Snowiest Place - CFS Reanalysis

Continuing with the discussion of normal snowfall patterns across interior Alaska, I thought it would be worthwhile to look at the CFS (Climate Forecast System) Reanalysis to see what kind of spatial variability of winter precipitation is represented in the model.  The CFS Reanalysis (CFSR) is NOAA's modern gridded historical analysis of atmospheric and oceanic conditions back to 1979; here's a paper describing the project.

The advantage of this data set for our purpose here is that the model - being run at fairly high resolution (about 38km grid spacing) - should be able to capture some localized features of the winter precipitation patterns across Alaska; it can "fill in the gaps" in the sparse observing network.  However, it's critical to bear in mind that the CFSR precipitation analysis is basically just a very-short-range model forecast and is not tied to observed precipitation, so it will have all the biases and errors that models are normally subject to.

Moving immediately to the results, the figure below shows the 1981-2010 mean November-March precipitation according to the CFSR over most of the state.  I apologize for the lurid color scheme, but it's designed to make it easy to pick out differences and specific values.  At first glance the precipitation patterns look very reasonable; it's good to see the dry zones along the Tanana River valley and immediately northwest of the Denali area.  However, the precipitation amounts are much too high in western Alaska; Bethel and Nome only see about 5" of precipitation in November-March, whereas the model says there is 10" or more.

Zooming in on the interior, and using a different color scale, we see numerous areas of enhanced precipitation in regions of elevated terrain; the black contours indicate the 500m elevation contour.  This fully supports reader Andy's comment that topography produces dramatic local variations in precipitation.  Andy mentioned the snowfall gradient from Takotna to Nikolai, and this is very clearly depicted in the model results; the marker labeled "2" shows the location of McGrath, which lies in the middle of a strong east-west precipitation gradient.  I've also added markers for several other locations, including those mentioned in the previous post.

The CFSR clearly shows the precipitation enhancement in the Nulato Hills to the west of Kaltag, and the model believes precipitation is even higher farther to the south, but as noted in Gary's comment, rain occurs more often as you go farther south.

From the point of view of "snowiest location", perhaps the most intriguing aspect of the CFSR data is the high precipitation amounts in the Kuskokwim mountains, and especially a bullseye of high precipitation at the northeast end of the range, about half way between Tanana and Minchumina.  It's interesting that reader Gary noted a preponderance of poor weather in this area north of Lake Minchumina, and I wonder if the region is particularly favored for upslope generation of snowfall during both westerly and easterly flow regimes.  According to the CFSR data, this might well be one of the snowiest places in interior Alaska, but unfortunately I'm not aware of any ground-level data to verify the claim.

The figure below compares observed 1981-2010 normal November-March precipitation to the CFSR data for 8 different locations.  Disregarding the pronounced wet bias, it's clear that the CFSR does have an ability to at least broadly discern relative differences in precipitation, and this gives a modest degree of confidence that the spatial patterns shown above probably do bear some resemblance to reality.

Deep Snow Above Fairbanks

I'll have more on the topic of "snowiest place in the interior" soon - and thanks to readers for comments - but today I just wanted to point out the remarkable accumulation of snow that's been reported from Keystone Ridge in recent weeks.

As of Wednesday evening, the cooperative observer on Keystone Ridge reported 32" of snow on the ground, which is greater than a normal winter's peak snow depth of 28".  Only 5 previous winters (in 1996-present) produced a snow depth greater than 32", and the all-time record is 48".

It's worth pointing out that the Keystone Ridge coop station changed hands this summer; the new location is about 3/4 mile west of the old location, but the elevation is nearly identical and both locations are close to the ridgeline, so it seems unlikely that there would be significant systematic differences in snowfall.  The November to-date liquid equivalent precipitation is 2.85", which is the highest on record for the month of November and therefore consistent with the record snow depth.

By way of explanation, I would point to the anomalous west-southwesterly flow that has affected interior Alaska this month.  The normal 850mb flow direction is southeasterly at this time of year (first figure below), but this year there has been a pronounced westerly anomaly bringing moist air inland (second figure below).

What's the Snowiest Place in the Interior?

November has been a very snowy month so far in Fairbanks, with 21.4" through midnight last night and measurable snowfall on 19 of 25 days including today.  Compared to the 1930-present history, these numbers are the 9th and 3rd highest respectively for the full month of November, with several days still to be counted before the month ends.

In terms of year-to-date snowfall, Fairbanks is at 46.1", which already exceeds last winter's total snowfall of 43.8".  It's only the 5th time that 46" has been reached by this date.  The earliest this snow accumulation was reached was 1992 (November 12), following the record September snows.  Approximately one quarter of winters in Fairbanks don't reach 46" of snow in the entire winter.

