Cold and Snow

As is only fitting, winter is coming to an end with another spell of unusual cold in much of Alaska; here are a few notable reports from the past several days.

-44°F  Umiat RAWS
-36°F  Eagle COOP
-34°F  Chicken COOP
-34°F  Tok 70SE CRN
-32°F  Salcha RAWS
-23°F  Goldstream Creek COOP

In Fairbanks, March is ending as the fourth month in a row with average temperature below the 1981-2010 normal, and it's the coldest first quarter of the year since 2007.

March was also a very snowy month in Fairbanks, as Rick Thoman illustrated nicely in the following plot (via Rick's Twitter feed); click to enlarge.

With 27.5 inches of snow, March was the snowiest month of the winter in Fairbanks.  Since the winter of 1929-30, this has happened 8 times before, so it's unusual but not rare.  Average (median) March snowfall is less than 5 inches, but it's a very skewed distribution, with occasionally much more occurring.

April is even more skewed, of course; the median is only 1.7", and yet even April has been the snowiest month of the winter on 4 occasions - most recently in 2008 (14.7").

Here's a frequency histogram of the snowiest month of the winter in Fairbanks, Bettles, and Anchorage.  Fairbanks stands out in terms of having October as the snowiest month rather frequently; I find this particularly surprising as rain is more common than snow in the first week or so of October in Fairbanks.

Other interesting features of the chart include the notable November-December peak in Bettles, and the enhanced frequency in late winter (February through April) in Anchorage.  Surprisingly, March has more often been the snowiest month than January in Anchorage, and average March snowfall lags only very slightly behind January.  I suspect part of this is an artifact of sampling variability, but there may well be a physical reason why January tends to underperform for snow; I don't think it's as simple as "it's too cold to snow", but if readers have any ideas, I'm all ears.

Why Cold Now - Part 3

In the first two installments of this discussion (here and here), I made a few comments about Alaska's surprisingly cold winter and the difficulty of finding a good explanation for it.  Of course from one perspective the reasons are obvious: the circulation pattern produced more northerly and westerly flow (and less southerly flow) than usual over the state, leading to a much-reduced influence of mild Pacific air.

The pattern is illustrated by the 500mb height anomaly map, which shows an unusual trough centered near the southeastern interior and a strong ridge over the central North Pacific.

A very strong westerly flow is implied between the ridge and the trough, and this strong and stable jet stream pushed warm air eastward into Canada and the Lower 48 rather than taking occasional (or frequent) northward excursions into Alaska.  Here's a map of the vector wind anomaly at 250mb; the westerly flow was more than 12 m/s stronger than normal to the south of Alaska.

For reference, here's the normal upper-level wind pattern in winter:

The map below shows the resulting temperature anomaly pattern, which really highlights that the configuration was "perfect" for bringing cold to the heart of Alaska; the state experienced an "island of cold" in a sea of unusual warmth (and I'm not showing Eurasia, which was much warmer still).

Another key aspect of the flow regime was a positive phase of the Arctic Oscillation (AO), indicating that (i) low pressure was unusually low in the Arctic, (ii) the mid-latitude westerly jet stream was unusually strong, and (iii) there was a strong contrast between unusual warmth in the mid-latitudes and relatively cold conditions in the Arctic.  The positive AO anomaly became increasingly pronounced as winter progressed, and the daily AO index reached all-time record positive values in February.

The AO phase has a modest connection to Alaska temperatures in winter, as illustrated by the chart below.  Note that I'm using detrended temperatures based on data since 1950, and yes, it was the coldest winter relative to trend since 1950.  However, the trend since 1950 is very steep (over 1°F/decade), and this winter's cold was less remarkable when compared to a longer-term trend or simply a long-term average.

The chart indicates that below-trend temperatures tend to be favored when the AO phase is positive, and this provides a reasonable connection between Alaska's cold winter and the larger Hemispheric flow pattern.  However, it's interesting to see that this winter's flow pattern over Alaska was not particularly well-aligned with the typical positive AO pattern.  The map below shows the average 500mb height departure from normal in the top 10 positive AO winters of the past, and while the same strong north-south gradient is evident across the North Pacific, the AO pattern usually has a trough axis near the Chukchi and Bering Seas rather than over eastern Alaska.

