« SIE 2011 update 11: the heat is on | Main | SIE 2011 update 12: fumbling in the dark »

Comments

Feed You can follow this conversation by subscribing to the comment feed for this post.

Ned Ward

Each of the contributors has a brief pdf with some notes about their forecast. I loved this line from the notes that Anthony Watts provided for his website's forecast:

"5. Estimate of Forecast Skill: none"

I would say that's a surprisingly realistic self-assessment on WUWT's part.

L. Hamilton

So here's a question for fellow denizens of the ice blog ...

I think the SEARCH SIO is a possibly transformative idea (not unlike Neven's blog) that opens public window on Arctic science as a process.

If the SEARCH SIO idea took new directions in the future, what might you hope for?

Rob Dekker

While we are talking about projections :

Intrade puts the likelyhood of 2011 Arctic minimum extent to break the 2007 record at 60 % :

http://www.intrade.com/v4/markets/contract/?contractId=744206

Where are these WUWT voters now ?
Actually I think the WUWT crowd may be a bunch of big-mounth yellers that don't even believe their own vote. I challenged the 371+ 5 million+ km^2 to a bet, and none of them showed up.
http://wattsupwiththat.com/2011/05/19/sea-ice-news-call-for-arctic-sea-ice-forecasts-plus-forecast-poll/#comment-667882

Scroll up and down, since the permalinks on WUWT don't work that well. Also check the end of the comments, where noone took my bet offer.

Incidentally, shouldn't our very own Neven's poll (4-4.5) not be part of the Sea Ice Outlook ? Did you apply with our poll results, Neven ?

crandles

>"If the SEARCH SIO idea took new directions in the future, what might you hope for?"

1. September minimum for next two years as well as current year to also be predicted.

2. Better clarity about what the ranges given by different contributions mean. I tried to contrast the RMSE which was overtuned to data not available at time of true prictiction from RMSE/Confidence interval without such tuning and also to contrast confidence interval from credible interval by discussing whether errors are growing over time. Asking all statistical methods to follow my approach may not be realistic but it may be sensible to ask for clarity on whether error range is too narrow because of overtuning and/or whether errors are growing over time or other confidence/credible interval differences.

3. Encourage discussion between contributors? Or does this reduce wisdom of crowds? Perhaps my next contribution will specifically encourage comments on my method or issues arising here on Neven's blog.

Piotr Djaków

Meanwhile Steven Goddard post a very scientific comment on Zhang numbers:
http://stevengoddard.wordpress.com/2011/07/15/finally-some-winners/
I've try to estimate minimum sea ice extent too. Based on Jaxa numbers ice extent on july 14 was equal to 7.52*10^6. Lowest melt between 14.07 and september minimum was about 2.4*10^6 km^2 and highest was 4.0*10^6 km^2. I'ts very likely that september average ice extent will be 4.4 ± 0.9*10^6 km^2.

Ned Ward

I think InTrade's 60% chance of breaking the 2007 record is a bit optimistic, as are the 39% of voters in this site's poll who predict < 4 million km2. On the other hand, the WUWT forecast is less plausible than either of those.

I did something similar to what Piotr Djaków describes.

Pick a given date (say, today) and I calculated each year's decline from that date to that year's minimum. I then calculated the mean and standard deviation of those declines. (There doesn't seem to be any interaction between the extent anomaly in mid-July and the magnitude of the eventual decline to the September minimum, which simplified things).

Thus, for each starting date (July 11, 12, 13...) you get a mean estimate for the minimum and a 95% confidence interval around that. As of 11 July, that was predicting a mean of 4.47 and a confidence interval of (3.53, 5.38).

So WUWT's prediction of 5.1 is within the 95% CI, but on the high side.

Ned Ward

Whoops. I should have read the other thread first -- didn't realize that the IJIS numerical data were still being updated, unlike the graph that's still stuck on 11 July.

Okay, so with the data through 14 July my model now predicts a mean of 4.33 and a confidence interval of (3.45, 5.17).

Wow -- what a difference three days can make. Suddenly WUWT's forecast is in danger of slipping out of the 95% CI.

L. Hamilton

Crandles, thanks for the ideas. The different estimation methods are too diverse to standardize their uncertainty calculations, but conversation among contributors on this and other points would be informative! Also, as you say, why not take shots at 2-year forecasts?

Neven
Incidentally, shouldn't our very own Neven's poll (4-4.5) not be part of the Sea Ice Outlook ? Did you apply with our poll results, Neven ?

Rob, I explained in my post on the SEARCH SIO June report why I wasn't submitting our poll.

More info in the post on Poll Results. I will do an update on that soon.

Neven
If the SEARCH SIO idea took new directions in the future, what might you hope for?

Good question, Larry! I'll have to think about this one, but first thing that comes to mind is that they do something during winter too.

L. Hamilton

"do something during winter too."

Another interesting idea. I'll be passing these and any others along.

The comments to this entry are closed.