The Jarrell tornado of May 27, 1997

F5  Bell and Williamson counties, Texas · surveyed by NWS Austin/San Antonio and a federal assessment team

On a hot, nearly windless afternoon in central Texas, a storm complex did something storms are not supposed to do: it moved backwards, south-southwest against the wind, building new cells along a boundary for hours. Its sixth tornado died in a field beside Interstate 35 and its seventh formed in the same place, grew to three-quarters of a mile wide, and walked south-southwest at five to ten miles an hour into the Double Creek Estates subdivision on the west side of Jarrell. Twenty-seven of the subdivision's roughly 132 residents died. The houses were swept from their slabs so completely that surveyors had trouble finding the debris from the air, the pavement was stripped from the roads, and a herd of cattle was killed in a pasture. The Granger radar, twenty miles away, lost power two minutes before the tornado arrived.

A dark, wide tornado fills the sky beyond a peach-colored house and a tree, seen over a wooden backyard fence, with a birdhouse on a pole at right.
The tornado from Jarrell. A wedge, not a funnel, photographed from a back yard on May 27, 1997. Photo: Charlotte Vancil, via Wikimedia Commons, CC BY-SA 4.0.
Rating
F5 NIST argued F3 winds could explain the damage
Path length
7.6 mi 2.4 in Bell County, 5.2 in Williamson
Max width
¾ mi at Double Creek Estates
Touchdown
~3:15–3:25 pm CDT near Prairie Dell, Bell County
Forward speed
5–10 mph witness estimates; toward the south-southwest
Deaths
27 all in Double Creek Estates
Injuries
12 one serious
Damage
$40M 1997 USD
Radar
KGRK, KEWX KGRK failed at 3:38 pm
Warning lead
~10 min warned 3:30 pm, at the subdivision 3:40
Environment
CAPE >6,500 J/kg shear below the 25th percentile
Outbreak
20 tornadoes 30 deaths across the day

The setup

By every measure of instability the air over central Texas on May 27, 1997 was extreme, and by every measure of wind shear it was unremarkable. A research sounding launched near Calvert at 2:45 pm found convective available potential energy of about 4,900 J/kg; adjusted for the boundary layer near Temple, Adam Houston and Robert Wilhelmson put it above 6,500. Dewpoints were in the mid 70s. But upper winds were "consistently out of the west-southwest at less than 20 m/s throughout the day," the 0–6 km shear fell "below the 25th percentile" of every supercell climatology they compared it to, and nearly every measure of low-level helicity fell below the 10th. Their verdict on the environment: "only marginally favorable for supercells" and "unfavorable for significant tornadoes."[7]

Stephen Corfidi of the Storm Prediction Center, writing the year after, put it more simply: "The 27 May event was unusual in that it occurred in an environment of comparatively weak vertical shear." Surface-based CAPE exceeded 5,500 J/kg across nearly all of central and east Texas; the boundary-layer winds "were quite light." What made the day was the geometry of the boundaries. A cold front sagging south from north Texas met a dryline near Waco, and the two merged slowly southward through the afternoon.[9]

The storm that moved backwards

The storms formed near Waco around 12:30 pm and then propagated the wrong way. "The storm complex was found to travel against the mean environmental wind (a process referred to as back building) along a dryline near its intersection with a cold front and the storm's gust front." Individual cells drifted east with the weak flow; the complex as a whole moved south-southwest because new cells kept forming at its southern end. Houston and Wilhelmson traced sixteen of them: six born on the cold front, five on the storm's own gust front, five on the dryline. None of the dryline cells contributed to the Jarrell storm.[7] Jon Davies later described the motion as "more than 100 degrees to the right of the 0-6 km mean wind, making this an extreme case in terms of both storm motion and thermodynamic factors."[12]

Why the boundaries kept the storms alive is argued over. Corfidi thought a gravity wave from a decaying overnight storm system, moving southwest "in-step with the surface wave," was "critical in realizing and maintaining supercellular convection," and added that "another event of the same magnitude is not likely to occur for some time to come." Houston and Wilhelmson, with the full radar record in hand, could not find a pressure signature of the wave where the cells were forming and concluded it "does not appear that it played a role." What they found instead was the "zippering" of the cold front and dryline, the front overtaking the dryline "and penetrating into the most unstable air to the east," which put new cells farther south each time.[7][9]

