Flames jet from damaged LPG spheres while, at right, a bullet tank burns. - Photo by David White.

Flames jet from damaged LPG spheres while, at right, a bullet tank burns.

Photo by David White.

Big fires draw attention. Flames rising from a nearby LPG storage facility brought residents of Mexico City's San Juan Ixhuatepec district out of their homes in the pre-dawn hours of November 19, 1984. What happened next turned the facility and the homes around it into a scorched landscape littered with hundreds of bodies.

Probably within 20 minutes of the first flames there was an explosion. A projectile the size of a tank truck shot through the air and struck a group of giant LPG spheres on the opposite side of the facility. Four ruptured, dumping about 12 million gallons of LPG simultaneously. A tremendous fireball ensued, leveling the facility and engulfing homes on either side of the surrounding valley.

Be advised that this is not the official explanation of what occurred that morning. Experts maintain that the destruction at San Juan Ixhuatepea was the result of a vapor cloud explosion. I do not agree with that conclusion. Finding those hundreds of bodies in the streets is one of several key pieces of evidence supporting my theory about what caused the blast.

Acting as an investigator for one of the companies involved, I arrived in Mexico City to find flames still jetting from ruptures atop the only two LPG spheres to survive intact. All that remained of the destroyed spheres was the twisted metal legs that once supported them. The rest of the facility had been burned almost to the ground. LPG bullet tanks lay tossed around like discarded beer cans. Homes hundreds of feet beyond the facility were also flattened. Bodies lying in the streets were still being removed.

There is some credibility to the vapor cloud theory. While pumping into one of the spheres from an LPG pipeline, operators could have overpressurized the sphere accidentally. Liquid propane spilling from relief vents would have ignited, flashing back to cause a BLEVE (boiling liquid expanding vapor explosion).However, south and east of the spheres lay a network of LPG bullet tanks arranged at close intervals. Thirty-six of those 48 tanks had BLEVEd as well, far too many to expect from a short-lived vapor cloud blast.

Firefighters survey the aftermath of a cataclysmic LPG explosion in Mexico. - Photo by David White.

Firefighters survey the aftermath of a cataclysmic LPG explosion in Mexico.

Photo by David White.

Located east of the bullet tanks had been a loading rack for filling LPG transport trucks. On the north end of that rack were the motor and axles front and rear of a transport truck. Nothing of the truck's LPG tank remained. The bodies of those killed in this area had already been removed. Lying near the loading rack and the scorched remains of the truck were 14 burned fire extinguishers.

My conclusion is that a fire broke out as a transport truck was being filled at the loading rack. There was no water spray protection, and it is unlikely there was an emergency shutdown. Personnel on duty attempted to deal with the fire themselves, explaining the burned fire extinguishers. The truck's LPG tank BLEVEd, with the blast hurdling it out of the loading rack area. That explained how the fire started, but discovering how the spheres ruptured and the bullet tanks BLEVEd required more detective work.

Examination of the destroyed homes raised further problems with the vapor cloud theory. The homes were located on uphill slopes on the north and south sides of the LPG facility. Those homes burned, but were not blown apart. A vapor cloud release would collect inside buildings and any other contained space, typically beneath the hood of a car. With ignition, roofs would have been blown off and car hoods sent flying. None of that happened at San Juan Ixhatepec. Also lacking blast damage was a facility warehouse much closer to the release than the homes.

Next is the question of the bodies found in the streets. Why were so many people standing outside hours before dawn? A vapor cloud explosion is a silent killer. These people would have died in their beds, or at least inside their homes. The logical conclusion is that some event drew them outside before they were killed. The condition of the bodies was another clue. A fireball from a vapor cloud explosion would have burned clothes and skin, but left the bodies intact. At San Juan Ixhuatepec, bodies were charred and, in some cases, all but incinerated. That indicates a hot, intense fire that burned for an extended period of time.

Finally, I interviewed eyewitnesses among the neighborhood survivors. Each told an identical story. First, the sound of an explosion woke them. Checking outside, they saw flames on the east side of the LPG facility, indicating the loading rack instead of the centrally located LPG spheres. There was then a second explosion, followed by a series of explosions before the deadly blast that engulfed the hillsides. For me, this was the final piece of the puzzle.

View of the demolished LPG complex with the circled area indicating where the truck-mounted tank launched. - Photo by David White.

View of the demolished LPG complex with the circled area indicating where the truck-mounted tank launched.

Photo by David White.

Even though bullet tanks are named for their shape, the same word could describe their velocity once either of the two round ends have blown out. When the tank on the transport truck BLEVE'd, it rocketed northwest across the facility and scored a direct hit on the cluster of four LPG spheres. Those spheres ruptured and spilled their entire contents at once.

Thankfully, experience with LPG spheres failing is limited. In quantities typical of bullet tanks, liquid propane rapidly vaporizes, turning into a gas looking for an ignition source. But when the immense quantity involved at San Juan Ixhuatepec emptied into a common dike only about 20 percent immediately turned to vapor and ignited. This caused the giant fireball that swept the populated hillsides. The rest only gradually warmed to vapor temperatures. I call this effect "self-refrigeration." In this state, the slow, sustained fire more akin to gasoline results.

In Europe, dikes beneath LPG spheres channel spilled liquid to an area away from the tanks. The same is not true in the U.S. and Mexico. In effect, the LPG spheres and bullet tanks were now sitting in a deep pool of slow burning fuel. The large number of BLEVEd bullet tanks confirms this. The blast knocked many of the bullet tanks loose from their concrete saddles. Steady flames impinging against the vapor space did the rest. Had the fixed monitors on the dike wall activated, the result would have been the same.

In the end, the dike failed. Burning LPG pouring into the surface drains spread the fire throughout the facility for some time. Due to poor communications, the refinery continued to pump LPG to the facility for hours after the blast.

How did the two remaining spheres survive relatively intact? The cluster of four spheres scored by the tank truck missile shielded the two slightly larger spheres on the west side of the containment dike. The two surviving spheres are not thought to have been as close to full as the others. Rather than a catastrophic failure, the two spheres formed large blisters near the top that split open and created a burning relief valve for the contents.

Officially, the final death toll was 490. Taking into account those who died later from severe burns, the death toll was probably closer to 2,000.

What happened at San Juan Ixhuatepec runs counter to what most firefighters would expect. Four giant LPG spheres dumped their contents. Evidence suggests that the fire that swept the facility was long and intense, not quick and hot. Accepting the official explanation on faith alone ignores a rare but dangerous phenomenon that could someday repeat itself.

Lessons Learned

  • A loading terminal with no water spray protection and no automatic shutdown system stripped this facility of defenses that could have saved hundreds of lives.
  • Physical evidence is not the only thing an accident investigator should base his opinion on. At San Juan Ixhuatepec, I spoke to the eyewitnesses and gained a special understanding of what happened.
  • The Europeans have the right idea about LPG dikes. Channeling the spilled LPG away from the tanks was another missed opportunity at San Juan Ixhuatepec that could have saved hundreds.

Editor's note: This after-action review appeared in "Disasters Man-Made" by David White and Anton Riecher that was published in 2011.