Apocalypse Now

Yesterday the rain guage at Onion Creek/Hwy.183 at Austin, TX, operated by the Lower Colorado River Authority (LCRA), recorded 14.39 inches of rain (365.5 mm). In one day.

A Texas man, whose car was overcome by flood, escaped by jumping out of his car and climbing a tree. You can see a report here:

That amount of rainfall in one day, at that location, is staggering. The question is, how staggering?

Unfortunately I haven’t found historical data for that specific location. But I have found data, from the Global Historical Climate Network (GHCN), for nearby stations in Austin, TX. One should be advised that they don’t represent the exact location measured by LCRA, and it’s possible that there’s something unusual and unique about the Onion Creek/Hwy.183 location that induces storms to drop excessive rainfall amounts at that exact spot. But … what are the odds?

Let’s look at daily precipitation amounts for Austin’s Camp Mabry:


These data cover Jun. 1, 1938, through Oct. 8, 2015. Within that time range the maximum daily rainfall was 191.8 mm (7.55 inches) on Nov. 15, 2001. The amount recorded at Onion Creek/Hwy.183 yesterday was nearly twice as much.

The “survival function” of daily rainfall in Austin, for large values, very nearly follows an exponential distribution. If we fit an exponential distribution to the tail of the distribution, we can estimate the likelihood of getting 14.39 inches in a single day. It turns out to be a once-in-170,000-years event. But we should really fit a generalized Pareto distribution (according to extreme value theory), in which case 14.39 inches of rain in one day would only be a once-in-40,000-years event.

As I say, that’s applying the statistics for the nearby Austin-Camp Mabry station to the observation at Onion Hill/Hwy.183. And it’s possible the statistics don’t really translate from one location to the other, no matter how nearby they may be.

But any way you look at it, the rainfall measured in Onion Hill, TX, near Austin, was apocalyptic. The only thing less likely, is that Texas congressman Lamar Smith or Texas senator Ted Cruz will “get it.”

25 responses to “Apocalypse Now

  1. This looks like an example of a “probable maximum precipitation” type event, the kind of event that is used to design the most critical infrastructure (like dams) … although I think in Texas a true PMP event would be much bigger. In the Canadian Orairies a 24 hour PMP is about 500 mm. In the early 2000s there was a 400+mm event in Saskatchewan.

    Now the odds of such a thing happening at any particular location are, of course, very long. But the odds of them happening somewhere are much higher.

    Although I know it is difficult thing to do, what would be most interesting is an alaysis of trends in PMP-ish events (> 200 mm in a day, for example) for a region (Texas, perhaps).

    I have seen analyses that looked at dew point temperature records as a proxy for potential PMP magnitude, but never seen it applied to analyze a PMP risk trend … Just planting ideas if you have the time.

    • I think this is the Saskatchewan storm that Ernst is referring to:

      The Vanguard Torrential Storm
      DOI: 10.4296/cwrj2702213

      Fraser G. Hunter, David B. Donald, Brian N. Johnson, Wayne D. Hyde, Ron F. Hopkinson, John M. Hanesiak, Markus O.B. Kellerhals & Bart W. Oegema

      Canadian Water Resources Journal / Revue canadienne des ressources hydriques, Volume 27, Issue 2, 2002



      On July 3, 2000, up to 375 mm of rain fell in eight hours from an intense and slow moving thunderstorm complex, inundating the Vanguard area of southern Saskatchewan. More than 100 mm of rain fell over an area of 1700 km2, making this the largest eight-hour storm ever documented on the Canadian prairies. Air-parcel back trajectories indicated that warm air from mid-continent North America, associated with a low-level jet stream, provided moisture for the storm. The storm was centred over the Notukeu Creek drainage basin. Discharge was 0.05 m3/s on July 2, but on July 4, peaked at 267 m3/s. The storm generated 353 million m3 of rain, of which an estimated 38% was discharged to Notukeu Creek and 62% or 219 million m3 was retained on the landscape.

    • Despite that there are long-standing (and pretty silly) published methods, it’s not possible to estimate PMP with any confidence from historical rainfall data; the record is always far too short.

      Modern practice prefers estimates from simple maximal storm models — e.g. of generic thunderstorms and tropical cyclones. Short duration (up to about 6 hour) PMPs for Texas latitudes would be estimated with a thunderstorm model. The principal parameter there is commonly an “extreme 24 hour persisting dew point” … which I guess is what you’re getting at. On our “water world”, I’d expect that would generally track somewhere close to the rise in extreme mean daily temperature. Yes, some numbers would be interesting — they have direct infrastructure cost impacts.

      BTW, Australian practice assigns notional probabilities to PMPs to enable interpolation of design values between them and historical extreme rainfall distribution fits like Tamino’s. Typical notional probabilities for 6 hour thunderstorm PMPs are in the 1:10⁶ – 1:10⁷ annual exceedence range — i.e. way way rarer than this Texas event.

    • Interesting… I wondered how this compared with the 2013 Toronto flood–a rather expensive event that many will remember. The headline number on that storm was just 126 mm in 8 hours.

      Still, the economic effects will continue to linger for quite some time in the form of elevated insurance premiums:


  2. Being a resident of Austin, I can tell you there is an official weather-data site at the Austin-Bergstrom Int’l Airport (ABIA), formerly Bergstrom Air Force Base. ABIA is much closer geographically to the LCRA’s Onion Creek/Hwy 183 site than is the Camp Mabry site. If you can acquire the ABIA rainfall data, that may be better for inferences re Onion Creek.

    Thanks for your work.

