Nenana Ice Classic 2016

Just a quick note since I’ve been tracking this statistic for a few years, but the Nenana Ice Classic tripod went down this afternoon (Apr 23, 3:39 Alaska Standard Time). See the earlier post for what this is and why it says something about the climate (see posts on 2014 and 2015 results).

With this unofficial time, this year places 4th earliest for the breakup of ice in the Tanana river. It is unsurprising that it was early given the exceptional warmth in Alaska this year.

The exact ranking of years depends a little on how one accounts for leap-year and other calendrical effects. The raw date is the 4th earliest, but given that this year is a leap year, it would be the 5th earliest counting Julian days from the start of the year. Tying the season to the vernal equinox is more stable, which again leads to the 4th earliest. But regardless of that detail, and consistent with local climate warming, the ice break-up date have advanced about 7 days over the last century.

As a side bet, I predict (based on previous years) that despite enormous attention in the skeptic-osphpere given the Nenana result in 2013 (when it was remarkably late), it won’t be mentioned there this year.

  1. Brian Gunter:

    It is interesting that the date of the Nenana breakup is not highly correlated with the maximum ice thickness. It seems as though river conditions, such as high flows caused by rainfall and/or snowmelt in upper catchments or due to the collapse of ice-jams either upstream or downstream of Nenana, may be major factors affecting the breakup date. Of course the breakup date is still dependent on the overall climate of Central Alaska. Have a look at some doodling that I have done with available data at:

  2. Kevin McKinney:

    Ah! Thanks. A colorful little sidebar item.

    I suspect that if you want takers on your side bet, though, you’re going to have to offer some darned attractive odds.

  3. Russell:

    A literally periphaeral question : the break-up depends on mechaical as well as thermal action. What does the illustrated trend tell us about the evolution of average sea state and wave height in the unforzen ocean adjacent to the propagating break-up zone?

  4. John West:

    Is that a statistically meaningful trend?

  5. Chase Stoudt:

    So I did two analyses. I looked at freezing degree days and did a linear regression then used that model to predict the day (~50% of the variance). I then did a histogram count (the bins were arbitrarily chosen) and used the largest bin counts for the time of day breakup.

    The day the linear model spit out was April 23rd, the time from the histogram was 3:40.

    Of course I tried to be clever and assumed that nature doesn’t like round numbers, so I changed my bet to 3:41.

    I can still win it if I’m the closest bet, it’s going to take them a couple of weeks to count all the tickets though.

  6. Henry:

    But aren’t the uncertainties on the gradient (which you don’t give here) so large as to make this basically an upper limit on the rate of change?

    [Response: The OLS uncertainty ±4 days/century (95% CI). So the upper limit would be ~11 days/century. – gavin]

  7. Steve Garvin:

    There is clearly a pause in the downward trend since 1993. The day has actually been getting later since then.

  8. L Hamilton:

    Nenana, like the ice-out dates on Lake Winnepesaukee, makes a nice stats-course example. Residuals from the OLS line that Gavin shows are not autocorrelated, so there is no advantage to fitting an ARIMA model instead. They are slightly non-Gaussian, mainly the two high outliers (1964 and 2013), so it’s interesting to note that if we use an outlier-resistant method that does not assume normality, such as robust or quantile regression, those see an even steeper decline (-9 or -11 days/century, respectively). For any of these trends, nominal t test probabilities are below 0.0005.

  9. Kevin McKinney:

    #3–Don’t have a properly quantitative answer to your question, Russell, but based on this Google Map image, I think the qualitative answer is “pretty darned little”:

  10. Kevin McKinney:

    Sorry, wrong link:


  11. Mark A. York:

    A friend from high school in Maine has lived in Fairbanks for decades. He says the winters are nothing like the’70s. I believe it as I was there then and beyond, too, anecdotal though that info is.

  12. Dana Mix:

    I have a way to solve the leap year problem. I’ve done it. You calculate the time between the tripod falling and the instant of the new year for 1917. Then use the Julian year of astronomy, 365.25 days from then on. So for 1918 the time used is Jan 1st at 6 a.m. I don’t have the spreadsheet with me, but I’ll post it. The change was 7.6 days per century.

  13. Shawn Hernan:

    Interesting data. Does there exist any data that could determine what, if any, effect that contaminants in the river could be contributing to freezing point depression of the water? Would it not be the case the great use of salt on the roads, for example, or petroleum byproducts seeping into the water would have an effect as well?

  14. Dan H.:

    Yes, that is correct. You can even go back a few more years to 1988, and still obtain a trend towards later break up dates. That can often happen when the dataset has only 30 points. However, any date prior to that will result in an advancement of the break up date. That is based on straight linear averaging, although I suspect you are using the 10-year averaging, as shown here:

  15. JKSmith:

    The system is super hard to predict with sparse data sets.
    Up river the temperatures were very high this year and melt off occurred early. It looked like melt water flowed out and did not lift the ice out. I expected a date prior to 4/20 based on heating degree days. Then some late snowfall occurred. Snow tends to act as insulation and slows impact solar heat input. Another big input to the system is the melt date / flow of the Nenana River. It joins the Tanana just downstream of the Tripod. When the Nenana river goes out the local ice elevation goes up – that floats the ice and it goes out. For what it’s worth my guess was within 67 seconds. #13 – I never saw salt or anything going into the watershed, #11 Every old timer in the Yukon had it harder than recent arrivals, #9 Yes, no tides in Nenana.