The temperature pattern associated with the positive AO phase reveals a cold signal over southwestern Alaska (see below), which is certainly consistent with what happened this winter; but it seems the positive AO is not typically associated with more widespread cold across the interior and north.  As an aside, it's interesting to see a negative PDO-like temperature pattern across the North Pacific in these winters, and indeed the PDO was negative in 7 of the 10 winters (as it was this winter).

A closely related, and significant, feature of the Northern Hemisphere winter was the unusual strength of the polar vortex up in the stratosphere.  The stratospheric polar vortex forms every winter and has a fairly close connection to the lower-atmosphere AO phase, but the two don't always behave in tandem over the course of a winter.  However, this winter the polar vortex was strongly coupled, especially in late winter, and the upper-level vortex became remarkably intense by late winter - see below.

Does the presence of a strong stratospheric vortex alter or magnify the AO temperature signal over Alaska?  Not really; the map below shows winters with both a positive AO phase AND a significantly stronger than normal stratospheric vortex, and it's a very similar signal to the AO pattern by itself.

So are there any winters in the historical record with a similar strong polar vortex/AO phase and more widespread and pronounced cold in Alaska?  The answer is yes; the winter of 1989-1990 was notably cold in Alaska, and the AO became increasingly positive (indeed extremely so) as winter wore on, similar to this winter.  The previous winter, 1988-1989, is probably more memorable to Alaskans because of the incredible cold snap in January 1989, but the subsequent winter of 1989-90 was actually colder overall, and February 1990 was bitterly cold (easily the coldest February on record for the state).

Here are the 500mb and surface temperature maps for 1989-1990: the match is by no means perfect, but it's a good deal better over the North Pacific and Alaska domains than with the positive AO composites.  Notice in particular the relative absence of cold in southeastern Alaska and the Aleutians, and the warm region over the northeastern North Pacific to the west of Oregon - all similar to this winter.

The similarities between these two winters, separated by exactly 30 years, are sufficiently intriguing - especially in regard to the dramatic strengthening of the AO phase in late winter - that I'll have to dig deeper and try to understand what was going on.  And that means readers can be on the lookout for Part 4 of this discussion in due course.

CRN Data Visualization

Spare time for new blog posts has been in short supply lately, but I'll mention a little project I've been working on that may be of some interest.  As many readers know, NOAA's Climate Reference Network of top-quality climate observing instruments has been gradually adding sites in Alaska over the past several years, and the state's network is now up to 22 sites.  Here's a map (click to enlarge):

All of the data is readily available from NCEI, but options to visualize the data appear to be very limited.  I'm aiming, therefore, to put together a simple interface that provides some charting and perhaps mapping capabilities to summarize the wealth of climate monitoring data that's flowing from the CRN instruments.

Here are a few prototype charts showing departures from normal of climate variables observed at the CRN site near Fairbanks over the past several months.  The seasonal normals are calculated from the full period of record, which is already nearly 18 years long for temperature, precipitation, and solar radiation, but only 10 years for wind speed and humidity.  Click to enlarge the images.

Here's precipitation since the beginning of the year.

Look for more posts on the CRN data in (hopefully) the not-too-distant future.

Why Cold Now - Part 2

Back in late January I raised the question of why this winter produced persistent cold over Alaska; it's such a dramatic change and stark contrast to other recent winters that it begs for attempts to explain it.  After all, if we can't explain these things even in hindsight, what hope do we have (speaking personally) of predicting them?

To confirm the point that this winter was indeed remarkably cold in Alaska, the chart below shows the December-February statewide average temperature according to NOAA's climate division data.  In absolute terms, the winter was the coldest since 1998-99, and compared to the trailing 30-year average it was the coldest in 50 years (since 1969-70).  One could argue that this winter was about a once-in-a-generation cold winter relative to the modern warmer climate; and if background warming continues (as seems inevitable), it may be a very long time before it's this cold again.  Click to enlarge:

In my earlier post I discussed the apparent lack of connection of the winter pattern with tropical ocean temperature signals.  Tropical oceanic and atmospheric phenomena are often closely followed by long-range forecasters because of their slow and predictable progression, together with known (or supposed) mechanisms for influencing higher latitude weather.

In addition to sea surface temperature (SST) patterns, forecasters often track the behavior of tropical convection, i.e. regions of disturbed weather, including large clusters of thunderstorms.  One tool for such tracking is the so-called velocity potential of the upper troposphere; the velocity potential simply isolates the divergent part of the flow, which reveals where convection is relatively active (divergence aloft) or inactive (convergence aloft).

