Sunday, January 11, 2015

Warmth Aloft and Wind Direction

After reading Richard's excellent post the other day, I though it would be interesting to see if cyclical patterns exist in the direction of winds coming into Alaska. Richard also looked at Barrow's upper level wind direction as it related to Autumnal warming along the North Slope. This is a complementary post to those articles.

Most of Alaska has seen substantial surface warming in the winter months during the last few decades. The warmth has been well documented in numerous booksarticles, and reports. The cause of this increase in warmth is multifaceted and is not the subject of this analysis. One of the frequent responses to observed temperature increases is to question station siting and the effects of urbanization and landcover changes. One way to see if the observed surface temperature increase is a "real" or an artifact of the measurement technique is to look at air temperatures recorded from weather balloons (RAOBs) just above the surface. This blog posts looks at the case of Anchorage, Alaska, and whether the surface temperature increase corresponds to a free-air temperature increase. I know that this blog generally focuses on interior conditions but since the data were handy for Anchorage and since the effects are fairly uniform over large areas, it seems appropriate to go with Southcentral data in this case. I will look at Fairbanks data as time permits.

Surface Temperatures

An inspection of annual temperatures in Anchorage (see Figure 1) shows an obvious increase over the course of the climate record. It is worth remembering that the temperatures since 1953 represents the Anchorage International Airport location and before 1953 it represents a combination of Merrill Field, the Park Strip, and Ship Creek locations. Note the shift in temperature regime in the mid-1970s.

Figure 1. Annual temperature in Anchorage, Alaska, from 1916 to 2014. Several years in the 1920s has too many missing observations for inclusion.

Since temperatures are much more variable in winter than in summer, annual swings in temperatures are often reflected in the temperatures during the cold months of the year. For example, December or January can be +/- 15°F compared to normal but July or August cannot be more than about  5°F from normal. Figure 2 shows the December-January temperature in Anchorage for the length of the climate record. Note the high degree of correlation between the lines in Figure 1 and Figure 2.

Figure 2. December-January temperature in Anchorage, Alaska, from 1916 to 2014. 

If we use standard 30-year periods to calculate long-term normals, as NCDC and WMO does, the difference in temperatures over time is readily apparent. Figure 3 shows all 30-year daily normals since the 1921-1950 time period. The large increase in winter temperatures are clearly evident. 

Figure 3. Daily normal temperature for Anchorage for all 30-year periods beginning with the 1921-1950 time period.

Upper Air Temperatures

At high latitudes, low level temperatures are often dominated by a cold-pocket of air at the surface. This is referred to as a temperature inversion. This cold pocket of air often masks the presence of a change in airmass. Since cold air at the surface is difficult to displace, you can have a situation where an old airmass is present at the surface and a new airmass is present a few thousand feet above the surface. Sun angle, wind speed, and proximity to mountains determine the time lag between the length of time that an airmass change is reflected at the surface. Many times, the airmass changes several times without being noticed at the surface. Therefore, any study of temperature changes must also look at upper air temperatures so as to eliminate the problem of temperature inversions. For this study, we are going to use the 850 mb level (~5,000') as a proxy for free-air temperatures. All weather balloons launched in Anchorage since 1948 have sampled the air temperature at 850 mb and reported the height, temperature, dewpoint depression, wind speed, and wind direction. Figure 4 shows the the annual 850 mb (~5,000') temperature at Anchorage since 1948 and Figure 5 shows December-January temperatures only.

Figure 4. Annual 850 mb temperature in Anchorage, Alaska, from 1948 to 2014. 

Figure 5. December-January 850 mb temperature in Anchorage, Alaska, from 1948 to 2014.

The temperature increase at 850 mb is both noticeable and statistically significant. However, the magnitude is slightly less than observed at the surface. Note that the units for Figures 1 and 2 are degrees Fahrenheit and the units for Figures 4 and 5 are degrees Celsius. As we did with surface temperatures, we show the daily normal 850 mb temperatures for all 30-year periods in Figure 6.

Figure 6. Daily normal temperature for Anchorage at 850 mb for all 30-year periods beginning with the 1951-1980 time period.

A side-by-side comparison of Figures 3 and 6 shows a similar, dramatic increase in wintertime temperatures in Anchorage. At the surface, January temperature have increased about 7°F since the 1950s and the 850 mb temp has increase by as much as 1.9°C (3.4°F). The 7°F surface increase from the 1950s is equivalent to 0.55 standard deviations above the normal value at the time. The 1.9°C increase in December-January 850 mb temps is equivalent to 0.35 standard deviations above the normal value at the time.

