Coolwire 10                24 August 2004

Alaskan Warming Threnody
by Willis Eschenbach

Many claims have been made lately about the "global warming" in Alaska. Fairbanks has been said to have been hit hard by global warming. Not knowing much about the temperature regimes of Alaska, despite having fished commercially in the Bering Sea for a number of seasons, I thought I should take a look to determine what's going on.

First, a look at Fairbanks. There are two overlapping long term records available for Fairbanks. The earliest is from the Experimental Station of the University of Alaska at Fairbanks. The other is from Fairbanks Airport, about 3 km away. Here are the records (17 year Gaussian averages):

The excessive rise in the Fairbanks record has long been recognized as being a result of the Urban Heat Island (UHI) effect (see for example Magee, N., Curtis J., Wendler, G.  "The Urban Heat Island Effect at Fairbanks, Alaska", Theoretical & Applied Climatology, Vol. 64, pages 39-47). Because of this distortion of the Fairbanks record, I have used the University record. The University record (orange) stops at 1990. Based on the close correspondence between the two records, in order to get a long record for the area I have taken the liberty of extending the last few years of the University record (extension shown in blue) by linear regression from the Airport record.

Here are two sets of independent data for Fairbanks, one from the Alaska Climate Center at the University of Alaska at Fairbanks, and the other from the NASA Goddard Institute of Space Scientists:

Note the agreement between the two independent data sets, both of which show that Fairbanks has been cooling since 1977. Also note that there have been four years since 1970 which were as warm or warmer than 2003 (1980, 1981, 1987, and 1994). In short, for nearly a quarter of a century, the temperature in Fairbanks has been dropping.

It has also been claimed that the ice cover of the Northern Hemisphere has dropped radically. Again, perception is outstripping reality in this area. Some areas have less ice, in particular the North coast of Alaska, but other areas have more ice. Net change? Here's the record of the last year, current to mid-August 2004:

SOURCE: University of Illinois Arctic Climate Group

Note that the Northern Hemisphere sea ice currently (mid August) covers exactly the same area that it covered on average for the 22 years from 1979-2000.

To determine the short-term trend for the entire state of Alaska, I plotted the Gaussian average of the data shown in the chart you get when you click the “Mean of all Stations: +2.69” text at the bottom of the top graphic on the page cited under the chart below. I also show the linear trend since the last PDO shift. There has been a slight decrease in temperature (-0.5 degrees/decade) since 1980.

This data clearly shows that overall, as the web page says, Alaska warmed +2.69 degrees since 1971. However, it also clearly shows that Alaska cooled for 23 years, from 1978 to 2000. The dangers of short temperature records, as well as the sensitivity of trend lines to starting and ending points, are both in play here.

SOURCE:  Alaska Climate Research Center, University of Alaska at Fairbanks

Like the Fairbanks temperature, the temperature of the entire state has been dropping since about 1978.

Other people have said that the date of the ice break up on the Nenana river shows that Alaska is warming drastically. Here's a typical analysis of the Nenana data claiming to show a weak trend whereby the warming is causing the ice to break up earlier:

SOURCE: Dominique Bachelet, personal communication

Again, this only shows the inadequacy of using trend lines with climate data. Climate is a chaotic system, subject to unpredictable changes. For climate data, trend lines are useless. Gaussian or other similar averaging is the only thing that will allow us to understand the data. In addition, it is necessary to have long term records to have any chance of understanding what is happening.

For example, here is the Gaussian average of several long term sets of Alaskan climate data, including the Nenana ice break up:

TEMPERATURE DATA SOURCE: NASA Goddard Institute of Space Sciences

We can see that the Nenana ice breakup is ruled by the same factor which rules all of Alaska's weather -- the Pacific Decadal Oscillation. All of these records have the same form -- rising to a high around 1940, dropping to 1950, level to 1970, rising to 1978, and basically level since then.

How do we know that Alaska is ruled by the PDO? The strength of the PDO can be measured by the Aleutian Low Pressure Index, the size of the low pressure area around the Aleutians. Here's some Alaskan temperatures plotted against the PDO:

TEMPERATURE DATA SOURCE: NASA Goddard Institute of Space Sciences
ALPI DATA SOURCE: Beamish, R.J., C.E. Neville and A.J. Cass. 1997. Production of Fraser River sockeye salmon (Oncorhynchus nerka) in relation to decadal-scale changes in the climate and the ocean. Can. J. Fish. Aquat. Sci. 54: 543-554.

We can see that the temperatures of cities as far apart and as climatically different as Fairbanks, Nome, and Anchorage slavishly follow the Pacific Daily Oscillation. As the PDO goes, so goes Alaska. Because of this, we can control the temperature data for the PDO, to see if there are any changes which are not attributable to the PDO. Here is the result of that analysis for the same cities:

Several things are of interest about this record. One is that there has been very little change in detrended temperatures over the period of the record. Only Anchorage has warmed significantly, about a degree since the start of the record.