All this snow led me to wonder what the snowiest location is in interior Alaska.  Some time ago Rick Thoman indicated on this blog that Kaltag has a normal annual snowfall of around 120", and after looking at the NCDC normals and GHCN data, I'm unable to find any location that exceeds this (excluding sites at elevation in the Alaska Range).  In the official 1981-2010 normals, which are rather sparse because of the stringent requirements for data completeness, the highest mean annual snowfall in the interior is a (to me) surprisingly high 97" in McGrath, with Bettles (91") coming in second.  Fairbanks comes in at 65" (mean, not median).

Looking at all GHCN data for Alaska, the highest mean annual snowfall for a valley-level interior station with at least 10 years of at least some data is at Ambler, with 113" on average from 1981-82 to 1991-92.  This includes a remarkably high 197" in the winter of 1988-89, of which 66" purportedly fell in April 1989.  At first glance this is implausible, but the daily observations do not look at all unreasonable, and there is no doubt it was a snowy month in the region; Kotzebue reported one of their snowiest Aprils on record.  The NCEP reanalysis, which provides an independent estimate of conditions (i.e. it is not influenced by surface station reports), indicates an unusually wet month across northwestern Alaska (see below).

Next in line after Ambler for mean annual snowfall are Coldfoot and Indian Mountain, both of which average 99" of snow according to my calculations; both locations are slightly above 1000' elevation.  Indian Mountain was the site of the largest annual snowfall on record in the interior, at 217" in 1966-67.  Interestingly this again included a very large 48" in April, and the reanalysis estimate of precipitation backs this up, with a swath of more than 4" liquid equivalent extending east of the Seward Peninsula to the vicinity of Indian Mountain.

Given these results, it seems likely that the snowiest region in the interior is the zone from Kaltag north to the Kobuk valley and northeast to the Koyukuk River area.  Within this broad area, the paucity of data makes it difficult to say anything with confidence, so I'd be glad to hear of any anecdotal evidence or additional data that might clarify the issue.

Summary of SST Impacts

Reader Gary suggested recently that it would be helpful to have a summary of the typical impacts of major ocean anomalies on Fairbanks climate.  There are many different ways this information could be presented, but I thought it might be useful to visualize the contrasts between different ocean phases in a chart format.

Beginning with impacts on winter (November-March) temperature, the chart below shows the median temperature anomaly for strongly negative, near-neutral, and strongly positive phases of 4 different large-scale sea surface temperature (SST) patterns.  I've included the Atlantic Multidecadal Oscillation (AMO) for good measure as I've mentioned it a few times recently.  Note that the data are from 1930-2015; strongly negative and strongly positive are defined as the bottom 10 and top 10 years respectively, while near-neutral consists of the 20 years in the middle of the 86-year distribution.  Also note that I de-trended the temperatures before calculating the anomalies so that relatively warm and cold anomalies are more evenly distributed through the history.

The PDO and ENSO (El Niño/La Niña) impacts on temperature are as expected, with a negative PDO and La Niña both bringing cold winters, and with El Niño showing a less reliable warm influence than a positive PDO phase.  The NPM appears to have little effect on temperature, but surprisingly the AMO phase is modestly influential, with a more positive phase correlated with warmer conditions.

Similar charts for winter precipitation and snowfall are shown below; the PDO and ENSO impacts are as expected, and the impact of the NPM is "opposite" to that of the PDO and ENSO, as I showed in my September talk.  It's interesting to note that the strongly positive NPM phase brings more than twice as much precipitation as the strongly negative phase (median 3.65" vs 1.44"), but the snowfall is only 78% higher (55" vs 31").

The charts below show parallel results for the other seasons of the year, defined as April-May for spring, June-August for summer, and September-October for autumn.  Vertical scales are left unchanged to highlight the relative magnitude of temperature anomalies and precipitation amounts in different seasons.  There's a lot that could be said about the different patterns, but I'll leave additional comments for another time.

Cold During Positive PDO

Last week I noted that strong El Niño conditions in the tropical Pacific Ocean do not preclude unusual cold in Fairbanks.  However, it's a little more surprising that below-normal temperatures have occurred statewide recently in conjunction with a positive PDO phase, and the recent cold in Anchorage stands out as an unusual event in this regard.  A daily record low temperature was tied on the 18th (-8°F), and record low maximum temperatures were tied or broken on 3 of 4 days from the 15th to the 18th.  If it weren't for the November 1989 cold spell, Wednesday's high temperature of +3°F would have been the earliest on record for such a cold day in Anchorage.

