Mexico City (MEX) Traffic Drop — a −32% Day Built Around a 4 h 47 min Civil Pause on Independence Day, September 2026

What fired, and why the shape matters more than the number. Our detection layer flagged MEX when 16 September landed at 524 observed movements against a rolling 7-day average of 761 — a −31% traffic anomaly at Mexico's busiest airport. The day has since settled at 556 movements, −34% against the Wednesday baseline and −32% against the settled 7-day average of 819: the same anomaly, a firmer number. On its own a daily percentage says little — it can be a slow erosion, a coverage hole, or a stop. The hourly profile answers that. MEX ran a normal night and morning bank through 15:15 UTC, logged nothing civil for 4 hours 47 minutes, then restarted at 20:02 UTC with a run of long-haul and US arrivals. The missing ~240 movements in that window account for about four-fifths of the day's ~291-movement shortfall against its Wednesday baseline. A clean stop and a clean restart at a single metro is the signature of a bounded, planned-looking pause — not of weather, which degrades gradually, and not of a receiver outage, which we check next.

It was not our receivers. Every traffic page we publish has to rule out its own measurement first — Monterrey in August was a coverage failure, not a closure. Here the evidence points the other way. First, the network was still listening over the city: between 15:30 and 19:59 UTC it heard seven aircraft below ~4,500 m inside the Mexico City area — none with a scheduled-airline callsign, three with Mexican Air Force (FAM) prefixes, some circling at roughly 2,750–2,900 m for hours. A dead receiver hears nothing; this one heard aircraft that were not airliners. Second, global coverage was normal: worldwide position reports in the 15:00–19:59 UTC hours totalled 257,433 on 16 September against 253,402 a week earlier, and the day's global movement rollup — now settled at 254,003 — sits inside its usual range. Third, an independent count agrees: the second ADS-B aggregate we cross-check puts MEX at 524 movements that day (274 departures, 250 arrivals), about 6% below our settled 556 and far under any normal Wednesday — a separate pipeline seeing the same collapse. What we did not see during the pause is just as telling: no aircraft reporting on the ground anywhere in the Mexico City area from 15:00 to 19:59 UTC, against 5–12 an hour a week earlier.

It was not the weather, and it was not Popocatépetl. MEX reported every hour through the pause: VFR, 8–9 statute miles in haze, winds calm to 6 kt, high overcast at 22,000 ft, no thunderstorm. Across the last 30 days, 15 of 1,018 hourly observations (1.5%) were IFR or LIFR. The airport brief shows MEX as ash-exposed because that flag is radius-based — Popocatépetl, 66.5 km from the field, had advisories within 48 hours — so we tested the polygons directly. VAAC Washington's 21:37 UTC advisory on 15 September reported ash to SFC/FL210 in an area whose nearest edge was ~18 km from MEX; the 03:25 UTC advisory on 16 September reported emissions not observed, possibly obscured by cloud; and the 09:21 UTC advisory reported ash not visible on satellite and carried no forecast ash area. No MEX or NLU position fell inside any reported or forecast area, and the pause began six hours after that last advisory. The geography closes it: Puebla (PBC), the airport nearest the vent and 28–29 km from the ash areas, kept moving traffic in the window (14 movements), while NLU — 51 km or more from every area and the farthest north — stopped alongside MEX. A plume does not skip the closest airport.

What the timing and the traffic suggest — labeled as inference. 16 September is Mexico's Independence Day, and the national military parade is held in central Mexico City that morning. The pause covers 09:15–14:02 local time, stops both commercial airports of the Valley of Mexico together, and the only low-level traffic our receivers heard over the city in those hours carried non-airline callsigns, three of them Mexican Air Force. That pattern is consistent with a planned suspension of civil operations for state flying over the capital. We have not seen the notice itself: our NOTAM feed contains no aerodrome notices for MEX at all, in any period, so its silence here is weak evidence either way. Our movement record also only begins on 28 April 2026, so we have no previous 16 September to compare against. We therefore publish the measured pause and the fitting explanation side by side, and do not present the explanation as confirmed.

