La Niña Ending Soon

La Niña has made its presence known in Alaska this winter, and winter certainly isn't over yet, but major changes are afoot in the tropical Pacific Ocean.  Here's the canonical Niño3.4 SST index, which is negative (cooler than normal) during La Niña and positive in El Niño.

A region of warmer than normal SSTs is expanding quickly westward from the coast of South America, leaving equatorial cool anomalies looking anemic in the central Pacific:

Compare with the situation just six weeks earlier:

The long-range computer models expect warming to continue through spring and into summer, resulting in El Niño by summer.  This month's forecasts are more aggressive than previously, and judging from the current trend, they may not be aggressive enough.

If El Niño does develop - as seems very likely - this will be a risk factor for increased wildfire activity in Alaska this summer.  However, the statistical connection hasn't always played out in recent years; 2023 was an analogous year with El Niño developing, but Alaska's fire acreage was very low (in contrast to Canada).

In the meantime, La Niña won't relinquish her influence on high-latitude patterns immediately, and the latest forecasts suggest a strong possibility of another La Niña-like persistent statewide cold spell starting the middle of next week.  The cold plunge next week seems to be locked in: here are two leading ensemble models for next Thursday morning.

Looking farther ahead, the details are less clear on the evolution into March, but there are strong hints of a persistent ridge near the Date Line and a trough over eastern Alaska and northwestern Canada - very similar to the persistent pattern in December.

Below is an example: compare the forecast portion (left side of the graphic below the dashed line) to this December 18 post.  The blue shading near the longitude of Alaska indicates the trough that will produce the cold northerly flow; this model shows the setup lingering into mid-March.  (Note that the previous post includes an explanation on interpreting these figures.)

January Climate Data

Looking back at Alaska climate data for January, the month was a story of two dramatically different halves, with the prolonged mid-winter cold snap ending (on a statewide basis) right at the month's midpoint.  It's worth recalling again just how persistent the cold was for that period of six weeks ending January 15; here's the UAF statewide temperature index since November 1st:

The contrast in the mid-atmosphere pressure pattern could hardly be greater between the first and second halves of the month.  The first half of January saw a strong trough planted right over the state, but this reversed to a high pressure ridge for the second half:

Owing to the dramatic turn-around, the monthly average temperature wasn't all that unusual, only 3.5°F below normal, and even the coldest climate division (Bristol Bay) was considerably warmer than January 2020 (and even more so than January 2012).  Southeast Alaska was warmer than normal, as unusual cold only lasted through the first few days of the month there.

Amazingly, the December 31st record 50-inch snowpack in Juneau melted out completely by the middle of January, with 30 inches disappearing in a rainy second week of the month.  All of Southeast Alaska and the eastern half of the state at large was considerably wetter than normal, while the west coast and western interior were dry, along with most of the Alaska Peninsula.

The dryness across the Seward Peninsula and interior northwest compounded a lack of snow from earlier in the winter (see figures below), and snowpack was far below normal by February 1 for places like Nome, Kotzebue, and Ambler - at least according to ERA5 data.

The ERA5 temperature map above shows that unusual cold was widespread over the eastern half of the Bering Sea, and so it's no surprise that sea ice expanded more quickly than normal, nearly reaching St Paul Island by the end of the month.

Sea ice did in fact reach St Paul Island a few days later, according to the NWS analysis:

Nevertheless, on a monthly basis, January Bering Sea ice extent was about 12% below the 1991-2020 normal, because of deficits in the western half of the basin.  The basin-wide total was in line with the past several years, and seems to have stabilized (for now) well above the 2015-2018 lows.

Alaska Landslide Inventory

Back in November a study was published that looked at historical news reports to create a new inventory of landslides in Alaska dating back to 1883.  Many other inventories have been produced over the years for specific purposes, but this approach focuses on the human exposure to landslides.

https://link.springer.com/article/10.1007/s10346-025-02663-z

As one would expect from increased population and better reporting, the number of reported landslides has increased tremendously over time, although the number of landslide-caused fatalities has not increased (largely because of the 1936 Juneau landslide).   The authors argue that a major part of the increase in landslide numbers is related to warming of the climate, which is hypothesized to cause a higher frequency of freeze-thaw events, rain-on-snow events, and rainfall extremes.

