Wednesday, October 7, 2015

The recurrence interval - a PR disaster for meteorology

The terrible floods that occurred recently in South Carolina have triggered the usual brouhaha over the notion of the "recurrence interval", with the SC event having been said by some to be even more rare than a "1000-year" event.  The general public mostly takes this to mean at least 1000 years should pass between such events, so it seems weather disasters of this sort must be "freak" occurrences, demanding some sort of special explanation.  Unfortunately, a ready "explanation" for this event has been that it was caused by anthropogenic (i.e., human-caused) global warming (AGW).  Even some people who should know better have jumped onto the AGW bandwagon with regard to this event.  I'll return to that shortly, but first I want to try to dispel some of the misunderstandings associated with recurrence intervals.

Atmospheric events are not "periodic" - that is, they don't occur at regular intervals in time.  If that were so, weather forecasting would be a heckuva lot easier and considerably more accurate.  Hence, the perception that a recurrence interval is based on some periodic atmospheric behavior is simply a misunderstanding of the term's meaning.  Forecasting is difficult, in part, precisely because weather is most definitely not periodic!

Recurrence intervals generally are calculated by fitting some sort of statistical distribution (e.g., a Poisson distribution) to the existing record of events.  It doesn't take much knowledge to realize that we don't have a record of heavy rainfall events longer than about 200 years (in the USA), so how can we come up with a meaningful definition of a "1000 year" event?  The answer is simple - we can't. To do so is an exercise in extrapolation, and extrapolation is well-known to be a risky thing to do.  If we had 10 000 years of data for every location in the USA, we might well be able to have a plausible definition for a 1000-year event at all those places, but such data simply don't exist.  What we can say, from our knowledge of the occurrences we have observed, is that the SC floods were an event that is outside of prior experience (in SC).  This doesn't mean it's a "freak event" for which some exotic explanation must be offered.  Curiously, given that weather events happen when the ingredients for such events are brought together, the chances for a similar event to occur soon after one event has already happened are relatively high.  If the weather brought those ingredients together on one day, there's an increased probability that it might happen again soon after.

For flash flooding, in particular, it's not at all uncommon for heavy rain to occur on two (or more) days in succession.  The first event saturates the soil, creating the hydrologic conditions that make it possible for a flash flood on the second day.  This has happened many times in the history of flash floods around the world.  It seems silly to refer to an event as a 100-year (or 1000-year) event when it happens on consecutive days!  It's quite acceptable to use recurrence interval terminology in the context of communication among scientists, because (hopefully) they understand what the term means.

Fortunately, the notion of recurrence intervals is almost never mentioned in the context of major tornado outbreaks or high-impact tropical cyclone landfalls.  Most people (at least in the USA) seem to understand that these are not "freak" events, but rather occur at irregular intervals when their ingredients come together.  They occur somewhere in the USA every few decades or so ... frequently enough that the public is reminded of the possibility that such things can happen.  Major rainfall events also occur at irregular intervals, and often enough that people should get the right message:  really big events can happen somewhere in any given year.  But for some reason I can't explain, the reference to recurrence intervals is common with respect to heavy rains, and so this issue comes up over and over again.  It's a public relations nightmare that we should stop inflicting on ourselves.  In scientific papers, discussion of recurrence intervals is more acceptable, but making public statements about them is just confusing and makes us look silly.

Finally ... was this event "caused" by AGW?  In a word ... NO!  The event happened because the ingredients for a flash flood-producing rainfall event were brought together.  Nothing particularly exotic was required and those ingredients were not, on their own, remarkable or unusual.  It's always rare for really heavy rainfalls to happen in any given location and this was an example of a somewhat atypical weather pattern in which the flash flood-producing rainfall ingredients to be brought together.  The threat of an additional event - landfall of Hurricane Joaquin - was never realized, fortunately.  Nevertheless, the numerical weather prediction models were quite accurate in anticipating the heavy rainfall event in SC days in advance; the forecasts for heavy rains were quite good, as a result. 

A somewhat more nuanced description of the role of AGW in this event is that AGW is thought by many climate scientists to make it likely that extreme flash flood events will become more frequent.  This event is consistent with that prediction but, on its own, doesn't provide "proof" that AGW was a contributing factor.  There are indeed scientific studies that suggest that heavy rainfall events are becoming more frequent.  Thus, it's appropriate to say that this SC flood case is one more piece of evidence to that effect.  But to say that this event was caused by AGW is simply not scientifically acceptable.