80 Cathay flights, 40% less contrail warming, and one route doing half the work

Google said on Monday that it has taken its contrail-forecasting system to Asia, running an operational trial with Cathay Pacific in which more than 80 flights flew routes designed to avoid creating warming contrails. Google's own satellite analysis puts the reduction in those contrails' warming impact at roughly 40%. Cathay is Google's first commercial airline partner in Asia to test the method, and the trial covers ultra-long-haul flights, which Google defines as routes lasting more than 16 hours and often spanning 13,000 kilometres or more.

The premise is unglamorous, which is part of the appeal. Contrails form when aircraft fly through cold, humid air. Most vanish in seconds, but some spread into cirrus-like sheets that trap outgoing heat. Google puts their share of aviation's total climate impact at roughly one third, a figure it has previously attributed to the 2022 IPCC report. Avoiding them requires no new aircraft and no new fuel — only knowing in advance which slices of atmosphere will produce them, and nudging the flight a few thousand feet around.

What the system actually does

Google says its system combines AI predictions with satellite imagery and weather data to map contrail-forming zones before departure, so dispatchers can build altitude shifts into the flight plan. In the air, Cathay Pacific pushes updated forecasts to the flight deck over in-flight Wi-Fi and surfaces them inside its Electronic Flight Folder, the tool crews already use, rather than as a separate application. That detail matters more than it sounds. An atmospheric forecast a pilot has to go and find somewhere else is a forecast that mostly does not get used.

The numbers, and what they rest on

Google says the trial targeted more than 100 flights across Cathay Pacific's network, that more than 80 of them flew contrail-avoidance routes, and that its analysis of satellite imagery estimates those flights reduced the warming impact of their contrails by roughly 40%. Two qualifications travel with that number.

The first is that it is a counterfactual. Nobody observes the contrail a flight did not make, so the benefit has to be inferred from a model of what would otherwise have happened — and here the model, the satellite pipeline and the claim all belong to the same party. The second is concentration. Google reports that interventions on the Hong Kong–Singapore corridor alone account for more than half of the trial's total reductions. The headline result is, to a large extent, one route's result, and nothing in the announcement establishes that it generalises to a network.

The comparison with Google's earlier work is also less direct than it looks. In 2023, a trial with American Airlines flew 70 test flights over six months and reported a 54% reduction in contrails. That is a count of contrails; the Cathay figure is a reduction in estimated warming impact. They are different measurements of different things, and neither has been independently audited.

The cost this announcement does not price

Steering around a humid layer generally means burning more fuel, and more fuel means more CO2. Whether contrail avoidance is a net climate gain therefore depends on the exchange rate between the warming avoided and the carbon added. Google published that number in 2023: the American Airlines flights that avoided contrails burned about 2% additional fuel. Monday's announcement reports no fuel-burn figure for the Cathay trial at all. Without one, the 40% is half a ledger.

Where this sits

The Cathay trial is the third step in a sequence rather than a standalone result. In August, Google announced Operation Blue Skies with the UK Government: a 30-month programme built around two operational trials of roughly four months each in Shanwick, the eastern half of the North Atlantic corridor. Google says that airspace accounts for about 5% of global contrail warming and that around 10,000 flights pass through it during trial hours each year. Measured against that, an 80-flight Asian trial is a scouting exercise — the point is operational feasibility on long Asian and transpacific sectors, not a settled effect size. Google says a larger second phase is coming, alongside a partnership with the nonprofit Contrails.org.

Why an AI newsletter should care

The forecasting model is not the interesting part here. Predicting ice-supersaturated air is a weather problem with a long research history behind it. The hard part is proving the intervention worked, when the thing you prevented is by definition unobservable and the instrument that measures it belongs to the organisation making the claim.

That is structurally the same problem as a vendor benchmark — a self-reported score on a self-selected sample, with the counterfactual supplied by the vendor's own tooling. It is worth naming as one. The trial may well be measuring something real; the point is that, on what has been published, a reader cannot tell.