Extreme rainfall events in mid-high latitude areas are often produced by "atmospheric rivers", and there's a lot of interest among climate scientists in historical and future trends in these events.  To cite one example, the following study indicates that mid-latitude atmospheric rivers have become more frequent in Northern Hemisphere winter (as defined by moisture transport in the ERA5 reanalysis data).

https://www.nature.com/articles/s41612-025-01191-w

The following figure from the paper (click to enlarge) shows a region of increased frequency in the Gulf of Alaska, adjacent to landslide-prone Southeast Alaska.

Extreme precipitation events have increased widely across southern and southeastern Alaska, according to this data:

Just for fun, I pulled out the ERA5 precipitation amounts for each landslide identified by Darrow and Jacobs as having been triggered by excessive rain.  The chart below shows the rain excess above normal for the 7 days ending on the landslide date, at the ERA5 grid cell closest to the landslide location.

Of course, ERA5 is thoroughly incapable of reproducing local rainfall variations in complex terrain, which is no doubt a critical factor in many cases, so this analysis is very crude.  Nevertheless, it's mildly interesting that the fraction of events with a rain excess above +4 inches more than doubled from pre-1990 (10%) to post-1990 (22%).

One other comment - the increase in atmospheric river events near southern Alaska in recent decades is related to a greater frequency and persistence of La Niña-like (and negative PDO-like) anomalies in the Pacific, with increased warming in the tropical West Pacific:

https://www.science.org/doi/epdf/10.1126/sciadv.adq0604

Cold and Snow Correlation

Last week I alluded to the linkage between the Pacific Decadal Oscillation and seasonal snowfall in Anchorage, with a negative PDO phase often producing increased snow relative to normal.  Of course the PDO also influences seasonal temperature variations - a negative PDO typically brings colder conditions to Alaska, especially in the south.

These twin correlations imply that seasonal mean temperature and snowfall are also correlated to each other in Anchorage: increased snowfall tends to accompany cold.  Here's a chart illustrating this relationship in the last half-century.

I thought it would be interesting to look at the snow/temperature relationship elsewhere around the state.  I started by examining ERA5 reanalysis data, but it's not particularly helpful: it actually suggests a positive correlation in the Anchorage area, presumably because the model's resolution is inadequate (i.e. the results reflect the influence of conditions at higher elevation).

However, one region where the model likely gives a good picture of the low-elevation snow/temperature relationship is in Arctic Alaska.  In the far north, warmer conditions are favorable for increased winter snow, simply because of increased availability of moisture.  This is supported by data from Utqiaġvik up until measurements ceased in 2019: warmer winters tend to be more snowy.

However, historical data from Kotzebue and Nome do not have a positive correlation, despite what the model says.

How about the interior?  The relationships are weak, with little correlation, although there's a suggestion that the warmest winters tend not to be very snowy in McGrath and Fairbanks (especially the latter).  Particularly for Fairbanks, this presumably reflects the drier weather that occurs when the pattern favors warm downsloping/chinook winds from the south and southeast.

Data from Bethel suggests a hint of a negative correlation like Anchorage: more often than not, colder-than-average winters are also snowy.

Turning to Southeast Alaska, Juneau shows a robust negative correlation, as we would expect for a much warmer climate where unusually warm winters tend to be more rainy than snowy.

These charts probably give a good sense of how seasonal snowfall is likely to trend over time if Alaska's climate continues to warm in future decades.  Broadly speaking, we would expect winter snow accumulation to diminish in the low-elevation south and to increase in the Arctic, while the interior may remain relatively unchanged.  But of course large annual and decadal variability will continue, and there may be long-term circulation changes (e.g. related to the PDO) that have significant impacts on long-term trends.

I'll be happy to add charts for other locations upon request (if the data is adequate).

More Anchorage Snow

Anchorage has seen another big snow event this week, occurring three weeks after the record (for January) snowstorm earlier in the month.  The new snow accumulation of 11 inches takes the January total to just over 40 inches, which is a record for the monthly total.  It's also the fifth snowiest calendar month in the modern Anchorage climate history (1953-present).

Here's a visualization of calendar month snow totals in Anchorage, with the top 10 events for the main snow months marked in red.  February has two huge outliers - 1996 and 1955 - but the most favored month for big totals is clearly December.

