Are cars really a safe option in tornadoes?

Image: FEMA photograph by C. Doswell

Posted: 11 January 2003 Updated: 12 May 2009: revised and added to the content.

As usual, the following represents only my opinions. Feel free to challenge me at (use the email link or cut and paste the address and replace the # with '@') if you're so inclined.

When I first posted this in 2003, there had been something of "media blitz" associated with publications by Prof. Tom Schmidlin of Kent State University concerning the issue of the safety in automobiles. He and his colleagues are questioning the validity of the existing safety recommendations regarding automobiles in tornadoes. Having been asked questions about this on several occasions, I feel the need to attempt to clarify the situation, at least as I see it. In a recent article about the 3 May 1999 tornado, Prof. Schmidlin and a colleague said:

The common notion, born in the aftermath of the 1979 Wichita Falls tornado, that being in a vehicle during a tornado warning is synonymous with injury and death can no longer be assumed or promoted. Although high numbers of deaths have occurred in vehicles during tornadoes, the occurrences are rare (Hammer and Schmidlin 2001 ).1 The use of vehicles to flee the path of the Oklahoma City tornado has shown that people can successfully make reasoned decisions that reduce the risk of injury and death. Many people recognized that it was safer to be in any location outside of the tornado path, including a vehicle, than to be in a house within the path.

Source: Hammer, B., and T.W. Schmidlin, 2002: Response to warnings during the 3 May 1999 Oklahoma City Tornado: Reasons and relative injury rates. Wea. Forecasting, 17, 577-581.

As discussed earlier2 by Prof. Schmidlin and another colleague, there is some cause to question the validity of the existing safety recommendations. Relatively little hard research has been done, especially with respect to the relatively safety of different shelter alternatives. There's a good reason for this lack of research, however: it's challenging to imagine an experiment whereby the relative safety of strategies can be tested under realistic conditions! There's no way to go back and find out after the fact what the outcome would have been if a person had made a different choice in the same situation. The winds and destructive power of a tornado are hard to simulate in a wind tunnel or other controlled physical experiments. And in this case, it's not clear just which among the various options is safest and by what margin the degree of safety is increased relative to the other options. Instead, inferences have to be made from existing data about the distribution of deaths and, when the relative safety differences are not a strong signal, those inferences can be questionable. I believe there are reasons to choose not to support changing the recommendations at this time.

Recently (01 May 2009 update), the American Red Cross has published some new safety rules for tornadoes. In particular, they say

If you are caught outdoors, seek shelter in a basement, shelter, or sturdy building. If you cannot get to shelter, a recent study* suggests doing the following:

Get into a vehicle, buckle your seat belt, and try to drive at right angles to the storm movement and out of the path.

If strong winds and flying debris occur while you are driving, pull over and park, keeping seat belts on and the engine running. Put your head down below the windows, covering with your hands and a blanket if possible.

If you are unable to get to a building or vehicle, as a last resort, lie in a ditch or depression and cover your head with your hands.

*Schmidlin T., et al, 2002: Unsafe at any (wind) speed? American Meteorological Society, 1821-30

This advice seems to have accepted Prof. Schmidlin's advice unilaterally. Apparently, The Weather Channel has followed suit. I disagree with these proposed changes - I continue to believe it's ill-advised to be changing our recommended safety precautions while the validity of Prof. Schmidlin's research is still at issue.

The distribution of deaths and injuries in tornadoes is reasonably well-known, of course. Consider the following chart:

