scientists” in my hometown paper, The Wichita Eagle. 

The headline was attached to the worst story science story of the week written by Seth Borenstein of the Associated Press. The story was printed in many newspapers across the United States, including the Tulsa World, Miami Herald, and others. 

Let me state — again — on this blog how wrong this story is: Of the 551 people killed by tornadoes in 2011, more than 99% were located in both a tornado watch and a tornado warning at the time the storm arrived! 

Mr. Borenstein cites Joplin. Here is the forecast of the Joplin tornado in the form of the tornado watch:

The watch was issued at 1:30pm, 4 hours and 11 minutes before the tornado reached Joplin! The watch (a forecast) further says there is a “high” probability of tornadoes and a “moderate” probability of a tornado of F-2 intensity or greater.

Did things go wrong later that afternoon in Joplin? Yes. Are there still further improvements to be made to the warning system? Yes, to that, too. The warning system is hardly perfect. But, to trash the science that got the major tornadoes right 99% of the time is ridiculous. This type of ignorant reporting ( ”Tornado Forecasting Eludes Weather Scientists”) does nothing but discourage people from taking warnings seriously — and that is dangerous

Because of stories defending the protagonist in Fakegate, there were many worthy contenders. Still, I hereby nominate Mr. Borenstein for the Dianne Sawyer Award for inaccurate reporting about weather and storms. 

"Error on the Side of ‘Safety’"

Note: In view of the National Weather Service’s decision to experiment with tiered tornado and severe thunderstorm warnings in Missouri and Kansas beginning April 1, the concerns raised in this posting become more relevant than when it was first published on February 4. I’ve added additional information in maroon type and elevated it to the top of the blog for today. If you are not aware of the NWS’s plan (they still have not made a public announcement) see the two postings bellow.


Last week, I criticized journalism is general, and ABC News in particular, about lazy journalism when it comes to tornadoes. I wrote, pertaining to their inaccurate reporting there was “no warning” of the Alabama pre-dawn tornadoes,

This seems to confirm my suspicion that there is a key on journalists’ word processors that says “there was no warning” and they simply press that key every time they have to do a story about storms.

There seems to be another group that suffers from either lack of knowledge about the rapid progress we have made in the field of storm warnings, inertia, or timidity: Emergency managers.

As I have been gathering data from around the nation for the purpose of reviewing last year’s tornado season, it seems emergency managers have a mantra:

Nick Crossley, director of emergency management and homeland security for Johnson County, said the warnings on Saturday night were necessary throughout the entire county.
“We erred on the side of caution,” Crossley said. “It’s much better to sound the sirens and warn people than not to sound the sirens.”

Crossley was responding to criticism about sounding the sirens — twice in one evening — in areas that were never under a warning.

Or, take a look at this video from KMOV-TV, St. Louis, after I criticized St. Louis County for sounding sirens in areas more than 25 miles (with the tornado moving away) from the tornado. They have the capability to sound sirens selectively (i.e., NWS polygons) if they wish to do so.

He said, “I’d rather be safe than sorry” and that we “never know” which way a storm is moving.

So, how bad was the overwarning he was defending?

May 25, 2012, St. Louis.

Above are two images I took during the storm. At left is the funnel cloud for which the warning was issued. At the bottom of the photo is the Mississippi River separating Missouri from Illinois and, at lower left, the south leg of the Gateway Arch.

At right is a photo of the local television storm coverage. “STL” is downtown St. Louis where my hotel room was located. The orange arrow denotes the “hook” echo which shows the tornado’s location and what prompted me to leave my room and go to a location in the hotel to take a photo of the funnel (note: it was past me, I was safe) as it moved northeast. There are no other storms to the west yet the sirens are going off as far away as Pacific, MO (purple arrow). 

Below is a Google Map image showing the location of the funnel (F) moving northeast (thick red arrow). The orange arrow from the above image is carried over. Pacific, MO is located with the purple arrow as above. Pacific is 35 miles behind the tornado threat which is moving northeast, away from Pacific!

I’m not talking about a mile or two safety buffer, I’m talking about tens of miles! St. Louis County has the technological capability to sound the sirens only in areas actually threatened but they choose not to use it. 

Now, take a look at this story from WFIE-TV in Indiana from January 18th (updated Jan. 25th) that came to my attention yesterday February 3:

The National Weather Service allows each county to decide which sirens to set off during a storm. 
14 News found some Tri-State counties are now choosing this option, while others say their policies won’t change.
40 sirens sounded Tuesday morning, getting attention all across Vanderburgh County for a warning that was only issued for the northern section.
Meteorologists clarified on Twitter that the warning did not include Evansville, even though the sirens were going off.
“Our policy is always to sound them off through the whole county,” said Vanderburgh County EMA director Sherman Greer.
Greer says it’s a policy that errs on the side of safety. [emphasis mine]
“Meteorology and the weather and everything is not an exact science,” Greer pointed out.
That is why he’s not comfortable with switching to that new option, from the National Weather Service, that would allow him to set off only those sirens located in the affected part of the county.
“If something strays a little bit further than that area, then we’ve got a problem.”
“I think it’s a good idea. I’m not convinced yet that they are quite as precise as we would like to see it,” said Henderson EMA Director Larry Koerber.
Across the river in Henderson, Koerber also fires all 32 of his sirens, no matter where the storm is in the county.
“We don’t want to miss something and say ‘Well,if the path is there in the southern part of the county’ and sure enough it makes a left turn and winds up in the middle of Henderson,” Koerber says.

