Issues With the Forecasts and Warnings of the 2025 EF-3 St. Louis Tornado; With Update II, There Was No Advance Warning

St. Louis is no stranger to major tornadoes and the deaths they can produce. It is the metropolitan area where the most tornado-related fatalities have occurred (Based on an AI search, more than 420). Friday, an EF-3 intensity tornado occurred in St. Louis and then moved across the Mississippi River into Illinois. Five were killed in St. Louis, 38 injured, and an estimated $1.6 billion in damage occurred. The tornado was a mile wide in places. 

As we have been reporting for nearly 15 years, major issues have developed in the National Weather Service's (NWS) tornado warning program. Unfortunately, some of those issues occurred again Friday. There were also serious issues with local emergency management. 

Unlike other nations, the U.S. does not have a National Disaster Review Board (modeled after the hugely successful National Transportation Safety Board) which would be able to do a comprehensive report. So, I am providing this more limited report in hopes that lessons can be learned in other communities with a goal of mitigating future injuries and losses. 

National Weather Service's Boast About Its Performance 

Because of writing so much about the NWS's problems, especially the acute staffing issues the past four months, I was originally going to let these latest issues pass. However, that changed when a NWS official boasted about his agency's performance Friday. It is concerning the NWS believes its performance was outstanding not only in St. Louis but in several other locations where issues arose. However, I am going to focus on St. Louis because of my familiarity with the area. 

The Forecast

Dr. Bieda's boast about the forecast isn't timely because he presents the Storm Prediction Center's 3pm outlook which is after the regional tornado outbreak had begun. Regardless, there was an issue with this forecast because, while close, it did not have the path of the St. Louis tornado in the "strong" tornado forecast area (strong is defined as EF-2 to EF-5 intensity) even though the St. Louis tornado was only about 40 minutes in the future. 

If he had chosen to discuss the tornado watch, that would have been fine as it was timely and included St. Louis. The tornado watch was issued about 2.5 hours before the tornado occurred. 

Since 2011, the NWS has been touting a system to make short-term forecasts of major tornadoes ("Warn on Forecast" or WoF). An experimental WoF product suggested a tornado would move across parts of the St. Louis Bi-State area that afternoon. It would have been available about an hour before the tornado began. While the tornado is depicted slightly too far north in Missouri (it was excellent in Illinois), this should have had the local office ready to issue a tornado warning at a moment's notice.

Issues With the Warning

In fairness, this was not the most obvious tornado warning situation. It is also likely the warning, such as it was, saved lives since it was a strong tornado in a densely populated area and it didn't look like a typical tornado to the general public. Note the lack of tornado photos with this event. This video shows an apartment being destroyed but you don't see a typical funnel cloud as the storm approaches, nor do you see the typical "funnel/tornado" shots from a mile or so away. Like Joplin, people could not warn themselves and were completely dependent on the warning system. 

Still, it is important that tornado warnings be issued before a tornado actually begins. The NWS's published goal is 13 minutes of "lead-time" (the interval between when the warning is issued and when the tornado begins). That was not achieved Friday afternoon.

Below is a copy of the tornado warning. It was issued at 2:34 pm CDT. 

Update II, 8:30pm Thursday, May 22:  Please go here. It shows dash cam video of the tornado in progress as the warning is being issued. Unfortunately, the evidence is pretty strong that there was no or, at best, very little lead-time on this tornado warning. Combine this the damage father southwest (about 2:30pm on TDWR image below near the suburb of Des Peres) and the tornado was clearly in progress prior to 2:34pm. 

AccuWeather issued a tornado warning for its business clients in the path of the St. Louis tornado at 2:16pm, that was 18 minutes before the NWS. 

Besides not achieving the 13 minute goal, the tornado "located over St. Ann, or over Overland...." is simply incorrect. Radar clearly shows insignificant rotation near St. Ann. The dangerous rotation was 7 miles to the south -- which was a near carbon copy of the NWS's Joplin error when they said the tornado was near Carl Junction when it was near Riverton (10 miles to the southwest). 

Also like Joplin, people were likely mislead by Friday's warning message stating the wrong location. It says the direction of movement was "east." The actual direction of movement was east northeast which is important because of the incorrect location of the incipient tornado -- different areas were actually affected rather than the areas that would have been affect based on the information in the warning message. 

Based on data from both the St. Louis NWS WSR-88D and the Federal Aviation Administration's (FAA) Terminal Doppler Weather Radar (TDWR), a warning on the dangerous rotation could have been issued as early as 2:29pm instead of the actual warning time five minutes later. The TDWR data below shows the difference.

TDWR data depicting the rotation centers at 2:29pm when a warning was first possible
and 2:34 when the warning was actually issued. 