Data Source: Storm Prediction Center

There has been some effort to establish the locations of fatalities, especially since the deadly tornadoes of 10 April 1979 in Texas and Oklahoma. In that event, it was found that many of the fatalities occurred in automobiles. Many people fled homes (in some instances, that were not struck by the tornadoes!) only to die in their vehicles. Some people drove right into the tornado, apparently not recognizing what they were encountering. Based on this experience,3 the NWS and other agencies decided to encourage the public to abandon their vehicles when encountering a tornado, to seek whatever shelter they could find. Since the 1979 disaster, no tornado has produced a similar high percentage of fatailities in automobiles. It can be seen from the graph that most people die in homes ... either mobile homes or "permanent" (site-built) homes ... see below for a brief discussion of the mobile home issue. These data should not be taken to mean that staying at home is a risky option! Rather, it means that most people experience tornadoes while in their homes. In only a few years during the period of record (1985-2008), a relatively large number of people died in in businesses, or schools, or churches, or even outdoors. In the Wichita Falls tornado of 1979, the majority of fatalities occurred in vehicles, but since then, that experience has not been repeated. Does this mean that the 1979 event was a fluke, or does it mean that people have been listening to the safety recommendations and so choosing to abandon their vehicles, as recommended? Frankly, I can't say for sure. But these findings do not make me ready to endorse a change to the "abandon your vehicle" recommendation and to recommend instead that people stay in their cars! If we encourage people to ride out tornadoes in their vehicles, we could see an increase in vehicle-located fatalities.

The existing FEMA safety recommendations (consistent with those from the NWS) concerning what to do in a car are:

If in a car:

I've never been completely in agreement with these recommendations - especially the part about not trying to escape the tornado in a vehicle - but I don't know exactly how to replace them with comparably simple statements. Simple statements in a "call to action" can never replace situation awareness - making the correct choice among your options depends on your specific situation! Storm chasers regularly evade tornadoes in open country without much difficulty, so I don't believe it's necessary to abandon your vehicle just because you see a tornado. We chasers routinely outrun tornadoes on open roads. An intelligent, responsible storm chaser knows enough about storms to evade tornadoes with relative ease. But not everyone knows enough to do what a knowledgeable storm chaser can do. Thus, I'd summarize my recommendations for people who encounter a storm while in cars (or other vehicles) in the following way:

Be aware of your situation!

Your options depend on the specifics of your situation, and an important issue is how much time you have. The fastest and safest thing to do is to avoid being in the tornado's path. A vehicle can take you out of a tornado's path in seconds, but you must know in which way the tornado is moving.4 Once you've determined its movement,

Like Prof. Schmidlin, I believe that simply abandoning a motor vehicle upon seeing a tornado is not necessarily good advice. It's too simplistic. However, I'm not ready to change the formal recommendations just yet. Perhaps further research will validate Prof. Schmidlin's point of view, but I'm quite reluctant to reverse our existing policies about vehicles. A change of the rule to recommend riding out a tornado in your vehicle is also too simplistic! My surveys of tornadoes have shown me that cars are not safe places to ride out a tornado (as shown dramatically by the photo leading off this essay). They might be marginally better than being in a ditch or under an overpass, but only marginally. They might also be marginally better than in a mobile home, but I have my doubts that you're much better off in a vehicle than being in a mobile home.

The mobile home issue

There's clear evidence that mobile homes are dangerous places to stay in a tornado. An F2 tornado will sweep away most mobile homes (and is likely to at least tumble most vehicles). Would those mobile home residents be much safer sitting in a car in their driveway? I have serious doubts about that. If the choice is that of mobile homes vs. site-built frame homes, there's no doubt about the relative safety, because mobile homes represent less than 10% of the homes in the United States (in most areas ... however, in the southeastern U.S., more than 10% of the homes are mobile homes ... see here for more discussion of the Southeast), and yet mobile homes now (as of 2002) account for about half (or more) of the annual tornado-caused fatalities. My colleage, Dr. Harold Brooks, and I have done some study of this and the results are compelling. Unlike the issue raised by Prof. Schmidlin, the existing data support strongly the assertion that mobile homes are many times more unsafe than "permanent" homes. But that says nothing about the relative safety in a vehicle. People living in a mobile home without access to a proper shelter need to understand that they need a tornado plan whereby they can get to an adequate shelter in a few minutes. A vehicle is not an adequate shelter, but if we give out the recommendation that it is, I'm concerned that mobile home residents would conclude that they can use it for their shelter option and abandon the search for a better alternative.