We keep hearing from emergency managers; and there are many more examples I can cite:

Error on the side of safety.
Activating sirens over and over and over in areas where there is no threat (deliberately sounding them in Evansville even though the tornado warning did not include Evansville). Is that really erring on the side of safety?

Or, is it really erring on the side of protecting the emergency manager from second guessing (i.e., fear of criticism if a tornado occurs without the sirens going off)?

There is also the complementary comment that meteorologists “don’t know” where the tornadoes are and/or don’t know where they are going to go. This was true 40 years ago during the early years of trying to warn people of tornadoes. We weren’t very good at it then. “Better safe than sorry” made sense in the 60′s and 70′s and, in some areas with poor radar coverage, even the 80′s.
Today: This deliberate and geographically exaggerated overwarning makes no sense in an era of Doppler and Dual-Polarization radars, debris balls, GPS storm reports, etc., etc. As I previously reported on the blog, 99+% of the tornado fatalities in 2011 occurred in areas that were under both a tornado watch and tornado warning before the fatality occurred. Meteorology does know where the storms are going to be.

So, here is the problem:  The evidence is rapidly accumulating that “erring on the side of safety” is doing nothing but training people to ignore warning sirens. 

I mentioned last week that I’m working on a project that pertains to last year’s tornadoes and I originally wanted to get into this subject when I was farther along. But, the news report from Evansville tweaked my conscience.
It is long past time to stop warning areas tens of miles away from the tornado threat. The polygon warnings, while not perfect, build in enough margin of safety to allow sirens to be sounded in and along the polygon. 
Heck, build in a 1-2 mile buffer (easy to do with today’s technology). But stop sounding sirens in areas tens of miles away from any threat!

Between the media inaccurately yet constantly telling people how bad the warnings are and emergency managers sounding the sirens 20 miles behind the tornado it is almost a wonder that anyone pays attention. But, with good television and radio reporting, many are able to intelligently respond and save their lives in spite of these handicaps. But, there is no reason for an environment where making the correct decision has to be so hard.

Based on the preliminary research I have done pertaining to 2011, there is no question that complacency cost lives. I’ll have more when I am finished with the work.

There is still time in many jurisdictions to implement a new policy in time for the 2012 tornado season. Great started, please. Otherwise, I fear we are going to more lose precious lives to complacency.

ADDITION Saturday 11am:  I’ve received some surprising (at least to me) feedback about this post. Apparently, a number of readers do not know that I have written a book documenting how accurate storm warnings have become. For those interested, it is Warnings: The True Story of How Science Tamed the Weather.  Warnings is a non-technical read that explains how courageous scientists built the system that managed the amazing feat of getting both a watch and warning in place in advance of 99% of the fatalities in 2011′s record tornado season.

While there is still work to be done, there is no question that meteorology has advanced storm warnings to where they should be accorded the level of respect that medical diagnosis receive.

If you doubt that is the case, please read the book (OK with me if you go to the library or buy the less expensive ebook version!) before 2012′s tornado and hurricane seasons. Doing so might save your life!

Addition (Feb. 18): I learned last week that Johnson County, Iowa became the fourth jurisdiction since Joplin to announce that it will start sounding sirens for severe thunderstorm warnings, including 1″ hail, which is why this posting is so important. The key to saving lives is getting people to take action and they will only do so if they feel comfortable with their decision. 

Combined with the National Weather Service’s plan to complicate the tornado warnings and give their local offices the option of adding the sentence, “A tornado is possible.” to severe thunderstorm warnings, we are setting ourselves for confusion, delays seeking shelter, and — potentially — the loss of additional lives. 


That is the sound of the National Weather Service blowing up the severe thunderstorm and tornado warning system that has served us so very well for so many years. Starting April 1, in the geographic areas served by the National Weather Service offices in Kansas City, Wichita, Springfield (MO), Dodge City, Topeka and Goodland (KS), there will be multi-tiered severe thunderstorm and tornado warnings.

The changes which I will describe below spring from the high death toll from U.S. tornadoes in general, and the Joplin tornado in particular, in 2011. Unfortunately, in my opinion, they will make the issues worse and will likely cost lives due to confusion.

The New Severe Thunderstorm Warnings

The NWS is going to put additional emphasis on hail size and wind speed, which is fine. While imperfect, the science exists to do this.

Unfortunately, they are going to allow a sentence to be added to severe thunderstorm warnings that states, “A tornado is possible.” What do you or a school principal do with that? Go halfway down the basement stairs?

Given the political pressure the National Weather Service seems to be under at the moment, I forecast that many severe thunderstorm warnings will contain that unfortunate sentence and the “overwarning” problem, which we know causes complacency, will get measurably worse.

The New Tornado Warnings

This is where it really gets bad. There will be, starting April 1, three types of tornado warnings:

  • The “ordinary” tornado warning
  • A tornado warning declaring a “particularly dangerous situation”
  • A “tornado emergency” for “catastrophic” damage.

The first problem is that the science does not exist to do this! We have no skill at short-term tornado strength forecasting. None.

