Table of Contents
March 16, 2026 was the worst day of the year for the NAS: 216 ATCSCC advisories (z=+3.69, the highest z-score in 141 days of data). Nine of those advisories mentioned ZBW (Boston Center), but no BOS ground stop was issued.
This is the interesting negative case: the system was under extreme stress, Boston Center was affected, but Logan itself stayed open.
1. What the Advisory Data Shows
| Metric | May 19 (storm) | Mar 16 (peak) |
|---|---|---|
| Total advisories | 167 | 216 |
| BOS ground stops | 4 (118, 146, 160, 167) | 0 |
| ZBW advisories | 14 | 9 |
| z-score vs baseline | +2.26 | +3.69 |
Source: FAA ATCSCC Advisories 2026-03-16
2. What the REPL Would Test
The hypothesis: if BOS wasn't ground-stopped, ADS-B message volume should be normal or slightly depressed — not the 37% drop seen on May 19. The REPL test:
;; Compare message volumes (def mar16-volume (hourly-counts (load-data "2026-03-16" (range 24)))) (def may19-volume (hourly-counts (load-data "2026-05-19" (range 24)))) (def baseline (hourly-counts (load-data "2026-05-15" (range 24)))) ;; Normalize to baseline (map (fn [h] {:hour h :mar16 (/ (get mar16-volume h 0) (get baseline h 1.0)) :may19 (/ (get may19-volume h 0) (get baseline h 1.0))}) (range 24)) ;; Expected: mar16 ≈ 1.0, may19 << 1.0 in hours 21-23 UTC
3. Why This Matters
The negative case strengthens the storm case study. If Mar 16 — a worse day by advisory count — shows normal ADS-B volume near BOS, then the May 19 drop is specifically attributable to the BOS ground stop, not to general NAS stress. Without the negative control, the May 19 analysis could be explained by "the whole system was busy."
4. What's Missing
- ADS-B archives for Mar 16 (not yet confirmed)
- The specific ZBW advisories on Mar 16 (route restrictions? flow constraints? not ground stops) — worth fetching for comparison
- If no ADS-B data: this case study demonstrates hypothesis formation from advisory data alone, with the REPL test as a specification of what we'd check