After the I-35 bridge in Minnesota came crashing down into the Mississippi River a decade ago, killing 13 people in rush-hour traffic, National Transportation Safety Board investigators raced to the scene to find much of their evidence underwater and being pulled apart by backhoes in a frantic search for victims.
In the painstaking months to come, they rebuilt portions of the bridge in a nearby park, using cranes and barges to move it downriver.
Ultimately, a grainy video image led them to the bridge's tragic point of failure, but not before they dug through years of inspection and maintenance records, constructed models, sampled the steel and concrete used to build the bridge and parsed 40-year-old design plans.
Two of the investigators on the case, Robert Accetta and Dan Walsh, are now applying that same methodical approach to unraveling the devastating collapse of the pedestrian bridge at Florida International University that killed six people last month.
Because the investigation remains open, Accetta, the lead investigator, declined to talk about specific details. But much of what he and Walsh learned during their 15-month investigation into what happened on a Minnesota summer day, along with other bridge accidents, will help them determine why 950 tons of concrete fell suddenly and seemingly without warning in a bridge touted as a feat in design, construction and safety.
"The evidence is probably the most important part," Accetta said in an interview with the Miami Herald. "Sometimes witnesses will report how they remember something ... but the evidence doesn't change."
When the Minnesota bridge collapsed, rush-hour traffic was bunching up on the bridge. Two southbound and two northbound lanes had been closed while workers were preparing to resurface the eight-lane bridge, which added tons of material to the bridge. Altogether, 111 vehicles were on the bridge, including a school bus carrying 62 students.
As the south span suddenly gave way, a malfunction eerily similar to the abrupt failure on the FIU span, a nearby motion-activated camera kicked on, showing the center span over the river drop nearly intact to the river below. In addition to the 13 killed, 145 people were injured.
As the search for victims continued over the next day, the local sheriff's office began pulling apart pieces of the bridge. They cut the concrete deck and a support beam with a backhoe and used blowtorches to pull apart steel trusses.
That left NTSB investigators with a mangled mess, not unlike the pile of debris strewn across Southwest Eighth Street when the FIU bridge, which was still under construction, failed.
"We had to move pieces downriver on barges and have cranes lift them into place," Accetta said. "We had all the structure laid out as if you'd cut it open and filleted it open in a public park that we commandeered."
In the FIU case, investigators have also preserved large sections of the bridge, storing them in a Florida Department of Transportation warehouse. They've also sent core samples of concrete pulled from the debris for testing and plan to examine tensioning cables, which were being adjusted by a worker when the bridge fell.
"All the facets of the investigation will be looked at exhaustively," Walsh said.
The agency's highway division has three teams of investigators who rotate on-call for two weeks at a time. Altogether, just 18 investigators are responsible for the entire country, and new issues are always arising, Accetta said — like self-driving cars. Three days after FIU's bridge collapsed, the agency began investigating the crash of a self-driving Uber that killed a woman in Arizona.
But bridge failures, and their potential for a high number of casualties or risk of occurring in other similarly built bridges, almost always get investigated, they said.
Even before they arrive on scene, Accetta said the teams are in touch with local authorities, usually consulting about how to preserve evidence.
"We don't hinder the rescue, but we do try to preserve evidence," he said. "And we don't always know what is evidence."
Accetta is a trained civil engineer, has worked for the agency for 18 years and previously worked as a police officer in Arlington, Texas, where he reconstructed accidents. Walsh is a licensed professional engineer who joined the agency 16 years ago after working on public highway projects in Texas and Maryland. Together, they've investigated bridge and tunnel collapses across the country.
They may also rely on outside experts. For the Minnesota bridge, the Federal Highway Administration's research lab outside Washington, D.C., and experts at the State University of New York helped construct models to simulate the collapse, including a 3D rendering of the bridge's deck and supporting trusses.
Like police investigators, the NTSB team will interview witnesses but for different information. A sound, like the one an FIU employee heard hours earlier and described as a 'cracking whip,' may or may not be related.
Among the simplest, but sometimes most telling clues, are pictures, although it's not yet clear if the multiple shots taken by drivers and nearby witnesses of the FIU bridge collapse reveal anything. They don't tell the whole story, but early in an investigation, they can point investigators in the right direction.
As they searched original Minnesota bridge designs and four decades worth of maintenance and inspection records, the team wandered through a warren of information, including repairs and changes and evidence of wear and tear. They also discovered the bridge had been deemed "structurally deficient" since 1991. But determining what mattered, and what didn't, required cross-checking and interviews, including one with an inspector about what he'd seen 10 years earlier.
A 2003 inspection by a company contracted by the Minnesota Department of Transportation finally provided the key they needed. It included a photograph of a gusset plate used to secure the bridge's trusses. The gusset appeared to be bowing.
"The plate was visible and we noticed the abnormal bending," Accetta said. "That helped us focus on that area when we laid it out in the park."
When they looked at changes in the bridge's design, they realized the steel used for trusses had been changed, but designers never recalculated the load on the gussets.
They also took a closer look at the maintenance work being done and realized that workers had loaded tons of sand and rock, along with concrete mixers and water tankers, onto the bridge, dramatically increasing the weight and stress. Ultimately, they found that the mistake in the design plan, combined with storing material for the concrete and construction equipment, led to the collapse.
While the FIU case won't have previous inspections to help investigators, Walsh said they have collected video, interviewed witnesses and have construction plans.
"A lot of that will apply toward the FIU bridge collapse," he said.
Investigators have also taken core samples from concrete.
And they've received numerous suggestions from engineers, and some amateur sleuths — just after the collapse, theories posted to YouTube, including a Canadian vlogger's dramatic re-enactment of tendon snapping, drew millions of views.
"Anybody can come up with a theory of what happened, and you will have people coming forward to tell you what they think happened," Accetta said. "But the hardest issue for us is to confirm it and document it with evidence so we can back it up."
Because the agency's focus is safety, the final report will not necessarily place blame. The reports instead recommend changes to improve safety. Based on the Minneapolis investigation, a series of changes were made nationwide, including how workers store and stage material on bridges, and inspections and design work for bridges that use gusset plates.
A 2006 investigation of the Boston tunnel collapse led to a national tunnel inspection program that did not previously exist, Walsh said. After a bridge across the Skagit River in Washington fell when it was struck by a truck with an oversized load, signs were changed. Rules on how to brace girders during construction were revised after a 40-ton steel girder buckled and fell onto a highway in Golden, Colorado, Walsh said.
Investigations normally take a year or more.. But Accetta said the FIU case may require more time because it is so complex.
"We have a lot of testing we want to do with parts of the bridge that we've taken core samples [from] or have large parts that need to be inspected," he said. "So that part will take time and until we have answers from that evidence, it's going to slow the process down. We're in no rush and we want to make sure we get it right."