Sea ice: how to describe the possibility of a record?

With all the wild excitement over 2007’s record low there is clearly room for noise to be made about summer sea ice, so predictions must be made! We all know, of course, that whether this year, or any other year, is going to be a record is going to be a matter of chance: on top of the long term negative trend (and we can argue about whether that trend is typified by the slope of 1979-2006, or has got steeper recently) there is a lot of interannual variation. Certainly I’m not aware of any sea ice modeller with enough confidence in their predictions to put any money on them.

But the next best thing is a press release, such as New record Arctic sea ice cover minimum? Climate researchers from Bremerhaven and Hamburg present new prognoses. from AWI. Now the heading is clear enough: they are interested in the question of a new minimum. What is objectionable is the next bit: The German researchers agree upon a continuing negative trend. Another critical minimum of Arctic sea ice is to be expected in the late summer of 2009.

Well, there you go: a clear prediction of a new minimum this summer. No? Well, no. Because the text then gets a bit smaller and says: We have computed in this year’s first prognosis that the ice cover of the Arctic Ocean will lie at the end of the summer with at least 28 % probability under that of 2007 – the year with the lowest-ever measured ice extension. Put simply, “is to be expected in the late summer of 2009” is wrong. It should be “has an approximately 1-in-4 chance in the late summer of 2009”, based on their model. Or you could read a bit further on and find that Hamburg say: We estimate a probability of 7 % that this year will fall below the negative record of 2007.

17 thoughts on “Sea ice: how to describe the possibility of a record?”

  1. I posted about this a few days ago (the full forecast doc is available – one of the links at the bottom of the page William links to). AWI’s technique is to force their sea-ice model with the summer atmospheric data from 1989 to 2008. They ran two ensembles. the second using “improved” initialisation with recent ice state information. The unimproved model ensemble gives these probabilities:

    probability to fall below 2007 (record minimum) is about 28%,
    probability to fall below 2008 (second lowest) is about 55%,
    probability to fall below 2005 (third lowest) is about 96%.

    The second ensemble:

    probability to fall below 2007 (record minimum) is about 49%,
    probability to fall below 2008 (second lowest) is about 75%,
    probability to fall below 2005 (third lowest) is about 99%.

    Guess which forecast I prefer?


  2. For 2009 they forced the ice model with atmospheric conditions from each year between 1989 to 2008. This suggests a slightly optimistic expectation on atmospheric conditions for this melt season programmed into the model?

    The individual model runs show that the 1990 atmospheric forcing gives the lowest ice extent result so perhaps this is not an issue. But if not an issue, how so? Did they add in extra warming for atmospheric conditions from 1990 to account for the extra 19 years of Co2 warming? About 0.2 a decade is +0.38 degrees global, and Arctic is higher so perhaps approaching a 1 degree difference? Or does a decade or two of Co2 warming simply not matter?

    And the differences between the initial state from the free run and the optimised run are an indicator of the accuracy of this ice model? That is the free run is the models guess of what the ice should be based on the last 20 years of atmosphere data, and the optimised run is based on actual measurements of ice data?


  3. The more interesting question is what will be the minimum ice coverage at the north pole. If the current trends continue (see Nick’s comments) it might even be ice free sometime this year.


  4. Possibly sloppy writing around the words “critical minimum”. There will be a minimum this year, after all. Will it be “critical”? Will it be a record? Who knows?


  5. Here’s another question: Whose numbers will be used?

    The graph we’re all watching at NSIDC has 2009 neck-and-neck with 2007. The IJIS-JAXA graph has 2009 toward the middle of the pack.

    Now look at the “tale of the tape” at Cryosphere. Over the first half of 2007 the anomaly averaged below -1 M km2 and by June was testing -2 M km2. The first half of 2008 averaged above -1 M km2. Extent anomaly did finally reach -2 M km2 last year but it happened too late in the season for a run at -3 M.

    2009 shows more ice than 2008 on average over the first six months and more ice at this date. What ice there is could be thinner now so there may be an outside chance for a record — but it’s gonna have to get moving!

    In all likelihood, we won’t make -2 M km2 by mid-July (which I think is critical for a record) and 2009 will go in the books as second or third place.

    Add in a decent El Nino and 2010 (and especially 2011) could easily establish new lows.


  6. I don’t completely understand the fascination with Arctic Sea Ice, relative to Global Sea Ice as I discuss here.

    [Because we all expect Arctic ice to decrease more rapidly than Antarctic. Partly this is based on experience, and partly on models. The recent fascination with beating 2007 is silly, though -W]

    Why would not one, but two different research institutes waste time trying to come up with short term models for Arctic Sea Ice coverage. Trying to corner the Sea Ice futures market?

    [Maybe to demonstrate that they know what is going on? Or possibly to prove that they don’t -W]

    Now it seems that ER wants to introduce a new micro metric called “ice coverage at the north pole”, which appears to not be actually be measured by anyone.

    Anyway I will go with the naive model. This year’s minimum will be higher than 2008 and lower than 2005 as measured by this.


  7. # 7
    The graph we’re all watching at NSIDC has 2009 neck-and-neck with 2007. The IJIS-JAXA graph has 2009 toward the middle of the pack.

    But in JAXA, 2007 is also in the middle of the pack for most of the spring (behind 2005 and 2006). And AFAIK, the late summer ranking is the same in both data sets (although the actual estimated extent is quite different).

    from post:
    We can argue about whether that trend is typified by the slope of 1979-2006, or has got steeper recently

    Which metric trend exactly should we argue about? I’d say it would make sense to use average sea ice extent for September (rather than the minimum value of daily estimates).

    Oh, yes, my prediction: I’ll go out on a limb and say the 2009 minimum will be closer to 2008 than 2007 or 2005.


  8. I wonder if the only really important Arctic Sea Ice issue, climatologically, is amount of ice cover (and therefore albedo) in June & July (period of greatest insolation).


  9. #3 Gareth: the German version’s here:

    According to my rusty German and to babelfish, the translation about “critical minimum” is accurate.

    More indication of the problem of science-by-press-release. Short of a more systematic solution, a small improvement would be to have named authorship of press releases. That might rein them in a bit.


  10. William — I’ve finally figured out what your prediction is for this September’s average extent (5.84 million km^2), and blogged it at my place. How about, then, 50 quatloos, even up, on over/under 5.38 million? Splits the difference between our respective estimates (mine being 4.92, both of us having about 0.5 million km^2 standard error).


Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s