Second, who is going to be able to keep straight whether a “tornado emergency” is better or worse than a “particularly dangerous situation”?

Third, even if #1 and #2 were not issues, what do you want the public to do differently?! Since we meteorologists want everyone to take shelter during a tornado warning, the two “tornado warnings on steroids” are superfluous.

Here is the National Weather Service’s hypothetical example based around the Joplin storm:



Fourth, they are going to elevate any accompanying wind and/or hail threat. To use the NWS’s example, which is built around the Joplin tornado,


The problem with this is that literally dozens of devices are now parsing these messages. During the Joplin tornado, two of my friends were trapped in the path of the storm because a message about hail (the least of their problems) overwrote a text message about the tornado.

What does a person do when hail is coming? Run outside and put the car in the garage….the last thing we want them to do when a violent tornado is approaching.

If the National Weather Service believes an F-5 tornado is approaching they should be urging people to take shelter and forget about the hail, lightning and other hazards.

Fifth, the chance of getting the people of Kansas, western Missouri, and adjacent areas educated by April 1 is extremely low. It has taken forty years to get the “watch” and “warning” concept to where they have widespread acceptance. I doubt this can be done in forty days.

This isn’t just my opinion. Dr. Laura Myers, a social scientist at Mississippi State University, wrote yesterday,

My conclusion: It would seem that more detail and more warning levels would help, but I think it just leads to confusion and [warning] fatigue.

When a tornado is bearing down, people need to act and act quickly. Having to think through warning types is counterproductive.

This experiment has the potential, through confusion, to undo a half-century of great progress in tornado warnings.  I urge the National Weather Service to reconsider and call the experiment off.


Because of the importance of this issue, I’m going to leave this on the top of the blog through Friday evening.

Really, Really Poor Journalism

I’m now in New Orleans (2.5 hours late, thanks Continental!) for the AMS Annual Meeting and several friends alerted me to Dianne Sawyer reporting, and emphasizing, earlier this evening there was “no warning” of the overnight tornadoes (see posting below). Here is the video:

video platformvideo managementvideo solutionsvideo player

This seems to confirm my suspicion that there is a key on journalists’ word processors that says “there was no warning” and they simply press that key every time they have to do a story about storms.

Dianne: There was plenty of warning, starting 21 hours before the tornado struck!

Meteorologists really face an uphill battle:  The media keeps inaccurately telling people that storm after storm occurs without warning.

The public, who — outside of the affected area — doesn’t know better, assumes the media reporting is correct.

So, when a tornado warning affects them, they don’t act on it. Why should they?! The warnings are no good!

There is a really good book on this subject I could recommend to Dianne….

UPDATE: 9AM Tuesday:  Another blog post with the step-by-step of the warning process in Alabama.

UPDATE II:  Thank you Capital Weather Gang for the link.

Local Leaders Allegedly Ignore Warnings: 1,200 Dead

When Typhoon Washi dumped a month’s worth of rain on Mindanao island in just 12 hours, the result was a series of surging flash floods that barreled straight down the slopes of denuded mountains to hit densely populated areas in Cagayan de Oro and Iligan cities. The local government’s failure to pay heed to national leaders’ warnings about the size of the storm heading their way was another problem, but the aftermath is focusing squarely on the consequences of Filipinos mistreating their environment.
President Benigno Aquino III captured the public mood when he said during a visit to Cagayan de Oro last week that the Philippines hasn’t learned the lessons of the past and must do better to safeguard its environment to avoid future calamities.

Story from the Philippines reported by the Wall Street Journal

How Good, or Bad, Were the Blizzard Forecasts? The Response, Part 2.

Satellite image of snow cover taken at 9:30am.
Light gray is snow except in eastern Oklahoma where it is low clouds.

In Part 1 of the forecast review (below), I discussed the forecasts of the blizzard that began on this blog Friday until the snow began falling in the area of interest Sunday evening.

Now, I’d like to discuss the reaction to the forecast.

Saturday afternoon, I was very confident this was going to be a major storm that would seriously affect holiday travel. So, in addition to posting the forecasts on my blog, I cross-posted that forecast on a railroading blog where I have several friends I knew would be interested. One reader posted a reaction immediately below my forecast:

While it is nice to wish for such storms, I doubt that this is anything more than wishful thinking by the author. 

I have seen nothing on any NWS sites that indicate that anything of this sort is forecast for the areas mentions. And, I live there!

This is a problem that was discussed at last week’s Weather Ready Nation Conference in Norman, OK, that people too often hear/see the warning and fail to appropriately respond. The question is “why”?

The rest of this posting refers to the forecasts of the storm in general via the media, internet, radio, etc., not just the forecasts on this blog (which, according to the traffic counter, reached about 5,000 people).

