Sunday, June 30, 2013

Rabett Does Hurricanes

Trends in hurricane damage are a sore spot hereabouts.  The problem is three fold, first the absolute damage varies wildly from storm to storm.  Second, the most studied storms are the ones that make landfall in the US, which are thankfully few and far between.  Third, there is no good way of accounting for many things that would mitigate damage such as early warning of storms from satellites, evacuation measures, higher standards for construction and more.

These thoughts were reinforced a couple of weeks ago talking to Eli Lehrer from R Street at a debate held in DC between Lehrer and Bob Inglis on one side and James Taylor (Heartland) and David Kreutzer,(Heritage) about how to deal with climate change.  As promised, more about that later.  Lehrer said that it was not possible to tease any meaningful trend out of his database for hurricanes.

Ethon has been thinking about this for a while, and it finally occurred to him that there was one unfortunate place in hurricane alley that might provide enough data, Hispanola, a large island in the Caribbean with two countries, the Dominican Republic and Haiti (bunnies might also toss in Cuba on one side and Puerto Rico on the other).  The NOAA historical hurricane track site can be used to see where hurricanes and tropical storms have passed


And population and GDP information is available from the Maddison Historical Statistics website.  Part of the reason for looking at Hispanola is that population in the DR and Haiti are about equal over the 1950-2012 time period while GDP per capita in the DR has steadily grown from $1400 (constant 1990 $) to about $5500 in 2011, in Haiti, per capita GDP remained constant  at about $1000 till ~1990, and since then has declined to  about $700.

Eli is but a poor bunny, and all he could do is troll the Wikipedia for the number of storm deaths.


Storm
Year
Hati
DR
Hazel
1954
400
0
Kati
1955
7
0
Ella
1958
30
0
Flora
1963
5000
400
Cleo
1964
192
7
Inez
1966
750
100
Beulah
1967

2
David
1979
0
2000
Allen
1980
220
7
Emily
1987
5
3
Gilbert
1988
30
5
Gordon
1994
1122
5
Hortense
1996

39
Georges
1998
209
380
Lili
2002
4
0
Odette
2003

8
Jeanne
2004
3006
18
Alpha
2005
17
9
Wilma
2005
12
0
Ernesto
2006
5
0
Noel
2007
73
87
Olga
2007
2
37
Dean
2007
9
6
Gustav
2008
77
8
Hanna
2008
529
1
Ike
2008
74
2
Thomas
2010
21
0
Sandy
2012
52
2
Isaac
2012
24
5
Irene
2011
6
9
1950-2011

11876
3140
2011-2000

3911
192
2005-1995

4358
459
2000-1990

1331
424

If the Rabett looks at the number of killer storms per decade 


1950-59 3
1960-69 4
1970-79 1
1980-89 3
1990-99 3
2000-09 12
2010-12 3

Suggestive that something is going on in hurricane alley.

What this needs, of course, is a serious study with more information about $ damages, numbers of houses destroyed, evacuations and storm tracks.  Including eastern Cuba and Puerto Rico would help as can be seen from the graph showing GDPs below.
To be honest, and Eli is always honest, all this is suggestive that more work might be worthwhile.

Volunteers?

6 comments:

Unknown said...

I'll put my hand up first.

It suggests (from my own crude analysis, grouping in 5-year buckets, except for 1954-1960 & 2010-2012):

- Deadly storms per year has increased from ~0.4/ yr up to 1990 to ~0.9/yr from then to now.
- Deaths per storm has remained approximately constant, though lower in the Dominican Republic.

The 2nd may be related to GDP, it is hard to see how the first can be.

Toby

Russell Seitz said...

There's some Category 5 censorship action over at Watt's , as he breaks wind in the general direction of critical comments, appended to his latest Two Minute Hate .



here's Willard's latest takedown.

Russell says:
June 30, 2013 at 1:59 pm
Among my college classmates was a French fellow whose father was still trying to live down a well-intentioned effot to improve his farming tenants and deer herd’s quality of life by reducing the surplus rabbit population.

To that end he imported some myxomatosis infected rabbits from Australia , and sure enough, by year’s end the neighborhood was bunny- free.

As, alas, was every rabbit hutch from Finnisterre to the pas du Calais a year later. It took a decade for the virus to burn itself out of hosts, and the rabbitry of France to recover.

EliRabett said...

Willard Tony does humor about as well as he does science tho. Excellent thing elst thee and Eli would be out of business.

bill said...

Oh, I don't know, W.Tony's efforts are at least as funny as Josh!

And at most. Talk about leaden...

Anonymous said...

Slightly interesting is that 1 in 3 "killer" storms occur with the letters G, H and I, over the period 1954 to 2012, rising to around 1 in 2.5 if you arbitrarily split the data set into two at the midpoint in terms of year (i.e. 1983).

Possibly of more interest is that whereas in the first arbitrary period there are no killer storms past K (the nominal midpoint in storm naming, unless you get more than 21 named storms in a season, such as 2005's Alpha), there are 8 killer storms with names past K in the second period.

Of course, there are obvious caveats to be borne in mind with this rather preliminary analysis, not least with using the storms' initials as a possible proxy for storm timing, so I wouldn't necessarily draw any conclusions from these observations. I just thought they were prima facie interesting.

Cymraeg llygoden

a_ray_in_dilbert_space said...

Cymraeg llygoden,

And your bonus assignment is to tell us why that is a consequence of the seasons + the Poisson distribution.