Betting on sea ice: $10,000

This year’s story so far: in May, I accepted some bets but was still trying to come to terms with Rob Dekker. In the comments there we came to agreement on the following:

If both NSIDC and IARC-JAXA September 2016 monthly average sea ice extent report are above 4.80 million km^2, RD pays WMC US$ 10,000. If both are below 3.10 million km^2, WMC pays RD US$ 10,000. In all other cases the bet is null and void

The numbers are a bit of a compromise, and of course the large “null gap” in the middle means a no-payment result is quite likely. Now is a sort-of good time to announce this, because this years ice has temporarily stopped falling off the cliff:


My previous post contains my reasoning, and also the other bets. At some point I’ll transfer that and those to here for completeness, but just for now its getting late so I’m going to put this post up.
Continue reading “Betting on sea ice: $10,000”

Sea ice, part 2

Or perhaps part 3. I’ve lost track. Sea ice – and now for something just a tiny bit different refers, as does the earlier This year’s sea ice. Yes, it looks like being part 3.


The above is the IJIS sea ice. Nothing very exciting at the moment (NSIDC have some nice pix. April 2011 was bang-on trend). As you recall, we’re trying to agree a bet or bets. I want to bet on monthly extent and some of the others want daily. After too much equivocation, I’ve decided to stick to my guns: monthly it is, and if you don’t like that, you’ll need to find someone else to bet with. Or, you can apply a handy adjustment. Rob reckons (based on IJIS) that the difference between monthly and daily is 0.17 (see this comment so you can adjust your expectation based on that, if you like.

Here is the pic I’m basing stuff on:


Green line is trend-to-year-using-all-data; so the 2011 point is the first extrapolated, and so on. Blue-dash is the trend, using last-10-years. And purple-dash is the same, but excluding 2007 (on the 2007-was-weird-theory; I’m not actually using that line). Red is the LS/Crandles exponential fit. What I find interesting is that every year since 1997 (except 2001) has been below the long-term trend “prediction” for that year. So I’m coming round to the view that things have changed. That means that for the pocket-money type bets, I don’t mind sticking to my pet theory (green line, so to speak) but for the serious money I’m afraid I’ll shift to the last-10-years theory (blue-dash). Where is the line dividing pocket from serious money? I don’t know, but $100 is P and $10k is S.

So to re-visit some earlier stuff: I said the chance of something beating 2007 and setting a new record low within the next 3 years (including 2011) seems quite good. So I’m not betting on that in response to Neven’s offer. But if I switch to the not-so-safe green line I think that just about fits the error bounds, so I’m now happy to take that on for the proposed E50. I’m basing that on the monthly values, of course. If you want to take up the offer based on daily values, then OK as long as we have a “buffer”: if the daily and monthly disagree about whether there is a record or not, the bet is void. Covering “the next 3 years” ie 2011, 12 and 13.

Another multi-year bet was Peter Ellis who says Allowing myself a reasonably wide fluff margin, like you’re doing, I’m prepared to bet £50 that the September monthly average will go below 2 million some time between now and the end of 2016. Bet voided if there’s one or more Pinatubo-scale eruptions between now and then. That looks OK; accepted.

Crandles offers 3 separate bets on the average of [2012, 2013 and 2014] (to be above/below 4.294, I take the high side), of [2013, 2014 and 2015] (4.119, ditto) and of [2014, 2015 and 2016] (3.94, ditto) for £67 each. That seems reasonable and is accepted.

That leaves me trying to satisfy Rob, which I think will be harder, since he wants to bet Real Money ($10k), so I’m going by the pale-blue-dashed-line. I think we might best off going for a bet on the single year where we disagree most, which (within an interesting timeframe) is probably 2016. So taking the 2016 point and subtracting my cowardly 0.5, I offer above/below 3.1, with me taking the high side. The data I’m using there is NSIDC (I’m fairly sure 🙂 monthly average extent. Having just checked, the IJIS seems to agree closely, as I’d expect (but I can’t find the monthly averages for that pre-made, so I checked 2007/09, which was within 0.1 of NSIDC). I suggest again that we buffer it: if the 2016 value for NSIDC or IJIS differs enough to affect the result, the bet could be void. To sweeten this not-very-good deal, I also offer Rob (but not me) the option to back out once the 2011 minimum data is in.

I think there are other people who have suggested they believe Maslowski’s “near total ice loss by 2016” stuff, and if you believe that, then I’d hope you’ll be happy with 3.1.

Did I forget anyone?

