Just before 3 a.m. on the morning of Oct. 31, a stream gauge installed alongside Onion Creek near Twin Creek Bridge in Manchaca, just south of Austin, failed.
Maintained by the U.S. Geological Survey, the sensor beams out information about the water level every 15 minutes. According to its final reading, Onion Creek was already far above its regular flow level of five feet. It had climbed 22 feet in the previous three hours, and was still rising rapidly.
Ninety minutes after the sensor stopped sending information, the city received the first flood-related call for help from neighbors whose homes were already filling with water two miles downstream. The emergency system was soon overwhelmed with pleas for assistance. Residents later complained they couldn’t get through to operators.
It wasn’t until several minutes after 7 a.m., however — nearly three hours after the 911 switchboard began lighting up — that Austin Fire Department officers authorized the first “Reverse 911” call to alert residents to the danger already swirling around them. Spanish language calls didn’t go out for an additional 2½ hours, even though the area boasts a large immigrant population.
Onion Creek eventually crested at more than 40 feet, the highest level ever recorded. Thanks in part to the broken gauge, the speed at which the water rose caught many by surprise.
“When we thought we had two hours to respond based on our readings, we actually had just minutes,” Austin Police Chief Art Acevedo told a crowd of residents at a community meeting Tuesday evening. “I wished we’d had more warning,” added Stacey Scheffel, flood plain administrator for Travis County, which includes some Onion Creek neighborhoods.
Over the following days, nature’s havoc became clear: about 660 structures damaged or destroyed, at least five lives lost, three of them in the Onion Creek area. By the most basic measurements, however, officials say the response to an historic flood was successful.
“We got everybody out of those houses without serious injuries or death, and all of our responders, without injury or death, so for me that is pretty good,” said Harry Evans, chief of staff for the Austin Fire Department.
Still, what is less obvious is how human-made efforts might have contributed to what many residents have complained was a tardy and unnecessarily chaotic response. Officials are starting to turn their attention to what, if anything, went wrong during the early morning hours of Halloween day.
An early examination of the response shows that there will be ample opportunity for second-guessing and close questions, from why early weather reports weren’t more accurate, to responders’ dependence on fallible technology, the timing of warnings and equipment failures — including a search helicopter that sat idling on the pad during the response.
Detection system sophisticated
Central Texas’s propensity for flooding can be traced to a topographical rise that runs roughly parallel to Interstate 35 from Fort Worth to San Antonio; and then turns west, following U.S. 90 toward Del Rio. Undramatic from the ground, the Balcones fault zone uplift is nevertheless sufficiently high to trap or slow rainstorms that brew over the Gulf of Mexico and churn their way north.
As rain falls from the stalled clouds and collects on the ground, the lift and tilt of the Hill Country sends it rushing toward Austin. With only a thin layer of soil to absorb the water, the result can be sudden torrents that quickly overwhelm local stream channels.
The phenomenon has given the area the label Flash Flood Alley. Texas has more sudden floods than any place in the country, and Austin sees its share. Disastrous floods have washed over the city with regularity, most recently in 1981, 1991, 1998 and 2001.
Often, the churning waters have claimed lives. Fifty-four people have died in the city’s floods since 1960 — 13 of them in the Memorial Day flood that hit Austin in 1981 after more than 10 inches of rain pounded the area in only four hours.
News reports at the time referred to the deluge as a “100-year flood,” yet that phrase is widely misunderstood. People tend to think it means such events will happen only once every century, but the phrase is more accurately an indication of probability; there is a 1 percent chance in a given year that a catastrophic flood will occur.
Measured over time, such odds become more ominous. “It means there is a 25 to 30 percent chance in a 30-year period,” said Roy Sedwick, executive director of the Texas Floodplain Management Association.
Because of its experience, the city’s response to its floods has evolved and improved over time. “The 1981 flood was a real wake-up call for the city of Austin,” said Sedwick. “We were not prepared to deal with that kind of event.”
Local and state governments initiated a series of measures to mitigate future floods, including storm runoff catchments to capture water and prevent flows from building into floods, as well as advances in technology that monitor weather and water movement in the watersheds to predict where and when floods will occur.
The city of Austin has one of the most sophisticated flood detection systems in the state — and, therefore, the country, Sedwick said. In 2010, it installed the Flood Early Warning System, designed to gather real-time information collected from water gauges and other sensors to model flood behavior in real time. About three-quarters of all Texas flooding fatalities deaths involve vehicles, and other sensors alert officials to when local low-water crossings begin to become submerged, automatically illuminating blinking warning lights in some areas.
“Austin has done just about everything they can do,” Sedwick said.
Flood plain spreads; maps outdated
Yet gaps remain. As much as technology has helped, in some ways it has also outpaced official efforts to harness it.
The Reverse 911 system still relies primarily on landlines, whose use is falling rapidly, to warn residents. Emergency responders have started using cellphone numbers to contact residents, but citizens must register to get on that list. Not many have; while Reverse 911 is connected to about a million landlines in the region, its cellphone registry has only about 68,400 numbers, said Ed Schaefer, homeland security director for the Capital Area Council of Governments, which operates the system.
