Fun with correlations

Posted by William M. Connolley on October 18, 2008
(4)
More »

Fergus is interested in the effects of El Nino on sea ice. So I looked at http://climexp.knmi.nl/. At first I thought it didn’t have sea ice, and it doesn’t in the indices, but it does have in the monthly data, which you can average. And then end up with:

i-bd7a0377118310aca4ec8e0f53560123-seaice-vs-sst.png

The shows Arctic sea ice correlated to SST over the sea. I used the Reynolds SST, so its 1982-2008. Oddly enough, more sea ice makes it colder in the Arctic. More interestingly, this applies to the Atlantic, but much less to the Pacific, though there is some effect (but outside Nino 3.4?). This is all months, as I failed to persuade it to do just September.

Though if you’re interested in the monthly anomaly plots, it will do them: click here. Very nice.

I’ve just realised that sea ice against Nino 3.4 is probably more interesting, in principle. Here it is:

i-ecefca23c1355725ed39a712e53c1b3d-seaice-vs-sst-1.png

Hmm, needs more work. Fergus asks, what about a 6 month lag? Not done: an exercise for the reader. The site supports it.

Disclaimer: I’m only playing.

(4)
More »

Comments

  1. #1 Fergus
    2008/10/18

    Have you allowed a delay in the link of 6-12 months?
    (ENSO preceding Arctic SSTA/sea ice loss)

    [Possible in principle. I haven't, but you can... -W]

  2. #2 Fergus
    2008/10/19

    Ha Ha! Does the search/ARCUS summary suggest something similar? : http://www.arcus.org/search/seaiceoutlook/index.php
    :)

  3. #3 mauri pelto
    2008/10/19

    In comparing glacier mass balance measured over the last 50 years on Lemon Creek Glacier Alaska and in the North Cascades, El Nino-SOI, is the primary climate circulation indices of importance to the North Cascades of Washington, but down the list below the Arctic Oscillation for the LC Glacier near Juneau, AK. http://www.nichols.edu/departments/Glacier/north%20american%20glacier%20mass%20balance.htm This is a local snapshot, but suggests to me that sea ice-el nino would be a weak enough link to be overshadowed by other variables. In terms of fun with correlation I related the mass balance to 10 climate indices but forgot to remove years in this least squares regression. Years came out to be the highest correlation for Lemon Creek Glacier, better than AO, PDO or SOI.

  4. #4 Fergus
    2008/10/19

    Mauri: the models have limited (!) skill for ENSO projection beyond six months, so my guess is just a guess. Carrying on in the spirit of fun, can you extrapolate a trend or forecast for sea-ice progression from the Lemon Creek data?

    Of course, it’s not strictly science…