Thursday, July 14, 2011

More Thoughts on the Tornado Outbreak of 2011

I've just completed  participating in a conference sponsored by the Geneva Association where I presented my thoughts on this years tornadoes.  In preparing for the presentation, I used our casualty models to see if this year suffered from any of the vulnerabilities we had identified that will inflate fatalities.  The answer to that question is yes.  Compared to previous outbreaks, this one occurred predominantly in the southeast which has the highest fatality rate from tornadoes than any other region.  As an example, the fatality rate for the US is .45 per million calculated using data from 1950-2010.  Mississippi has a fatality rate of 2.8, six times the US average.  Secondly, there is a greater proportion of mobile homes than previous outbreaks.  As a comparison, consider the tornadoes of 1953, the last year we suffered more than 500 deaths.  For the counties affected by those tornadoes, the percentage of mobile homes in the housing stock was 6 tenths of one percent.  It was 10 percent for the counties affected by this years tornadoes.  Our research confirms the observation that you are 10 times more likely to die in a mobile home struck by a tornado than a permanent home.  As more mobile homes are in the path of the tornado, higher casualties will likely follow.  Finally, one third of this years fatalities occurred on a weekend including the tornado in Joplin.  We have found that casualties are significantly higher on either Saturday or Sunday than during the week.  People are safer in the buildings they work or go to school in than they are in their own homes.  I used our casualty model to estimate the change in fatalities for the Joplin tornado if it had occurred on a Monday rather than a Sunday and found the expected fatalities would have been reduced by 23%.

So it's no wonder that we have now exceeded the largest annual death toll since records were kept beginning in 1950.  Or is it that simple?  I also spoke at the AMS Broadcast Meteorology Conference last month.  Some of the comments I heard from people expressed shock at the casualties but also questioning the value of the warnings that have been celebrated for minimizing casualties.  That started me thinking about how I could test whether or not the warnings were effective.  Keep in mind that the Joplin tornado had 24 minutes lead time.  One way to examine this notion is to compare the fatality rate from this years tornadoes with another outbreak.  Again, consider 1953 which had a similar number of killer tornadoes, 41 versus 56 this year, had a similar number of affected counties, 72 versus 75 this year, a similar percent of F-3 plus tornadoes, 78% versus 79% this year, and a similar number of fatalities, 519 versus 537 this year.  The population of the affected counties in 1953 was 3.9 million giving that outbreak a fatality rate of 167 per million.  Affected counties in this years tornadoes is 8.2 million giving this year a fatality rate of 35 per million, half the rate of  1953.  Can the entire difference be attributed to better warnings?  No, I don't believe so, but it is clear that something profound has changed in the last 60 years and the obvious candidate is the investment we have made in our warning system as well as the enhancement in communication technology that we enjoy today.

There is no argument that we have experienced a terrible tragedy.  But it's my belief that the death toll could have much higher if not for the investment by and the efforts of the National Weather Service and the forecasters that have trained to warn residents of the danger of approaching tornadoes.