Amazon 767 Recorder Captures Speed Concerns Before Fatal Miami Runway Overrun
A pilot repeatedly raised concerns about excessive speed before an Amazon Prime Air Boeing 767 overran a Miami runway, struck two vehicles and killed five people. The National Transportation Safety Board’s Sept. 9 investigative update also documents terrain warnings and a brief attempt to abort the landing, but it does not establish why the sequence ended in an overrun.
The 21 Air-operated Flight 7598 was arriving at Miami International Airport from San Juan on Sept. 6 when it left Runway 30. Five people were killed and five others were injured. The consequences extended beyond the cargo aircraft: people in vehicles were among those caught in the runway excursion, underscoring the exposure of airport workers and motorists near a runway end.
What the recorders establish
The cockpit voice recorder captured one pilot cautioning that the aircraft was too fast after flaps were called to 20 degrees. The NTSB did not identify which pilot made those comments. The same pilot continued commenting about excessive speed during the remaining recording, without what investigators described as a consistent verbal response.
The recording also captured two automated “sink rate” alerts and five “too low, terrain” warnings before sounds similar to touchdown. Those alerts are confirmed parts of the recorded sequence, but they do not independently establish what the pilots saw, how they assessed the approach or why particular control inputs followed.
Flight-data information adds the aircraft’s physical response. The nose and right main landing gear touched down at a groundspeed of 158 knots. Braking began at 146 knots, and the left main gear touched down at 134 knots. With the aircraft at 120 knots, the brakes were released and the throttles moved to a value consistent with go-around thrust. Four seconds later, at 117 knots, the throttles returned to idle and braking resumed.
The cockpit recording captured a pilot calling for a go-around 15 seconds before that recording ended. Because the cockpit and flight-data recorder timelines have not yet been precisely synchronized, the two preliminary timelines should not be treated as a frame-by-frame reconstruction. The NTSB said all of the recorder information remains subject to change.
Why the transition matters
A go-around is intended to discontinue a landing and establish another approach rather than continue toward a stop on the remaining runway. In this case, the recorded increase in thrust was followed quickly by a return to idle. That reversal makes the transition between landing and go-around an important investigative question, but it does not reveal the crew’s reasoning or prove that one decision caused the overrun.
Investigators will need to connect cockpit communication with aircraft configuration, control movement and the remaining runway available at each stage. Acknowledgment between pilots matters because it helps establish whether both crew members share the same understanding of speed, aircraft state and intended action. The absence of a consistent recorded verbal response is therefore relevant to crew coordination, although silence on the recording cannot by itself establish inattention, disagreement or loss of control.
The deceleration sequence presents a separate systems question. Speed brakes increase aerodynamic drag and reduce wing lift after touchdown, helping place more aircraft weight on the wheels for braking. Thrust reversers redirect engine thrust to provide additional deceleration. The NTSB found no indication in the recorded data that either system deployed.
That finding does not yet explain why the systems did not appear in the data. It does not establish whether they were selected, available, inhibited by aircraft conditions or affected by another factor. Nor does it mean wheel braking was absent: the flight-data recorder specifically showed brakes being applied, released during the go-around-thrust interval and then reapplied until the end of the recording.
The investigation’s next evidentiary step
The recorder data provide more than 400 flight parameters and approximately 54 hours of information, while the cockpit recorder contains more than two hours of good-quality audio across four channels. An NTSB cockpit-voice-recorder group is expected to review the accident flight and produce a written transcript. That work should refine timing and context, but the agency will still need to integrate recorder data with crew interviews, aircraft examinations and operational records before reaching causal conclusions.
A wrongful-death lawsuit filed by Yaraisi Santiso Morejon over the death of her husband, Yoel Rodriguez Naranjo, alleges pilot error, inadequate training, unsafe practices and aircraft-system concerns. Those are civil allegations, not NTSB findings. Amazon and 21 Air have said they are cooperating with investigators.
For now, the strongest confirmed tension remains narrow but consequential: repeated speed concerns and automated warnings preceded touchdown, go-around thrust was applied and then removed four seconds later, and the recorded data show no speed-brake or thrust-reverser deployment. Why that sequence occurred and how its elements relate to the five deaths remains undetermined.
| More aerospace and engineering stories, right in your MSN feed. Follow AMI on MSN |
By Thomas Caldwell — AMI’s senior editor for mechanical and mobility engineering, covering vehicle electronics, systems integration, electrification, chassis systems, propulsion, and safety policy.