The climatological drop-off in January is actually rather significant; December reaches 20" over a third of the time (26 of 73 years), while January sees 20" in less than 15% of years (10 of 73, including 2026).  So this month's big snow total is a notable outlier for the time of year.

Another interesting tidbit is the preponderance of high monthly totals in recent years: 2023 saw the snowiest November, 2022 and 2023 had the 2nd and 3rd snowiest Decembers, and now we have the snowiest January.  February 2023 gets an honorable mention with the 4th snowiest February (32").  These are good times for snow lovers in Anchorage.

The statistics are at least as striking when we look at liquid equivalent precipitation in Anchorage, i.e. it's been wet in recent winters.

2022 - 2nd wettest February

2022 - wettest December

2023 - wettest November

2025 - 3rd wettest January

2026 - wettest January

What changed in 2022?  The PDO (Pacific Decadal Oscillation), for one thing:

See this 2022 post for some comments on the correlation between the PDO and Anchorage snowfall:

https://ak-wx.blogspot.com/2022/12/anchorage-snow.html

North Slope Accident

Many readers will have heard that an oil field accident occurred on the North Slope on Friday, as a very large mobile oil drilling rig tipped over and was destroyed.  According to the following article, "Several people with North Slope experience have said the rig likely tipped due to being moved in warm 30 degree weather, which would have caused the road to be softer".

https://alaskalandmine.com/landmines/doyon-26-largest-land-drilling-rig-in-north-america-tipped-over-and-caught-fire-on-north-slope/

Here's a temperature chart for nearby Deadhorse, showing the temperature rising to 31°F on Friday.

Is it reasonable to believe that the road (presumably an ice road build on top of the tundra) became soft even though the temperature did not rise above freezing?

Yes - the properties of ice change significantly with temperature, and liquid water can be found in tundra soils at temperatures well below freezing.  The closer to the melting point, the softer the ice, and the greater the proportion of liquid.

Here's Google's Gemini (AI) response to a query about this situation.  We could quibble about a couple of details in the response, but it nicely lays out the issue that seems to explain the accident.

---------------------

In the context of the January 2026 incident on Alaska's North Slope, the "softening" of the road is a result of the unique mechanical properties of frozen water and permafrost near their phase-change threshold.

1. The "Warm Ice" Effect

Ice is not a static material; its strength is highly dependent on its temperature. As ice approaches its melting point (32°F / 0°C), it undergoes a process called pre-melting.

Reduced Compressive Strength: Engineering data shows that the compressive strength of ice at 30°F is significantly lower than at 0°F. At colder temperatures, ice is brittle and rigid. Near 30°F, it becomes more "plastic" or ductile, meaning it deforms under heavy loads rather than holding its shape.

Liquid Water Films: At 30°F, microscopic films of liquid water begin to form between ice crystals (intergranular boundaries). This acts as a lubricant, allowing the ice grains to slide past each other more easily, which reduces the overall load-bearing capacity of an ice road or frozen tundra.

2. Permafrost and Unfrozen Water Content

The soil in the Arctic (permafrost) is not just frozen dirt; it is a matrix of minerals, organics, and ice.

Unfrozen Water Content (UWC): Even below 32°F, a portion of the water in the soil remains liquid due to salts, pressure, and surface tension within soil pores. As the temperature rises from 0°F toward 30°F, the percentage of liquid water in the soil increases exponentially.

Loss of Shear Strength: Research from the US Army Cold Regions Research and Engineering Laboratory (CRREL) indicates that the shear strength of frozen soil drops dramatically as it warms toward the freezing point. At 30°F, the soil may still look frozen, but its ability to support the massive, concentrated weight of a mobile drilling rig (which can weigh several million pounds) is severely compromised.

3. Impact on Ice Road Infrastructure

On the North Slope, "roads" used to move rigs are often built by packing snow and freezing water over the tundra.

Creep Deformity: Under the immense weight of a drilling rig, ice and frozen soil undergo "creep"—a slow, permanent deformation. The rate of creep increases significantly as temperatures rise toward 30°F.

The "Punch-Through" Risk: If a rig is moved during a "warm" spell, the road may suffer from localized failure. If one side of the road yields or "squishes" even a few inches more than the other due to uneven softening, the high center of gravity on a tall drilling rig can cause it to tip past its point of recovery.