To illustrate, here's the departure from normal of this winter's velocity potential (VP) at 200mb.  Blue shading (negative VP) indicates unusual divergence aloft, implying more rain than normal (hence the locust outbreaks in East Africa), and yellow/orange indicates unusual convergence aloft (less rain than normal).  Note that VP is only useful in this sense in the tropics, because large-scale weather disturbances outside the tropics are not typically dominated by convective processes.

Now to the point of this discussion: it's worthwhile examining past cold winters in Alaska to see if there are similarities in the VP patterns.  If there are, then we might argue that the tropical convection was linked to the outcome in Alaska; and such a finding would not be controversial.

Here are the VP anomaly maps for Alaska's three coldest winters relative to trend since 1980; the lack of good satellite data prior to 1980 makes me reluctant to look at the VP analyses from earlier years.

Unfortunately there are no obvious similarities between the VP patterns in these years.  1989-90 perhaps comes closest to 2019-20, with enhanced rainfall in the Indian Ocean and suppressed convection over the Maritime Continent, but 1989-90 did not have enhanced activity over the central Pacific or over most of Africa.

Here's what a good match looks like: the winter of 2002-03 had similar tropical VP patterns to 2019-20, but it was one of the very warmest winters on record in Alaska.

To pursue a more objective approach, I calculated the similarity to 2019-20 of the VP anomalies along the equator, and none of the 3 cold years is in the top 10 matches among 40 years of data (although 1989-90 comes closest).  And just to put a nail in the coffin of this hypothesis, here is a map of winter patterns in those top 10 matches, i.e. the 10 winters with the most similar VP to 2019-20.

The pattern is about as opposite as it could be for northwestern North America; contrast the 2019-20 maps below.  So even if we had known the VP patterns perfectly in advance, an analog approach like this would have predicted another warm winter for Alaska.

In summary, this cursory analysis provides absolutely no evidence that Alaska's cold winter was directly linked to patterns of tropical rainfall activity, and this is a disappointment.  It's not entirely surprising, though, because the SST patterns provided no help either, and there are very close links between SST anomalies and tropical convective regimes.

In Part 3 of this analysis, I'll look at another aspect of the winter that became increasingly unusual as the winter progressed, and that's the strongly positive Arctic Oscillation and increasingly intense stratospheric polar vortex.  In this case there is definitely some demonstrable correlation to Alaska's winter weather, but I think we'll find that it still leaves many questions unresolved.

Winter Hangs On

Unusual cold is hanging on with remarkable tenacity in Alaska this winter, and last night saw temperatures worthy of the depths of winter in many locations.  Here are a few examples of notable low temperatures today:

-50°F  Bettles
-45°F  Tanana
-44°F  Chalkyitsik RAWS
-42°F  Nenana

And in and around Fairbanks:
-46°F  Salcha RAWS
-41°F  North Pole
-39°F  Goldstream Creek
-39°F  Smith Lake at UAF
-38°F  Fairbanks airport

Here's the temperature trace from the past two weeks at Smith Lake (UAF's North Campus).  The temperature did not break -10°F today, which is pretty impressive for the time of year.

Of course clear skies at this time of year produce large diurnal temperature ranges, as seen in the Smith Lake data.  The Chalkyitsik RAWS (about 20 miles east of Fort Yukon) often produces spectacular examples, and just the other day this site saw temperatures vary from a low of -49°F on two consecutive nights to a high of 0°F in the intervening day.  When looking at days with a high temperature of 0°F or below, this is the largest diurnal range on record for the site (data since 1997).

The chart below shows daily Chalkyitsik temperatures so far this winter, as compared to seasonal normals (obtained from the 1997-2018 history).

Two features are striking: the persistence of unusual cold since mid-December, and the expansion of the daily temperature range since early February.  The latter is entirely in keeping with climatology, which shows a pronounced rise in the average diurnal range in late winter; in fact it's remarkable to see the normal high temperature bottom out very soon after winter solstice, while the normal low temperature doesn't show any meaningful rise until well into February.

Solar Power in Alaska

Last week an article about solar farms in Alaska caught my eye:

https://www.bbc.com/future/article/20200219-the-solar-farms-fighting-climate-change-in-alaska

The discussion indicates that despite the obvious shortcomings of solar power generation in a place that receives so little sunshine during the season of peak energy demand, the cost of solar panels has come down enough to make it worthwhile to install solar farms anyway.

