The Nenana Ice Classic: 
Betting on Warming

A Critique of - "Climate Change in Non-traditional Data Sets",
by Segarin R., & Micheli, F., Science v.294, p.811, 26 Oct 2001

by John L. Daly

(Comments on this report shown at the end)

The above short paper appeared in Science, the title itself suggesting the author's intent, that of demonstrating `climate change' (code for global warming) in the data of the Nenana Ice Classic.

The `Nenana Ice Classic' (which was featured on this site early this year [2] ) is a betting lottery which began at Nenana, Alaska, back in 1917 and has been run annually ever since.   In January each year the townspeople mount a tripod out on the frozen Tanana River which runs through the town.  The tripod is secured firmly to the ice and a rope run from the tripod to a clock mechanism at the shore.  When the ice breaks up in late April or early May, the tripod gives way, gets carried only a short distance, finally tripping the clock.  The time is then recorded.

Fig.1 - The moment of breakup, May 8th, 2001 (Photo: Julie Coghill)

The lottery organisers take bets from people all over the world, the winner being the person who can make the closest prediction of the actual date and time of breakup.  Because there is big money involved, there is strict adherence to rules. This year (2001), the winner was a Japanese resident.

Fig 2 - Map of the area, Nenana at bottom left

This map shows the location of Nenana on the southern bank of the Tanana River which flows down past Fairbanks to Nenana, eventually joining the Yukon River.

There is a continuous  record of ice breakup going back to 1917 and this has prompted some global warming proponents to examine  the Nenana Ice Classic record for indications of global warming, perhaps to find a `smoking gun'.

The Segarin & Micheli paper suggested two mechanisms which would determine the timing of the ice breakup, namely `thermal effects' (in other words, the ice melting beneath the tripod), and `dynamic effects' (the ice being undercut and broken by mechanical forces from upstream).

While both these mechanisms would be climatic in origin (temperature in the case of melting, precipitation in the case of dynamic breakup), the authors did not explore the possibility that the city of Fairbanks upstream of the tripod might be artificially warming the river as a result of runoff and water usage in the city, all of which would end up in the Tanana River. The effect of an urban warming of the river would be subtle, but could make a significant thermal difference over time as Fairbanks expanded in population.

The record of ice breakup is shown below. The record for earliest breakup (suggesting climate warmth) was on 20th April 1940 at 3.27 p.m.  In 1998, that record was almost matched, but fell short by just 1½ hours, 20th April 1998 at 4.54 p.m.  The latest breakup occurred on 20th May 1964 at 11.41 p.m. during a cold period globally.

Fig.3 - Graph of ice breakup dates on the Tanana River at Nenana, 1917-2001

The total record from Nenana suggests either a slight warming during the 20th century (thus melting the ice) or an increase in precipitation (thus breaking the ice dynamically). This may be climatic, driven by temperature or precipitation, and could be caused in part by urban warming of the river in Fairbanks.  The authors cite Nenana residents observations that thermal breakups have been more common than dynamic ones in recent years, although they offer no actual data or record to confirm this. 

At this point the authors' stated aim of demonstrating `climate change' turned an interesting story into a statistical puzzle piece designed to make more of the trend than in fact existed. They sought to compare the ice breakup times and dates with air temperature and precipitation for the region.

"We examined the entire record of ice breakup to look for long-term trends in breakup. We compared the ice breakup record from 1949-2000 with available climatic data for Nenana and Fairbanks, Alaska (90 km away)."

This was an extraordinary step to take, to use only part of the record for statistical analysis, and to use a cold year (1949) as the start point.  In statistics, this is known as `end date distortion', where a long-term trend can be profoundly affected by the choice of start and end dates, particularly where those terminal dates are outside the mean. Significantly, this decision meant that the record early breakup year, 1940, did not affect the analysis.  Also significant was their omission of the 2001 breakup date (8th May), from the analysis, even though their paper was not published until late October 2001. This left ample time to include it in such a superficial statistical analysis which could be done on any PC, or even to delay publication of the paper a month or so until the 2001 date was included.

They did not state a reason for selecting 1949 as their start point, or for omitting 2001, but it was implied that lack of air temperature data for Nenana itself may have prompted them to do so.  However, the Nenana temperature record had large breaks and so the authors mainly relied on data for Fairbanks 56 miles away.  Using Fairbanks climatic data leaves no excuses for selecting 1949, as its data for both temperature, precipitation, and snowfall, goes back further than 1917.