Nor is there any indication of "greenhouse warming" -- Alaskan temperatures have not been increasing as the CO2 increased; in fact they have been decreasing since 1980. This lack of warming in the post-1980 record is shown in both the raw temperature record as well as the record corrected for effect of the PDO.

Finally, the actual temperature rise for Anchorage is sure to be less than shown, because this data is not corrected for the Urban Heat Island (UHI) effect. The increased size of the Urban Heat Island effect in the Arctic regions is not widely appreciated. Because the UHI is dependent in part on the the difference in temperature between the heat sources and the air, the UHI is enhanced in cold regions. See, for example, "THE URBAN HEAT ISLAND IN WINTER AT BARROW, ALASKA", Kenneth Hinkel, Frederick Nelson, Anna Klene and Julianne Bell, available at They show that a full two degree UHI occurs in wintertime in Barrow, a town of only 2600 people, and that the effect can be measured for miles outside of the town itself. Thus the real temperature rise in these cities must be less than is shown above.

As an estimate of the amount of the Anchorage UHI, here is the Anchorage temperature compared to the nearest rural station, Matanuska (60 km distant):

Anchorage's urban development took off after World War II. This is clearly reflected in the graph, which shows Anchorage steadily warming with respect to Matanuska since that time. In 1990, the UHI for Anchorage was about one full degree, with an increase of about +0.021 degrees per year since 1945. Knowing this, we can adjust the Anchorage temperature by that amount and re-run our residual analysis. Here, then, is the final temperature record for the three cities, detrended for the PDO and with the Anchorage UHI removed:

No warming overall, and cooling since 1978, with all three stations showing a swing of only +/- 0.5 degrees ... pretty tame stuff, no headlines here.

Having investigated the past, and having seen that the state is not currently overheating, what can a computer General Circulation Model (GCM) tell us about the future? Well ... not much. Here are runs from a couple of computer GCMs, one from the Canadian Climate Center (CGCM1) and one from the UK Met Office Hadley Center (HADCM2SUL), trying to reproduce just five years of Alaskan temperatures:

SOURCE: Dominique Bachelet, personal communication

There are four colors of lines in these graphs -- dark blue (averaged data), light blue (HADCM2SUL), green-gray (unlisted), and red (CGCM1). However, only three of them are listed. I assume the fourth is the raw Tmax or Tmin data.

The GCM results do not inspire confidence ... the two computer runs started in 1995, and by the year 2000, the two projections are two full degrees apart. In other words, at least one (and possibly both) of these GCMs has an error of two degrees in five years (forty degrees per century?). This does not exactly make me want to believe them ... if these are the best tools available, then I'll take no tool at all over these.

If these GCM models actually worked, we would be able to give them the conditions (CO2, temperature, etc.) for 1920, and they would be able to predict the current temperature, as well as the global climate shift of 1976/77, the global decline in temperature from 1940 to 1970, and the other climate features of the years from 1920 to the present. To my knowledge, however, there is not one GCM which can come anywhere near to doing that. None.

And even if a GCM could do that, unfortunately this would still not mean that it could correctly predict the future climate. That's the problem with chaotic systems -- unlike deterministic systems, they don't repeat the past. Another example of a chaotic system is the stock market. This is an arena where, as the US government forces the brokers to state in their advertisements, "Prior performance is no guarantee of future success." In other words, just because a GCM might be able to predict the past (and to date no GCM can even do that), it's certainly no guarantee that it can predict the future.

(As an aside, I note that the historical data used by the GCMs, like my data above, shows that the temperature of Alaska has dropped since 1980 ...)

The problem with GCMs is that we are trying to predict a chaotic phenomena. The PDO is a very good example of this difficulty, as it even more complex -- it is a bi-stable chaotic phenomenon. It flips from a warm state to a cool state at unpredictable intervals. Simply put, we don't have a clue what flips it from one state to another, and the length of the cycle is so great that we have very little data upon which to even hazard a guess.

Nor does the general agreement of the CGMs about the future climate of Alaska increase the confidence we can place in their answers. They are generally based on the same principles, they generally include the same variables, and they generally exclude the same variables (including the most important greenhouse gas, water vapour in all its forms). But this does not mean that they are right -- all that this means is that it is very likely that they will all make the same mistakes ...

While GCMs may well be the "best tools available" to predict climate as many claim, this does not mean that they are good enough to make the kind of predictions that people are using them to make. For example, when will the PDO flip again to a cool state from its current warm state? No one knows. Period.

It could flip tomorrow, or in ten years. It could could flip to a cool state and stay there for fifty years, or never flip to a cool state again. Despite our hubris, despite our computers that can do teraflops of calculations per second, despite all our education and insight and understanding, despite our scientific understanding of weather phenomena, despite our satellite records, despite all of this and more, a rude and ugly fact remains -- we cannot predict either the weather or the climate out for any distance into the future.