z Facts.com
 KNOW THE FACTS.  GET THE SOURCE.
About Printable
 
 
  Home
Global Warming
Warming: Effects
Atlantic Storms
Makes Sense
Theory vs. Data
Katrina Cause?
Stats & Data ♦
NOAA Decoded
 
  Don’t Miss:
 
 National Debt Graph

US National Government Debt

A Social Security Crisis?

Iraq War Reasons

Hurricanes & Global Warming

Crude Oil Price

Gas Prices

Corn Ethanol
 
   

   Statistics on Sea Warming vs. Hurricane Intensity

  How did zFacts conduct it's analysis of the sea-warming impact?

First zFacts thanks Dr. Kerry Emanuel of MIT for generously sharing his 61.5 years of data on Atlantic hurricanes and sea temperatures. This has had minor corrections made to it since he used it in his August letter to Nature. For each year he has summed the power of each hurricane in the season at many evenly spaced times during the hurricane's life to find the total amount of wind-energy dissipated during the hurricane season. He has also averaged many sea temperature readings for September in the part of the Atlantic where hurricanes spawn. The result is one summary point for each hurricane season—total wind energy and average September temperature.

These data fluctuate from year to year for several reasons and so hurricane energy does not correspond perfectly to sea temperature. Curve-fitting statistics can discover what is caused by temperature and what is not. If E stands for annual energy and T for September temperature, then the best-fit curve is:

E = 39.80 – 1.52 T + 0.80 exp(T – 26)

The exponential term allows the “curve” to actually curve and follow the data. This was found by “running a regression” on the annual data (not the smoothed data shown in the graphs). That regression also included the change in energy from one to two years ago and the change from two to three years ago. These helped “control” for the effect of the part of the short-term fluctuations that are not caused by water temperature.

Including the energy terms also nearly eliminates “auto-correlation” in the regression residual and which means the F-test performed next will be quite accurate. That test asks what is that chance that our fitted curve could, just through bad luck, be twice as steep as it should be. The answer is “less than a 5% chance.” The test is performed by fitting  a line with half the slope as well as possible and comparing how poorly it predicts with how poorly our best-fit curve predicts each year. This is a standard F-test for constraints on regression coefficients and can be found in the zFacts spreadsheet.

Just to double check, zFacts used a more rigorous but complex method of eliminating auto-correlation and found that the test was even more conclusive.

All of the data, graphs and statistics discussed here are contained in this spread sheet, except for the final test.
 
 
 
poppy-s
poppy-s
poppy-s
poppy-s
poppy-s
 
 


http://zfacts.com/p/127.html | 01/18/12 07:18 GMT
Modified: Mon, 17 Apr 2006 18:32:56 GMT
  Bookmark and Share  
 
.