Prof. Schmidlin's advice for people in mobile homes is to drive to a place of shelter - on the face of it, that's not inconsistent with what I'm saying. His work suggests that F2 tornadoes that have swept away mobile homes often leave vehicles in the driveways, implying that the residents would be safer in the vehicle than the mobile home. I'm not so certain about such an interpretation of his findings. I don't know what winds those vehicles actually experienced, and I don't believe he does, either. If a mobile home residents don't have a shelter available in their immediate vicinity, then staying in the mobile home or getting in a vehicle parked in the driveway are roughly equal options in my opinion ... both are bad options! Lacking a shelter and given enough lead time, I believe it's reasonable to leave the mobile home and drive to a nearby place of shelter (but see below). If the lead time is long enough to provide enough time to drive to a shelter location, there is some non-trivial possibility that the tornado could miss the mobile home, of course. Mobile home residents without shelter in the immediate vicinity of their mobile home need to be much more cautious in tornado warning situations than residents of site-built homes - they should be prepared to drive to shelter early, before it becomes highly probable that they'll be struck. Waiting for a high degree of certainty of being in the path of an approaching tornado means that they likely will lose the opportunity to leave at all, and be forced to choose only between bad alternatives.

Another issue raised by Prof. Schmidlin is leaving a mobile home to take shelter in a ditch. I agree that being outside is not a good option. Any difference in safety between staying inside a mobile home and lying in a nearby ditch is going to be minor. Prof. Schmidlin suggests that riding out a tornado in a car is probably safer than either staying in a mobile home or lying in a nearby ditch. Although I acknowledge he might be right, I don't believe the difference is as much as the difference between a mobile home and a permanent home. As already noted, I have my doubts that the car is significantly safer than the other two options, unless that car can be used to get the mobile home residents safely out of the tornado path.

Driving to safety?

The data above suggest that the relative safety in motor vehicles compared to that in mobile homes is greater (roughly 10 percent of fatalities in vehicles annually compared to more than 50 percent in "homes") ... but keep in mind that the data reflect the fact that most people experience tornadoes in homes, not in vehicles. More people die in homes than in vehicles because the overall number of people caught in vehicles is much lower than the number caught in homes, in general. This might not be the case for every event ... a tornado that tracked down a freeway during rush hour traffic "gridlock" might well cause hundreds of casualties in vehicles. As already noted, the advantage to a vehicle is that, under the right circumstances, it can get you entirely out of the tornado path. But consider the problem if everyone that might be in the path attempts to flee in a vehicle. How can people in a threatened city be certain whether or not they're truly in the path? Generally, most people will not be in the path, but we have no way of telling well in advance just who will and who won't have to move. If everyone that might be in the path gets in their car and drives somewhere, this could create a repeat of the 1979 Wichita Falls tornado disaster. Recommending that, as a rule, people shouild flee in their cars is going to put most of a city on the move in a tornado warning situation. Isn't it possible that all those vehicles, with many of them virtually certain to be speeding and otherwise not driving safely, would generate accidents and possible "gridlock" in the streets of a city? Then the occupants of those vehicles would truly be trapped with no shelter other than their vehicles.

The Oklahoma City metropolitian area tornado case of 03 May 1999 represents a special situation:

Yes, many people chose to drive to safety, because they had no adequate shelter for the oncoming tornado. Yes, in that case, fleeing the tornado worked successfully for most who made that choice. Unfortunately, abandoned vehicles blocking the highway near overpasses did create traffic problems - that's a different issue. I have no idea what traffic jams and associated accidents were created as a result of people trying to escape the tornado in vehicles. It's not hard to imagine scenarios where, if the recommendation were made that people should flee in their cars, the outcome would not be so favorable!

When time permits, if there's doubt about the safety of staying at home (i.e., no basement or tornado shelter), then it's certainly possible to get in your car and drive to some pre-arranged point of adequate shelter. The key is pre-arrangement of a safe location nearby. If you just drive away without any sense of where to go to reach safety, you do two things:

  1. Contribute to potential "gridlock" traffic that could trap all of you in your cars within the tornado path
  2. Create the chance to drive right into the path of an oncoming tornado, whereas your home might not even be touched

My recommendation: If you don't know where to go to find better shelter than your home (site-built),5 and it isn't nearby (5 min or less), you should stay home (out of your vehicle) and seek the best shelter available. Contrary to some public pronouncements, it is possible to survive even a violent tornado strike "above ground" (i.e., without a "safe room" or an underground shelter or basement), but you're depending on luck. I'm amazed at how many people survived even the F4-F5 parts of the path during the 3 May 1999 event. Survivors in such events are more common than fatalities, fortunately. Nevertheless, it's best to be prepared for the worst and to not have to depend on good luck.