Evidence in Industry Practice

The Alaska Department of Natural Resources (DNR) and North Slope operators (like ConocoPhillips or Hilcorp) monitor soil temperatures strictly.

Standard Operating Procedures: Most North Slope permits require a minimum "frost depth" and a maximum surface temperature before heavy equipment can move.

Historical Precedent: The "January Thaw" or "Warm Spells" are well-known hazards in Arctic logistics. Moving a rig at 30°F is considered high-risk because the margin of safety provided by the "structural rigidity" of deep-frozen ground is largely lost.

Conclusion: Even though 30°F is below freezing, it represents the "danger zone" for Arctic engineering. The combination of increased unfrozen water content in the soil and the increased plasticity of ice makes the ground "soft" enough to fail under the extreme pressures exerted by heavy industrial machinery.

Cold Revisited

Now that the weather has calmed down, it's a good time to look back at a couple of aspects of the cold spell in Fairbanks.

First, I've been meaning to compare the cold to previous episodes on various time scales.  From a historical standpoint, the most significant aspect of the cold snap was arguably its long duration, rather than its extreme intensity at any given time.  The widely-cited headline statistic is certainly the most notable: the temperature at the airport stayed below 0°F for 32 consecutive days, the most in over a century.  The record is 41 such days in 1917-18.

In terms of individual days, the daily maximum temperature of -46°F at the airport on January 4th was the lowest since 1977; even the cold snaps of 1989 and 1999-2000 didn't have a day that cold.

A lot more statistics could be compiled, but Rick Thoman saved me a lot of work with this summary (click to enlarge);

A curious aspect of the cold spell was that the international airport was colder than any other reporting site in the Fairbanks area except for North Pole.  This is unusual: the two longstanding co-op sites in the Goldstream Valley ("Ester 5NE" and "Goldstream Creek") are typically significantly colder than the airport.  The discrepancy was large enough that a question has been nagging at me: is it possible that the airport thermometer has been running too cold, so that the cold snap wasn't actually quite as severe as reported?

To illustrate the issue, here's a look at the joint distribution of daily mean temperatures from Fairbanks airport and Goldstream Creek since 2012, for the date window from December 1 through January 15:

Based on data from 2012-2024, Goldstream Creek is typically colder than the airport on more than 75% of days in December and January, but the chart shows that most of the coldest days in the recent cold spell were colder at the airport.  The airport's cold "advantage" emerged in mid-December as the cold snap deepened:

I decided to dig into this further by calculating the 2012-2024 December average temperature difference between the airport and several other valley-level sites around the area, and then examining how these differences vary from year to year.  The idea is to see if last month's cold anomaly at the airport is outside the bounds of usual variability.

The following figure shows the result of my calculations.  Points above the horizontal zero line indicate Decembers that were relatively warmer at the airport compared to the usual difference for the site indicated; for example, in December 2021 the airport was warmer than would normally be expected based on any of the other sites (yet that month wasn't particularly warm overall).  And in 2025, the airport was colder than expected.

The result confirms the unusual situation last month, but it also reveals that there's quite a lot of variability from year to year (e.g. 2021), so the fact that the airport was colder doesn't necessarily indicate a problem with the thermometer.

Interestingly, the temperature differences between the airport and Goldstream Creek show the least variability from year to year between 2012 and 2024.  In other words, Goldstream Creek has the most similar temperature variations to the airport, and the most consistent temperature differences.  However, last month the difference was way off, and this is what drew my attention to the topic.

Honing in on Goldstream Creek, then, the same calculation with temperature differences relative to other sites shows that December 2025 stands out as an apparently significant outlier - more so than using the airport as the reference point.

My conclusion is that if something is awry with the temperature measurements, then it's more likely at the Goldstream Creek co-op than at the airport; Goldstream may have been running too warm.  Consider that since 2012, the Goldstream Creek has never had fewer days reaching -40° than the airport in a winter; but this winter the airport has 21 such days, while Goldstream has only 13.

Of course, it's also possible that the Goldstream valley was actually less cold than would normally be expected for legitimate meteorological reasons.  Perhaps drainage flows from the hills into that valley kept the winds up a little bit, and this would explain why the nearby Ester 5NE site was also warmer than would be expected.