wal.sh
home > research > 2026-repl-driven-flight-tracking > includes > case-003-winter-storm · May 21, 2026

Table of Contents

On January 26-27, 2026, a winter storm hit the northeast. The FAA issued BOS ground stops across both days (7 BOS advisories on Jan 26, 4 on Jan 27, within 95 and 69 total advisories respectively).

Unlike the May 19 thunderstorm, winter storms produce a different ADS-B signature: extended ground stops (hours, not the 2.5-hour thunderstorm window), deicing delays visible as long gate holds, and reduced but not eliminated traffic — airlines cancel flights proactively rather than holding airborne.

1. What the REPL Would Show

The first REPL session would check data availability: 

;; Do we have ADS-B archives for Jan 26?
(def jan26-files
  (filter #(.exists (io/file %))
    (map #(format "/mnt/usb/adsb/raw/2026/01/26/sbs-2026-01-26T%02d.csv.gz" %)
         (range 24))))

(count jan26-files)
;; => ? — determines whether this case study has data or is FAA-only

If archives exist, the analysis mirrors the storm case study: message volume by hour, altitude band distribution, holding pattern detection. The hypothesis is different: winter storms should show sustained low volume rather than a sharp drop, and fewer holdings because aircraft are cancelled on the ground rather than held airborne.

2. FAA Timeline

Advisory Time (UTC) Facility Action
Jan 26 — BOS/ZBW 7 BOS advisories within 95 total
Jan 27 — BOS/ZBW 4 BOS advisories within 69 total

3. What's Missing

  • ADS-B receiver archive availability for Jan 26-27 (not yet confirmed)
  • If no archive: this case study stays FAA-only, demonstrating that the ground stop pattern is recoverable from public advisory data alone — the REPL exploration starts at a different layer
  • Comparison with May 19: thunderstorm (sharp, short) vs winter storm (sustained, scheduled)

4. Open Question

Do winter ground stops produce the same altitude-band signature as thunderstorms? The hypothesis: no. Winter operations keep aircraft on the ground (deicing, cancellations), so the approach band shouldn't show holding patterns. The REPL test: 

;; If archives exist: compare altitude distributions
(def winter-bands (altitude-histogram (load-data "2026-01-26" (range 24))))
(def storm-bands  (altitude-histogram (load-data "2026-05-19" [21 22 23])))

;; Are the distributions different?
(chi-squared-test winter-bands storm-bands)