RealClimate logo

The Nenana Ice Classic and climate

Filed under: — gavin @ 7 March 2014

I am always interested in non-traditional data sets that can shed some light on climate changes. Ones that I’ve discussed previously are the frequency of closing of the Thames Barrier and the number of vineyards in England. With the exceptional warmth in Alaska last month (which of course was coupled with colder temperatures elsewhere), I was reminded of another one, the Nenana Ice Classic.

For those that don’t know what the ‘Classic’ is, it is lottery competition that has been running since 1917 to guess the date on which the Nenana river ice breaks up in the spring. The Nenana river is outside of Fairbanks, Alaska and can be relied on to freeze over every year. The locals put up a tripod on the ice, and when the ice breaks up in the spring, the tripod gets swept away. The closest estimate to the exact time this happens wins the lottery, which can have a quite substantial pot.

Due to the cold spring in Alaska last year, the ice break up date was the latest since 1917, consistent with the spring temperature anomaly state-wide being one of the coldest on record (unsurprisingly the Nenana break up date is quite closely correlated to spring Alaskan temperatures). This year is shaping up to be quite warm (though current temperatures in Nenana (as of March 7) are still quite brisk!).

Since there is now an almost century-long record of these break up dates, it makes sense to look at them as potential indicators of climate change (and interannual variability). The paper by Sagarin and Micheli (2001) was perhaps the first such study, and it has been alluded to many times since (for instance, in the Wall Street Journal and Physics Today in 2008).

The figure below shows the break up date in terms of days after a nominal March 21, or more precisely time from the vernal equinox (the small correction is so that the data don’t get confused by non-climatic calendar issues). The long term linear trend (which is negative and has a slope of roughly 6 days per century) indicates that on average the break up dates have been coming earlier in the season. This is clear despite a lot of year-to-year variability:

Figure: Break up dates at Nenana in Julian days (either from a nominal March 21 (JD-80), or specifically tied to the Vernal Equinox). Linear trend in the VE-corrected data is ~6 days/century (1917-2013, ±4 days/century, 95% range).

In the 2008 WSJ article Martin Jeffries, a local geophysicist, said:

The Nenana Ice Classic is a pretty good proxy for climate change in the 20th century.

And indeed it is. The break-up dates are strongly correlated to regional spring temperatures, which have warmed over the century, tracking the Nenana trend. But as with the cool weather January 2014 in parts of the lower 48, or warm weather in Europe or Alaska, the expected very large variability in winter weather can be relied on to produce outliers on a regular basis.

Given that year-to-year variability, it is predictable that whenever the annual result is above trend, it often gets cherry-picked to suggest that climate change is not happening (despite the persistent long-term trend). There are therefore no prizes for guessing which years’ results got a lot of attention from the ‘climate dismissives’*. This is the same phenomenon that happens every winter whenever there is some cold weather or snow somewhere. Indeed, it is so predictable** that it even gets its own xkcd cartoon:

(Climate data sourced from Climate Central).
The fact remains that winters have been getting warmer on average. While scientists are very interested in potential influences on the variability (whether from volcanoes, solar effects, greenhouse gases, or Arctic sea ice loss), it remains the case that this is much harder and more uncertain than attributing trends in the mean, which as should be clear, are much more robust. As yet there is no truly convincing evidence that any change in variance has been detected, though there are a lot of ideas out there, and some very interesting discussions (see for instance, Francis and Vavrus (2012) and Barnes (2013)).

For fun, I calculated some of the odds (Monte-Carlo simulations using observed mean, a distribution of trends based on the linear fit and the standard deviation of the residuals). This suggests that a date as late as May 20 (as in 2013) is very unexpected even without any climate trends (<0.7%) and even more so with (<0.2%), but that the odds of a date before April 29 have more than doubled (from 10% to 22%) with the trend. The most favored date is May 3rd (with no trend it would have been May 6th), but the odds of the break-up happening in that single 24 hour period are only around 1 in 14.

So, the Nenana ice Classic – unlike the other two examples I mentioned in the opening paragraph – does appear to be a useful climate metric. That isn’t to say that every year is going to follow the long-term trend (clearly it doesn’t), but you’d probably want to factor that in to (ever so slightly) improve your odds of winning.

* Yup. 2001, after 2008, 2011, and 2013.
** It is so predictable, I am thinking about opening a derivative market on whether this year’s Nenana result will get mentioned.


  1. R. Sagarin, "Climate Change in Nontraditional Data Sets", Science, vol. 294, pp. 811-811, 2001.
  2. J.A. Francis, and S.J. Vavrus, "Evidence linking Arctic amplification to extreme weather in mid-latitudes", Geophysical Research Letters, vol. 39, pp. n/a-n/a, 2012.
  3. E.A. Barnes, "Revisiting the evidence linking Arctic amplification to extreme weather in midlatitudes", Geophysical Research Letters, vol. 40, pp. 4734-4739, 2013.