The tornadoes that did damage all formed the same way. "Each of the six observed tornadoes that produced greater than F0 damage formed along the storm-generated gust front, not along preexisting boundaries." The three strongest, at Moody, Lake Belton and Jarrell, formed at the "gust front cusp" and were preceded by a mesocyclone aloft, at 1.5 to 3 km, eleven to eighteen minutes before touchdown. The Jarrell tornado itself formed "as a nearly stationary convective cell became collocated with a south-southwestward-moving low-level mesocyclone near the gust front cusp--a behavior that resembles the formation of nonsupercell tornadoes."[8]

The warning

At 2:15 pm the Austin/San Antonio office telephoned the Georgetown police and the Williamson County sheriff to say severe storms were moving south through Bell County and would arrive within an hour and a half. At 3:30 pm it issued a tornado warning for Williamson County: "AT 325 PM A TORNADIC THUNDERSTORM WAS LOCATED ABOUT 5 MILES WEST OF JARRAL MOVING SOUTHEAST AT 10 MPH. THIS STORM HAS HAD A HISTORY OF PRODUCING TORNADOES AND LARGE HAIL. THE CITY OF JARREL IS IN THE PATH OF THIS STORM." The town's name is misspelled twice.[1][5] The federal assessment counted the lead time: "This gave a lead time of at least 10 minutes for the town of Jarrell but no lead time for the northern portion of Williamson County," where the tornado was already on the ground.[1]

The warning reached people. "Almost everyone interviewed said they were aware of the tornado warnings. Most said they first learned of the warnings through commercial television. Others said they received telephone calls from relatives or friends about the approaching storms." Nobody mentioned NOAA Weather Radio. "Due to the slow movement and high visibility of the tornadoes, most of the residents interviewed said they watched the approach of the tornadoes prior to taking shelter. Most said they knew to go to the center of their houses, to avoid staying in mobile homes, and to seek shelter rather than trying to flee the tornadoes. These actions definitely saved lives. However, in the case of the F5 tornado at Jarrell, people were killed even though they took the appropriate safety measures."[1]

The assessment was harder on the office than on the warning. Its findings record that the forecaster-in-charge procedure "was not implemented effectively," that severe weather operations fell to "a meteorologist intern and a radar operator," and that staff had been "focusing upon the potential for a heavy rain episode" and were "concerned with getting the routine afternoon forecast package 'out on time.'" The Fort Worth office to the north, which had watched the storms come south all afternoon, issued ten tornado warnings between 1 and 5 pm.[1]

Chronology

Times are CDT and follow the federal assessment; the official database records the same events an hour earlier under a Central Standard label.[1][4]

Double Creek Estates

The subdivision on the west side of Jarrell had "approximately 38 single-family structures and several mobile homes," three businesses beside it, and an estimated population of 132. All of it was destroyed. "Twenty-seven people perished in Jarrell. There were relatively few injuries reported--one serious and less than ten minor, an inference to the small probability of survival." Ten families lost two or more members; one family of five died together. The dead were 7 percent of the town's population.[1]

A debarked tree with all its leaves stripped stands in a scoured brown field beside a dirt track; searchers pick through debris at left and a dead black-and-white cow lies at center.
The edge of Double Creek Estates. Debarked tree, scoured ground, searchers, and one of the cattle. About 300 head were killed in a pasture beside the subdivision, "with many of them tossed and blown for over 1/4 mile." Photo: National Weather Service, via Wikimedia Commons, public domain.[2]

"None of the decedents lived in residences with basements, and most of the dead were recovered various distances from their homes. Several Double Creek Estate residents survived in bathtubs, while others successfully sheltered in a neighbor's storm cellar. This was apparently the only underground shelter in this subdivision." The ground itself was changed: "From the air, the ground appearance changed abruptly in the vicinity of CR 308 and continued until very near the end of the path."[1] Long Phan and Emil Simiu of the National Institute of Standards and Technology, who surveyed the site from the ground and the air: "The houses, which consisted mainly of wood frame construction, were completely swept off the concrete foundations. The debris of the houses were so small and scattered that it was difficult to spot them from the air."[10]

Aerial photograph of a bare concrete slab in a scoured field with a small heap of debris and a single person standing beside it, dated 5-31-97.
A slab from the air, May 31. One person for scale. Photo: Long Phan, NIST, from the federal service assessment, public domain.[1]
The remains of a house on the edge of the path: mud-spattered wall sections and a bathroom's walls still standing amid debris.
The edge of the path. "Five people survived the storm in a bathroom (center of picture)." Photo: E. L. McIntyre, NOAA/NWS, public domain.[1]

Gary Lohman, who watched it come, told UPI the next day: "We watched it for 10 or 15 minutes. At first it was thin and narrow and we thought it was going to dissipate; then it came up from the ground as a huge mass." Governor George W. Bush flew over the path by helicopter and called the devastation "mind boggling." That evening the count stood at 27 dead and 23 missing.[13] The final toll stayed at 27.

