The Precursor Was a Vehicle

What a long record of ground incursions at one airport should have told us

On March 22, 2026, a Jazz Aviation CRJ-900 operating as Air Canada Express collided with an Oshkosh Striker 1500 rescue vehicle while landing on runway 4 at LaGuardia Airport. Both pilots were killed and thirty-nine people were taken to hospital. The National Transportation Safety Board deployed a go-team under Chair Jennifer Homendy, and the investigation (DCA26MA161) is ongoing.

Nothing in this article addresses the cause of that accident. The Board has not determined one, and it would be improper for anyone outside the investigation to speculate. What can be discussed, because it was already a matter of public record before March 22, is the class of event it belongs to — and how long that class had been visible at that airport.

A record that was not hidden

Reporting after the accident assembled a sequence stretching back three decades at LaGuardia alone: in 2015, an Embraer E190 came within about 1,300 feet of an unauthorized vehicle, and snow plows crossed hold markings roughly 180 feet from a Boeing 737 cleared for takeoff; in 2012, a ground vehicle forced an aircraft on final approach to go around; in 2000, an A320 overflew a snow plow by an estimated 50 feet on departure from runway 4 — the same runway; in 1996, an airliner braked hard on landing to avoid a conflict. Crews reported further events involving fuel trucks and airport authority vehicles in 2014 and 2021. In May 2025, a Republic Airways flight was cleared for takeoff on runway 13 while a United aircraft was still taxiing on it.

None of these were secret. Each was reported, logged, and in most cases investigated. And across the National Airspace System, the trend line has been flat: there has been no sustained improvement in runway incursions since fiscal year 2023.

So the question this accident puts to the industry is not whether the precursors were visible. They were. The question is why visibility did not become action.

What I learned looking at the same class of event from the regulator’s side

In 2013 I was seconded for a year to the Shenzhen Safety Supervision and Administration Bureau of the Civil Aviation Administration of China, from an airline safety department. One of the matters I worked on was a ground operational collision between an airline ground vehicle and an aircraft. The work consisted of interviewing the personnel involved, reviewing operational and ground-support documentation, analysing contributing factors, and identifying procedural weaknesses, ground-support coordination deficiencies, and oversight gaps associated with the occurrence.

Two things from that year have stayed with me, and both bear on the pattern above.

The first is that aircraft-and-vehicle events are almost never a failure of a single driver. They sit at a seam — between the movement area and the ramp, between air traffic control and airport operations, between the operator’s procedures and the airport’s. Each organisation on either side of that seam can be fully compliant with its own procedures while the seam itself goes unowned. When an investigation stops at “the driver crossed without clearance,” it has described the last link and left the seam intact.

The second is that the regulator and the operator were looking at different data about the same airport. The operator saw its own events. The regulator saw events across operators but in less operational detail. Neither side held the complete picture, and neither side knew it. The pattern lived in the gap between two partial views.

Why inspection does not close this gap on its own

I spent six years on the other side of that relationship, organising the regulator-mandated special safety inspection campaigns at an airline. From that vantage, three structural limits are visible, and they explain a great deal about how a visible pattern can survive repeated inspection.

First, inspection checklists are built looking backwards. Their two sources are the current regulations and the last accident. A hazard that has produced only near misses has, by definition, not yet generated either. It therefore appears on no checklist — which is precisely the situation at an airport with a long record of close calls and no fatal outcome.

Second, the inspected party’s dominant incentive is to pass. This is not a moral failing; it is an incentive structure. Where inspection results feed assessment, rating or penalty, the rational preparation is to present what is ready and to defer what is uncertain. I have organised that preparation myself, and what struck me at the time was how natural it felt — natural enough that nobody involved experienced it as concealment. The remedy is not exhortation. It is to record findings surfaced by the operator separately from findings surfaced by the inspector, and to stop penalising the first category. When self-reported findings rise in the quarter after such a change, that is visibility improving, not safety declining — and somebody has to be prepared to explain that to senior management before the numbers move, or the policy will be reversed.

Third, the loop breaks between writing a corrective action and verifying it. Corrective-action reports are usually complete: finding, cause, action, owner, due date. Verification is usually a documentary check — a response was received, so the item is closed. Whether behaviour changed is not visible on paper. The discipline I would apply is that every corrective action must specify, at the moment it is written, how it will be verified, by whom, and when — and that the verification method may not be “review the response.” It must be observation of the activity the next time it occurs. This produces fewer actions. That is the point: three actions confirmed to have changed behaviour are worth more than ten closed on paper.

The practical question for operators

An operator reading about LaGuardia may reasonably conclude that the matter belongs to large airports and to the agencies. For the ground-vehicle interface specifically, that is a mistake. Any operator whose aircraft are serviced on a ramp shares the seam, and in most cases shares it with contracted providers whose procedures the operator has never read.

Three questions are worth asking this week, and none of them requires a formal programme to answer. Who owns the interface between your aircraft and every vehicle that approaches them, by name? When one of your crews reports a vehicle conflict, where does that report go, and who sees it alongside the other reports from the same station? And when a corrective action was closed last year, what would you look at today to establish that the behaviour actually changed?

A fatal accident makes a hazard undeniable. The work of safety management is to make it actionable before that.