We learn more on the response from USA Today:

 In Hays [Kansas], drivers who managed to get ahead of the closing still left the interstate earlier than planned, booking three dozen rooms at the Fairfield Inn in a mere 20 minutes Monday night. Greg Boughton, a hydrologist from Cheyenne, Wyo., and his family quit traveling in the afternoon after their SUV nearly slid into a ditch…

Heather Haltli, 29, and her husband were traveling from their home at Hill Air Force Base in Utah to attend a family funeral in Abilene, Texas, but the storm slowed them down so badly that they had to take refuge at the Comfort Inn in Garden City, Kan.
“We’ve been traveling about 20 miles per hour all the way from Denver,” Haltli said Tuesday. She said they had passed up to 15 wrecks including rollovers, upside down cars and jackknifed trucks as they drove through Colorado.
“I don’t think we’ll be able to make the funeral, but we’ll keep going,” she said.
The storm was blamed for at least six deaths Monday, authorities said. Four people were killed when their vehicle collided with a pickup truck in part of eastern New Mexico where blizzard-like conditions are rare, and a prison guard and inmate died when a prison van crashed on an icy road in eastern Colorado…
In northern New Mexico, snow and ice closed all the roads from Raton to the Texas and Oklahoma borders about 90 miles away. Hotels in Clayton, N.M., just east of where the three states touch, filled up. Multiple highways remained closed early Tuesday.
Bill Cook, who works at the Best Western in Clayton, said he hadn’t seen such a storm since the 1970s, when cattle had to be airlifted with helicopters and the National Guard was called in to help out…

The storm Mr. Cook was referring to was in February, 1971.

[second article, same source]

At least 40 people were stranded at the Longhorn Motel in Boise City, Okla., where manager Pedro Segovia said blowing snow had created drifts 2- and 3-feet high and closed the main road.
The Colorado Army National Guard said it rescued two stranded motorists early Tuesday in eastern Las Animas County, in the state’s southeast corner, using a special vehicle designed to move on snow. Smaller highways in that area remained closed.

By putting the numbers in various articles together, the order of magnitude of people stranded was well into the hundreds or low thousands.

Is this too high? Low? We don’t know because this type of research has not been done in meteorology to the extent I, and many other meteorologists, would like. I suspect the number of people stranded was higher than it needed to be.

I guess I’m perplexed as to the apparent “wistful thinking” response and the fact that thousands of people drove into a well-forecast blizzard putting themselves — and rescuers — in peril. Why? It can’t be because they believe their cars can handle the measured 7 to 10 ft. drifts…no car or truck can. Is it because they don’t believe the forecast? Is it they are not aware of the geography (i.e., their route of travel was right through the center of the blizzard forecast)? They don’t think to check the weather for a road hundreds of miles away if the sky is fair at home when they depart? What?

Is it because television weathercasters are hurting, rather than helping, our image? This incident occurred in Los Angeles this morning:

I also suspect an element of the problem is too many are unaware of the amazing progress that has been made in storm warnings during the last decade (like the forecast skeptic cited at the top of this posting). Because people are unaware of how accurate the warnings have become, they are disinclined to act on them.

Breakthroughs in medicine are routine news. And, just yesterday, we learned that astronomers have found two new planets. It was worldwide news. Yet, great progress in meteorology rarely makes news. So, we are still viewed by many as the people “who can keep our jobs while being wrong half the time.”

I certainly welcome the introduction of social science into the field of forecast and warning response. The sooner we get some answers to these questions, the sooner we can save more lives.

If you have any thoughts, please feel free to post them in the Comments.

Another meteorologist chimes in here.

Posted in Uncategorized | Tagged

Have We Solved the Tornado Problem (For Residents of Conventional Housing)?

Today was the second day of the NOAA Weather Ready Nation Conference in Norman, OK.

Yesterday, I talked about the fact that 99% of 2011 tornado-related fatalities occurred in locations that were in both a tornado watch and a tornado warning at the time the tornado arrived. I called that a remarkable scientific accomplishment that virtually no one — inside or outside of meteorology — is aware of.

Want to know how big an accomplishment the 99% is? Consider: It was recently announced that Albert Pujols will receive $250 million dollars over the next ten years to strike out two-thirds of the time*! Pujols’ lifetime batting average is .328, meaning he gets a hit slightly less than one-third of his at bats.

As I have learned more today, I have convinced the “tornado problem” is not homogeneous. So, I want to propose an idea that will be controversial: Weather science has solved the tornado problem for those who live in conventional housing.

Take a look at this graph I presented yesterday. It is a logarithmic graph from NOAA’s Dr. Harold Brooks that depicts the death rate from tornadoes (deaths per million population):

click to enlarge

If you take out 2011, the death rate in conventional homes has been cut more than 99.3% as compared to what it was before the tornado warning system was created.

Unfortunately, the death rate in conventional homes will never be zero, especially in F-4 and F-5 tornadoes.

Since zero is impossible, is it worth spending finite dollars and research resources to try to “move the needle” from 99.3% closer to 100%? Before you answer yes, let me go on.

If you look at the same graph above, the black squares show the death rate in mobile homes. For mobile homes, which house more and more American families, the needle has hardly moved at all. The death rate in mobile homes now is about the same as in conventional housing before the tornado warning system was created.

So, here is what I conclude: While we should strive to improve the accuracy, timeliness, and geographic specificity of all tornado warnings we should stop putting effort into specifically improving the safety of people in conventional homes and apartments. Instead, put that money into mobile home safety.

When resources are finite, good management dictates putting those resources where they will do the most good. In the case of tornado safety, that appears to be mobile homes.

Comments are welcome!

* Thank you, Dave Freeman.

70th Anniversary of Pearl Harbor

It is now December 7 in Hawaii. Today is the 70th anniversary of the surprise attack on Pearl Harbor. The anniversary is being recognized in solemn fashion on Oahu with several survivors in attendance.

