Assessing Record High Temperatures

I came across an interesting piece of information from the National Weather Service office in Minneapolis-St. Paul (MSP), showing that a new record low of 21 was set on April 20.  A single record temperature, high or low, is not necessarily notable, but part of what makes this story interesting is that the new record low set on April 20 was the first new record low set temperature set in the Twin Cities since August 2004.  For some perspective, the frequency in the occurrence of new daily record temperatures varies from location to location.  For stations with a long record history, setting new record temperatures is relatively rare.  MSP has a long record history, with observations dating back to 1871.  So, we wouldn't necessarily expect the Twin Cities to experience record temperatures very frequently.

What is really astonishing about this simple record low temperature is that while looking through daily records at Minneapolis-St. Paul since 2004, I found nearly 40 daily record high temperatures (see table).  In a static climate scenario, one would expect a roughly equal number of daily record high and low temperatures to be set over a given time period.  The fact that MSP has set nearly 40 record highs for just one record low since August 2004 is statistically quite unlikely.  Many of these record highs occurred in March 2012, an anomalously warm month for the Midwest that I have discussed in previous entries.  Data in the table at right are from the University of Minnesota Climatology Working Group.



Setting a single record high or low temperature it not especially useful in itself for assessing climate trends.  However, a series of records, or a much greater number of records in one direction versus another, is quite telling, and an increased frequency of record highs is expected under our current climate warming scenario.  If we plotted all of the observed high temperatures for a given date at a given site, we would expect a normal, or Bell shaped, distribution, with the majority of days clustered around the daily mean temperature (blue series on chart).  Record temperatures would be represented by the tails of this curve and would occur infrequently.  If we shift the distribution just a few degrees to the right, as is expected under a climate warming scenario, record high temperatures under the old scenario become much more common (red series on chart).  The chart is purposefully exaggerated for effect, but the unfortunately, depicts the pattern of record temperatures over recent years at MSP quite well.