Adventures in Statistics

The following graph was prepared by NOAA and presented on the National Public Radio website.  The gist of the accompanying article was OMG!!  WE'RE ALL GONNA DIE!!! Or something like that.  In particular, weather ("which is not climate") has suddenly become climate again.   

at least until the last line of the article:

Scientists blame an unlucky combination of global warming and freak chance. They say even with the long-predicted increase in weather extremes triggered by manmade climate change, 2011 in the U.S. was wilder than they predicted. For example, the six large outbreaks of twisters can't be attributed to global warming, scientists say.

(emph.add. Note the bald assertion "manmade."  No indication that since the earth had been
warming for upwards of 400 years, at least some of that increase had to be natural.  We are
left to wonder if there would still be weather extremes if the climate were changing naturally.)

Of course, the long-predicted weather extremes were not all that long-predicted.  I can remember when it was increased temperatures, pure and simple, that would kill us.  (Of course, I also remember when it was going to be the Coming Ice Age.)  But about ten or so years ago the secular trend changed from the 1980-2000 increase to a more stationary series 2000-2011.  With the globe no longer warming quite so much, the mantra changed to "climate change" and this led to "global climate disruption."  As more and more snowfalls stalked Al Gore across his speaking itinerary, the claim escalated that "extreme weather," and not simply warmer weather, would be the consequence of nefarious Western technology. 

Now go back up to the graph.  I'll wait. 

Back?  Okay.  Notice that NOAA is not measuring physics but economics.  First, they are only counting weather events that caused more than a billion dollars damage.  Do we expect more of those today than 30 years ago?  Well, my handy-dandy inflation calculator tells me that one billion 2011-dollars is equivalent to $364 million in 1980-dollars, so there will surely be fewer billion dollar events in the early part of the graph.  Also, there are more people now than then, so the amount of infrastructure in constant dollars or not will have increased, too.   

Do not be deceived by the dotted line, which is the total damages in 2011 dollars.  That is the normed value of the events that already cost a billion dollars in current-year value.  IOW, first we select events that cost a billion or more (blue bars), then we add up their costs in current-year dollars (bluish line), and only then do we convert that to 2011-dollars (red dotted line).  Question is: were they deceiving themselves as well? 

Now, that being said, take another look at the graph of damages.  Discount the Katrina spike in 2005.  Is there really an upward trend in the dotted line, or are the data consistent with a stationary series?  How much can be attributed to population growth?  New Orleans in 2005 simply had more stuff to damage than New Orleans in 1969 (Hurricane Camille, strangely prior to the start of this graph) and her levees were in disrepair.  That ain't the weather's fault.  It's much easier to have a billion dollar storm than it was thirty years ago, because there are more billions worth of infrastructure built since then. There were 227 million Americans in 1980, 309 million in 2010, so we ought to expect at least an increase of 36% in the amount of Stuff. 

He Has Never Seen a Year Like This

But the money quote is this one:

The AP spoke to National Weather Service Director Jack Hayes, who said he has never seen a year like this in terms of extreme weather. Hayes has been a meteorologist since 1970 and calls the waning year, "the deadly, destructive and relentless 2011."

Ever since 1970, hunh?  And Hayes has never personally seen a year like this?  Anyone remember when short time frames were said to be insufficient to judge climate change?  Or at least insufficient to drop heavy hints about it? 

How about the 19th century?  (Or would that make the "billion-dollar" business a little too obvious?  Log cabins and soddies don't add up to much, dollar wise, even allowing for inflation.)  The great storm that started in Nov 1886 and lasted into January destroyed the open range cattle industry.  An 1888 blizzard paralyzed the Northeast railroads and put New York City only days away from running out of food and coal.  Another blizzard on the plains trapped schoolkids in their schools, and many died.  During the Great Blizzard of 1899, it snowed in Florida. 

19th century: Where is Global Warming when you really need it?

Look at the wood cuts by Currier and Ives to see what winter was like before global warming.  (Heck, I can remember bad weather from the 1950s.)  Then, too, we find that Niagara Falls largely frozen in this image from the Niagara Falls Public Library.  The date is supposed to be 1911. 

The interval from 1881-1900 captures four of the ten most deadly hurricanes in US history.  And there were a lot fewer people back then.  Of recent hurricanes, only Katrina in 2005 makes the list. 

Rank	Hurricane	Season	Fatalities
1	"Galveston"	1900	8,000–12,000†
2	"Okeechobee"	1928	2,500+†
3	Katrina	2005	1,836
4	"Cheniere Caminada"	1893	1,100–1,400*
5	"Sea Islands"	1893	1,000–2,000†
6	"Florida Keys"	1919	778
7	"Georgia"	1881	700†
8	Audrey	1957	416
9	"Labor Day"	1935	408
10	"Last Island"	1856	400†

We could look at tornadoes, too; but you get the idea.  Then there was the Dust Bowl of the 1930s, also prior to NOAA's graph, and to Hayes meteorological experience.  And we could go on.  Pfui, sez I.  

Hillaire Belloc once wrote that we tend to be overly impressed with our most recent experiences, and expect the recent past to continue.  Hence, the short-term perspectives give us global cooling or global warming, depending on the decade. 

The problem with the long-term, of course, is that the data are not so good and one must make do with surrogates; and when the surrogates fail to match the actuals where they overlap, we must "hide the decline."  Bummer.