Fig.3 above shows the effect of this statistical selectivity, the red portion being the period of comparison, which, in the absence of the earlier warm period, produced an upward trend. The failure to include 2001 only magnified that trend. If the whole record is taken, there is a slight trend to climatic warming (urban warming of the river notwithstanding), but much less than would be suggested by using the red portion of the chart.

The authors found that Fairbanks had warmed, and that this warmth correlated with the trends in ice breakup dates. But they also pointed out - "Heat-island effects due to urbanization and movement of the weather station may affect the Fairbanks record."  For `may' read `do', because the Fairbanks heat island is well documented [4].

Fig.4 shows the temperature record for Fairbanks and the university a few miles away.  The university is outside the main urban heat island. Taken together, it is clear that there is an insufficient trend to claim long-term warming. However, if only the post-1949 part of the data is considered, then a warming trend would be evident, due again to `end date distortion', not to any underlying climate change.

Fig.4 - Temperature record for Fairbanks and University, Alaska

Since the Nenana temperature record was too fragmented to mean anything, with a big 15-year gap around 1980, the authors really had no excuse for using 1949 as a start date for determining trends since the Fairbanks and University record above covers the full period of ice breakup data.

Fig 5 - Max, Min and Mean Temperature at University

Fig.5 (left) is a chart for Minimum, Maximum, and Mean temperatures at the University in Fairbanks, produced by the Alaska Climate Research Center [3]. The authors claimed a correlation between ice breakup dates and Min-Max temperature data, but again this only results from `end-date distortion' since even the largely neutral trend for the university shown here would become a warming trend if only the post-1949 data is considered.

Yet this paper was fully peer-reviewed and published in a major journal, the evident statistical flaws escaping the notice of the reviewers or editors.

The authors also looked at precipitation and snowfall as a possible cause of the slightly earlier breakup dates in recent decades, stating categorically - "Trends for precipitation and snowfall were non-significant for both cities." 


That's not how the Alaska Climate Research Center sees it.  Here is their plot of snowfall during the 20th century at Fairbanks.

Fig.6 - Seasonal snowfall totals at Fairbanks, Alaska [3]

As we can see, there has been a significant increase in total snowfall during the 20th century, the inevitable effect of which would be greater mechanical forces acting upon the river ice from upstream, causing it to break earlier than would be the case with thermal melt alone.  But the authors claim -

"Although the record does not reveal the mechanisms of breakup, long-time Nenana residents observe that thermal breakups, where the ice under the tripod "rots" away, have been more common than dynamic breakups in recent years."

Apart from the fact that it must be difficult to tell from visual observation alone whether river ice ice is being broken up through melt or is breaking up under the strain of mechanical forces from upstream, the authors appear to have been only listening to what they wanted to hear from the residents. They do not even say how many residents they spoke to.  It was a most unscientific approach to adopt in what was purportedly a scientific study.  The above snowfall data strongly suggests that dynamic breakup was at work, whatever the authors think a few residents might have told them.



It is clear from the Nenana Ice Classic data that there has been a slight trend toward earlier breakup dates, but nothing nearly as significant as that claimed by the authors - and hardly justifying a paper about it in a major journal. 

The authors attribute recent earlier breakups to climatic warming, but it is clear from Alaska Climate Research Center data that there has been little or no warming to speak of over the full period, but that snowfall has increased significantly, contrary to what the authors claim.  The authors also make no allowance for the possibility of urban warming of the river itself.

Their claims of correlations between river breakup dates and climatic warming are based primarily on selecting only part of the available data, significantly excluding the peak warm year of 1940. They also excluded the data for 2001, a colder than average year, even though there was ample time to include it.  Had it been included, it would have made a significant difference to their comparative analysis.

But then, they might not have got it published in time for the COP7 climate conference in Marrakech, Morocco, in early November. And it is that political imperative which seems to have underpinned what was really very sloppy science.



[1] Segarin R., & Micheli, F., "Climate Change in Non-traditional Data Sets", Science v.294, p.811,
      26 Oct 2001

[2] `Nenana Ice Classic' - Alaska's Coolest Lottery !', 25 April 2001,

[3] Alaska Climate Research Center

[4] Magee, N., Curtis J., Wendler, G.  "The Urban Heat Island Effect at Fairbanks, Alaska",
     Theoretical & Applied Climatology, Vol. 64, pages 39-47

Comments Received 

From "Miceal O'Ronain" <>, 28 Oct 2001

The Nenana Ice Classic: Betting on Warming

A fantastic detective job! I never occurred to me that a river could be an extended conduit for an urban heat sink. In fact, I can't recall ever seeing such a conjecture in print before. Any large city, with a river flowing through it, would have its urban heat sink altered by the presence of the river. As water is a much better conductor of heat, compared to air, a river would greatly extend and focus the area of the urban heat sink down river from the city. We are too accustom to thinking of urban heat sinks in terms 
of air flow.

I was playing with the data this week-end and it never occurred to me to look at a map! I am still troubled by some of the calculations which were performed in the paper, but your conjecture makes these issues irrelevant.

An absolutely brilliant analysis!

P.S. Any published papers of the impact of rivers on urban heat sinks?

From: "George (Jim) Hebbard" <>, 28 Oct 2001

It has likely occurred to you by now that thermal warming of the Tanana River by Fairbanks should obviously cause an increase in water temperature from upstream to downstream of the town.

I am in no position to check this out (I live in Florida) but doubtlessly you will receive much correspondence of this topic in the next few days. Please suggest to someone connected to the Fairbanks University that they measure a series of stream lines in the river and see if the average data reflects significant liquid thermal island effect.

I await eagerly their input.

George (Jim) Hebbard, PE, ChE
Lithia, FL, USA

From Jerry Brennan  <>  29 Oct 2001

It took me quite a while to realize why, but 1998 was the earliest.

1940 was a leap year, so April 20, 1940, was the 111th day of the year, while April 20, 1998, was the 110th day of the year.

Otherwise, the article is excellent (although in discussions of Alaskan temperatures through that period, I like to see mentions of the PDO shift in the mid 1940's, and especially the 1976-1977 shift). 

Regards,        Jerry B

Response from John Daly - 29 Oct 2001

Jerry -

Point taken.  The official Nenana data is based only on calendar dates and times, and this is the basis for Fig.3. However, if an astronomical calendar is used instead, some times/dates would need to be moved slightly. It would not make much difference to the overall pattern of breakup, but as you say, it would make 1998 the record year. 1998 was also the year of the big El Niño (a warming effect) and was in the positive phase of the Pacific Decadal Oscillation (PDO) which has also had a warming effect post-1976.

In the last two years, the indications are that the PDO has now moved into its negative phase which would explain the later breakup date in 2001.   -  John Daly 

From: Jan Curtis < >  29 Oct 2001

Just as important as the homogeneity of weather stations when conducting climate studies, the issue of uniformity of erecting the Nenana River tripod should be taken under consideration. While the current method is to cut the ice in the river, then refreeze the tripod's legs about 4 to 6 weeks prior to the actual river's breakup, I am not certain if this procedure was always employed.

My limited experience while working at the Alaska Climate Research Center was that snowfall served to insulate the river ice from the extreme cold, allowing the relatively warmer river to help melt the ice faster from beneath. Light snow cover on the river could possibly result in delayed breakup. The bottom line is that river ice dynamics/thermodynamics is very complex and cannot be systematically used as a on/off switch for determining climate change.

Jan Curtis
Wyoming State Climatologist

From: "John S. Baltutis" < >, 29 Oct 2001

Superb analysis! Keep up the great work. I hope you sent your findings to Science as a rebuttal to the paper, though I doubt they'd publish it.

John Baltutis

Sat, 3 Nov 2001 14:14:08 -0000
From: Ian Tilbury <>

As usual another excellent demolition job on on the suit's-us Scientists.
The most alarming thing is the way this is presented in a supposedly 'scientific' journal, as an Engineer I look at this 'study' and the first thing that strikes me is the attempt to make connections from data with so many interconnected variables, ie

1/ Thickness of ice.
2/ Failure mode of ice, Thermal/Dynamic or more likely a combination.
3/ River temperature.
4/ Urban heat sink effect.
5/ Thermal insulation effect of documented, increasing snowfall.
6/ Uniformity of Tripod erection.
7/ Air temperature.
8/ Possible outside interference from gamblers ? 

With all these variables how can anyone claim to be able to produce a 'scientific' result ?

