Pilot Nick - 9 min read
What happened
On September 6, 2026, at 17:53 UTC, a Boeing 767-300 freighter registered N1997A, operated by 21 Air on behalf of Amazon Air, landed on runway 30 at Miami International Airport at the end of a flight from San Juan, Puerto Rico. Two crew members were on board. It was the aircraft’s third rotation of the day, after Cincinnati to Miami and Miami to San Juan.
It did not stop on the runway. According to the sequence established by the NTSB, the aircraft went past the northwest end of the runway, struck navigation aids, then hit a van belonging to a contract cleaning company, parked inside the airport boundary with seven people on board. It then went through a fence and struck a car travelling on Northwest 67th Avenue before coming to rest beside a parking lot. A fire broke out.
Five people were killed, all of them occupants of the cleaning-company van. Five others were injured, including both pilots. By September 8, the pilots had been treated and released; three people remained hospitalized, two in critical condition and one stable.
The FAA announced an investigation. The NTSB sent a team of thirty-two and held a first press briefing on September 7 at 4 p.m. local time, without releasing any technical findings. Nothing below prejudges their conclusions. It is only what the public ADS-B trace, made available for download by Flightradar24, allows us to establish once it is re-referenced to the official geometry of the airfield.
The runway
Runway 30/12 is 9,360 feet long. That is the full pavement length. Landing-performance calculations use a different number.
For a landing on runway 30 the threshold is displaced by 945 feet, meaning the usable landing area begins that far beyond the physical start of the runway. The FAA declared distances apply a further reduction at the far end. The landing distance available, or LDA, is 7,913 feet, or 2,412 metres. Runway 12/30 is also the only runway at the airport not aligned with the other three, whose full pavement lengths are 8,600, 10,506 and 13,016 feet.
The 7,913-foot figure is the basis for the landing-distance comparison here. If the estimated touchdown position is correct, roughly 4,100 feet remained within the declared landing distance, with about 4,600 feet of pavement to the runway end. The stopping margin also depended on landing weight, speed, runway condition and the deceleration achieved after touchdown..
What the trace shows
The file contains 537 timestamped position points, with barometric altitude, ground speed and heading. The sampling is irregular, from several minutes apart in cruise down to two or three seconds over the last two minutes of the flight.
The analysis that follows was done from the trace downloaded on the morning of September 7 and re-referenced to the official runway geometry, using the FAA-published coordinates of the two runway ends as the axis and the origin. It is not my work. I have checked the numbers against the approach plate, the METAR and the ADS-B Exchange track, and I present them here with my own reading as a line pilot. The data is public, and anyone can download the same files and repeat the same steps.
The profile reads in three stages, and that breakdown matters a great deal more than the simple observation of a high approach.
From 11 NM to 5 NM from the threshold, the aircraft descends on a 3-degree path with a deviation between 13 and 41 feet, and a lateral deviation of less than 0.06 degrees from the runway centreline. That is instrument-guidance precision, and nothing on this segment announces what comes next.
Between 4.6 and 3 NM, the aircraft stops descending and holds between 1,550 and 1,600 feet. The 3-degree path keeps going down and passes underneath it. In about thirty seconds the deviation grows from 123 to 501 feet, with the aircraft even regaining 50 feet of altitude along the way. This is where the vertical margin was lost, and nowhere else. The level segment coincides with a fall in ground speed, from 227 knots shortly before it to 180 during it, followed by a recovery to 195.
The ILS or LOC RWY 30 approach chart in force that day gives this segment its full meaning. The glideslope is set at 3.00 degrees and crosses the threshold at 51 feet. The published intercept fix, BIRDD, carries an altitude of 1,600 feet and sits 4.8 NM from the threshold. The aircraft reaches 1,600 feet at 4.8 NM, exactly where it was supposed to, and that is the point from which it stops descending. So it did not fall off a path it was holding. It arrived at the right place, at the right altitude, and did not begin the final descent. The glideslope passed beneath it.
From 2.2 NM the descent resumes at 5.4 and then 6.7 degrees, more than double the published slope. This is a catch-up dive rather than an approach running parallel to the path and offset above it, and it never converges. The slope only eases back toward 4.4 degrees over the last 2,000 feet, when there is no longer enough height left to recover the speed as well. The aircraft crosses the threshold at 110 feet above the ground, instead of the 50-foot reference, at 177 knots ground speed.
Laterally, centreline tracking stays excellent until 1.4 NM, then the aircraft drifts up to 129 feet right of centreline near 1 NM before converging again. The drift begins at the precise moment the catch-up dive starts.
Touchdown is around 3,800 feet past the displaced threshold, with an uncertainty of about 400 feet due to the quantisation step of the transmitted altitude. The regulatory touchdown zone ends at 3,000 feet. Ground speed at contact is 159 knots.
Touchdown, seen from the ground. The trace puts it around 3,800 feet past the displaced threshold.
The weather
The sequence of hourly observations at Miami says a lot. At 17:13, wind variable at 3 knots. At 17:53, the time of touchdown, wind from 190 at 17 knots gusting 26, with two decisive entries in the coded remarks: a wind shift recorded at 17:37, and a peak wind measured from 180 at 26 knots at 17:52, one minute before the landing. By 18:53 the wind had dropped back to calm. A thunderstorm was over the field, and the cell passed through during the landing. On runway 30, the reported wind implies a tailwind component of about six knots, increasing to roughly nine knots if the 26-knot gust came from the same direction. However, the remarks record a separate peak wind from 180° at 26 knots at 17:52 UTC. That corresponds to a tailwind component of approximately thirteen knots. These figures are calculated from airport weather observations; they do not establish the exact wind the aircraft encountered during the flare and touchdown.
METAR KMIA 061753Z 19017G26KT 10SM TS SCT020CB BKN048 BKN150 BKN250 30/22 A2994 RMK AO2 PK WND 18026/1752 WSHFT 1737 SLP137 TSB12 OCNL LTGICCG E-S TS E-S MOV E CB DSNT NW T03000222 10333 20261 55003
The 17:53 observation. The wind shift (WSHFT 1737) and the peak gust (PK WND 18026/1752) are in the remarks, one minute before touchdown.
Two other aircraft flew the same approach that afternoon, one twenty minutes before, one two minutes after. Their traces are what turn a single profile into evidence, and they're below.