The radar record

Two WSR-88Ds mattered. The Department of Defense radar at Granger, KGRK, "was located only 20 miles from the Jarrell tornado," and its data "showed a detailed view of the developing circulation at 1,500 ft elevation and above," with "a small hook" once the circulation formed. Then: "This radar ceased operation at 3:38 p.m. and was no longer available to the EWX forecasters." It had lost commercial power, and its emergency power failed to come on "because a control switch had been left in the improper position during a previous maintenance visit."[1] The Austin/San Antonio office's own radar, KEWX at New Braunfels, was 70 miles away, and "early data on the storms from the NWSFO EWX WSR-88D were significantly rangefolded due to second trip ground clutter."[1]

The archive reflects both facts. It holds the full day from KEWX and from Fort Worth's KFWS, and nothing at all from KGRK, whose Level II data from that day are absent from the national archive this site draws on; the research that reconstructed the storm used KGRK only in the coarser Level III form, from a copy kept by an NWS meteorologist.[7] What the archive can show is the storm as the warning office saw it.

Radar reflectivity from KEWX showing a large red storm core over Bell and Williamson counties, Texas, with a magenta core between Jarrell and Holland and a second storm over Georgetown.
Reflectivity, 3:43 pm CDT (20:43 UTC), KEWX, 0.5°. Seen from 70 miles south, with the beam more than a mile above the ground. The core above 70 dBZ, magenta, sits three miles northeast of Jarrell; at this range and resolution the radar sees the hail and rain around the mesocyclone, not the tornado, which is on the storm's southwest flank at the subdivision. The KGRK marker at right is the Granger radar, dark for five minutes.
Radar reflectivity from KEWX six minutes later, the magenta core now directly east of Jarrell and the storm's southwest edge over the town.
Reflectivity, 3:48 pm CDT, KEWX, 0.5°. Six minutes on, the core has moved south-southwest to sit directly east of Jarrell, the direction the whole complex had been traveling since noon. The velocity field from this radar is almost entirely range-folded over the storm, as the assessment found, so no velocity figure is shown.

Replay the storm's backward march on KEWX, the radar the warning office had, from Lake Belton to Jarrell.

Open in the archive

The published science adds structure the operational radars did not resolve. Houston and Wilhelmson found the low-level mesocyclone that preceded the tornado was "less than 4 km in diameter," and that the tornado's rapid expansion, to "nearly 1/2 mile in width" in the words of Storm Data, coincided with a cell merger at 3:37 pm, which "may indicate that mergers occurring during ongoing tornadoes might have resulted in tornado intensification." No rotational velocity for the tornado has been published, and no wind was measured.[8]

What the F5 rests on

The National Weather Service rated the tornado F5 "based on early visual observations of the damage," and by Fujita's definition the damage qualified: houses swept clean from their foundations. The official database carries it as F5 to this day.[4][10] Within a year the National Institute of Standards and Technology published a critique of the Fujita scale built on this tornado. Its conclusion: "The strongest damage caused by the Jarrell tornado can be explained by wind speeds corresponding to an F3 rating (i.e., 71 m/s to 92 m/s). An F4 rating, or the F5 rating officially issued by NWS, need not be assumed to explain the observed damage."[10]

F5 or F3 the National Weather Service's rating from the swept foundations, against the wind speed NIST's engineers thought sufficient to produce them, given how the houses were built Phan and Simiu, NIST Technical Note 1426, July 1998. The scale's wind ranges are estimates; no wind was measured in this tornado.[10]

The engineers' argument was about construction. The design wind speed for central Texas was 40 m/s (90 mph); with a typical safety margin, houses "would be expected to collapse under wind speeds in excess of 56.6 m/s (125 mph)," and winds of 155 mph "would leave no buildings standing, especially if the construction is mediocre or poor, as appears to have been the case for buildings destroyed by the Jarrell tornado." What they found on the slabs: "in many cases, even the sill plates that connected the wood frames to the concrete foundations were blown away," and where sill plates remained, "there was evidence that the sill plates had been connected to the foundations by nails" spaced about three feet apart. "There was no evidence of anchor bolts or steel straps used at the few random foundations where close inspections were performed." Williamson County had adopted no building code.[10]

A road in the subdivision with its asphalt surface stripped away in patches, a government sedan in the foreground and a surveyor crouching beyond.
Stripped asphalt. Road crossings "were clearly marked by strips of scoured asphalt road surface," from five feet to more than 300 feet long; the top layer was about 0.8 inches thick. Photo: I. Hakkarinen, NOAA/NWS, from the federal service assessment, public domain.[1][10]

The tornado's slow motion is often given as the other half of the argument, and it is worth being precise about who says so. The 2025 review of the EF5 "drought" summarizes NIST as suggesting the damage "could have been attributed to winds in the F3 range, given the long duration of tornado wind exposure at any given point and the quality of attachment of homes to their foundations." The NIST report itself records the witnesses' 5 to 10 mph but rests its conclusion on design wind speed and attachment, not duration. NIST's authors noted their conclusion "is consistent with an independent assessment by Texas Tech University," and recommended that the documentation of all 150 or so F5 tornadoes then on record "should be revisited."[10][11] The 2025 review lists Jarrell, with the 1998 Birmingham and 1999 Moore tornadoes, among the cases that "reignited discussions about the likelihood of overestimation of wind speeds necessary to cause damage at the higher ratings of the F scale," a debate that produced the Enhanced Fujita scale in 2007.[11]

The safety rules, questioned

The assessment recorded a public reaction it had not seen before. "After the Jarrell tornado, which moved unusually slowly, there was a widespread belief by the general population that persons who took appropriate safety actions were killed, while persons who took inappropriate safety actions lived. This has prompted questions from the public as well as the media regarding the viability of tornado safety rules the NWS promotes." News coverage "highlighted people in the path of the Jarrell tornado who fled the area and survived while many others who obviously followed prescribed safety measures perished," a "mixed message" the assessment feared would send people into their cars next time.[1]

Its answer was to hold the line, with a caveat: "In the vast majority of situations, people who take shelter in sturdy, well-built structures will survive most tornadoes whereas people who try to flee in automobiles will place themselves at much greater risk," together with a call for "significant public and industry education on the meaning of well-built tornado shelters." The NIST engineers drew a sharper conclusion: "Since, as pointed out earlier, even an F3 tornado could flatten most residential buildings, NWS should in our opinion consider revising its safety recommendations, which are based on the assumption that almost all tornadoes, including F4 events, are survivable."[1][10]

The rest of the day

Jarrell's figures are its own. The outbreak's are larger: twenty tornadoes across the Fort Worth and Austin/San Antonio forecast areas, and, by the assessment's count, 30 deaths, of which 27 were at Jarrell, one at Cedar Park, one in the Pedernales Valley and one a drowning in Austin. The CDC's own bulletin counts 29. Damage across the region exceeded $125 million.[1][3][6]

The same back-building complex had produced the Moody tornado, an F3 that "tossed a car and a pickup truck several hundred feet," and the Lake Belton F3, which crossed the lake and came ashore for a third of a mile. A different storm sent an F3 through the Cedar Park business district at 4 pm, where an Albertsons manager put customers in the walk-in refrigerator and a 69-year-old man died of a heart attack waiting out the storm in his truck. At about 4:30 the Pedernales Valley tornado formed on Lake Travis, reached F4, swept a stone house from its slab, and killed a 25-year-old man who fled his mobile home.[1][4]

Aftermath

Gabriel's Funeral Chapel in Georgetown "took care of the 27 people who had passed away in the tornado."[15] Jarrell was then a town of about 400 people, a figure it had held since the 1960s. By 2000 it had 1,000, and by the 2020 census 1,753, on the growing edge of Austin's suburbs. Double Creek Estates was rebuilt; a property listing today sells it on its "well-constructed homes."[16][18]

A granite memorial with a cross and a heart, names inscribed on side panels, under a flagpole, with a Texas historical marker for Jarrell at right and green fields behind.
Jarrell Memorial Park, 2020. The monument carries the names of the dead; the state historical marker beside it tells the town's story. The park is at 1651 County Road 305, on the subdivision's side of town. Photo: Larry D. Moore, via Wikimedia Commons, CC BY 4.0.[17]

No F5 has been rated in Texas since. The tornado's place in the science is the argument it started, about what a rating measures, and in the public mind it is the image of a storm that arrived slowly enough to watch and left nothing.