The BBC article also states that, "perhaps surprisingly, Alaska is a sunny place", and cites a 2015 piece by Brian Brettschneider to back up this claim.  However, the BBC author unfortunately made an unwarranted jump from daylight to sunshine; while it's true that year-round daylight totals are actually greater in Alaska than points farther south, the total amount of solar energy is nowhere near as great.

https://www.adn.com/science/article/sunniest-day-year-look-why-alaska-has-most-daylight/2015/06/20/

This distinction raises an interesting question, however: do the long daylight hours and relatively clear skies of early summer produce enough solar energy to be competitive with locations in the Lower 48?  To take a quick look at this, I used data from the CRN sites that have been running since 2002 near Fairbanks and since 2010 on the Kenai Peninsula, and I compared total solar energy in May through July to 3 sites in the central and eastern U.S.

First, here's the average rate of solar energy input available at the surface, based on the full period of record at each site (15-17 years except for 9 years at Kenai).  The two Alaska sites receive less energy despite having much longer daylight hours, but the difference is not huge, and this reflects the point of the BBC article - that there's plenty of solar energy to be harnessed in Alaska for part of the year.  Click to enlarge:

The shortfall in total energy despite longer daylight hours is a function of both cloudiness and solar elevation angle.  The chart below illustrates the cloudiness aspect by showing how much of the clear-sky maximum is received at each location.  On the Kenai Peninsula, most summer days are fairly cloudy, and even in the relatively sunny interior, 6 out of 10 days have enough cloud to keep solar radiation at less than 70% of maximum.  In contrast, 6 out of 10 days have more than 70% of maximum in my neck of the woods (Georgia).

So while solar power does have fairly good seasonal potential in Alaska, even the long daylight hours of summer are not enough to make it fully competitive with the rest of the U.S.

Yukon Frozen at Dawson

Long-time readers will recall that in the past few winters I've drawn attention to the interesting failure of the Yukon River to freeze over properly at Dawson in the Yukon Territory.  For the past three years the Yukon government has been unable to construct an ice bridge across the river to West Dawson, and local residents have resorted to alternative routes to make the crossing.  Here are some posts from previous years:

https://ak-wx.blogspot.com/2017/02/yukon-river-at-dawson.html

https://ak-wx.blogspot.com/2018/02/yukon-river-at-dawson.html

https://ak-wx.blogspot.com/2019/02/dawson-ice-bridge-problems.html

Happily, this winter is a return to normal, as the government-sanctioned ice bridge was open by Christmas and was available for heavy traffic a month ago.

https://www.cbc.ca/news/canada/north/dawson-city-ice-bridge-opens-2019-1.5407399

https://www.whitehorsestar.com/News/ice-bridge-enters-wider-thicker-phase

One could be forgiven for thinking that colder weather this winter is the reason for the more normal freeze-up, but in fact it wasn't particularly cold at all during the freeze-up period.  The average temperature in November and December was -2°F, compared to -5°F and -7°F in 2016 and 2017, respectively; the 1995-2015 average was -5.1°F for these months.  Nor was there a pronounced cold spell; only 6 days before the turn of the year dropped to -30°F, compared to a 1995-2015 median of 10 such days.

The accumulation of freezing degree days shows the same thing: this winter was apparently no more favorable for freeze-up in terms of thermal conditions, although the January cold provided a good boost after freeze-up.  Click to enlarge the chart below.

This simply confirms what we noted in prior years: unusual warmth did not explain the persistent lack of freeze-up in the last 3 years, and unusual cold can't explain the river's return to normal behavior this winter.

As noted before, there are many possibilities for potential causes of the abnormality in recent years, but one new clue comes from a look at warm-season precipitation in the upstream Yukon drainage.  The chart below shows May-September precipitation at four sites for the past 5 years, and clearly last summer was drier than any of the preceding 4 years.

The 1981-2010 normal for Mayo and Dawson is about 195mm for May-September precipitation, so last summer's deficit was fairly substantial.  (Whitehorse was close to its normal of 155mm, and I don't have a normal value for Carmacks.)  This is a tiny sample across a very large area, of course, but the river level data from Dawson support the idea of reduced flow late last year.  Here are the September mean level values from the past few years (data obtained here):

2015   2.83m
2016   2.02m
2017   1.56m
2018   1.85m
2019   1.03m

To my mind it makes sense that lower flows would freeze over more easily, but I don't think this is all there is to it; I believe there was an ice bridge in 2015-16, but that year apparently had a very high flow rate going into autumn.  At any rate, it's nice to see something more normal this winter.  Here's the webcam view from a couple of weeks ago.

