Wednesday, February 21, 2024

More Context for Cold

Another round of downslope/chinook warming brought another widespread thaw to the Tanana River valley and Fairbanks-land yesterday.  Temperatures above 40°F were widespread, including 45°F at Fairbanks airport, the highest February temperature in 20 years.  The year-to-date average temperature in Fairbanks is now only 2.2°F below normal.

Looking back again at the late January and early February cold snap, I was curious to see how the lower atmosphere (not surface) temperatures compared to past cold spells on various time scales.  However, with NWS weather balloons no longer going up during severe cold, there's some missing data for the recent event; so I extracted data from the ERA5 reanalysis.  I looked at Fairbanks 850mb temperatures and 1000-500mb thickness, with the latter being an excellent measure of the average temperature of the lower half of the atmosphere.

Here's the annual (winter) minimum for 3-day average 1000-500mb thickness:

On this time scale, the recent event was one of the coldest since 2000, but it was nowhere near as cold as the more extreme events of earlier decades.  The complete absence of seriously cold events after 2000 is quite remarkable, and suggests that the cold tail of the distribution has been dramatically curtailed since then.

Below is another perspective, showing annual minimum values of 3-day 850mb temperature (y-axis) versus 1000-500mb thickness (x-axis).  The values are quite highly correlated, of course, but there's some variability in the extent to which cold is concentrated at lower atmosphere levels like 850mb.  On a 3-day basis, the recent event was relatively less unusual in terms of 850mb temperature, with several other events being colder in recent years.

The 7-day and 14-day charts below show quite similar findings on these time scales, with the recent event being a bit more anomalous overall on a 14-day basis, as I noted before; but it was still not in the least unusual compared to earlier decades.

The two "granddaddy" cold spells of 1989 and 1999 stand out clearly here; those two events were much more extreme than anything in recent years, and also worse than anything that happened in the otherwise colder 1950s, 60s, and 70s.

Finally, it's interesting to observe that two quite extreme cold spells occurred in the last 20 years when we look at a 30-day average: late February and early/mid March of 2007, and January of 2012.

I hadn't previously realized how anomalous that 2007 event was, but in fact ERA5 shows the event having the lowest 30-day average temperature since 1950, by both metrics (thickness and 850mb temp above Fairbanks).  Given that it was so late in the winter, the surface temperatures were nowhere near as low as they would have been a few weeks earlier; but nevertheless March 2007 was the second coldest on record in Fairbanks, and 27 of 31 days had a low of -10°F or lower (the highest such number for the month).

Here's a look at the 500mb height pattern for the 30-day period: what a classic!

The 500mb height anomaly:

And the 850mb temperature anomaly, according to the older (but still basically reliable) NCEP reanalysis:


  1. I ran the Iditarod in 2007. All I can say is, yup, it was. Haha!

    1. Wow! Looks like it was very chilly especially through the Alaska Range and starting into the interior. Congrats on finishing, that's an incredible achievement!

  2. 850 mb temps falling this week again. More cold coming for me and my friends.

  3. Getting to the bottom time of year for temps on the north slope. Late Feb/early March. I'd imagine the delay is complex but is associated with thickest sea ice, longest time for the arctic basin with no solar input, and overall mid atmospheric coolest time of year. Mike

    1. Yep, that sounds like the correct mix of factors that determine the seasonal cycle up there. Possibly also reduced cloud cover and water vapor (drier atmosphere) allowing more radiational cooling vs earlier in the winter.

      Would be interesting to determine where in the Arctic the seasonal minimum is most delayed. Presumably near the pole.