Record Cold? Not so fast

Recent news headlines consuming much of the US have spouted that it is “the coldest in decades” and in fact “colder than Mars” due to a sinister and seemingly novel fandangled “Polar Vortex”.

It certainly has been way below normal across the eastern 2/3rd of the country due to a highly amplified trough that brought cold air into the deep south (and warm air far north). This unusually cold air mass coincides with the climatological coldest time of the year for much of the US, setting the stage for all time records, right? While plenty of daily low temperature records were broken, zero all time low records were set in the past week in the US (through Jan 6th). Moreover, in 2013 all-time record high temperatures outpaced all-time record low temperatures at a 7:1 clip.

macfig 2014-01-07 at 8.13.30 PM

Despite all the hoopla in the media, the only ALL TIME RECORD LOW temperature set for stations that I could find was in Van Buren, Maine (records back to 1964) located just about as far north and east in Maine as you can get.

Nonetheless, it has been brutally cold across the eastern 2/3rds of the US, especially relative to winters of the past decade plus.  Interestingly enough, we’ve all surely heard about what is being referred to as a hiatus or pause in the increase in global temperatures since 1998, even though science has identified processes that help explain the phenomenon.

Have all time record low temperatures waned in the past decade?

It has been widely reported daily high temperature records have outpaced daily low temperature records across the US in recent decades.  What about for all-time records such as the coldest nighttime low for any day of the year?

For this exercise I took the US Historical Climate Network stations as they are composed of a set of high-quality long-term stations across the US and have been used in countless studies. Most of these stations are located in rural areas, thus free from some of the non-climatic effects on observed temperature (like urban heat islands).

For fair play, I restricted the data to 1920-2012 since the number of stations with records dating back prior to 1920 drops off rapidly.  I also tossed out any stations that were missing more than 5% of their data, leaving a total of 639 stations dispersed across the US, albeit with much more limited coverage in the wide open western US and a particularly high concentration in the Corn Belt of the country.

The game is pretty simple: using the 93-year record I found the year when the absolute highest and lowest temperatures were set. In the case of tied records, I only counted the last year. If temperature extremes were completely random, we’d expect an average of 11% of stations to set an absolute record low, for example, in any given decade.

Below I plot the year during which the absolute minimum and maximum of daily low temperatures was last set for the 1920-2012 time period. For emphasis, records set after 1998 are shown by large symbols.

Screen Shot 2014-01-07 at 3.44.46 PM

Aside from a large cluster of stations in Iowa having record cold temperatures in January 2009, hardly any stations set their coldest overnight temperature in the last 14 years. Conversely, around 30% of the stations, with a pretty widespread coverage, had their warmest night ever in the last 14 years. Clearly, far more than might be expected if were left to chance. So much for that lack of warming since 1998.

The story for daytime high temperatures is similar, with 15% of stations setting or tying their all-time record high since 1998, particularly in the western US, compared to around 6% of stations setting their all time lowest daytime high record. Another lopsided comparison.

The NCDC has a nice interface for examining temperature records for these stations.

How do we expect all-time record temperatures to evolve in a warming climate?

Moving away from real-world observations and into the world of climate modeling we can run a similar experiment, here substituting “stations” for 1-degree by 1-degree grid cells covering the contiguous US. I looked at simulated daily maximum and minimum temperatures from 20 different models participating in the 5th Coupled Model Intercomparison Project and considered both a historical forcing run (1950-2005, though not meant to be the same as what we observed during these years) and a future run with basically unmitigated climate change through 2100 (also known as RCP8.5).

Rather than show the results for each model, I pooled the year (from 1950-2100) during which the record low maximum and minimum temperature were set for each of the pixels spanning the lower-48. The results are displayed below, binned by decade.

recordtmin

Some immediate points:

(1) All time record low temperatures can still be set in a warming climate, albeit at diminishing numbers. That’s natural variability for you.

(2) Even in the modeling world, there is a noted reduction in the number of all time low temperature records in the 1990s and 2000s

The story for all time record highs looks substantially different. With models projecting increases in temperature over the continental US during the summer months approaching 10 degrees F, most of the records get set during the last two decades of the 21st century in the simulation.

So enjoy the cold!