Sea ice – and now for something just a tiny bit different

[Update: see comments. We’re having some dispute about whether to bet on the monthly averages (the scientifically respectable thing to do) or daily min (the wildly exciting popular choice). I need to bother work out the numbers. Until then, you’ll have to be patient (2011/3/31; I’ve adjusted the posting date from 2011-03-22 to push this to the head of the queue)]

But not very different. Neven reminds me, again, that I promised to put up a slightly longer-term bet; see This year’s sea ice and in particular this comment and reply.

Neven offers:

I’d be willing to bet 50 euros on a record low within the next three melting seasons. And with record low I mean absolute daily minimum extent based on data as reported by IJIS.

Well, we need not worry about the amount for the moment, instead I’m trying to set the terms of the bet (for amounts, it will be word-of-honour for anything down at the “trivial” end and something a little more formal if you want to go above £1,000).

I dislike daily values; monthly means are better, and are less likely to be influenced by oddities, and more likely to be consistent between series. I’m happy to use IJIS though.

Otherwhere, Gareth predicts the end of summer sea ice in 5 years, based on extrapolating thickness trends, and I’ve taken the liberty of ripping off his image. This I don’t believe at all, and if anyone is prepared to bet on no (or essentially no) summer ice by 2016, I’ll be very interested.

Anyway, here is the pic I’m basing my judgement on:


(this is a mod of one from a previous post). Black line is just the obs. Green line is trendline using all the data that far (so it is a straight line into the future). The other lines are a bit more complicated, and I keep forgetting what they are: the blue line is the trend, using data only as far as the point on the x-axis, but taking the trend to where it would be in 2010. The red line is the same as the blue, but using only the 10 years before that point. And lastly the one that matters is the purple line, which is the trend to 2013, again using the previous 10 years, though displaced off (so the point at 2013 is drawn from 2001-2010 data; the point at 2012 from 2000-2009, etc).

Using last-10-years is being pessimistic (or from my point of view, erring on the safe side) but based on that, the chance of something beating 2007 and setting a new record low within the next 3 years (including 2011) seems quite good. So I’m not betting on that; after all, remember that what we’re aiming for is a bet that both sides regard as a sure-fire win for them :-). So, based on those lines, and needing the usual error margin of 0.5, which I’ll multiply by sqrt(3) since there are 3 years, I think a “fair” (ish) bet would be on the ice going below 3 (whatever the units are, I forget) any time during the next 3 years. Anyone interested, or care to dispute my numbers?

Days vs Months

[Added in update, 2011/3/24 22:00]

Clearly, people would rather be betting on a record *daily* low, whereas I’d rather think in terms of monthly averages. My justification for that is that it reduces the statistical noise, and so allows for smaller uncertainty margins. Though I admit I haven’t looked at the day/month differences. I suspect that really it shouldn’t matter too much: ice can’t vary much on a daily basis; it isn’t possible to have some vast down-spike on just one day (barring an unlikely compression event). Conversely, if you believe in a record daily low, then you ought to believe in a record monthly low, too.


* IJIS / JAXA ice extent which DA notes (prematurely, but probably correctly) has just peaked this year.

Sea ice: and the winner is… no-one!

Hurrah. That saves lots of effort paying :-). Not long ago it was looking bad for the good guys (i.e., me) with a “douple dip” recession of sea ice. But a strong perforcance from the boys up north in the mushy white stuff stakes saw a sharp rebound at the end of the month, leading to a monthly average for september of 4.90 (thanks for C for vigilance). As a reminder, recent years have been:

2000  9      Goddard      N   6.32   4.31
2001  9      Goddard      N   6.75   4.55
2002  9      Goddard      N   5.96   3.98
2003  9      Goddard      N   6.15   4.01
2004  9      Goddard      N   6.05   4.35
2005  9      Goddard      N   5.57   4.03
2006  9      Goddard      N   5.92   3.97
2007  9      Goddard      N   4.30   2.78
2008  9       PRELIM      N   4.68   2.93
2009  9      NRTSI-G      N   5.36   3.42
2010  9      NRTSI-G      N   4.90   3.02

Or so says Other datasets will give you slightly different answers, of course. Note taht 2010 gets the coveted number 3 spot in terms of both September average and absolute minimum.


The bet for this year was (from Three views of sea ice).