Areas considered prone to flooding also have continued to grow. The first federal flood map of the Onion Creek area near William Cannon and Pleasant Valley Road in Southeast Austin, from 1981, shows the identified flood plain snaking tightly on either side of the channel like a band of pipe insulation. (The early maps were also inaccurate in places and not very useful, Travis County’s Scheffel said: “It was hard to tell if properties were in or out of the flood plain.”)
A decade later, after the Army Corps of Engineers redrew the maps, the new flood plain spread dramatically farther way from the creek channel, like an ink stain seeping across a paper towel. The 1997 map added large swaths of land to the area considered prone to flooding. A 2001 update added still more.
One reason is that the updated calculations of historic rainfall data added a whopping 10 feet to the potential water elevations that could result from drenching rains. The higher level pushed the identified flood zone farther from the creek bed.
Yet even the latest maps become outdated with new and future development. Although storm runoff systems are supposed to account for new building projects, experts say new roofs, driveways, streets and parking lots still reduce the ground’s ability to soak up rainfall.
It “means more water is going to run off into the stream,” and will do so faster, said Kevin Shunk, Austin’s flood plain administrator. That’s true for the hard-hit Onion Creek area, which lies downstream from Hays County, home to some of the fastest-growing communities in the country.
“Maps today are fairly accurate,” Sedwick said. “But they don’t show every area that might flood.” In a given year, he said, about a third of the claims made to the National Flood Insurance Program come from people whose property lies outside a flood plain. Although her evaluation of the Halloween flood isn’t final, Scheffel said typically about 20 percent of Travis County homes damaged by rising water are outside a mapped flood plain.
The expanding flood plain isn’t just a line on a map. It is used to decide where homes can be built out of harm’s way. Around Onion Creek, that has meant that as many as 500 homes built in the 1970s, in an area initially considered safe, are now squarely in the high water’s path. In the area around William Cannon and Pleasant Valley Road, “the entire neighborhood is now in the flood plain,” Shunk said.
After the 1998 flood inundated the area, the city and the Army Corps of Engineers concluded the best way to prevent future danger was to buy out nearly 500 vulnerable homes. Since then, using city and federal money, the program has purchased and demolished 322 of the structures.
More than 160 of the homes inside the danger zone remain; there has been no money available to cover the anticipated $33 million it would cost to buy them. On Thursday, however, city officials announced their intention to purchase 115 homes damaged on Halloween day.
Effect of gauge failures debated
Flood predictions begin with weather information, and forecasters predicted heavy rain potential in the days leading up to the Halloween storm. A flash flood watch was posted at 4:11 a.m. on Oct. 30 — more than 15 hours before the heavy rains arrived in Travis County. The watches were converted to more serious “warnings” by that evening.
Yet some forecasters say they were still caught off guard by the actual amount of rain that fell in the region. “What we did not expect was to have 14 inches of rain,” said Paul Yura, warning coordination meteorologist for the National Weather Service in New Braunfels. “That was not something that was forecast.”
While hurricanes or severe tropical storms can be studied and monitored for days in advance of an approaching storm, experts concede the impact of lesser events is more difficult to predict. Forecasting improvements in recent decades allow meteorologists to warn the public of general flood potentials, but Yura said it can still be a guessing game to pinpoint exactly where and how much rain will fall.
Yura said the Halloween storms posed a particular forecasting challenge as they pounded the same areas repeatedly, causing more water to collect and begin running off. “Obviously, we can learn from this event, but a lot of it is technology — trying to develop models about where that heavy rain will develop,” Yura said.
The impact of the storms was heightened by the ground already being saturated from a soaking rain two weeks earlier. As the rain continued to fall, the two circumstances together led to more water rushing to streambeds.
To help anticipate where and how to respond to rising floodwaters, emergency managers rely on data relayed to them by a series of stream and rain gauges. The U.S. Geological Survey maintains half of the 55 total stream gauges, said Robert Joseph, director of the survey’s Texas Water Science Center.
Yet, he said, the sensors typically are installed to measure common flows, not historic events. Installed on the bridge over Onion Creek, for example, the battery- and solar-powered Twin Creek gauge failed when water overtopped the structure. At times during the flood, the bridge was under 5 feet of water.
Near U.S. 183, another Geological Survey Onion Creek stream gauge failed Halloween morning; Joseph said his agency is still trying to determine why. According to recording data, the sensor stopped sending information at about 6:15 a.m. At the time, the water level was at 21 feet and rising fast. By the time gauge resumed sending data, 2½ hours later, the creek had nearly reached 40 feet. (The Twin Creek gauge didn’t resume working until the afternoon of Nov. 1.)
Officials disagree on how crucial the data gaps were to the rescue response. Despite the complaints from Acevedo and Scheffel, Shunk, who coordinates local flood response, said that emergency officials collect such gauge information early on, but it is heavily supplemented from on-the-ground reports arriving from on-site workers. As a result, he said, the early morning gauge failures didn’t have a significant impact on how city workers responded to the rising waters.