Incidentally on BBC radio 4 recently there was a programme called 'Captains Log'.  This highly amusing offering told how 'scientists' are using the Log books from the days when Britain ruled the waves to try to find other 'smoking gun' indications of 'Global warming'.  This not surprisingly brought forward the great storm in the 18th Century and many other 'exceptional' weather events. Unperturbed by this contradictory data they have now gone further back to, youv'e guessed it - the Bible and Babylonian writings of pre-history. I am currently waiting for a programme showing how man's thoughtless use of oil lamps caused the biblical flood!

Ian Tilbury (GB)

From: Ross McKitrick <>   Wed, 07 Nov 2001 16:51

The attached file nenana.doc lists the number of days past March 1st of the ice breakup each year since 1917, in case anyone wants to work on it. I haven't adjusted for leap years. One of my students asked about the story because he saw it written up in the Toronto Globe and Mail. I presented the story to my class as an example why they must always read sources critically. 

The role of the Pacific Climate Shift is important here, not surprisingly since Alaska is where it is. Over the whole sample the break up occurs 0.068 days per year earlier or 1 day every 14.7 years. The trend is significant (t=2.70). The Science (1949-2000) subsample shows a breakup of 0.16 days per year earlier or one day every 6.4 years. The sample up to the PCS (1917-1976) shows no trend: a breakup .009 days earlier per year or one day every 111 years and the t stat is only 0.22. The post PCS sample (1979-2001) has a trend of .19 days earlier per year (1 day per 5.3 years) but it is also insignificant  (t=0.98). 

At my course page (see entry for Oct 31) I have a pdf file with some more calculations and some surface air temperature data showing the striking effect of the PCS in nearby stations, as well as the boring temperature record from Tanana itself, all from GISS.  

Cheers Ross
Ross McKitrick

Associate Professor
Department of Economics
The University of Guelph
Guelph Ontario Canada, N1G 2W1
tel (519) 824-4120 ext 2532
fax (519) 763-8497

From:  "Dr Wilson Flood" <> Wed, 7 Nov 2001 10:18

Take a look at the graph you have drawn in Fig3. The blue line could quite legitimately extend to about 1970.It should not be red. The temperature clearly falls until 1965 so why red? This would match the drop in temperature between 1940 and 1970 seen elsewhere followed by the slight rise from 1970 to 2000. Once again the 1998 El Nino year is showing up. Without 1998 most arguments for global warming would look pretty silly. Most scientists discount a rogue result like 1998 unless of course you can use it to justify your case. The whole global warming scenario rests on 1998.

Wilson Flood PhD 

Response by John Daly - I marked the record from 1949 to 2000 in red because this was the period of time represented by the authors statistical comparisons with Fairbanks and Nenana. It was their selection, not mine.

From:  Hiram Jacobs <>  Fri, 9 Nov 2001 17:35

With regard to the Nenana Ice Classic study, I noticed that you did not reference or discuss the study by Magnuson et al. that examined 39 freeze-thaw data records and found trends both for later freeze-up and earlier spring thaw. Magnuson et al. found that spring thaw was occurring  approximately 10 days earlier over the 150 years encompassed by the records (freeze-up was occurring about 9 days later). I can't tell from the graph how many days earlier (from 1940 to present) the Nenana ice break-up is, but the earlier trend is consistent with Magnuson's data. Some of Magnuson's data is for a 
lake in Switzerland (Lake Constance) that would essentially integrate processes in the Alpine region. Since you have indicated that recession of mountain glaciers is not indicative of a global warming trend, the earlier spring thaw dates for this lake are an interesting question.

URL for a layman's level discussion of the Magnuson et al. paper: 

Reference for the Magnuson et al. paper:   -   Historical Trends in Lake and River Ice Cover in the Northern Hemisphere, John J. Magnuson, Dale M. Robertson, Barbara J. Benson, Randolph H. Wynne, David M. Livingstone, Tadashi Arai, Raymond A. Assel, Roger G. Barry, Virginia Card, Esko Kuusisto, Nick G. Granin, Terry D. Prowse, Kenton M. Stewart, and Valery S. Vuglinski, Science 2000 September 8; 289: 1743-1746.

Hiram Jacobs


FastCounter by bCentral

Return to `The Greenhouse Hall of Shame' page

Return to `Still Waiting for Greenhouse' main page