Amazon Prime Air Boeing 767 Crash in Miami: Flight 7598 ADS-B Analysis and More Accidents
A Boeing 767 freighter operating for Amazon Prime Air crashed after overrunning Runway 30 at Miami International Airport on Sunday, September 6—and in this episode, we take a detailed look at ADS-B data that reveals how a seemingly normal approach became dramatically unstable during its final miles.
The aircraft was N1997A, operating as Prime Air Flight 7598 from San Juan, Puerto Rico, to Miami. The Boeing 767-300 freighter was operated by 21 Air on behalf of Amazon Prime Air. At least five people were reported killed following the accident.
The approach initially looked remarkably normal. From approximately ten miles until five miles from the runway, the 767 tracked almost exactly along a conventional three-degree glidepath. The accuracy of that track strongly suggests the autopilot was being used.
Then something changed.
At about five miles from the runway, Flight 7598 leveled at approximately 1,600 feet. It remained around that altitude, or slightly above it, for roughly the next mile and a half.
That level-off rapidly put the airplane above the normal vertical profile. At four miles it was already more than 200 feet high. At three miles, it was nearly 500 feet above the glidepath. At two miles, it remained more than 500 feet high.
Then the crew apparently tried to salvage the approach.
Between two miles and one mile from the runway, the aircraft descended on approximately a five-degree flightpath. Inside one mile, portions of the descent approached six degrees—roughly twice a normal three-degree approach angle.
Runway 30 at Miami also has a 945-foot displaced threshold. According to the ADS-B data, the Boeing crossed the physical beginning of the pavement about 165 feet above runway elevation. By the actual displaced landing threshold, it was still approximately 100 feet high, about 50 feet higher than a typical threshold crossing.
Speed compounded the problem.
During the level segment around 1,600 feet, the aircraft slowed to roughly 180 knots. Once the descent resumed, the speed increased again to about 195 knots. Despite the steep descent, the airplane managed to slow only to approximately 175 knots crossing the displaced threshold.
The data suggest the 767 touched down at roughly 160 knots. When it went off the end of the runway, it was still traveling about 110 knots.
But one of the most interesting clues appears much earlier in the flight.
ADS-B data include the altitude selected by the flight crew. During the descent, the selected altitude had initially been 3,000 feet. But as the airplane descended through approximately 6,700 feet, the selected altitude was changed to 1,600 feet.
That number is significant because 1,600 feet is a published altitude associated with the final approach segment for approaches to Runway 30.
Normally, when an autopilot is properly configured to capture and follow an approach, it can descend through the altitude selected in the altitude window once the appropriate approach mode has captured the vertical path.
So why did this airplane level at almost exactly the selected 1,600 feet?
One possibility is an automation surprise. Perhaps the autopilot was not configured in the mode the pilots expected, causing it to capture the selected altitude rather than continue down the intended vertical path. The crew may then have disconnected the autopilot after realizing the airplane had leveled.
There is another possibility. The pilots may have intentionally disconnected the autopilot to begin hand flying, and during that transition inadvertently leveled the airplane instead of continuing down the glidepath.
At this stage, neither explanation has been established as the cause. The NTSB investigation has only begun.
But regardless of why the level-off occurred, an even larger safety question emerges: Why continue the approach?
By the final two miles, the airplane was both high and fast and required an extraordinarily steep descent to reach the runway. A go-around would have reset the situation and given the crew another opportunity to fly a stabilized approach.
The Miami Boeing 767 crash is only one of several major accidents we examine in this episode.
We also discuss the final report on Pilatus PC-12 N357HE, which crashed near Recluse, Wyoming, killing all seven aboard. The airplane encountered convective weather at approximately 26,000 feet. Its autopilot disconnected, apparently after encountering an updraft that modeling indicated could have exceeded 3,300 feet per minute. The pilot then struggled to control the airplane manually in IMC, became spatially disoriented, entered an extreme descent, and the aircraft partially broke up before impact.
There is also important new evidence in the Carlisle, Pennsylvania, collision between a Bell 407 helicopter and a Cessna 150. Surveillance video shows the Cessna’s wings rocking dramatically only seconds after following the helicopter over the runway. I believe the evidence strongly suggests the small airplane encountered the Bell’s powerful downwash and wake vortices.
Finally, we examine the fatal flight of Cessna 150 N8110S. A 65-hour student pilot departed after civil twilight on a long cross-country flight, continued into dark-night conditions and likely IMC, and struck rising terrain. Surprisingly, the NTSB docket confirms that a CFI had given the student the required cross-country endorsement.
These accidents are very different, but several share a common lesson: once a flight starts deviating significantly from the plan, continuing can turn a manageable problem into an accident. Whether it is an unstable airline approach, an autopilot disconnect in IMC, helicopter wake turbulence, or a student pressing on into darkness and weather, recognizing when to abandon the original plan may be the most important decision a pilot makes.