The Role of Wind

When the airmass south of Alaska is advected northward from the sub-tropics, warm conditions ensue. Indeed, this winter has seen a never ending stream of warm air from the central North Pacific Ocean move into Alaska bringing warm air with it. Is this the case for all warm winters? The reason we spent so long introducing the upper air conditions in the previous section is because an airmass moving from the south to the north is often not apparent in the surface wind observations – but it clearly shows up in the 850 mb wind observation. Therefore, let us take a look at how the winds at 850 mb blow during warm or cold years.

Figure 7. Frequency distribution of December-January wind direction at 850 mb by decade. Inset chart shows the average 850 mb temperature by decade.

Figure 7 shows the cardinal wind direction by decade for the Anchorage balloon station at 850 mb for only the months of December and January. Since those two months showed the most dramatic surface and 850 mb warming, it is entirely possible that southerly winds might be correlated with the increase in winter temperatures. The chart shows a notable increase in southeasterly and southerly winds at 850 mb over time. There is a corresponding drop in the amount of time with northerly and northeasterly winds during these two months. The R-square value between 850 mb temp and the percentage of 850 mb southeasterly winds by decade (Figure 7) is 0.64. Therefore, we can conclude that a fair amount of the upper air warming during the winter months is due to an increase in southerly winds which advect warmer air from lower latitudes.

Figure 8. Scatter plot showing December-January 850 mb temps with the percentage of time December-January 850 mb winds are from the southeast at Anchorage, Alaska.

Looking at individual seasons instead of decades, the relationship between wind direction and 850 mb temperatures in December and January is easier to see. Figure 8 shows a scatter plot of December-January 850 mb temps and the percentage of observations with southeasterly winds. The R-square correlation is still very high. Similarly, there is a strong negative correlation between 850 mb temperatures and the percentage of time winds are coming from the north. Figure 9 displays the strength of this correlation.

Figure 9. Scatter plot showing December-January 850 mb temps with the percentage of time December-January 850 mb winds are from the north at Anchorage, Alaska.


There is a clearly defined relationship between warm winter temperatures in Anchorage (at the surface and at 850 mb) and the prevailing wind direction. This should not come as a surprise to anyone. Air from south of Alaska is obviously warmer than air from north of Alaska.

Figure 10. Percentage of December-January 850 mb wind observations from the southeast at Anchorage, Alaska..

What is surprising is the increased frequency of 850 mb winds originating from the south and southeast. Figure 10 shows the long-term trend of the percentage of 850 mb observations with a southeast wind. The question becomes why? The answer to why is not so easy. Southeast winds at 850 mb imply a low pressure center in the Gulf of Alaska. Are Gulf lows becoming more common? Are they becoming more intense? One of the more talked about climate drivers in Alaska is the Pacific Decadal Oscillation. Using University of Washington data, we can show that years with positive PDO values indeed have a larger percentage of 850 millibar southeast wind observations.

Figure 11. Scatter plot showing December-January Pacific Decadal Oscillation (PDO) index values with the percentage of time December-January 850 mb winds are from the southeast at Anchorage, Alaska.

Figure 12. ESRL Reanalysis of sea level pressure change between the 1981-2010 period and the 1951-1980 period.

How does all of this affect air pressure patterns? Since the pressure-gradient force drives winds, the air pressure regime should tell us a lot about the wind magnitude and direction. Figure 12 clearly shows that air pressure has been lower in the Bering Sea and the Aleutian Islands during the more recent 30-year normal period. This confirms the hypothesis that lower pressures are more frequent or more intense in a geographical position that promotes southerly air flow. Still, this does not answer the why question. Does the decrease in sea ice affect the pressure patterns? What about the Pacific Decadal Oscillation? This is obvious a question of great importance to Alaskans and an active area of research.

Supplemental Figures for Fairbanks

Figure S1. Annual 850 mb temperature in Fairbanks, Alaska, from 1948 to 2014.

Figure S2. December-January 850 mb temperature in Fairbanks, Alaska, from 1948 to 2014.

Figure S3. Percentage of December-January 850 mb wind observations from the southeast at Fairbanks, Alaska.

Figure S4. Percentage of December-January 850 mb wind observations from the south at Fairbanks, Alaska.

Figure S5. Scatter plot showing December-January 850 mb temps with the percentage of time December-January 850 mb winds are from the southeast at Fairbanks, Alaska.

