Sunday, April 7, 2013

Interior Alaska Temperature Anomalies: Hills vs. Valleys

I was talking with a colleague this past week about standardized anomalies as a way of grouping locations into broad climate regimes (e.g. the Alaska climate divisions work in Bieniek 2012).  A primary question is to what extant different elevations anomalies are correlated. This is hard to answer for Alaska because there are so few stations above their respective valley floors with long periods of record. However, I thought I'd give it a go for Fairbanks-land.

Here's a plot of the difference in monthly mean temperature standardized anomalies (anomalies computed from the NCDC 1981-2010 normals) between Keystone Ridge and Fairbanks Airport (14 miles apart) for entire period of record for Keystone Ridge (June 1996 to present). This is not an ideal test as Keystone Ridge is only 1200' higher elevation than the Fairbanks Airport. However, data for both sites are complete and there is no differing time of observations to confound the question. The way this plot is constructed, positive values indicate the higher elevation station is warmer than expected compared to the valley, and vice versa. In winter, this is straightforwardly interpretable as variations in mean monthly inversion strength. In summer, the situation is more complex but certainly differences in the number of rainy days will effect the differences in anomalies.

The overall correlation between monthly temperature anomalies at the valley and higher elevation site is an very impressive 0.94. Looking at individuals months, May and October have correlations of 0.99, while December has the lowest correlation at 0.88. In the above plot there are some interesting outliers. The largest negative anomaly is November 2003. This was a wet and cloudy month, with only minimal inversions most of the time. The largest positive anomaly is December 2005, with exceptionally strong inversions (and hardly any precipitation) prevailed the second half of the month. Just eyballing the plot, there may be some longer term signal. Since late 2007 there have been a (seemingly) large number of months with negative differences, while 2004 and 2005 had a large number of months with positive differences.

Variations in daily anomalies are greater, as the plot below since 2011 shows, I've presented a slightly different view in this case, with the actual Fairbanks Airport daily anomalies in green and the Keystone Ridge difference in red. Again, positive values indicate that Keystone Ridge is warmer than expected compared to the valley, and vice versa.

Note: it would be better to use the anomalies calculated from  means and standard deviations for the overlapping period of record (again, June 1996 to present), but it turns out that the difference from using the 1981-2010 normals is, in this case, negligible, and since I otherwise use the current NCDC normals for calculating standardized anomalies, I chose in this case to remain consistent. 


  1. Rick,

    The apparent long-term signal in the anomaly difference is interesting. It sure looks like it's correlated with the PDO, which was generally positive in 2004 and 2005, and negative in more recent years. Since the PDO is also correlated with temperature, does this mean the PDO effect shows up more strongly at higher elevation? For example in recent years it's been colder with the negative PDO, but also relatively colder at Keystone Ridge.

    It would be interesting to run the PDO correlations by month at both stations...

    The climate division paper is very nice, thanks for the link.


  2. Brian BrettschneiderApril 8, 2013 at 9:13 PM

    As you noted, precipitation can confound those relationships. It would be interesting to see the same analysis with days having similar "weather" grouped together. If you are using the surface temperature observations as a proxy for measuring inversions, or lack thereof, the radiosonde values would of course contain that information.