Tuesday, November 25, 2014

Cold This Week - Or Not

The coldest temperature observed at Fairbanks airport so far this cold season is only -6 °F, which is equivalent to the normal low temperature on November 14, i.e. more than 10 days ago.  The long-term normal for lowest temperature through November 25 is -26 °F, and only two early winter seasons (1979 and 2002) were warmer by this metric.  The lowest daily high temperature so far is a relatively balmy 7 °F, compared to a long-term normal of -10 °F for coldest day season-to-date.  There just haven't been any significantly below-normal temperatures since early October.

The computer forecast is suggesting that a change is imminent, as the latest computer MOS (model output statistics) forecast indicates low temperatures of -24 °F and -27 °F on Thursday and Friday of this week respectively.  The computer forecast also shows highs well below zero.  However, the National Weather Service forecasters apparently aren't buying it, as the coldest low temperature in the official forecast is -11 °F and the coldest high is 3 °F.  Here's the NWS forecast and the last two iterations of the MOS forecast:

The NWS forecast discussion doesn't offer any hints as to why colder conditions aren't expected.  The GFS model forecasts of the 500 mb height and the sea-level pressure on Thursday afternoon are shown below; with a surface high retreating to the east and a pressure gradient persisting, the set-up doesn't look ideal for cold, but if winds go calm under clear skies it will undoubtedly get chilly.

The current forecast discrepancy reminds me of the result I showed last month concerning the temperature bias in the winter high temperature forecasts from the NWS.  In the past three winters the NWS high temperature forecasts have been too cold on average for Fairbanks, even at shorter lead times; perhaps the forecasters have also recognized this and are now trying to correct for it.  The chart below shows the daily error of the day-1 high temperature forecasts from the past 3 winters; the day-1 forecast would be (for example) the forecast issued on Tuesday morning for the high temperature on Wednesday.

The forecast bias was particularly persistent last winter, when 87% of all days ended up warmer than the forecast indicated.  It was a warm winter overall, so it's not surprising that the bias was greater than in the previous two winters, which were slightly colder than normal.

Readers should note that none of this discussion should be construed as criticism of the NWS forecasts or personnel; their services are tremendously valuable in a myriad of ways and I applaud their efforts and dedication.

[Update November 29: the NWS forecast ended up closer to the truth than the MOS, as we would hope.  The lowest temperature observed at the airport was -18 °F and the lowest "daytime" high temperature was -1 °F.  Here are the recent day-1 high temperature forecasts and the observed maximum temperatures from 3am to 9pm AKST:

Forecast for Wednesday: +7 °F
Observed: +12 °F

Forecast for Thursday: -2 °F
Observed: -1 °F

Forecast for Friday: -2 °F
Observed: +5 °F

Remarkably we see that the day-1 high temperature forecasts were too cold again.]


  1. There is also a strong tendency to keep the long range forecast close to climatology farther out into the forecast cycle.

    1. Agreed. But MOS very much exhibits the same tendency, as it is derived from a linear regression. MOS is often a first guess for forecasters, so in this case I think it's clear that NWS is actively discounting the guidance.

  2. Perhaps some NWS meteorologist have read the post on the forecasting and realized that they really needed some correction.

    The forecast disscusion I read this morning mentioned that the models were on agreement for the next few days. So I'm suprised that the temps were ignored.

  3. Oh good! A guessing game is afoot. My old thrown turkey bones say too much SE flow for real cold and the "H's"/ridge are displaced by a "L" ~ 60 hrs out.

    Ultimately the degree cloud cover will determine the surface temps. Dem' bones never lie too much.

    But next week? These folks might use the same bones: http://www.cpc.ncep.noaa.gov/products/predictions/610day/610temp.new.gif

    Best to keep an eye on any cold over the NWT moving our way this winter.


    1. Fun, isn't it? It appears the difference this time is the expected wind speed, with NWS calling for 10mph winds tonight, decreasing to 5mph on Friday, and lowest temperature of -13F. Latest GFS MOS shows 0-1 knot winds and -27F (or -30 per the NAM MOS).

      If the computers are right we'll probably see the first -40 show up somewhere.

    2. So the idea is that more winds means weaker inversion means higher temp. Still might need to plug the car in.

      Perhaps you could look at wind speed forecast lags and how NWS handles them differently than the models - like the temp analysis. If NWS forecasts for lighter winds than projected then that would influence their temp forecast via inversion breaking.

    3. Eric, yes... more wind means more mixing of warmer air from above. It appears this is what happened, with calm not persisting for more than a few hours at a time at the airport; also we had some high clouds to provide some warming. Lowest temperature -18F and lowest tmax +2!