Figure S6. Scatter plot showing December-January Pacific Decadal Oscillation (PDO) index values with the percentage of time December-January 850 mb winds are from the southeast at Fairbanks, Alaska.


  1. Very nice presentation Brian and a good learning tool. I wonder how Fairbanks' 850 mb data follows this pattern?

    I tend to look at the forecast of Polar Jetstream patterns and their expected flow when anticipating winter weather. There's a suggested (and debated) association of the Polar Jet with the Arctic Amplification theory and Rossby waves...slowing of the W>E Jetstream due to a reduction in temperature gradients between the Arctic and equatorial latitudes.

    The resultant unusual N-S looping of the slowed Jet could bring warm air from the central Pacific when properly aimed at Alaska. Seems to be reasonable "Why", but as usual well above my pay grade.


  2. I added some figure for Fairbanks at the end of the post Gary. Same general trends but slightly lower magnitude. The exception being a decrease in SE winds in the last 15 or so years. However, there has continued to be an increase in southerly winds.

  3. Brian, great discussion, thanks. Figure 10 in particular seems very useful to explain the modern warmth - although I see the PDO step change in 1976 with little or no trend since then. I suggest the PDO is the main culprit for long-term changes in winter, leading to circulation changes which directly modify airmass frequency. Presumably this would show up in a scatter plot of PDO index versus frequency of southeast wind.

    1. Thanks Richard, I added a figure (Fig. 11) showing the PDO correlation with 850 mb southeast winds in Anchorage and also for Fairbanks (Fig S6).

  4. Another question please. Given a 850 mb (and presumably lower) S/SE wind regime during the positive PDO phase, for Fairbanks could that result in more or less precipitation during winter?

    I'm thinking the mountain ranges to the south may dry the incoming air when the flow is from that direction, and substantial snow may require a SW/W flow pattern undisturbed by elevated terrain.

    Another contribution to warm air above Fairbanks during S/SE flow may be adiabatic heating north of the Alaska Range during strong wind events. That pattern of flow is often accompanied by the E>W Tanana River Jet. Fairbanks is frequently shielded from the Jet at the surface, but may not be overhead at altitude.

    Anchorage may experience similar effects from bordering and channeled terrain.


  5. To get a feel for how 850 and surface temps have changed in Fairbanks, we can go back to Richard's post on There's been plenty of discussion on temps, winds at various levels on this blog and how they relate to PDO, ENSO, blocking patterns, etc. Perhaps some lengthy blog review is needed to summarize what has been found the last couple years. Otherwise some duplication might start occurring. :)

    The difference line on Fig. 6 reminds me of the temperature deviations discussed in the blog post I linked to. It's like all of the seasons have been shifted earlier by a few weeks. I mentioned this earlier but I couldn't find the blog post. I think if you find the reason why the seasons have done this you will have found the reason for the weather regime change.

    1. Eric, I had actually forgot about the post you referenced. You are right that with so many posts in the last 1.5 years, they tend to overlap at times. Regarding this post, I had noticed that the upper level flow was out of the south or southeast nearly every day since November 1st this winter and I wanted to see how correlated warm temps are to southerly winds. That may sound obvious but there could be a situation where a high PDO index or warm Gulf of Alaska waters cause an increase in low-level clouds; which cause the positive temperature anomalies independent of any wind pattern change. It's just one more piece in the puzzle.

    2. If it was low clouds and not pervasive winds you would think that Fairbanks would not see signs of winds too. But better to double check.

      Would you think that the winds being more southerly in Fairbanks as compared to Anchorage's SE is more a function of the strengthen Aleutian low? And there is the blog posts about blocking highs from a little while. So I think it's more than PDO that is affecting things.

  6. This is a very interesting look at the data. I think so many of these indicators (meandering jet stream, blocking ridge over the western CONUS, persistent southerly flow over Alaska, positive phase PDO, El Nino, warm 850 mb temps) are closely related, and the trick is to find the leading indicators, or the forcing mechanisms, as opposed to the lagging indicators. After 6 winters of more negative-PDO conditions, I was convinced we were finally into the alternating 30 year pattern to the 1976 change, but this winter has blown the roof off of most theories. It could just be a blip in the long-term, I suppose.

    1. Perhaps the alternating 30 year pattern is changing to a more unsettled regime due to changes in Earth's climate.

      Drop a bouncing ball and notice how the longer it bounces the shorter the events and lower the amplitudes. Balls aren't climate but it's worth noting the potential.


  7. On January 15 the CPC noted a record + PDO for December in their outlook for February weather: