A meteorologist stares at a pressure map, trying to figure out where the cyclone will go. They have a model, but it is flawed, and fresh measurements, but they are incomplete. Merging the two so that the forecast remains physically meaningful is harder than it seems: classical methods sometimes produce negative temperatures or impossible probabilities.

The new approach views the problem through the lens of quantum mechanics. Instead of ordinary numbers, operators describe the state of the system, and their transformation rules guarantee that positive quantities stay positive. It is like a translator who condenses a long text but never flips its meaning into the opposite.

To test the method, they took the chaotic atmospheric model Lorenz-96. Assimilating data on the slow variables produced forecasts with an error of about 0.35 and a correlation of 0.95 at zero lead time. As the lead time increased, the error grew, but the uncertainty estimate honestly tracked the actual error, much like an experienced navigator marks a fog zone on the chart.

Next, they applied the method to El Niño in a climate model with 44 414 ocean temperature observation points. Useful forecasts of the Niño-3.4 index persisted up to 12 months, comparable to the best existing approaches, but here they also obtained full probability distributions, not just point estimates.

A translator that never loses the meaning turns out to be useful not only in theory. Already today, such an approach can be applied in weather, climate, and ocean forecasting, where preserving positivity and estimating uncertainty are crucial. And in the future, it could migrate to quantum computers, where such operations are especially natural.