Alaska Climate Divisions

Last week a UAF news article highlighted the value of the Alaska climate division analysis that was developed a few years ago by Peter Bieniek and others from UAF and other universities, along with NOAA collaborators such as Rick Thoman.  NOAA has long used so-called climate divisions in the lower 48 to keep track of climate variations in climatically similar regions, but nothing comparable was available for Alaska until this work by Peter et al.

https://news.uaf.edu/taking-a-deep-dive-into-alaskas-record-breaking-warm-year/

The journal article describing the new climate division work was published way back in 2012, but as the article explains, it took a few years for NOAA to adopt the divisions for "official" monitoring.

https://journals.ametsoc.org/doi/full/10.1175/JAMC-D-11-0168.1

I'm a big fan of the Alaska climate divisions, but one of the potential shortcomings is the relative scarcity of ground-truth station data; only 42 sites (including some Canadian) were used to determine 13 climatically similar regions, and some divisions had far more sites than others.  The Northeast Interior division, for example, contains only one station (Fort Yukon), and the North Slope division has only one non-coastal site (Umiat).  Such is the world of historical Alaska climate analysis.

For reference, here are the Alaska climate divisions:

After reading the UAF news piece, I started wondering if modern reanalysis data would produce similar climate divisions to the Bieniek results.  To address this, I used monthly mean temperature data from the ERA5-Land reanalysis, now available from 1981 through most of 2019.  ERA5-Land is a higher-resolution version of ERA5 (9km vs 31km grid spacing) that models only surface variables such as 2m temperature, 10m wind, humidity, snow cover, and so on; it does not deal with oceans or the atmosphere aloft.  I'm hopeful that ERA5-Land may be an improvement over ERA5 for Alaska in winter (see this post from a few weeks ago), although I haven't done any investigation on this yet.

Regardless of the possible deficiencies of ERA5-Land, it's interesting to see what the climate division analysis produces.  I ran cluster analysis on the gridded monthly mean temperature anomalies (standardized) from 1981-2018, and the following maps show the results, ranging from 3 to 10 clusters, based on two alternative methods.  Bieniek et al tested these two methods and a third, but they focused on results from Ward's method (right column below).

K-means method	Ward's method

There are a number of interesting aspects to the results.  First, the K-means cluster boundaries tend to jump around somewhat, because the method starts with a random choice each time and iterates to a solution.  For this reason it is also not 100% reproducible, i.e. you can get different results when you run it again.  In contrast, a hierarchical method like Ward's is reproducible, and the boundaries don't move around as the clusters are progressively sub-divided.

Despite the differences in the results, certain features are similar: the North Slope division emerges quickly and remains very well-defined throughout; a Panhandle division emerges at k=6 for both methods; and the clusters are really quite similar for k=5,6,7, and 9.

Perhaps most interesting, in my view, is the absence of some of the distinctions that are found in the Bieniek results.  For example, even if we go all the way up to 15 clusters (see below), there is no sub-division within the Panhandle, whereas Bieniek has three Panhandle divisions and another for the Northeast Gulf.  Similarly, the ERA5-Land clusters give no separation between Aleutians and Northwest Gulf (e.g. Kodiak Island).  As the number of clusters increases, the sub-dividing mostly takes place in the interior and eventually on the North Slope.


On the other hand, the ERA5-Land clusters quickly break apart the West Coast region, rather than keeping it together as Bieniek does.

I mention these differences out of curiosity, not to suggest that the Bieniek divisions are wrong.  It's very likely that ERA5-Land has certain deficiencies that would hamper the assessment of climate similarity - for instance, the reanalysis may be wholly inadequate in the very complex terrain of the Panhandle.  More investigation would be needed to see how well ERA5-Land reproduces climate in the vicinity of the stations used by Bieniek et al.

Lastly, it's not clear to me whether there is an optimal number of clusters based on the ERA5-Land analysis.  Traditionally one looks at the distribution of within-cluster variance and seeks to find a threshold beyond which (i.e. for smaller numbers of clusters) the variance starts to increase more quickly; but the results from ERA5-Land show no obvious stopping point.  Bieniek also found that using gridded data made it impossible to tell where to stop.

Personally I like the look of the K-means solution with 9 divisions, but it's purely a personal preference.  I'd be glad to hear any comments from readers.