Wednesday, August 19, 2015

Pace of Temperature Records

Last summer we looked at changes in the rate at which temperature records have been set over the last 60 years or so; see here and here.  I've often pondered the same topic since then, and I thought it would be interesting to update the results for 2014 and 2015 to-date, and to look at which stations have warmed the fastest (and slowest).

First, the two charts below show a 19-station average of annual numbers of daily temperature records (scroll down to see the stations listed on another chart below).  The data are taken from 1954-present and records are defined relative to this period only.  The first chart shows occurrences of record warmth, whether for daily maximum (red) or daily minimum (blue) temperature; and the second chart shows occurrences of record cold.  As we noted before, cold records have diminished somewhat more rapidly than warm records have increased.  Among the 4 categories of records, the most rapid changes have occurred for low daily maximum records, followed by low minimum records, then high minimum records, and the least rapid change has occurred for high maximum records.

The 1976 PDO shift is clearly evident on the charts, and the extreme warmth of the past two years is also a striking feature.  2014 saw the highest number of warm records (both maximum and minimum temperatures) and also the least number of low maximum records.  Interestingly the number of low minimum records was slightly lower in 1978.  Remarkably, 2015 looks very likely to outdo 2014, as the number of high minimum records is already almost equal to 2014, and so far there have been just a handful of days with daily cold records at a few stations in 2015.

One minor concern I had with the results above is that there is a higher concentration of stations in southern and southeastern Alaska, so I repeated the analysis after excluding Kodiak, Homer, Talkeetna, Juneau, and Big Delta.  The results are little different, although 2015 to-date looks a little less extreme - see below.

It's interesting to compare the overall pace of change between the different stations.  I attempted to do this by adding the slopes of the linear trends for the warm record counts and subtracting the trend slopes of the cold record counts - see the results below.  Based on this metric, the long-term rate of warming has been greatest at Homer and Talkeetna, but interestingly the relatively nearby stations of Kodiak and Gulkana have seen some of the smallest warming rates.  The difference between Kodiak and Homer is particularly interesting as both stations are heavily influenced by North Pacific sea surface temperatures.  The second chart below shows that there is a long-term trend in the temperature difference between Kodiak and Homer, and 2014 was the first year in which Homer was the warmer of the two locations.  This looks a bit suspicious to me; there might be some station siting issues that have affected the reported temperatures from either or both of these places.

Lastly I'll show the record counts for several individual stations.  The situation at Homer has been very extreme during this year and last:

At Talkeetna the rapid warming pace comes from a very large number of warm records in 2002 and 2003:

The slowest-warming site, Gulkana, still shows a long-term warming trend in all 4 record categories:

Lastly, if we look at Fairbanks with data back to 1930 included, the picture changes slightly, as the high maximum records no longer show a long-term warming trend; however, the long-term decrease in cold records is still very much evident.


  1. Another factor going into this is the number of temperature records hit will decrease as more data is produced. This will not effect the data that shows the warming trend when high records continue to be broken but could exaggerate the effect on the cold end.

  2. Tyler, for the purpose of this study I defined the records relative to the entirety of the period shown, so this removes the effect you're describing. For example, if the Feb 10 max temp record (1954-present) is 50F, this is the level that must be attained in any of the years to add a count on the chart. 50F is the record to be beaten in 1960 even if it didn't occur until 2010; so the records look forward in time as well as back.

    In day-to-day experience with evolving historical records, you're of course correct that the rate of setting records would generally diminish over time if the climate were unchanging.

    Thanks for reading.

  3. I wouldn't consider Talkeetna and Gakona as close to each other. They are seperated by over 100 air miles. And they do have different climates: Gakona consistently gets colder temperatures and has some rain shadowing from the Chugach mountains. The only similarity is that both are in river valleys and both are at about the same latitude.

    I would compare Talkeetna to Willow, Trapper Creek, or maybe Cantwell.

    1. Fair enough - they are not close and they do have very different climates. I just thought it was interesting that they were at opposite ends of the warming-rate spectrum.

    2. I would whole heartily agree with you here. Even though they are not too close to each other, they still have huge differences that seem more related to regional climate than the general pattern. It really brings up the question as to "why?"

  4. Richard, how did you handle ties? Just curious.

    1. Ties counted the same as stand-alone records, i.e. adding a full count to the year in question. It might be a bit more rigorous to divide the tie counts among the number of years in the tie. However, the 1954-present total number of records (by my method) is not greatly different between stations: it varies from 1670 at Bettles (fewest ties) to 1989 at St Paul (most ties). Presumably my method correspondingly overweights the more maritime stations in the aggregate analysis (top 4 figures).