  16. Kevin McKinney:

    “Does there exist any data that could determine what, if any, effect that contaminants in the river could be contributing to freezing point depression of the water?”

    Good question. Have no idea about such data, but take a look at the geography involved–the sizes of the river and of the town (population 378 in 2010). I seriously doubt that there’s anything like enough contamination to make much difference one way or another. (It is true that Nenana is around 60 miles downstream of Fairbanks, which boasts better than 30,000 inhabitants.)

  17. Richard Caldwell:

    Dana: The change was 7.6 days per century.

    Richard: Excellent work! Now, the diurnal cycle still throws a wrench, (since you’re starting at different times of day) but I suspect that it’s smaller than the wrenches other techniques throw.

  18. Fred Simpson:

    More proof of a little to a bit of a cooling trend over the last 25 years in spite of CO2. Thanks for sharing.

  19. Titus:

    A quick comment to your quick note.

    Looking at the graph it appears to me that the ‘Break-Up Dates’ have got later in the last 20 years or so. Definite upward trend since mid 1990’s.

    I’m pretty neutral on this topic but thought it worth a mention.

  20. Hank Roberts:

    Poor Fred Simpson sees a blue line, sloping down from left to right, and mistakes it for a cooling trend.

    Sorry, Fred.

  21. Ray Ladbury:

    Uh, Fred Simpson, what kind of hallucinogenics do you have to consume to see a cooling trend in that data?

  22. Kevin McKinney:

    Fred Simpson, #18–More proof that you know which cherries you prefer.

  23. Greg Simpson:

    Fred, it’s extremely noisy data since it’s about one small area. It’s going to take more than 25 years to reliably see a trend.

  24. Edward Greisch:

    “Arctic sea ice is falling off a cliff and it may not survive the summer”

    “Temperatures for the Arctic are expected to range between 2.5 and 3.5 C above average over the next seven days”

    “Such low sea ice measures will also tend to wrench weather patterns around the globe — providing zones for extreme heatwaves and droughts along the ridge lines and related warm wind invasions of the Arctic that will tend to develop all while generating risk of record precipitation events in the trough zones.”

    So Nenana isn’t alone.

  25. Hank Roberts:

    > It’s going to take more than 25 years to reliably see a trend.

    Anyone done that arithmetic? Got error bars?

  26. Dan H.:

    You may be the one on hallucinogens. There is a (very) slight cooling trend over the past 25 years.

    You may wish to re-read Fred’s post to accurate understand that he said 25 years, not 89 years.

    Yes, it is extremely noisy, but that does not negate the existences of a trend. Personally, I would say that the trend over that time frame is flat, as the slope is extremely small compared to the scatter in the data.

    Trends change, and claims that one simple, linear trend exists over the entire dataset may not be the best interpretation.

  27. Hank Roberts:

    > re-read Fred’s post to accurate understand that he said 25 years
    Cherries ….

  28. WebHubTelescope (@WHUT):

    This set of data of historical Minnesota ice-out dates I prepared is un-cherry-picked for the Curry disciple Dan H.

    Preparation H, lol.

    Been a couple years since I did this chart. With this year’s near record early ice-outs, it should strengthen the correlation.

  29. Ray Ladbury:

    Dan H.,

    Hmmm. On Dan’s planet it is possible for a cooling trend to give rise to 2 and perhaps 3 back-to-back record hottest years! Neat trick.

    Dan, do you practice saying these things in front of a mirror to suppress the grin that must inevitably tug at the corners of your mouth when you say anything so absurd?

  30. Dan H.:

    I do not know where you get the idea that we are in a cooling trend. Certainly not from anything I have said. Perhaps you are just unaware that we are discussing the Nenana Ice Classic, rather than global temperatures. In this instance, we have not witnessed 3, or 2, or even 1 record year. Indeed, the year was the 4th earliest. It was close to the 3rd, but far distant from numbers 1 and 2. While a simple, linear trend is nice to see how the break-up date has changed over history, it is rather limited in its details. A running average can be more telling, and shows that the entire trend towards earlier break-up dates occurred between ~1970-2000. There has been no change since. y the way, Nenana is on this planet, not whatever planet you think might be cooling.

  31. Kevin McKinney:

    #26, Dan H–

    “Yes, it is extremely noisy, but that does not negate the existences [sic] of a trend. ”

    Yes. Including a continued “warming” one. (Quotes since the data are a proxy for temp, at best.)

    Hence, “cherries.”

  32. Hank Roberts:

    > A running average … shows that the entire trend towards earlier break-up dates
    > occurred between ~1970-2000. There has been no change since.


  33. Martin Bernstein:

    Umm, Dan H., I think I might know where Ray got the idea that you thought there was a cooling trend: Post 26 in which you typed, and I quote, “There is a (very) slight cooling trend over the past 25 yearts.” Pretty hard to misinterpret that! This is almost as funny as the time you reversed the polarity on the drought index in a discussion on Tamino’s blog.

  34. Dan H.:

    I think you are confusing the local Nenana trend with the discussion of a global trend. [edit – tedious]

  35. Cody Nichols:

    To Brian Gunter – Thanks for pointing that out! It adds a really interesting dynamic to the subject. Though, if you think about it, rainfall and snowmelt are both significantly impacted by change in the climate. So, even if it is one step removed, the correlation is still well intact.

  36. jyyh:

    Still the winter radiative cooling will keep the amoc running at some level.