That the september mean ice *extent* be below 4.835; but with a “buffer” where we call it a draw: between 4.735 and 4.935, no one wins. I’m taking the “high” side of this; anyone interested in the “low” side let me know. For my part, 4.835 is arrived at as the 1979-2009 trend extrapolated, minus 0.5 which is the SD. It seems to have become tradiational for people to bet small amounts, which is fair enough if we’re just playing. But this was intended to flush out the “the sea ice is in catastrophic decline” people. OTOH, if there are any “the sea ice will certainly recover this year” people then you can get odds on trend-plus-SD, i.e. ice being above 5.835 if you like (note that those are all spuriously precise but never mind)]

So, if we’d played for just-the-number (without the buffer) I’d have won; but I agree, it is better to include the buffer. Assuming I (or ws it C?) did the calcs right, the trend line was for 5.335, and we’re clearly below that, but by less than the SD, so it doesn’t matter.


* Three views of sea ice – defn of this years contest.
* 2009
* 2008
* Lab Lemming’s pool (I think I’m the very broad green line)
* Tamino got lucky

Update: pressed by C in the comments, I’ve now calculated the trend for myself, and so I’ve added this pic:


which shows the extents, the trend-to-2010 calculated using all previous years, the trend-to-year using all previous years, and the trend-to-2010 only using 10 previous years. I make the baseline prediction for next year 5.235, which agrees with NB, so all is well.

Yet more sea ice

Continuing from Three views of sea ice. Well, tis now mid-June, so the futurology aspect of the prediction is closing rapidly. Or so you would have thought. I’ve just taken £50 against CR for the ice being below 4.735 (he gets the low side) or above 4.935 (I get the high side). But my principal debt on sea ice is failing to write anything more about it. so, to remedy that!

I was going to suggest that the most interesting way of doing the pool was via Intrade. Unfortunately their Arctic sea ice pool doesn’t look very interesting. The bet is “2010 greater than 2009” and is trading at around 43%, and hasn’t had a trade in a while. Since I’d say ~50% is fair odds, neither buy nor sell is very interesting. Here is a pic:

<img src="" height="225" width="460"

I suspect that I might want to put in some “sell” bids at alower price but I’ve yet to work out exactly how that goes. Maybe later.

It is hard to deny that the current AMSR pic is looking bad for the good guys; but never mind. I’d stick with my old opinions for the moment (I mean, just look at the variability on that chart! One month is hardly a guide to the next) I ought to offer my apologies to those who suggest I look at some sea ice images: I’m afraid I haven’t found the time to do so and this largely reflects a lack of interest on my part; perhaps with a bigger computer and a faster internet connection I might. As it is, I’ll just have to go forward blind.

I thought I’d do a quick “news” search on sea ice and was pleased to find The Economist with a sensible story and, for a bonus, plugging my humble blog. I look forward to hordes of punters turning up with wads of cash ready to throw down. The Economist notes the Arcus sea ice outlook series, which in my mind is pegged as “not doing very well” in the last two years. However, they have learnt one thing from their disastrous May forecast of 2008 – don’t do a forecast in May :-). We’re waiting on the June forecast, which should be interesting.

While I’m here… another link worth reading is RMG’s take on “when will the Arctic first be ice free in summer”. I say worth reading, and I think it is, but I also think it is entirely wrong.

And for those who have forgotten the default bet, and can’t be bothered to read the old post, it is:

That the september mean ice *extent* be below 4.835; but with a “buffer” where we call it a draw: between 4.735 and 4.935, no one wins. I’m taking the “high” side of this; anyone interested in the “low” side let me know.

[Update: ARCUS 2010 is now out; New Boy Nevin gets credit for being the first I saw to blog it -W]

Three views of sea ice

The troops are getting restive. What wil happen to this year’s sea ice? Rumours abound. Let’s look at some pictures.


Probably the least interesting is this one from NSIDC. But it looks exciting, doesn’t it. Woo-hoo, look, the sea ice now is lower than it was in 2007, that means it will be at minimum, too, doesn’t it? Well no, of course not. Look at this one from AWI:


2006 was well below 2007 at this point, and turned out to be uninteresting. As far as I know, no-one believes in predicting the minimum (which, of course, is the only number anyone gives a toss about) based on previous months. And when I say no-one cares about anything else that isn’t really true; various people have started caring about the ice volume instead, but I’m a bit dubious about how exciting that is.

But for no reason that I can strongly justify, I happen to prefer the IJIS AMSR pix, reproduced below:


Which is similar too, but clearly not identical too, the AWI pic. Which shows the obvious: that retrieval from space isn’t exact.