I found this article very interesting:

In their youth, on the eve of disaster, they belonged to the Signal Corps Aircraft Warning Service on Oahu. Pvts. Joseph Lockard and Robert McKenney worked at the Opana mobile radar station on the northern tip of the island. Opana had gotten a radar set Thanksgiving Day 1941. Its operators could look out over the Pacific from a height of more than 500 feet…
The remote Opana site had no quarters, so the soldiers who manned it camped several miles away at Kawailoa. On Saturday, Dec. 6, they got a call to operate the radar set early the next morning.
“Joe Lockard and I were the only experienced so-called crew chiefs there,” McKenney said in a 1991 video interview with the National Park Service. “I tossed a coin to see who would draw that duty, to be there to operate from 4-7 a.m. .. I tossed with Lockard and he lost, so he got the job.”
“We spent the night at the site and turned on the equipment and were on line and in contact with the information center at 4 a.m.,” Lockard said. “George was at the plotting table; I was the operator at the scope.
“After the exercise, we didn’t shut down the unit at 7 a.m. because we didn’t have any transportation back to Kawailoa. The truck hadn’t arrived. So I decided to give George some training.
“I started to put him in front of the scope and there it was — this huge echo on the screen. I had never seen any kind of response on the equipment that was so large.
“At first I thought there might have been some glitch with the equipment. So I checked everything I could and everything operated OK, so it had to be real. There had to be something out there.”
The blip was 136 miles out and closing fast. It was 7:02 a.m.
Elliott tried to call the information center but couldn’t raise anyone on the plotters line because the plotters had all gone to breakfast at 7. He used the administrative line to call the switchboard, and Pvt. Joseph McDonald answered. McDonald, from Archbald, Lackawanna County, near Scranton, and Lockard were friends.
“Joe told us that everyone had left the building,” Lockard said. “We asked him to look around and see if he could find anybody, and he did. He found a young Air Corps lieutenant, Kermit Tyler, and brought him to the phone.
“I talked to Kermit Tyler and tried to convey my excitement at the fact that we had never seen anything like this on radar, and that it obviously had to be planes. . I didn’t have any idea how many. I pushed it as far as I could, but you can only argue with an officer so long.
“He just said, ‘Don’t worry about it.’ “

Very Dangerous Tornado Situation

Radar at 5:13pm. Tornado located near arrow’s tip, moving ENE toward Kingston but Holly and Frierson are both in danger. Strong rotation indicated.

To help orient you, I’ve highlighted Shreveport at the top.

Note: This is the storm that was the subject of the item two postings down.

UPDATE: RADAR from 5:31pm, tornado passed over or just west of Kingston, now crossing I-49 and moving toward or just south of Frierson.  Take cover in these areas.

More On the Extraordinary Tornadoes of 2011

The Birmingham News did a well-balanced story on my theory that the power failures caused by the line of thunderstorms that moved across Alabama the morning of the 27th played a role on the large death toll from the tornadoes later in the day. I found this paragraph to be interesting:

The Alabama Emergency Management Agency expressed doubts about the deadly impact of that day’s power outages because the potential for particularly nasty weather had been widely publicized throughout the week. “It wasn’t like it shocked all of us,” said Yasamie August, a spokeswoman for the agency.

There is a very big difference between, “severe thunderstorms will likely occur on Wednesday” (spoken by a weathercaster on, say, Tuesday evening) and, “A tornado is headed for your neighborhood! Take cover now!”, particularly when you have been distracted by dealing with an all-day power outage.

There is a second article from yesterday’s Wichita Eagle about the use of sirens in a wake of the National Weather Service’s report on the Joplin tornado. I’m quoted about using sirens only for tornadoes and only in areas actually in the path of the storm. Jim Schmidt, emergency manager for Butler Co., KS, says,

But Schmidt said he understands and endorses the use of tornado sirens when strong winds threaten areas where camping and boating are popular.

Butler County sounds the sirens in El Dorado and next to El Dorado Lake any time confirmed winds of at least 80 mph are moving into the area.

“We can have 40,000 people at the lake,” Schmidt said. “If there are 80-mile-an-hour winds coming and you’re in a camper, it is a whole lot different than if you’re in a nice brick house.”

Jim makes a good point and I agree provided the sirens are only sounded near the lake and that signs are erected at the Lake’s entrances to explain the policy. Elsewhere, I believe it is bad idea to sound sirens for severe thunderstorm warnings.

In the comments to the Eagle’s article a man name Jeff Johnson wrote,

I live in Joplin, a few blocks from the EF-4 destruction when the tornado first started. I had been watching the storm on radar for at least 2 hours, had the TV on as well. I heard the first tornado warning sirens go off, but the rotation was to the north and it would pass to the north of my location and the couplet wasn’t all that impressive. Some of the neighbors had gathered outside and were talking at the time. I wasn’t prepared for what happened after that though. A new cell popped up to the south of the first cell. I caught this on the radar and noticed that it was forming really strong rotation. A meteorologist on TV cut in and circled where the rotation was and it was almost directly over my area. I was able to see that rotation on radar and it was directly west of my area. I decided to look out a west window but couldn’t see much other then the sky being pitch black with a wall cloud in front of it. It looked like night. A few seconds later I heard what sounded like rolling thunder, only it didn’t get quieter, it got louder. Then a few seconds after that, the second tornado sirens sounded and I knew for sure it was a tornado and took shelter in a closet. Looking back on it now, if the tornado had been a few blocks closer, I’m not sure I would be here. I’m getting a tornado shelter later this fall/winter though, that’s for sure.  [emphasis mine]

In the series of tornado education seminars for business that I have been doing since May, I say,

Don’t Believe Your Eyes, Believe the Tornado Warning!