Interestingly, a friend of ours who lives in Des Peres (city name not shown), near the south edge of the rotation's path, had tree and roof damage; so clearly a tornado or damaging winds were in progress by around this time.  

The "confirmation" of a tornado statement -- and emphasizing the life-threatening nature of the storm -- was not issued until 2:50pm which is later than I would like to have seen. 

One of the likely reasons for this delay was the radar was not in 80-second "tornado mode." It was set to an inappropriate 3-minute (180 second) mode. This is a frequent error by local National Weather Service offices these days.

I spoke with a meteorologist at the St. Louis NWS who told me how they determined the time of the beginning of the tornado. They said they compared the radar location of the tornado to where the damage first occurred. Below is the radar at 2:41pm, the beginning time of the tornado per the NWS.

Based on a new examination of the radar data, I'm confident the tornado was in existence prior to 2:41pm. At 2:39pm, a "debris ball" -- caused by lofted debris -- is visible on the TDWR (circled, left panel).

Even earlier, lofted debris is visible on the NWS's WSR-88D.

The dark blue area is where the NWS's radar shows the start location of the tornado -- which matches the NWS's ground survey --  although the path may have begun a little farther southwest; and that the tornado began about 2:37. This means the tornado warning's lead-time was only three minutes. 

To summarize, the warning messages:

• Initially highlighted the wrong location.
• Was issued later than it could have been. 
• Had just a three-minute lead-time. 
• Had the wrong direction of movement. So, like Joplin, the area actually affected was different than the area that would have been affected based on the information in the warning 
• The Severe Weather Statement said the tornado at 2:50 was over "Venice." I was a meteorologist/resident of St. Louis for five years (TV and radio) and I don't recall a city named Venice in St. Louis County. So, I looked it up and neither Apple Maps nor RadarScope have a city named Venice in that area. The NWS should be using well-known locations.

This is not a terrible performance based on overall NWS tornado warnings of the past decade or so. However, it certainly is not one the NWS should bragging about. 

St. Louis Emergency Management's Major Issues

Local emergency management completely dropped the ball.

While the tornado was still in progress, social media was filling with "I didn't hear sirens" or "where were the sirens?" reports. Why not? The sirens were never activated!! And, they weren't activated because the emergency management staff -- in spite of the tornado watch out for 2 hours and 40 minutes -- was offsite attending a seminar. 

I placed a phone call to the CEMA (the name of the EM agency in St. Louis) and they asked me to send an email, which I did. They evidently have chosen not to answer. However, I have managed to confirm they were out of the office based on one news report and a second person who has requested our conversation be off-the-record. 

The mayor has expressed her extreme displeasure and has taken siren activation away from emergency management and given it to the fire department. 

This is similar to the awful situation on Maui when the NWS went over and above the call of duty. Four days before the catastrophic wildfire in 2023, they warned the emergency management department and other officials of an "unprecedented" wildfire risk. They updated that warning.

What did Maui Emergency Management do? All eight employees got on a plane to Honolulu for meetings and none of them were on Maui when the fires broke out. The sirens were not sounded in Maui, either.

There are similar documented similar cases all the way back to Katrina (2005). 

One can legitimately ask, "What do emergency managers think they are getting paid for?"

Television Storm Coverage

Via the internet, I was watching some of the television coverage of the storms as they occurred. I have also talked with one of the meteorologists after the fact. I have frequently referred to TV mets as "backstopping" the NWS. In this case, like many others recently, the TV coverage was out ahead of the NWS. 

On numerous occasions the past nine months, we have talked about the major changes needed in the NWS and, especially, NOAA. Perhaps it is time to face fact that the NWS is often not the lead meteorological agency to warn of tornadoes. 

Personal Responsibility

There is a story of golfers "trapped" on the Forest Park Golf Course as the tornado passed by. They complained (validly) about the sirens being silent. However, with a tornado watch in effect, a severe thunderstorm warning in effect, and high amounts of lightning to the west what were they doing on the golf course to begin with?!  

Yes, the NWS was subpar. Someone should probably be fired in emergency management. But, it is past time to consider that people should take Mother Nature's obvious cues and take responsibility for themselves when they are outdoors and it is obviously unsafe. 

This incident cries out for a detailed, expert investigation -- which can only be done by a National Disaster Review Board. Congress? Trump Administration? Are you paying attention?

Update #1, Additional information 11:50pm May 20:

Via John Farley: "Venice" is in Illinois near Granite City. The NWS should have used GC to describe the location of the tornado as it is much better known. 

Illinois never occurred to me because they were still touting the tornado warning behind the tornado in Missouri! Parts of the still-warned area are more than 12 miles behind the tornado. 

The CEMA EM Administrator is in serious trouble. Via KSDK TV.