Friday, September 20, 2013

Winter Trends in Early Snow Years

Wednesday's early accumulating snow in Fairbanks raises the question of whether similar occurrences in the past offer any hints as to the climate patterns in the upcoming winter.  I'm always on the lookout for analogs like this, because years ago I found that unusual early-season snowfall in certain parts of the lower 48 can be a useful predictor for winter conditions.  It's conceivable that the same might be true in interior Alaska.

To take a first look at the data, I pulled out the years in which measurable (0.1" or greater) snow was observed in Fairbanks on or before September 20.  Since 1930, there have been 10 of these years prior to this year: 1930, 1933, 1948, 1962, 1972, 1978, 1980, 1989, 1992, and 1993.  Looking at the temperature and snowfall data from the subsequent winters, it seems there is a tendency for warmer than normal and snowier than normal conditions in each of November, December, and January.  For the three-month average, 2 of the 10 years had below-normal temperature, 2 had near-normal temperature, and 6 had above-normal temperature (based on the 1930-2012 tercile categories).  The distribution was the same for snowfall: 2 below-normal, 2 near-normal, and 6 above-normal.

Below are histograms of the November-December snowfall and temperature, with the columns split between the early-snow and no-early-snow years.  Although a subset of 10 years is rather small to draw a robust conclusion, the distribution does seem to be shifted towards relatively warm and snowy conditions in the years when early snow occurred.

In conclusion, the early Fairbanks snowfall seems to give us a very tentative hint that November through January might be warmer and snowier than average in Fairbanks this winter.  It's not clear why this might be the case, although the early snowfall might be signaling something about the oceanic or atmospheric pattern that is likely to persist or influence the winter conditions.  There is a lot more that could be said about the outlook for the winter, based on other factors, and I hope to put up some more posts in the weeks ahead.


  1. Very interesting info as always. I better service my snowblower just in case.

    The PDO temp trend ( seems somewhat cool compared to the '80's-mid-'90's when some relatively warm winters and snow were experienced.

    I also see the CPC's looking at a neutral ENSO forecast ( so that's better than cold perhaps.

    Other factors are at play of course like jetstream looping and arctic ocean temps. We'll just have to pop some corn and watch it play out.


  2. Thanks Richard. As I commented in a previous post, I do like analogs. They provide a good starting point - especially if you assume that climate is not entirely stochastic.

  3. An initial way to evaluate the observed distribution would be to ask: what is the chances of getting six out of ten "warm or snow" in a three-category outcome when the category chances are equal. So this is a straightforward binomial probability, which works out to 7% chance of six or more occurrences. Not bad.

    So the next question is: why would a single weather event have some future predictive value of an independent season? Hmmm...

    1. Rick,

      I've wondered the same thing many times. Here's one way I have thought about it: conditions have to be "just right" to get snowfall well outside the usual season, so perhaps there is a large-scale phenomenon, or set of conditions, that shifts the distribution to make early snowfall much more likely. Then we can think of the event itself as a symptom of something much bigger; and the much bigger thing, whatever it is, either persists or influences the weather pattern for months to come. The early snowfall is a fingerprint of the important driver.

      I'd like to go back and look in more detail at the analog years to see what similarities can be found in major oceanic and atmospheric anomalies around the time the early snow occurred.