Friday, December 2, 2011

unscientific extrapolation

somewhat to my surprise 12000 lines wasn't enough to crash the computer

So, what happens on summer solstice a bit over 5,5 years from now? I've got no (EDIT:proofed) idea.


Oale said...

using a hand-drawn ellipse to estimate the end date of some unknown process is not at all scientific, but it is quite simple to do and it looks stylish. or not.

Gail said...

Can you please explain exactly what that graph depicts? For us computer illiterates?

Oale said...

ok, why not explain it away,

This is an example of the simplest way of curve fitting.

The blue wiggly line is the Northern Hemisphere Sea Ice (~365-day) running average (about from July 4th? 1979 to May 17? 2011), so the seasonal variation is mostly removed.

I then moved the thing on to MsPaint to see what sort of curve would fit in to that (no maths involved here at all). Found out the normal ellipse fitted pretty neatly on the curve.

As the ellipse starts returning back at 17,4760 (this is the fraction of year 20XX,XXXX, calculated from the pixels) it cannot anymore represent any on-going physical process, so it might be an end date for something.

This is of course already assuming there is some process in the Arctic that will end, and responds to an elliptical curve rather than, say parabolical or sine (or Gompertz). I have no proof there is one. As this is areal data, the only such process that might do that (likely doesn't, at least exactly) I can come up with is the thickness of multiyear ice. So, the date could be a guess of the date multiyear ice reaches the thickness of first year ice (it floats deeper so the possible warm current underneath the ice hits it better), but I haven't the foggiest why this would happen at the area of 7.6Mkm2 (calculated from pixels).

In addition there are no error bars and the labels are left out so one can take the image and easily make it misrepresent anything (it's not science.)

Gail said...

Ha, that is odd, I had guessed Arctic sea ice! At least, it IS stylish!

Jim said...

Tamino gets a similar result with a quadratic curve fit: Go Ice Go! … Going … Going … Gone!!!