Skip to content

Topic

AI Weather and Atmospheric Forecasting

Use of satellite imagery, flight data and machine learning to forecast atmospheric conditions with enough skill to drive operational decisions.

Current stories

productOne report1 publisher

DeepMind's extra day of hurricane warning rests on three years of historical storms

Google DeepMind's WeatherNext Cyclones gives hurricane forecasters about an extra day of accuracy on average, according to a Nature study of past storms. So far it has one live season on record, and that season is what forecasters will lean on when the next storm strengthens fast.

Publishers:gizmodo.com

Reality

Evidence55
Adoption60
Hype gap+25
Incentives60
Confidence50
scienceOne report1 publisher

Scarce storm-scale data limits AI forecasts of hurricane intensity, one researcher argues

AI weather models that rival physics systems globally lack the storm data to foresee jumps like Hurricane Polo's 24-hour rise to Category 5, a researcher says. The author's own research suggests chaos may still cap long-range forecasts even with better data.

Publishers:phys.org

Reality

Evidence40
Adoption
Not scored
Hype gap0
Incentives
Not scored
Confidence35
productConfirmed4 publishers

Contrail avoidance gets a budget: a £5m trial to see if AI forecasts can move aircraft

Operation Blue Skies will spend 30 months testing whether Google's contrail predictions are good enough for NATS controllers to shift aircraft 2,000ft over the North Atlantic.

Perspective Coverage

4 publishers
Builder
Builder 30%
Operator
Operator 44%
Investor
Investor 26%

Reality

Evidence58
Adoption10
Hype gap+20
Incentives60
Confidence55
buildConfirmed3 publishers

WeatherNext 3 refreshes a 5 km global forecast every hour off live satellite mosaics

Google's new model trains on sparse station observations instead of six-hour-old physics output, which is what buys the 5 km surface grid. Wind aloft still comes out at 25 km, the same as WeatherNext 2.

Perspective Coverage

3 publishers
Builder
Builder 37%
Operator
Operator 45%
Investor
Investor 18%

Reality

Evidence55
Adoption45
Hype gap+25
Incentives70
Confidence60
scienceOne report1 publisher

ETH Zurich spent a year copying 100 petabytes of NASA data onto servers next to Alps

The Swiss team now has the two things an AI hazard model needs, a full Earth observation archive and a top-tier machine, in the same room. Nobody has reported the forecast skill that pairing is meant to buy.

Publishers:phys.org

Reality

Evidence48
Adoption25
Hype gap+42
Incentives60
Confidence50
productOne report1 publisher

Cathay flew more than 80 flights on Google contrail forecasts sent over in-flight Wi-Fi

Google's satellite analysis puts the avoided warming at about 40%, with more than half of it coming from the Hong Kong to Singapore corridor, which is a better guide to where this pays than the trial's ultra-long-haul billing.

Publishers:thenextweb.com

Reality

Evidence55
Adoption35
Hype gap+22
Incentives78
Confidence55
scienceOne report1 publisher

DeepMind's WeatherNext 3 adds live satellite mosaics alongside traditional analyses

Google DeepMind says the model resolves temperature at 5 kilometres and updates every hour, and it hands the headline accuracy claim to an outside evaluator, Brightband, whose scores the announcement does not publish.

Publishers:deepmind.google

Reality

Evidence32
Adoption15
Hype gap+45
Incentives88
Confidence42
productOne report1 publisher

WeatherNext 3 moves Google from weather research to weather supplier

The model feeds Search, Maps and Gemini and ships on Google Cloud, so anyone whose product depends on a forecast now has a supplier to choose and an accuracy claim from a startup's benchmark to read closely.

Publishers:techcrunch.com

Reality

Evidence52
Adoption38
Hype gap+19
Incentives74
Confidence55
scienceOne report1 publisher

ECMWF's AI and physics ensembles swap places between five days out and two

A five-season Atlantic study led from the University of Miami found ECMWF's AI ensemble carrying more tropical cyclogenesis signal at 84 to 120 hours and less at 36 to 48, which makes it guidance rather than a substitute.

Publishers:phys.org

Reality

Evidence60
Adoption55
Hype gap−5
Incentives60
Confidence58
scienceOne report1 publisher

NOAA's 69% record El Nino call lands where forecast models are weakest

The odds now favour a record season. The AI forecasting being pitched at the countries with the least margin is the least tested on exactly that kind of event.

Publishers:phys.org

Reality

Evidence38
Adoption24
Hype gap+14
Incentives66
Confidence42