  3. michael sweet

    Many of the other high rainfall events would have been associated with Tropical Storms. Since no such storm was associated with this event, shouldn’t an adjustment be made to the likelyhood of occurance? Perhaps you could delete the previous events associated with a tropical storm and than calculate the probability of the rainfall. It would obviously be less likely.

    • David B. Benson

      My understanding is that this is the leftovers from the tropical cyclone which made landfall in Mexico.

      • michael sweet

        From Jeff Masters http://www.wunderground.com/blog/JeffMasters/category-4-chapala-on-its-way-to-yemen-texas-gasping-after-more-recor
        “The absence of a tropical cyclone makes this event across central and south Texas all the more remarkable.”

        It is my understanding that the flooding in Texas started with a lot of rain that was not associated with a tropical storm. Then the Mexican storm went through but with only a moderate amount of rain. The current storm is also not a tropical storm. Texas often sees high rain during El Nino years. If they keep getting rain like this month it will be interesting to see how deniers explain it.

        People here and in the recent South Carolina flooding have attributed flooding to Tropical Storms when the flooding had started before the tropical storm was in the area (or after it was gone). These record rains seem to me to be just what scientists have described as problems caused by AGW. Attributing them to Tropical Storms delays recognition that they are AGW. Jeff Masters blog provides daily discriptions of heavy weather events as they develop.

      • I don’t think so: that was reported to have happened on the 26th, with remnants of the system already moving into the Southeast. This event happened four days later.


      • Michael, thanks for that interesting link.

        It’s striking that the same part of Texas hit back in May–some will recall that those floods were described as unprecedented, with one campground that had stood undisturbed on the Blanco River for decades being destroyed–was seriously affected again.

        “At 8:45 a.m., the Blanco River in Wimberley was at 17.6 feet and flood stage is 13 feet. At 26 feet, major flooding impacts several lowest homes on the Blanco River near Wimberley. At 2 p.m. on Friday, the Blanco River crested at 42.5 feet, which is a half-foot shy of the Memorial Day weekend floods.”


        Wonder what the stats on that have to say?

      • The remnants of Patricia passed through last Monday; this was a completely different system. That probably made the flooding a little worse in that the ground was still pretty saturated from the earlier storm. This felt a *lot* like the 1998 flood, which also occurred in October, with many of the same elements in place. Note that between the Memorial day floods and this past week, we got almost no rain; we were back in drought conditions (Bastrop and Smithville were on fire again). Then somebody turned the spigot on full blast.

        I’ve lived in or around Austin for 25 years, San Antonio and the surrounding area before that. This year has been *weird*. We’ve had freak weather before, but we haven’t gone from flood to drought back to flood again, at least not that I can remember, at at least not to this severity.

  4. By the way, Kerry Packer, the man whose car was swept away by Onion Creek, was eventually rescued via helicopter and is fine:


    Pre cell phone, he quite likely wouldn’t have been so fortunate.

  5. The excess rains that occurred over central and eastern Texas, both before and after the remains of Patricia went by, were due to a persistent low pressure trough that’s been generally stuck over the plains for a few weeks now. Most of the rain that fell with Patricia occurred over eastern Texas closer to Arkansas and Louisiana since that’s where the remains of that particular system went. These other events are due to a continuing flow of tropical moisture that’s being picked up by that stagnant low pressure trough and funneled across the southern plains, especially over central Texas. This is exactly what is expected with a warming atmosphere and the focus of the studies of sluggish jet stream flow and stuck blocking patterns done by Jennifer Francis and her team at Rutgers. Of course, the still building El Nino in the Pacific is contributing to this particular pattern at this particular time. But all the El Nino is probably doing is focusing this particular wet pattern over Texas. We can expect other areas to experience similar extreme patterns as time goes on.

  6. Camp Mabry received 15.00″ / 38.1 cm of rainfall on September 9, 1921

    [Response: Source?]

  7. The 1921 event was associated with a Cat. 1 Hurricane.

  8. George BAilley

    There is a reasonable probability of a significant rain event spanning the measurement boundary – so I prefer to do my stats looking at 2 day rainfall totals – rather than 1 day rainfall totals (assuming that the measurement interval is 24 hours).

  9. Weather events in a world of lower CO2 concentrations are not predictive of the weather in worlds at much higher CO2 concentrations.

    The high CO2 system is does not play by the same rules, and has not yet disclosed the rules.

  10. The principle problem associated with human-caused climate change, I think, is the velocity of the change. We seem to be changing the climate at a faster rate (perhaps 10 to 100 times faster) than is natural (albeit humans are natural) except for bolide collisions such as 66,000,000 years ago. Humans might be able to live in a Cretaceous or Eocene climate but not if we got there in a geologically brief span of time. Have to ever discussed this in any of your blog posts?

    thank you

    Tony Noerpel

  11. It turns out that 13.09″ of the precipitation at that gauge fell in a 3-hour period. That will probably turn out to be the most exceptional duration. The maximum 1-hour total was 7.02″.

    I’ve been asked if this is a record for the state. The answer is no: on May 31, 1935, at a ranch north of D’Hanis (about 120 miles from Austin), 22″-24″ fell in a 3-hour period. A couple of nearby ranches received 16″ in the same interval. This is one of the world’s all-time 3-hour rainfall totals. In the resulting flooding, Seco Creek belied its name and actually merged with the next creek to the east.

    • nielsengammon

      And the one-day record for Travis County is 16.02″, set in the 1921 event at the Hill’s Ranch COOP station a few miles from Austin-Bergstrom Airport.

  12. Am I the only one having trouble to watch the video? I get the message that it has been removed.