28 Responses to “The Nenana Ice Classic and climate”

  1. 1
    Uma Bhatt says:

    Nenana (and Yukon) ice break up has been investigated in the context of the large scale climate and is closely related to what is going on in the North Pacific (Bieniek, P. A., U. S. Bhatt, L. A. Rundquist, S. D. Lindsey, X. Zhang, R. L. Thoman, 2011: Large-Scale Climate Controls of Interior Alaska River Ice Breakup. J. Climate, 24, 286–297).

    I taught a climate statistics class at the University of Alaska Fairbanks, Dept. of Atmospheric Sciences in 2011 (I think) during spring semester and as a class team exercise we ‘forecast’ the Nenana ice breakup date. We tried numerous parameters using simple multiple regressions and we even bought tickets to the ice classic (giving an oath that the money would be shared evenly if any of our times won). In the end, our forecasts did not win but just about all of our forecasts were within two days of the actual breakup, which is not bad for a forecast but not good enough to win the jackpot which requires minute accuracy (

  2. 2
    tamino says:

    I blogged about the ice classic three years ago, here:

    Yes there’s correlation with other variables, including the PDO, but that correlation is in the short-term variations. Even after removing the PDO correlation the trend remains, and is statistically significant. What could be the cause? Hmmm…

  3. 3
    Magma says:

    But obviously the two 1940 & 2013 data points prove that global warming is a fraud, Al Gore is fat and we should be preparing for a new ice age.

    Or something like that.

  4. 4
    Radge Havers says:

    Ice-Break on the Santa Cruz:

    No comment.

  5. 5
    Pete Dunkelberg says:

    Even for the Arctic overall winter temperature variation is striking, as you can see by clicking through the years. Is there any small area outside the tropics where winter (for the appropriate hemisphere) temperatures are not volatile?

  6. 6
    jyyh says:

    nice post illustrating the effect of statistical outliers immediately after they’ve happened, thanks.

  7. 7
    Caroline says:

    Has anyone looked at the changing distribution of guesses for Nenana? I mean, obviously the actual time and date of ice out reflects changes in (local) climate, but the guesses would reflect what people who live in the region perceive as “normal” which would also be important data.

  8. 8
    Dan H. says:

    Yes Magma,
    Selecting those dates would suggest that. Just like using 2014 contiguous U.S. winter temps. Conversely, choosing 1998 or the U.S. winter of 2012 would indicate the opposite. There is a definitive long-term trend, and a good correlation with the pdo on shorter intervals.

  9. 9
    John Mashey says:

    This is a nice reminder of a long-ago visit to Nenana, on way from Denali to Fairbanks. I do believe British viticulture has progressed since 2006, and Richard Selley’s 2 editions of the The Winelands of Britain are delightful, but sadly, not easy to buy. However, from examinination of current data, I think there are enough vineyards in Yorkshire to support a wine tour on our next visit there.

    As Gavin notes, UK vineyard count is not necessarily a good temperature proxy, for various reasons, such as varying geology, buildover of earlier vineyards, different grapes, etc. Although I don’t know of academic studies, a better vineyard proxy, with fewer confounding factors than uk, might be the recent history of Okananagan, B.C. wine region, in which vinifera plantings have generally spread northward around a 70-mile long lake. Obviously, this could only be a recent proxy for mid-20th century temperatures, as I doubt there was a medieval wine industry here.

  10. 10

    #9–Ontario, too:

    Especially the intrepid souls on the Bruce Peninsula!

  11. 11

    On the ice page there is a lot more info (than just the breakout date). They have several ice thickness measurements varying from just 6 points (1990) to more than dozen. Don’t you think a graph based on this data would be more robust?

  12. 12
    Bob says:

    Tamino, well that must be it, because correlation is always causation. Hmm!

  13. 13
    Ray Ladbury says:

    Bob, you want to try that comment again, but this time try making sense?

  14. 14
    Bob Maginnis says:


    It is the Tanana River that breaks up, at the town of Nenana.

  15. 15
    BillS says:

    In Maine, USA, ice-out, the day that a local lake sheds its ice has been a wagering date for 100 years or more. Ice out day is both a source of amusement and hope — hope that spring is not that far off. Data for some of the lakes goes back to about 1850 or earlier. The US Geological Survey has compiled data on 29 New England lakes ice out dates. Data can be found here:‎

    Unfortunately, the USGS hasn’t updated the data since about 2000. This year may be an outlier.

  16. 16
    Steve Metzler says:

    @BillS #15:

    tamino had this one covered too, about 3 years ago in response to yet another lame attempt at WUWT to correlate ice out dates with anything but AGW:

    Ice Out

  17. 17
    BillS says:

    @16 Steve Metzler:

    Thanks for the link. Nice to see someone else bothered to plot the dates. While the USGS has stopped compiling the dates Maine’s Dept. of Parks & Lands has ice-out dates from 2003-2013 but the list of lakes & ponds varies from year to year.

    I suppose an alternative explanation to AGW is that the ice departs when the ice fishermen run out of beer.

  18. 18
  19. 19
    Walter says:

    “non-traditional data sets that can shed some light on climate changes.” ???

    A 6 day per century trend …. You have got to be kidding, or you have waaaaaaaaaaay too much time on your hands to be playing with this, and then to also spend the time to write it up as an “real article” about climate “science” here or anywhere else.

    Maybe giving a drunk presentation at the xmas party would fit.

  20. 20

    Minnesota is the land of 10,000 lakes and I have a blog post looking at the ice-out trends as well:

    The last year and likely this year are upticks in the ice-out date but that is not surprising as it does fluctuate quite a bit. Minnesota is large enough that I separated the regression curves out in 1degree latitude zones.

  21. 21
    Gonzo says:

    Interesting if you invert Nenana data it looks a lot like pre-adjusted land temp records. If I was a betting man I’d the trend will be going down (later break up dates)as the trend since the 98 El Nino (i know a cherry pick) is certainly down.

  22. 22
    Steve Metzler says:

    @Walter #18:

    Sure, the Tanana river at Nenana is *one data point*. Have a look at the 3 Maine lakes that tamino examined in the ‘Ice Out’ link from my post at #15. Just since 1970, those lakes are icing out ~8 – 10 days earlier. That’s a pretty significant trend. And worrying for those that appreciate the implications of such.

    Liked the “science” in scare quotes, BTW. Classy.

  23. 23
    Magma says:

    Between the USGS report and the Maine State government link that BillS provide it looks like there could be a continuous series for ice-out dates for up to ~15 different lakes in Maine through to 2012. Be a bit of a pain to put it into spreadsheet form, but the data is there.

  24. 24
    scott Halama says:

    you have to throw the first 60 guesses out and all the passed in the passed

  25. 25
    Dan H. says:

    Web hub,
    That data does not surprise. According to NOAA, the growing season in Minnesota has increased by 9 days in the past half century. Those two datasets appear to agree.

  26. 26
    scott Halama says:

    pass that and throw short this year…

  27. 27
    Carbomontanus says:


    I also make my own proxies of that kind and for several reasons. First of all maybe in order to have som personal amateur training at least, with what the big guys are dealing with, in order to understand it better and to see when they cheat.

    For instance: When did King Halvdan Svarte plunge through the ice of Randsfjorden with horses and all his men in the warm mideival period?

    Next to there is 4 mideival churches at least, of stone masonry with burnt lime mortar and plaster with Fresco paintings.

    Which is an extreemly important document of mideival summer warmth and length of summer season. Because that lime mortar is quite extreemly sensitive to low temperatures when set up. I learnt that from a swedish church mason, and I have looked further into it. The limit between stone and lime mortar masonry in the north and northwest Europe is quite a reliable document for judging summer warmth i the area.

    You have to be able to guarantee mild weather with absolutely no frost one month onward, for that mortar to form Calcit CaCO3 and not Ikait CaCO3 . 6H2O. Ikait is stable below +5 Celsius, and if that forms, the very plaster will fall down like dry chalk powder next summer, and all that burning of wood was in vain.

    It takes a quite tremendous heat. Think of an endoterm process that goes on at 900 Celsius which is high orange near yellow hot. Only the King and the Church could afford it. Redbrick is burnt at 600 thus quite much cheaper.

    But to burn brick and lime requires huge thimbers and forests that can be harvested, logged and chopped. That was not the case on Island for instance, not even on Færøerne in the warm mideival period, but on the Orkeneys, St.Magnus!

    But in Norway you find it all the way down at the sea and a bit up the walleys hardly higher than 200m, and all around the coast, Oslo Stavanger, Bergen, Selje, Nidaros, and Alstadhaug .

    Above that and north of that you rather find the extreemly fameous Norwegian Stave Churches, for which the Pope had to give Dispensatione for building in wood. And you find the same limit and phaenomena in the Novgorod Arch Bishop See area, the extreemly fameous russian wooden cathedrals from the warm mideival period.

    Thus notice Calcit- Ikait, and lime mortar and fresco paintings and the limits of that in the warm mideival period.

    We have a Vitis vinifera in the garden at 120 m above sea level. It was stolen right down here in Oslo down at the sea in Norways warmest hole, probably planted there by Swedish Railway workers in the warm thirties. Now it stands 100 m higher up and gets ripe at Eqvinox.

    That is in scale, but neither Vitis vinifera nor Barley are good indicators, because they are cared for artificially by quite extreme, we dare to say fanatic interest by some people. Both origine from Kurdistan and have spread worldwide by human interest. Reliable botanic indicators and proxies should rather be looked for in the wild and very well adapted local vegetation. That is rather good old gardener and peasant wisdom.

    I found a survey of eastern European wines. That was quite incredible. Some sorts can survive -30 celsius and get ripe in the first week of September. Probably for clergical reasons, because the russians rather drink beer and vodka, just as the English ratker drink Scotch and brown ale and just a glass of port for their nerves.

    But those grapes in the sunny wall is not for wine. It is for eating and to have a personal mark of what is possible in the garden. Thus wine growing in Scotland is obvious and tells us only of the scots.

    If you manage sweet plums then you can also try a hardy wine.

  28. 28
    Sceptical Wombat says:

    It may not signify anything but the Australian air safety regulator has just announced that last summer had an extremely high number of air turbulence incidents affecting large passenger planes. Maybe there is an opening for yet another index?