Friday, July 20, 2018

Hourly Rainfall History

Recently I have spent a good bit of time digging into the history of hourly precipitation data from Fairbanks, with a focus on the warm season months of May through September.  NOAA's hourly precipitation data set ostensibly covers 1949-2011, although there is inexplicably a long gap from 1952-1962 in Fairbanks; so I have filled this in by transcribing numbers from the historical LCD forms (a somewhat painstaking effort).  I also brought the data up to the present using hourly ASOS observations.

Now that I'm in possession of a reasonably complete data set, there are a number of questions that can be answered.  For instance, is there evidence that short-duration heavy rain events have become more common in Fairbanks?  We might expect this in a warming climate with increasing moisture content.  The chart below shows annual counts of days (May through September) with at least 0.5" of rain within a two-hour interval.  I used a two-hour period so that the analysis captures events that crossed the top of the hour, and the chart starts in 1956 because we have essentially complete data since then.  (But see a note at the bottom on the 1997 event.)

A striking result is the remarkable cluster of heavy rain events in 2005-2010: there were 8 separate events in 6 years, but the previous 50 years had seen only 10 such days in total.  If we assume a Poisson distribution based on the first 50 years, it is extremely unlikely (p<0.0001) that 8 events would occur in 2005-2010 by random chance, so "something" in the climate was different (perhaps a remote climate influence).  Interestingly, however, the last 7 years have seen only one of these heavy rain events.

The overall 62-year mean frequency of these events is 0.31/year, or a return period of 3.3 years, and this is nicely consistent with the NOAA precipitation atlas (

For more analysis on heavy precipitation events in Fairbanks, but at a daily time scale, see this post from last October:

And regarding the 1997 event that is included in the chart above: 0.75" of rain was reported for the daily total on June 8, 1997, but the hourly totals were not reported owing to a rain gauge malfunction.  There's no way of knowing for sure, then, whether 0.5" fell in two hours at the airport, but it turns out this event was a well-documented severe storm and there's little doubt that it qualifies for the analysis here.

Here's a write-up on the storm by one of the NWS staff in Fairbanks:

Looking at the hourly METAR reports, a key observation was made at 8:58pm on the 8th, stating that hail and rain had begun at 8:13 along with a peak wind of 50 knots (perhaps the rain gauge was blown over!).  Apparently hail fell for 18 minutes and rain for another 22 minutes.  None of the other hourly observations from that day reported more than light rain, so it appears that most of the 0.75" fell in less than one hour.

It just so happens that the Pedro Dome radar was offline for most of the day, but a scan from 8:55pm showed a strong storm southwest of Fairbanks - see below (image courtesy of  As the storm was reported to be moving south in the METAR data, this is almost certainly the cell that brought chaos to the city.

1 comment:

  1. As your radar image suggests, these heavy precip events are spotty and storm totals represented by streaks. Maybe that “something” could be local - some disturbance (fire, etc.) that set up a convective engine upstream from the airport station causing repetitive precip streaks across the station. The right persistent disturbance could upset the independence assumption of the poisson test. Hard to test though.