Videos

Footage and retrospectives, all via YouTube's privacy-enhanced embeds.

The man with the camera. KVUE on the footage shot from Jarrell that day.
Twenty-five years. KXAN with survivors for the 2022 anniversary.
The meteorology. KVUE breaks down the setup with an expert.
Texas Terror. Overcast's documentary treatment.

The other tornado whose rating the engineers questioned within the year is Bridge Creek-Moore, the only one with a measured wind. For the scale that came out of the argument and the first tornado rated on it, see Greensburg. For a modern rating built on what a tornado did to heavy objects rather than to houses, see Enderlin.

All tornado profiles · Search the radar archive by date and time · How to read reflectivity and velocity

Sources

Path, width, deaths and the warning chronology follow the federal service assessment of April 1998; the official database's figures are given where they differ. Times are CDT (UTC−5).

  1. NOAA National Weather Service, The Central Texas Tornadoes of May 27, 1997, Service Assessment, April 1998, including Appendix E by the Centers for Disease Control and the Texas Department of Health.
  2. NWS Austin/San Antonio, May 27, 1997 Tornado Outbreak, event summary with 20th anniversary report (PDF).
  3. NWS Fort Worth, The Tornadoes of May 27, 1997.
  4. NCEI Storm Events Database, 1997 bulk details file, events 5598913 (Jarrell F5), 5598911, 5597318 (Prairie Dell), 5599220 (Moody), 5597314 (Lake Belton), 5598914 (Cedar Park) and 5598918 (Pedernales Valley), from ncei.noaa.gov/pub/data/swdi/stormevents/csvfiles/.
  5. Iowa Environmental Mesonet, archived NWS warning text: NWS Austin/San Antonio Tornado Warning for Williamson County, 3:30 pm CDT.
  6. Centers for Disease Control and Prevention, Tornado Disaster, Texas, May 1997, Morbidity and Mortality Weekly Report, November 14, 1997.
  7. A. L. Houston and R. B. Wilhelmson, Observational Analysis of the 27 May 1997 Central Texas Tornadic Event. Part I: Prestorm Environment and Storm Maintenance/Propagation, Monthly Weather Review 135(3), 2007.
  8. A. L. Houston and R. B. Wilhelmson, Observational Analysis of the 27 May 1997 Central Texas Tornadic Event. Part II: Tornadoes, Monthly Weather Review 135(3), 2007.
  9. S. F. Corfidi, Some Thoughts on the Role Mesoscale Features Played in the 27 May 1997 Central Texas Tornado Outbreak, 19th Conference on Severe Local Storms, 1998, updated 2007.
  10. L. T. Phan and E. Simiu, The Fujita Tornado Intensity Scale: A Critique Based on Observations of the Jarrell Tornado of May 27, 1997, NIST Technical Note 1426, July 1998.
  11. A. W. Lyza, H. E. Brooks and M. J. Krocak, Where Have the EF5s Gone? A Closer Look at the "Drought" of the Most Violent Tornadoes in the United States, Bulletin of the American Meteorological Society 106, 2025.
  12. J. M. Davies, Significant Tornadoes in Environments with Relatively Weak Shear, 21st Conference on Severe Local Storms, 2002.
  13. UPI, Bush visits Jarrell, search continues, May 28, 1997.
  14. C. Osborn, The Jarrell tornado, 20 years later, Austin American-Statesman, May 26, 2017 (summary only; the article is paywalled).
  15. A. Archer, Gabriel's Funeral Chapel and Crematory: honoring our past 30 years, Williamson County Sun, October 16, 2023.
  16. Texas State Historical Association, Handbook of Texas, Jarrell, TX.
  17. City of Jarrell, Parks.
  18. Weather Underground, Twenty Years: A Look Back at the Jarrell Tornado Catastrophe, May 26, 2017; a property listing for Double Creek Estates; and Wikipedia, 1997 Jarrell tornado, consulted September 2026, used as a finding aid.

Radar imagery on this page was rendered by WeatherOverTime from NOAA Level II data recorded by KEWX. Where sources disagree on a figure, the difference is noted in the text. Corrections: contact@weatherovertime.com.