Joseph added his federal agency was well aware of the rapidly rising waters and the potential for gauge failure early on. “So we had someone on-site (at the 183 location) by 6:30” taking manual measurements and texting them to emergency planners, he said. Yet the bridge was downstream of the Onion Creek neighborhoods already taking on water; Joseph said nobody was able to get to the Twin Creek location to measure flows just upstream.
Alerts sent in Spanish delayed
Also unclear is how well the city’s various warning systems worked. Officials said they received their first flood-related emergency call around 4:30 a.m. By 9 a.m., 911 operators had received about 1,200 calls — four times the amount they typically get during that period.
Although some residents complained about a slow response, the fire department’s Evans said crews had been responding to calls throughout Travis County all night and had mobilized in the Onion Creek area by 4:30 a.m. Yet Evans added the department didn’t issue the first order to use the city’s Reverse 911 system, which robo-calls homes in a designated area, until 7:09 a.m. and again at 7:19 a.m.
A little more than two hours later, he ordered another round of calls to additional homes. That was the first time the alerts went out in Spanish. Demographic data from the 2010 U.S. Census shows that in the tracts where the worst flooding occurred, just over half of the residents are Hispanic. When the area directly surrounding those tracts is included, the number rises to two-thirds.
Evans said about 3,000 homes in all were called, mostly in the neighborhoods near Onion Creek and Interstate 35. In retrospect, he said, he should have ordered the Spanish language warnings earlier. “Dealing with all the things we had, we probably should have done Spanish, but at the same time, we were making 100 decisions a minute,” he said.
Evans said a review of the timing of the emergency calls will be part of the city’s overall performance review of its response. “We want to know how we can do better, and that may take a while for us to figure out,” he said. “But we will.”
First responders also suffered at least one major equipment glitch. The city of Austin and Travis County combined own seven helicopters. However, at the height of the storm response only four were in the air.
One each from the city and county were down for maintenance. “The reason we have three is that it takes three to keep two in service,” said Casey Ping, Travis County’s STAR Flight program director.
Two of the helicopters were limited in their ability to help. One of the county choppers was equipped for firefighting, not water rescues. The city’s new $3.75 million police chopper, meanwhile, unveiled to the public in September, has limited ability to engage in swift-water rescues, officials said. Police used that helicopter to provide aerial reports, including where they saw victims stuck atop cars or homes, and to direct emergency crews on the ground.
Yet another chopper sat on the tarmac at Austin-Bergstrom International Airport, blades spinning, for more than 30 minutes. But it was never granted takeoff clearance because of communication issues with air traffic controllers, officials said. FAA officials said the tower at the airport lost all radio communication for 15 minutes, but they weren’t contacted by the police helicopter pilot through some other ways, including the telephone.
“As a general rule, we give priority to law enforcement aircraft, particularly during emergencies,” said FAA spokesman Lynn Lunsford.
Cmdr. Nick Wright, who supervises the police department special operations unit, said that, after sitting on the ground for 30 minutes, the crew shut the engine off and went inside. They were then ordered home to rest and returned hours later for an overnight air patrol shift.
“They were here, they were ready to go, and we knew they were needed up there,” Wright said. “But we weren’t able to fulfill that.”
TODAY IN THE STATESMAN,
MONDAY ON KVUE
Reporter Tony Plohetski spent several days examining the emergency response through the use of helicopters. Watch his report Monday on KVUE News at 6 p.m.
HOW TO HELP
Several agencies in the Austin area are collecting donations of items and money for flood victims:
Austin Disaster Relief Network
Collecting financial donations for gift cards, shelter, housing, rebuilding, transportation and other needs. To donate, visit www.adrntx.org or mail to Austin Disaster Relief Network (ADRN), P.O. Box 3817, Cedar Park, TX 78630.
The network also is organizing volunteers to help with the cleanup. Volunteer at www.adrntx.org/index.php/central-tx-flood-news or call 512-331-2600 for more information.
American Red Cross of Central Texas
Collecting financial donations. Visit www.redcross.org/tx/austin, call 512-928-4271 or mail to 2218 Pershing Drive, Austin, TX 78723
Salvation Army of Central Texas
Collecting financial donations. www.salvationarmytexas.org/austin
River City Youth Foundation
Seeking clothing for children and adults, nonperishable food, water, diapers, toiletries, baby formula. Dove Springs Recreation Center, 5801 Ainez Drive
United Way for Greater Austin
Has launched a text-to-give campaign and online donation page for flood victims. To make a $10 donation to support long-term recovery efforts for flood victims, text UWATX to 85944, or, to give online, visit uwatx.org/flood.
Rainfall map online
Find links to a detailed map of Oct. 30-31 rainfall totals, creek levels and previous coverage of the floods with this story at mystatesman.com.
This story reflects the American-Statesman’s focus on broader public safety issues raised by specific events. Previous investigations examined gaps in wildfire defenses for homes on Central Texas urban fringes, toppling light poles at school sports fields and splintering glass panels falling from downtown hotel balconies.