Looking at weekly numbers, the week ending November 19 was 16.3°F below normal in Anchorage, which has only been observed a handful of times in the past 20 years in November (most recently in 2011).  In view of the fact that the PDO was significantly positive in October, and a positive PDO favors unusual warmth, this is surprising.

The chart below shows the mean temperature anomaly in Anchorage for all weeks ending in November since 1952, with the horizontal position of the points corresponding to the October mean PDO index in the same year.  The recent cold week is highlighted; in the past, the October PDO index was always below +0.40 when a week this cold was observed in November.  The closest analog was in 2003, when the October PDO index was +1.1 and the week ending November 20 was 14.2°F colder than normal.

Here's a comparison between recent weekly SST anomaly maps, showing that the positive PDO pattern has largely disappeared in the past month, so the November PDO index will be lower than October, but almost certainly still positive.  North Pacific waters remain warmer than normal overall.

In the interior, the first -40 temperatures of the season were observed with the recent cold spell.  These included:

-43°F   Kanuti Lake in the Kanuti NWR southeast of Allakaket
-42°F   Clear Creek RAWS northwest of Hughes
-42°F   Chicken COOP
-40°F   Huslia

Record Temperatures

The subject of record temperatures came up the other day when Anchorage set a record low temperature for the first time in 2.5 years. In Anchorage, there is a strong downward trend in record low temperatures during the period of record and a strong upward trend in record high temperatures. Does the same hold true for Fairbanks? Last year I wrote a post that summarized record frequencies for 14 stations; however, it used a standardized starting point so many early years were excluded.

When just looking at the Fairbanks International Airport, the observations for this station go back to December 1929 in the GHCN database. For this analysis, I used January 1, 1930 as a starting point.

Methodology

Every time a record temperature occurs in a year, that year receives one tally for records. For example, the record high maximum for January 8th was set in 1930; therefore, the year 1930 receives one point for that record. The record high for January 9th was set in 1937 and then tied in 2002. In the case of ties, each year receives a partial tally for the record (0.5 points when shared between two years. The sum total for all 86 years (for each record category) is 366 – one tally for each day of the year. If the data were truly random, we expect that each year contains 4.25 records for each category.

Charts

• Figure 1 shows the annual count of record high maximums.
• Figure 2 shows the annual count of record high minimums.
• Figure 3 shows the annual count of record low maximums.
• Figure 4 shows the annual count of record low minimums.
• Figure 5 shows a stacked bar chart of all categories combined.

Analysis

I must admit to being surprised that the rate of record high maximums and high minimums (Figures 1 and 2) were not more strongly positive. Figure 1 shows no trend whatsoever and Figure 2 shows only a very weakly positive trend.

For the cold records, a pronounced negative trend exists.In Figure 3 and Figure 4, the trend toward low instances of record cold events begins in the early 1970s. Perhaps most interesting is the very pronounced, record cold snaps of Fall 1992 and Spring 2013. In both cases, many records were established. Without these two anomalous events, the long-term cold record trend would be very strongly negative.

Interestingly, the trend of all record occurrences (see Figure 5) is moving downward. In the first half of the climate record, there was a ~40% chance that any given year was responsible for 20 or more records. In the second half of the climate record, only ~15% of years have seen 20 or more records attributed to them. This is not unexpected though from looking at the first four figures. Since record high temperatures are nearly level and record low temperatures are less common, the total for more recent years should be lower – which it is.

Figure 1. Record high maximum temperatures by year for Fairbanks International Airport station (1930-2015).

Figure 2. Record high minimum temperatures by year for Fairbanks International Airport station (1930-2015).

Figure 3. Record low maximum temperatures by year for Fairbanks International Airport station (1930-2015).

Figure 3. Record low minimum temperatures by year for Fairbanks International Airport station (1930-2015).

Figure 5. Stacked bar chart showing all records by year for Fairbanks International Airport station (1930-2015).

Getting Cold

The first temperatures below -30°F were observed last night in two different parts of Alaska: on the central North Slope, where -35°F was observed in at least a couple of spots (Prince Creek near Umiat and Otuk Creek near Ivotuk), and in the Fortymile Country, where Chicken also reached -35°F.  It's only about a week early for Chicken to reach such a temperature; it happened in October in 2008.

Clouds have kept the central interior warmer in the past couple of days, but continued cooling aloft will allow for plummeting temperatures at some point in the next few days.  A time-height cross-section of temperature from this morning's GFS forecast shows almost the entire column below -24°C (-11°F) by Wednesday.  An airmass this cold (completely below -23°C) is fairly unusual this early in the season, having been observed only 8 times in 65 years prior to November 20 (but 3 times since 2003).

What does it mean at the surface?  One answer is to look at the GFS MOS forecast, which shows -41°F in North Pole and -38°F at the Ester 5NE COOP on Tuesday.  This is an objective statistical forecast based on multiple regression with the model variables; in theory it's an unbiased estimate.

Judging from webcam images today, it appeared that the Yukon River was still moving at Dawson - albeit choked with ice - and the Tanana River was still flowing at Nenana.  Here's a video of steam fog over the Tanana River on Thursday morning, when the low temperature was -12°F.  (The resolution's not too good, I know - a couple of static shots are below.)

Cold During El Niño

With the advent of significantly below-normal temperatures in the interior (-26°F at the Salcha RAWS this morning), and a forecast for still colder weather, the question arises as to whether this is unusual during a strong El Niño episode.  As I'm sure most readers are aware, one of the strongest El Niño events on record is nearing peak intensity in the equatorial Pacific Ocean, and as we saw here El Niño often coincides with warmth in the eastern interior, especially when the PDO phase is positive (as it is now).  But what about the very strongest El Niño events?  Do they fit the typical pattern?

The charts below show the daily temperature anomalies from October through March for the top 3 El Niño winters on record, based on October-December mean SST in the Niño3.4 region.  These El Niño events were considerably ahead of any other years based on this metric.

Interestingly we see that each winter had a very significant cold spell in January, and 1982-83 had several cold spells earlier in the winter.  1997-1998 was characterized by unusually low variance, with temperatures often not far from normal, but the other winters were more variable.  None of the winters was overwhelmingly warm, and it's clear that unusual winter cold is entirely possible in Fairbanks during strong El Niño events.  Note that the PDO was positive in 1997-98, increasingly positive in late winter of 1982-83, and slightly negative in 1972-73.

As for precipitation, 1982-83 was very wet during October and November, but the other two strong El Niño winters had near-normal precipitation in early winter.  Both 1997-98 and 1982-83 turned dry in the new year as the North Pacific Mode became negative (more so in 1998).

Recent Snowfall

Frequent light snowfall over the past 12 days has brought the snow depth up to 6 inches in Fairbanks, which while still modest, is actually above the long-term normal of 5" for this date.  A more substantial snowpack is in place now in the hills; for example, the Keystone Ridge COOP (1600' elevation) is reporting 16" on the ground, which is the most for this date since 2004.  The SNOTEL sites to the northeast of Fairbanks are reporting between 11" and 20", as seen in the map below (click to enlarge and view the small icons).

Farther afield, McGrath also has 6" (normal 4") and Bettles has 14" (normal 9").

The onset of relatively snowy weather in the interior is not too surprising in view of the highly positive phase of the North Pacific Mode.  The NPM index in October was the second highest on record since 1900, and according to station data since 1950, a positive NPM phase tends to bring above-normal precipitation to interior Alaska.  The map below shows the percentage of years with above-normal precipitation in October through December when the October NPM index was strongly positive.  More details on the NPM influence can be found in the slideshow from my UAF talk back in September:

https://accap.uaf.edu/sites/default/files/The_BLOB_AK_Weather.pdf

[Update November 10: another 2.3" of snow yesterday brings the snow depth to 7" in Fairbanks as of midnight.  This is now the snowiest start to November since 1990, and 4th snowiest on record.  The season-to-date snowfall of 33.5" is the 3rd highest on record.]

Another Warm October in Barrow

To nobody's surprise, October was substantially warmer than normal in Barrow again this year.  Barrow's persistent October warm anomaly has been discussed here several times before, including:

http://ak-wx.blogspot.com/2011/11/barrow-in-october.html

http://ak-wx.blogspot.com/2012/11/record-warm-ocotber-in-barrow.html

http://ak-wx.blogspot.com/2013/11/another-very-warm-october-in-barrow.html

http://ak-wx.blogspot.com/2014/10/barrow-temperature-update.html

The chart below shows the October mean temperature in Barrow each year since 1930, with markers colored according to which (long-term) tercile the temperature fell into (credit to Rick Thoman for devising this illuminating chart presentation).  This year was the 14th consecutive year in which the October temperature fell within the upper tercile.  Interestingly the past 2 Octobers (2014, 2015) have been the coolest of this 14-year sequence; I wonder if Barrow might experience a "near-normal" October sometime soon.

Prior to 2002, the highest frequency of warm Octobers in Barrow occurred in the 1930s, 40s, and early 50s; but the climate regime back then was nowhere near comparable to the modern state, as much colder Octobers regularly occurred in between warm years.