Anyway, so what? So, it is not too late to join in the great predict-sea-ice-this-year party. My entry this year is “the same as” last year: linear trend since 1979, this time *not* omitting 2007 or 2008. As I recall, last year RMG was kind enough to work out what that came to in real numbers; perhaps he will again. As before, the possibility of a bet only really makes sense if people actually think very differently from me: which would be, either they think that there will be substantial ice growth this year, or some sudden collapse.

So, there you go: usual money on offer, which is to say “unlimited”, in the sense that so far no-one has wanted to take me close to my limit. If you manage to get there I’ll let you know.

I’d also be interested in the other betting pools that are presumably around this year.

* Me declaring victory in 2008
* The 2009 results
* Me wondering, in January, what we might bet on this year and various people making impractical suggestions 🙂

[A couple of updates: firstly, I should mention that Gareth has been pushing the ice volume stuff and I admit that picture does look interesting.

and second, NB says “The ice is looking uncommonly mushy on MODIS” which I think I’d pay attention to, and look at myself, if I had the time. Maybe during tomorrows’s batch of compilations.

And third, the default bet is:

That the september mean ice *extent* be below 4.835; but with a “buffer” where we call it a draw: between 4.735 and 4.935, no one wins. I’m taking the “high” side of this; anyone interested in the “low” side let me know. For my part, 4.835 is arrived at as the 1979-2009 trend extrapolated, minus 0.5 which is the SD. It seems to have become tradiational for people to bet small amounts, which is fair enough if we’re just playing. But this was intended to flush out the “the sea ice is in catastrophic decline” people. OTOH, if there are any “the sea ice will certainly recover this year” people then you can get odds on trend-plus-SD, i.e. ice being above 5.835 if you like (note that those are all spuriously precise but never mind) -W]


Sea ice, briefly

I doubt I’ll be running the ever-exciting competition again this year, due to a lack of people who strongly disagree with me (i.e., the decline will be on the long term-trend, plus some error margin).

But While I’m here there appears to be some excitement from Romm over a Grauniad study about a GRL study about the role of wind forcing in sea ice loss, in particular in 2007.

The paper says

The unprecedented retreat of first-year ice during summer 2007 was enhanced by strong poleward drift over the western Arctic induced by anomalously high sea-level pressure (SLP) over the Beaufort Sea that persisted throughout much of the summer. Comparison of the tracks of drifting buoys with monthly mean SLP charts shows a substantial Ekman drift. By means of linear regression analysis it is shown that Ekman drift during summer has played an important role in regulating annual minimum Arctic sea-ice extent in prior years as well.

which seems fair enough (I don’t guarantee it is true, but it looks like a perfectly reasonable thing to say). By the time this has reached the Grauniad the sub-headings have got a little dodgy Wind contributing to Arctic sea ice loss, study finds New research does not question climate change is also melting ice in the Arctic, but finds wind patterns explain steep decline but only a little bit dodgy. By sub-heading standards, this is pretty good (Romm says “especially misleading subhed” which looks over the top to me). Even the Daily Mail isn’t too bad.

For those not familair with the basic idea, sea ice either melts and grows in situ or is pushed about by the winds. Strong winds can either make it advance further than it “ought” to be able to get, or by pushing it too far south melt more of it than would be normal, or push it back north (see, I can’t even tell from the paper which one of those it is. Never mind, I don’t need to know). If you pinned your hopes on 2007 being the harbinger of a new trend, well, you’ve already been disappointed by 2008 and 2009 so this isn’t exactly a hammer blow. It just confirms the obvious: that there was something other than GW that made 2007 an exceptional minimum.

As Romm points out, we (well me, JA and Brian) still have $1000 between us on an ice-free Arctic by 2020, and oddly both sides still seem quite comfortable with that. Romm’s post links to an ice-volume “prediction” of zero by 2012-2016 which I think is absurd. But time will tell. 2012 certainly isn’t far off.

Sea ice again?

At some point I need to decide if I’ll re-start the by-now-traditional sea ice bet for the summer. Before we start arguing over the details, remember that there is a lot of inter-annual variablity so we need to disagree *a lot* to have a meaningful bet. But at least one person has said in the comments that they feel “worried” by this years ice, so that suggests pessimism that I can exploit!

Any thoughts about what you might want to bet about, please leave a comment. My default position is going to be “will fit the 1979-2009 slope extrapolated to 2010”. Bear in mind that this isn’t a forecast – it is some kind of rough idea. If I frecall, RMG thought ~0.5 Msqkm was about the right uncertainty.

Oh, and here is a pointer to to get you started.