Thursday, September 17, 2015

How Unusual Was the Storm That Produced the Hildale Flood?

A critical question pertaining to the meteorology of the Hildale flood is how unusual was the storm that triggered it.

I'm going to provide some preliminary analysis here based on radar precipitation estimates.  I want to emphasize preliminary because there are a host of issues regarding radar precipitation estimates that will ultimately need to be explored in the coming weeks.

The image below shows the estimated storm-total precipitation in the Hildale area for the period ending at 2301 UTC (5:01 PM MDT).  All but about 0.1" of this total was produced during the flood period from 2000–2300 UTC (2–5 PM MDT).  The black square is located in the upper reaches of Maxwell Canyon along the axis of maximum precipitation and just north of Hildale.  The greatest estimated precipitation accumulation is actually southwest of Hildale.  Amounts in the area generating the flood are around 1.5 inches.

The bar chart below shows the storm-total estimated precipitation on a scan-by-scan basis.  There are two periods of heavy precipitation.  The first from 2007–2029 UTC (2:07–2:29 PM), which produced 0.88 inches of estimated precipitation, the second from 2218–2237 UTC (4:18–4:37 PM), which produced 0.52 inches of estimated precipitation.  In the case of the latter, most of that precipitation fell in a 10 minute period from 2227–2237 UTC (4:27–4:37 PM).

From this we can extract the maximum precipitation rates, with some adjustments when necessary to account for the mismatch with the time between radar scans:

5 min: 0.4 inches (2227–2232 UTC/4:27–4:32 PM)
10 min: 0.53 inches (2007–2016 UTC/2:07–2:16 PM), adjusted from 9 min accumulation of 0.477 inches
15 min: 0.72 inches (2011-2025 UTC/2:11–2:25 PM), adjusted from 14 min accumulation of .67 inches
30 min: 1.21 inches (2007–2029 UTC/2:07–2:29 PM), adjusted from 22 min accumulation of .885 inches

More difficult is adjusting the total precipitation.  Collectively, the total precipitation is 1.4 inch over a period of 3 h and 30 min.  Thus, I'll use 1.2 inches and 1.4 inches as a range for the 3-hour accumulation.

We can then compare these amounts to return intervals for the Hildale area that I obtained from the NOAA Precipitation Data Frequency Server.  WIthin the ranges of statistical confidence, these amounts have return intervals of 10-50 years depending on duration period.  Keep in mind that return interval is a misleading term.  For example, a 1 in 25 year accumulation doesn't happen every 25 years, but instead has an estimated likelihood of occurring in any given year of 1 in 25.

These numbers are a bit lower than you may have heard in the news where the rainfall was described a 100 year event by the National Weather Service.  They may have good reason for a rarer event estimate. First, there are a number of algorithms to estimate precipitation using radar data and they may be using a different approach than the one I am using above that they feel is more appropriate.    Second, the data I'm using above is somewhat lower resolution than what they use in real-time.  It could be that a more intense maximum would be evident in higher resolution data.  Finally, they could be using a different approach for calculating return intervals as this is another source of uncertainty.  These are issues to address in the coming weeks.

Some additional food for thought.

First, while we can calculate return intervals for precipitation rates, the real question is how anomalous was the flooding that occurred in Hildale.  That's a more difficult question to answer given the paucity of streamflow data at that location and the rare nature of these events.  This might be a topic for a future post, but I'll probably move on to other things.

Second, I've focussed on the peak precipitation rates above, but the event featured two periods of high intensity separated by a short period.  One can use a 3-hour total accumulation as I've done above to account for this, but how unusual is it to have to short periods of intense precipitation occurring within 3 hours and how does that ultimately affect the runoff?

Third, it is important to remember that the return intervals (or better put the odds of an event in any given year) are for a point.  Monsoon convection is highly localized.  Precipitation rates like those above might be rare at any given location, but they are quite likely to happen somewhere in southern Utah in any given monsoon season.  For this reason, I don't consider the Hildale precipitation intensities to be all that unusual, even if they led to an unusual flood by recent human experience in the Hildale area.


  1. Very sad. All 7 canyoneer bodies have now been recovered from the Keyhole canyon drainage, based on a timeline reported by the NPS, it appears it was the second wave of precipitation that caught the canyoneers.

  2. Keyhole canyon is very close to Hilldale, was it the same cell that caused both floods?

    1. It appears that Keyhole would have had only light precipitation from the first cell, which was what meteorologists call a right mover and moved strongly across the flow toward the east with just some blowoff toward Keyhole. In the 2nd period of precipitation, the cell that moved across Hildale moved northward and appears to have tracked over Keyhole. if the commenter above is correct about the timeline, it probably produced the Keyhole flood.

      I'm still unclear on if the victims in Hildale were caught by flooding produced in the first period of precipitation or with the addition of the second period of precipitation.

  3. A good visual analysis of the precipitation that caused the floods can be found here: