Tuesday, October 4, 2016

Gradually Cooling Off

The coolest temperatures of the season so far were observed in the interior last night, with several locations dipping into the single digits - as befits the time of year.  The Chalkyitsik RAWS dropped to +4°F, which is the lowest temperature observed so far in Alaska this season as far as I'm aware.  Other lows included:

7°F   Beaver Creek RAWS
8°F   Circle Hot Springs COOP
9°F   Eagle airport and RAWS
9°F   Chicken COOP

At Fairbanks airport the first indisputably hard freeze occurred the night before, with a low of 22°F.  Yesterday was also the first day with a daily mean temperature of 32°F.  If these conditions were to continue, it wouldn't be long before some ice would be running in the rivers, but the forecast shows a warming trend.

[Update October 5: this morning was colder, with a low of +1°F at the Chalkyitsik RAWS and +4°F at Chicken. Closer to Fairbanks, the Goldstream Creek COOP had 11°F, and it was 9°F at the Salcha RAWS.]

Speaking of the forecast, I couldn't help noticing a major discrepancy today between the GFS MOS (automated computer) forecast and the rest of the guidance for Fairbanks.  Here's this morning's MOS bulletin, with the "N/X" line showing low temperatures well down in the teens from Thursday through Monday, and with high temperatures staying near or below freezing for several days despite clear skies.

A quick look at this evening's NWS forecast for the airport shows a dramatically different picture, with high temperatures above 40°F and lows not far below freezing.

The GFS model forecast itself (from which the MOS output is derived) shows increasing warmth aloft in the next week, with 850mb temperatures rising well above freezing by Friday, and with very low relative humidity suggesting clear skies.

The raw model 2m temperature forecast shows conditions similar to the NWS forecast and drastically different from MOS.

So what's going on with MOS?  Frankly I have no idea, but it's clearly wrong - and this is interesting, because MOS usually serves as a baseline or "first guess" for forecast systems; it is generally much better than raw model output, because it's a statistical regression based on historical observations.  I briefly wondered if perhaps MOS is assuming there is snow on the ground already, but this is most unlikely as the historical probability of having snowcover in Fairbanks at this date is less than 20%.

To show how unlikely it is that Fairbanks will see a daily high temperature below freezing any time soon, we can look at how cold the upper-air temperatures need to be to produce a sub-freezing day.  The chart below shows the afternoon 850mb temperatures on all such days from September 20 through November 19, beginning in 1957; we can see that 850mb temperatures are always well below freezing when daytime conditions remain below freezing in September or early October.  As the autumn advances and the sun sinks, nights lengthen, and snow cover becomes more likely, it becomes possible to have sub-freezing days when 850mb temperatures are well above freezing, but early October is too early for that to happen.


  1. This is sometimes the time of year for this stuff which can warm and wet things up:


    Source: http://forecast.weather.gov/product.php?site=CRH&product=PMD&issuedby=AK

    I wonder how frequently and maybe to what extent such storms affect our weather this time of year? That may have been previously addressed, if so please link.


    1. Gary, we had some discussion about the November 2014 event:


      I haven't looked at the frequency of this kind of thing, but it would be a most interesting study.

    2. Good memory link, thanks Richard. Like a goose, every day is a new one for me.


  2. Remember the typhoon remnants in November 2014? It built a very strong and warm ridge over eastern Canada and Alaska. It was very warm aloft (10C or more at 850mb if I remember correctly. It's also brought very dry weather. Looks like next week is the perfect setup for pollution inversions in North Pole. Light winds and clear skies. (Except Tuesday)

    1. I meant western Canada and eastern Alaska. 😅

    2. Mike it would appear to be what we'll again experience in the coming days as the ex-Typhoon collaborates with increased heights/ridging to warm us up.

      The Northern Lights have been spectacular under clear nighttime skies but ham radio propagation is almost dead as a consequence.


  3. From Richard in the text above:

    "I briefly wondered if perhaps MOS is assuming there is snow on the ground already, but this is most unlikely as the historical probability of having snowcover in Fairbanks at this date is less than 20%."

    From today's NWS Northern Alaska Area Forecast Discussion:

    "Model guidance continues to be too cool with the high temperatures, and expect that to continue until we have snow on the ground."

    Why would a particular model assume what's obviously not the case? Why can't it adapt or be improved?


    1. Gary, thanks for pointing to that statement in the NWS discussion. I'm sure the forecaster is referring to MOS ("model guidance"), which evidently does NOT use observed snow cover (or lack of it) as a predictor. Here's a description of MOS:


      The underlying fluid-dynamical models (GFS, NAM etc) certainly do include some estimate or prediction of actual snow cover, and their raw temperature forecasts will include that influence, but the MOS forecasts are only (it would seem) affected by upper-air variables (850mb temperature, humidity, etc). So there's no way for the model snow cover effects to make their way into the MOS forecasts. But I'll do some digging on this to see if I can find authoritative answers.