As Mr. Johnson confirms, Doppler radar is a powerful tool and is far superior to the untrained eye. Please take shelter whenever the sirens sound!

NWS Report on the Joplin Tornado

After major weather-related disasters, the National Weather Service conducts a “Service Assessment” to understand what can be learned and what can be improved with regard to their forecasts and warnings.
The NWS just released its Joplin Service Assessment pertaining to the storm that killed 162 on May 22.   
Several of their key findings include (my words in bold, assessment language in italics): 
False Alarms of Severe Storms are a Real Problem
It was common in the interviews to hear residents refer to storms always blowing over and missing Joplin, or that there seemed like there was a protective bubble around Joplin, or ―there is rotation all the time, but never in Joplin…
Sounding Sirens for Severe Thunderstorm Warnings Caused Complacency  
the perceived frequency of siren activation (false alarms) led an overwhelming number of participants to become desensitized or complacent to this method of warning. Many noted that they hear sirens all the time[sirens] go off for dark clouds, they are bombarded with [sirens] so often that we don‘t pay attention, the sirens have gone off so many times before, sirens are sounded even for thunderstorms, and all sirens mean is there is a little more water in the gutter.
Because People Want to “Confirm the Threat for Themselves” – the Invisible Nature of the Rain-Wrapped Tornado Misled People into Believing the Siren Was Sounding for a Severe Thunderstorm Warning.
While searching for additional information concerning the severe weather threat constitutes ―taking an action,‖ the actions many residents described taking were not the immediate life- saving measures desired with the issuance of a tornado warning. In most cases, these life-saving actions, or the decision to find shelter, were associated with additional extraordinary risk signals. This was generally achieved in different ways, including:
a. Physical observation of the environment (seeing the tornado approach). 
While significant numbers of people actually did this, the approach was complicated by having a ―rain-wrapped‖ tornado that made the tornado more difficult to recognize until it was very close. There were numerous accounts of people running to shelter in their homes just as the tornado struck, despite significant advance warning of the risk.
The report confirms my recommendations that sirens should only be sounded in extraordinary circumstances (tornado) and then only in the areas directly at risk. The multicounty siren activations in areas like St Louis lead to complacency and the opposite public response from what emergency management would like.
To individuals, take shelter when the warning is issued. Don’t run outside trying to figure out the threat.

How Good Are Tornado Warnings Becoming?

Before Doppler radar, it was virtually impossible to track and warn of the small tornadoes created by tropical storms and hurricanes.

How far we have come!

Today, as an experiment, I did some tweeting on the rotating thunderstorm that moved across the northwest suburbs of Atlanta. The radars (NWS WSR-88D and FAA’s TDWR) both did a great job locating the storm and I posted this image at 2pm CDT.

Doppler wind data showing rotating winds over
the northeast part of the Atlanta Metro region.

Note that the arrow is pointed right at Woodstock.

So, I tweeted the warning based on this and other data (not shown).

As best we can tell at this point, the tornado touched down several times and damaged homes, trees, and power lines.

Chuck Blevins, Atlanta-Journal Constitution

More information from the Atlanta Journal-Constitution here.

According to the NWS, the tornado damaged the homes at 2:15pm CDT (3:15pm EDT). That means, I (or another meteorologist looking at the situation) was able to specifically name Woodstock and the immediately surrounding area 15-minutes of advance warning.

A dozen years ago there would have been zero warning. Weather science has come a very long way. 

"Was Vermont Warned?"

There are now two news stories (one yesterday, one today) complaining that Vermont was not warned of the danger of Irene. As I state below, my head was down in the data for days and so I cannot speak for anyone else. But, I can state that Vermont was included in the warnings Tweets that I was sending out.

How explicit were the warnings?

For four days, I — and others — were warning of the likelihood of flooding in New England. I posted this Saturday along with map outlining, in red, an area of “major flooding” that included Vermont.

Calling, in advance, for evacuation preparations for a 200-year flood plain width around rivers is indicative of an extreme hazard. I know the NWS had a flash flood watch out for the area. When I wrote about all of the hazards meteorologists were trying to think through, this was one of the hazards I was talking about.

So, as tragic and devastating as the record flooding has been, it is not accurate or fair to say there was “no warning.”

Comment on "Tornado Warning Fatigue"

Welcome, Instapundit Readers!  Please feel free to look around while you are here.
Thanks for the link, Glenn.

Glenn Reynolds over at the ubiquitous Instapundit writes:

We were awakened in the middle of the night last night by a tornado warning for a storm that never even came close to our house, but that hit north Knox County pretty hard. I’m happy to have the weather radio to warn us, but we’re beginning to suffer tornado-warning fatigue.

What Glenn is referring to is that if the weather radio is not programmed correctly, it will go off throughout the night for storms you don’t care about in areas where you do not live causing you to be awakened in the middle of the night unnecessarily.

