22 September 2013

Revisiting the "Consistent With" Canard

Over the past few days I've been engaged with a lively debate with a colleague over whether it is meaningful to proclaim that the extreme rainfall observed in Colorado several weeks ago are "consistent with" predictions of more intense rainfall associated with human-caused climate change.

Long-time readers will know that I believe the use of the phrase "consistent with" in this context is a canard and devoid of substance. Here is an example using exactly such a construction related to the Colorado floods.

The analogy I'd suggest is a 52-card deck stacked with an extra ace. After being dealt a blackjack (i.e., an ace and a face card) it would indeed be appropriate to proclaim that "such a hand is consistent with expectations for hands dealt from this stacked deck."  Of course, being deal a 5 and a 2 would also be "consistent with" the stacked deck. For that matter, being dealt a blackjack would also be "consistent with" hands coming from an unstacked deck.

Motivated by this discussion, I downloaded precipitation records from NOAA for Boulder (here in .dat), which covers a period from May 1, 1897 to August 31, 2013.

In the first half of that time period, covering 58 years or so, Boulder experienced 24 days with measured rainfall of 2 inches or more. In the second half of that period (also 58 years or so) Boulder experienced 20 days with rainfall of 2 inches or more.

How about really extreme, say 4 inches or more?  There is only one data point in the record at that level, July 31, 1919. Now, there is a second.

So what is this data "consistent with"? Pretty much anything, and that is the point.

Advice: just don't use "consistent with" in stretching to link particular extremes with human-caused climate change. It is pure spin.

Further, don't try associating a single event with human-caused climate change -- whether it is an extreme snowstorm in Washington, DC (leading an enthusiastic Senator to build an igloo and name it "Al Gore's New Home," below) or rainfall in Colorado. If you want to detect changes in climate you have to look at long-term records. In Boulder at least, the long-term records do not indicate much change at all in the incidence of extreme rainfall exceeding 2 inches in one day.