I thought it would be worthwhile to look at a few other climate observing stations around the Arctic, to see if extreme October warmth has also been observed elsewhere in recent years.  The map below shows the stations I looked at.

Proceeding westward from Barrow, the following charts show the time series of October mean temperatures at each location.  Svalbard and Eureka have shorter periods of record, so I didn't create the long-term terciles for those locations.  The Russian stations on Kotelny Island and at Dikson show similar results to Barrow, with uninterrupted persistent warmth in recent Octobers and a dramatic reduction in variance compared to the earlier climate; however, the extreme warmth seems to have started a bit later than in Barrow, with some cooler years persisting until 2006.  The Russian stations also show more pronounced warmth than Barrow in the early years; the lack of cold at Dikson in the 1940s almost rivals the modern warm pattern.  Both Russian stations had their warmest October of record in 1947.

Svalbard and Eureka both show an absence of unusually cold Octobers in recent years, although almost a decade has passed since the warmest October of record at Eureka; no recent October has come close to the (relative!) warmth of 2002 and 2006.

A comparison of 10-year trailing mean October temperatures (see below) shows strong similarities between the trends in recent years, confirming that the autumn warming is an Arctic-wide phenomenon; it's particularly notable that the rate of warming has been similar since 2000 at Dikson, Kotelny, and Barrow.  However, temperatures at Svalbard and Eureka have not increased on average for some years now; these differences may be attributable to decadal shifts in the October circulation patterns across the Arctic.

The data from Barrow and the Russian stations clearly show the multi-decadal variability that characterizes Arctic climate, as the earliest years (prior to 1955) were relatively warm compared to the middle part of the historical record.  In my view the multi-decadal variations are probably influenced by the Atlantic Multidecadal Oscillation, and so it seems likely that the next downturn in the AMO (which is still significantly positive) will also affect the Arctic temperatures.  However, it remains to be seen how significant this temporary cooling influence will be compared to the overall long-term warming trend.

Cold Forecast

The medium-range forecast is starting to look pretty chilly for interior Alaska, suggesting that true winter conditions may not be far off.  Here are the latest 6-10 day and 8-14 day temperature forecasts from NOAA's CPC, showing about a 60% chance of lower-tercile (significantly below-normal) temperatures in both periods.

Today's computer model runs show the circulation feature that is expected to be responsible for the cold: a big trough over northwestern North America.  The deterministic GFS forecast shows some impressive cold, with 850mb temperatures dropping below -30°C over the far eastern interior by November 15.  The last time 850mb temperatures dropped below -25°C in Fairbanks before mid-month was in 2006 (-26.5°C on the 13th).  But that is one of the more extreme scenarios that might happen, and probably not the most likely.

Here's a nice webcam photo from Dawson, YT, today.  Ice is still moving on the mighty Yukon River, but freeze-up appears to be imminent.  The temperature was +5°F at Dawson airport this morning, but reached -9°F on Sunday.

Temperature Drop

As befits the time of year, sharply colder temperatures have occurred throughout the eastern interior in the past couple of days.  Here are some of the colder observations I've noted:

-20°F  Chicken COOP  (only 2 days ahead of the normal first occurrence of this temperature)
-18°F  Mosquito Fork HADS near Chicken 19W
-17°F  Beaver RAWS
-17°F  Chisana RAWS (3318' elevation)
-16°F  Bolio RAWS near Fort Greely
-16°F  Fort Yukon RAWS
-16°F  Bettles SNOTEL
-14°F  Chalkyitsik RAWS

North Slope locations were even a little colder, with Deadhorse dropping below the -30°C mark (-22°F); the CRN site just south of Deadhorse made it to -26°F, which appears to be the lowest temperature observed in the state so far this season.

Fairbanks managed to drop to -1°F late on Saturday evening, which was not only the first sub-zero reading but the first temperature below 19°F.  After being unusually warm for most of the month, October ended up with a sub-zero temperature after all.  It was the 8th warmest October since 1930, but was much less warm than 2013 (mean temperature 31.8°F vs 36.1°F).  Snowfall in the last 3 days of the month produced an October total of 3.8", which is the 9th lowest since 1930.

The persistent warmth of October gave the perception of a much-lengthened autumn and delayed onset of winter in Fairbanks, but certain other parts of the interior experienced a more wintry scene throughout most of October.  For example, in Circle Hot Springs (elevation 860') the late September snow never melted off, and temperatures remained almost continuously below freezing after the 8th of the month.  The chart below shows the contrasting conditions between these two locations that are separated by only about 100 miles and 400' elevation.

On another note, webcam photos reveal that the Chena River froze up in the past 24 hours in Fairbanks, but the Tanana is still open at Nenana.