Before you purchase a weather radio, make sure the store will program it for you and have them do it before you leave the store! FIPS and S.A.M.E. codes — some of the exotic languages of meteorology — are things you shouldn’t have to worry about, let the store do that. If the store says they cannot program the radio, purchase it somewhere else. I’ve found Radio Shack does the job well.

If you already have a weather radio, take it back where you purchased it and have them program it for your specific location.

Otherwise, you’ll be like so many others that get tired of losing sleep and will unplug it or throw it away and that is dangerous when the “real thing” occurs. 

Here is a story about the storms.

Addition to the above posting: Glenn says he has his radio set for all but the highest level but there have been too many tornado warnings this year.

I’m currently in Oklahoma City attending the American Meteorological Society’s first-ever conference entirely devoted to storm warning techniques. This has been the worst tornado season since 1952 in terms of deaths with well above normal numbers of violent tornadoes. That is, more than anything, what accounts for so many warnings this year. There has been a huge number of storms.

All of us are hoping things get back to normal soon.

What Should Meteorologists Do?

I had not even gotten home from watching last night’s storm when the first critical email about the tornado warning in Sedgwick County hit my inbox. I’d like to spend a moment on the subject of the meteorologist’s dilemma in these situations.

Yes, it was a “false alarm.” A tornado never made it to the ground.

The storm was a “right mover” (tornado indicator), had a hook echo (tornado indicator, image below) and had rotation as measured by Doppler radar (tornado indicator). To the naked eye, there was lots of organized, sustained rotation.

Yet, for reasons meteorologists do not understand, the funnel cloud never made it to the ground. That is good — no tornado damage and no one injured. But, it left people grumping about having to interrupt their activities because of the warning.

In a 2003 paper (available here) I differentiate between “unavoidable” and “unnecessary” false alarms.  The former is caused when Mother Nature throws us a curve ball. The latter is when meteorologists know a given area is not threatened but limitations in the warning system cause non-threatened areas to be warned. Last night was some of each.

Take a look at the red polygon which is the National Weather Service’s path-based tornado warning. The hook (funnel cloud) location is within the polygon. The area inside the polygon experienced an “unavoidable” (from my point of view) false alarm because the science of meteorology cannot determine why this situation was not a genuine danger. So, to be safe, we warn.

But, look the radar image again and you see a dot and the word, “Wichita.” That is downtown Wichita where the annual Wichita River Festival was in progress. Because Sedgwick County sounds the sirens countywide, the River Festival and its patrons were subjected to an unnecessary false alarm due to the structure of the county’s siren network. It is “all or nothing.” That put River Festival officials in a no-win situation. They made the right decision (in my opinion) to shelter.

Fortunately, Sedgwick Co. is in the process of changing the configuration of its siren network so that by tornado season 2012 the sirens will only go off in the path of the storm.

The initial tornado warnings last night were absolutely justified given the imperfect state-of-the-art. I’m sure there are plenty of residents of Joplin, Tuscaloosa, Birmingham, Minneapolis, Raleigh, St. Louis, etc., that wish their tornado warnings this year had been false alarms. This is life-and-death business.

That said, I’m empathetic to the people who took shelter for what seems like no good reason. I’m hopeful we will achieve some breakthrough in the science that will allow us to cut down on the unavoidable false alarms.

How Do We Track a Massachusetts Tornado? With the Same Techniques We Use in Kansas

Already today, I have seen one comment that the Massachusetts tornado was “unpredictable.” While this has been a horrible tornado season (worst in more than a half century), the tornadoes have been extremely well predicted.

That said, I have also received a number of questions about how the long-track Massachusetts tornado was forecast and tracked and I’m happy to answer them. The graphic below was posted on this blog the evening before the tornado. Western Massachusetts is in the 30% relative probability area for tornadoes and severe thunderstorms which is a high number for the day before an event.

The graphic below was posted on my AccuWeather Pro blog (where I go into the deeper meteorology than I do in this public blog) about three hours before the tornado formed. I call the probability of tornadoes “very high.” The oval was where I was forecasting the high tornado risk. A tornado watch was issued for the public about twenty minutes later.

The tornado’s formation, from a “supercell” thunderstorm, evolved as it would in a Kansas tornado. Here is the radar at the touchdown point of West Springfield complete with “debris ball” (circled) which is a signature of a tornado on the ground.

The radar’s Doppler wind display was very helpful throughout the afternoon showing the tremendous value the public has received from the National Weather Service’s investment in Doppler radar in the 1990′s.

The rotation track of the Massachusetts tornado. The radar, south of Boston,
showed high-speed winds away from the radar (pink) next to high-speed
winds toward the radar (curved arrows). I have marked the
previous centers of rotation. There was even a “hail spike” indicating
the very large hail that fell with this storm. 

The last well-defined debris ball was east of Fiskdale at 5:18pm EDT.

AccuWeather’s SkyGuard® meteorologists were tracking the tornado and informing clients throughout the afternoon. The National Weather Service issued warnings for the public, a number of which were the topic of special reports at AccuWeather.com and on this blog.