Some final thoughts

We Americans have a tendency to feel much more secure in our vehicles than the facts warrant. I'm reluctant to reinforce that tendency with recommendations that people use their vehicles to flee tornadoes or as last-resort shelters. Although I agree with Prof.. Schmidlin that using a car to drive out of a tornado's path is possible, it seems to me that many people who do so will make bad choices that can put them in greater danger than if they had stayed put (as on 09 April 1979 in Wichita Falls). The publications and campaigns by Prof. Schmidlin and his colleagues include a number of reasonable caveats - his publications make no unreasonable statements. But I'm concerned that many of the details will be forgotten or ignored in the "media blitz" and the message that gets across to the public might well be that it's now considered safe to use cars (and other vehicles) to escape or ride out a tornado. There was nothing in the infamous "overpass" video that said in so many words that overpasses are recommended places of shelter, but the unverbalized message that stuck with many people was precisely that, and three people died under overpasses (and many more were seriously injured) as a result on 3 May 1999. I see constant evidence in tornado videos that overpasses are still widely used as "shelters" so the message from that video is likely to continue to claim victims. I don't pretend to know how "the public" will respond to Prof. Schmidlin's message, but I fear the consequences of a "green light" message to use cars as a "shelter" option or to evade tornadoes.

Doing research to resolve these important questions is what we need. I just don't believe in the applicability of Prof. Schmidlin's existing results as much as he apparently does. But I applaud the notion of seeking evidence to support our recommendations. Much, much more research is needed and at this point it's not clear that resource support for that research has been forthcoming. A big challenge is that proper controlled experiments replicating what happens in real tornadoes are not yet possible. As a result, simply studying the results of tornado cases is not capable of providing compelling evidence of what would have happened in that particular situation had people made different choices. This is a research topic where there are many uncontrolled variables in the data, leaving the interpretation of the results pretty ambiguous.

A huge confounding factor is the impact of human behavior in all of this. This represents another vast reservoir of currently undone research. If such research isn't eventually done, then we could be doing more harm than good by tinkering with our recommendations for public safety. What's the most effective mechanism for providing information to the public regarding weather hazards? I'm confident it's not confined to a few "bullet point" simple recommendations in a safety pamphlet. Whatever flaws exist with the current system, it's apparently resulted in a consistent decline of tornado fatalities since 1925, so I'd hope we can avoid interrupting that decline with well-intentioned but inadequately-researched changes.


1 Reference: Hammer, B. O., and T. W. Schmidlin, 2001: Vehicle-occupant deaths caused by tornadoes in the United States, 1900-1998. Environ. Hazards, 2, 105-118.

2 See: Schmidlin, T.W. and P.S. King. 1996. Cars and tornadoes: Where is the research? Bull. Amer. Meteor. Soc. ,77, 963-964.

3 And on the recommendations from an epidemiological study ... see: Glass, R.I., R.B. Craven, D.J. Bergman, et al., 1980: Injuries from the Wichita Falls tornado: Implications for prevention. Science, 207, 734-738.

4 I'm always puzzled by the difficulty most people seem to have in recognizing which way a tornado is moving. The audio tracks of "citizen" video often reveal that many people believe that a tornado is moving right at them, even when it's obvious (to me) that the tornado is moving toward their right or left. I wish I understood why it's so common for people to make this error because it can cause them to make bad decisions, such as moving so as to put themselves in the path, whereas they were not originally in the path!

5 I can't overemphasize the fact that mobile home residents need to have a plan worked out in advance, if there's no adequate shelter in their immediate vicinity. This likely means they need to leave their mobile home and move to some pre-arranged shelter location by whatever means is necessary. Residents in mobile home parks (at least in the tornado-prone parts of the U.S.) should seek to have proper community tornado shelters built within the park area.