The rest of Mexico had a quieter day too — and we have not measured why. Stripping out the pause does not make 16 September an ordinary Wednesday. Outside 15:00–19:59 UTC, MEX ran 2–8% below the same hours on 2 and 9 September. Across Mexico, flights touching a Mexican airport but none of the four Valley of Mexico–area airports (MEX, NLU, Toluca, Puebla) totalled 808 on 16 September against 1,031–1,064 on 2, 9 and 15 September — roughly −23%, spread across the day rather than concentrated in the pause window (301 vs 373–410 in it) — and 1,003 on 17 September. Guatemala City, next door, was normal (389 vs 378 and 397 on the two previous Wednesdays). A national public holiday is the obvious candidate for a softer schedule, but we have not measured it and some Mexican airports on this feed have patchy coverage, so we report the country-wide number as observed, not explained. Two further honesty notes. The night before the pause (00:00–05:59 UTC, the evening of 15 September local) MEX carried 160 movements against 189–207 on comparison days — a dip we cannot separate between the holiday eve and the Popocatépetl ash area current in those hours. And every count here is a floor from a crowdsourced receiver network, not the airport's official total.

What the data showed — the closing number. This brief is closed on the recovery, not on a conclusion about the cause. The closing measurement: MEX averaged 841 observed movements a day across the six complete days from 17 to 22 September — 99% of its 847 same-weekday baseline, and 3% above the 819 rolling average the drop was measured against — after the 556 of 16 September (−34%). Day by day: 901, 865, 828, 757, 867 and 825, five of the six inside the pre-event 752–874 band and the sixth (17 September) just above its top. The departure/arrival mix is the airport's ordinary one on every day (469–563 departures against 288–338 arrivals), and the clock is ordinary too: the 15:00–19:59 UTC window carried 255, 261, 256, 253, 256 and 243 movements — a mean of 254 against the ~248 of the twenty days before the event, and against 8 on 16 September. Not one of the six days has a hole at any hour. The second ADS-B aggregate reads the same six days at a mean of 805, its usual few per cent below ours, and agrees on the shape. Felipe Ángeles (NLU), which stopped on the same clock, came back on the same clock: 113, 93, 89, 82, 109 and 77 movements for the day, with 27–33 in the five-hour window on five of the six days and 17 on 22 September, against 4 during the pause. Two bookkeeping caveats travel into the archive with these numbers: 21 September's daily rollup is flagged as a coverage gap (we plot the flight-detection recount of 867, and a second feed reads 820), and 23 September is still arriving, so it is not counted here. In one line: a bounded 4 h 47 min pause of civil traffic at both of the Valley of Mexico's commercial airports, measured end to end, followed by six ordinary days — with a fitting explanation we could never source.

What would reopen this page. Two things, and neither has happened. An official statement or notice appearing retrospectively — that would move the framing from 'measured pause, cause inferred' to a named event, and it is the one gap this brief carries into the archive; our NOTAM feed contains no aerodrome notices for MEX in any period, so it will not be the source. Or a repeat: a second pause on the same clock, at both capital airports at once, would turn a one-off into a pattern, and the simultaneous stop at NLU is the fastest way to confirm one. Note what is not yet settled: 21 September sits under a coverage-gap flag on the daily rollup (two other counts read it as an ordinary Monday), 23 September is still filling in, and days inside the trailing three-day window can still move — 19 September went 798 → 828 and 20 September 725 → 757 after we first published them. Automated multi-signal analysis keeps re-checking MEX every three hours after this page stops: Popocatépetl has kept issuing advisories through the recovery (8 since 17 September, the latest at 02:59 UTC on 24 September) and is tracked on its own page; GNSS interference is at its floor (7-day ratio 0.001); the last hourly observation was VFR, with 1% of the past 998 observations IFR or LIFR. The MEX brief and the NLU brief keep updating from the same cycle; this page stops here unless one of those two things appears. Whether a given flight operates is decided by the airline, the airport and Mexico's aviation authorities, not by this page.