Not only did we use “Kansas” techniques on this tornado, a friend of mine from Wichita was vacationing with his family in Boston while this was going on. He wrote,

Thanks, Mike. As luck would have it, we passed on a side trip to Sturbridge and are somewhat north of the primary storm. Glad I looked in on your blog and moved my family out of harm’s way…

In June, 1953, before a tornado warning program existed in the U.S., a tornado struck Worcester, MA killing 94. It is likely that yesterday’s forecasts and warnings kept the death toll (tragically, four) lower than it otherwise would have been.

"You Never Know What Path the Storm is Going To Take"

This report from KMOV TV aired yesterday illustrates the apparent gap between meteorologists and emergency managers on the topic of selectively warning of the path of tornadoes. Another example of the gap is here.

I don’t ever wish to sound like I am “picking” on emergency managers. They have a difficult, “darned if you do, darned if you don’t” job. That said, it does appear that there is a significant gap between meteorologists and emergency managers with regard to the the NWS “storm-based” (or path-based) warnings that became official more than five years ago.

Mr. Mark Diedrich, of St. Louis County Emergency Management, says in the report, “You never know what path the storm is going to take.” That is simply not true. To illustrate, below is the storm-based warning (purple) issued by the National Weather Service on the St. Louis Good Friday Tornado when it was approaching Lambert Airport (an earlier warning covered the tornado prior to this time). I have added the location of the tornado at the time of the image (circled) and the path it actually took (arrow).

The tornado itself is just under the “rid” in Bridgeton Terrace and is
purple in color along I-70. 

Note that the polygon is far, far wider than the path of the tornado itself so that a margin of safety is built in. Even with sounding the siren only in the areas covered by the polygon the sirens would be sounding far to the south in Clayton and downtown St. Louis. I don’t mind that because it is part of the “margin of safety” built-in by the National Weather Service. But to sound the sirens every time any part of a county (or adjacent county!) is threatened is unnecessary and breeds complacency.

Selective Siren Activation, Part 3

Shawna and I drove through downtown Joplin minutes before the tornado hit. Sirens were going, but we could not see the tornado, and many people were out and about seemingly unaware or unconcerned. We had been following this storm complex for a couple hours without observing any tornadoes, and though we were a little nervous, we weren’t anticipating anything of the magnitude that buried Joplin around 5:45 pm CDT.
Storm chaser and meteorologist Jon Davies, from his blog here.


Guest, an eye-witness I spoke to said people at a driving range kept right on hitting golf balls even as the tornado sirens were blaring.
Stan Finger, Wichita Eagle web chat, Monday.

a CNN Wire report quoted Alexa Wattelet, in Joplin at the time the storm hit, as saying that “the sirens always go off, so no one thought anything of it.”

Joplin resident Rick Morgan thought about it before taking cover. 

“They go off, and it’s like, you know, tornado never comes, it seems like,” he told CNN on Monday.

Kansas City Star

What might cause this seemingly irrational behavior?  I have a theory: We have sounded the sirens so often that people have come to ignore them. I’ve written two posts in the last six days on this topic. Go here and here to read them.

From 1957 to 2005, the National Weather Service issued tornado warnings by whole counties. This was appropriate because our skill at locating and tracking tornadoes was not very good. So, alerting relatively large areas made sense.

With the advent of Doppler radar in the 1990′s and more accurate short term prediction tools, the National Weather Service switched to storm-based warnings, i.e., warning only the area(s) in the path of the storm. An illustration of the advantages is below:
Click to enlarge.
While visiting St. Louis earlier this week, I learned that they have a policy of sounding tornado sirens not only in all of St. Louis County when a warning is issued for any part of the county, but sounding the tornado sirens in all of St. Louis County when a tornado warning issued for an adjacent county!
In this hypothetical example, sirens would be sounding in Mehlville and Oakville
nearly 50 miles away from the tornado!
On Wednesday, a tornado warning was issued by the National Weather Service for a funnel cloud that followed the path indicated by the arrow. Photo is the actual funnel (located at the arrowhead) as viewed from our hotel (purple pin). The NWS issued an accurate warning for the path of the storm. 
Path of the funnel cloud Wednesday with actual
photo of the funnel cloud we tracks from South St. Louis County
Yet, tornado sirens went off over all of St. Louis County. I’ve circled some of the cities in which the sirens were sounded (there were many more) that were not threatened at any time.  
One would think this would prompt an examination of procedures since a false alarm occurred that affected more than 500,000 people. Unfortunately, it did not. KMOX radio interviewed me and interviewed St. Louis County officials.  The story is here. Here is what the emergency managers had to say,

Defending the policy, the Acting Director of St. Louis County Emergency Management,  Bill Roach, says the sirens mean one thing.

“Those sirens are an early warning device,” Roach said, “They don’t mean people should immediately run down in their basement and seek shelter.   What they mean is you should seek additional information.”
Roach says the current policy saves lives, because tornados can change directions, and trying to guess where to selectively warn people could lead to deaths. 

Trying to guess? This isn’t 1964. Meteorologists know the direction of movement of these storms. Plus, the National Weather Service adds a margin of safety. There is absolutely no scenario where (to pick one example) Ballwin was at risk from that storm. 
Saves lives? I believe this gross overwarning risks doing the opposite. When you sound the sirens time and time again in areas where there is no threat, people stop paying attention. I saw it in Lawrence and Overland Park Saturday and again in St. Louis on Wednesday.
It is long past time to rethink countywide and multi-county tornado siren activations. Sirens should only be activated in areas where there is a genuine threat.