For 36 days, tropical cyclone Freddy raged over the Indian Ocean, the longest-lived cyclone on record. Meteorologists tracked it from satellites, aircraft, and ground stations. But one detail eluded all instruments: the precise air density deep within the vortex. And that density determines where the cyclone will turn tomorrow.

The muon flux behaves like a thermometer that measures the temperature of the whole body, not just under the armpit. Cosmic rays slam into the atmosphere and produce showers of particles. Among them are muons, heavy relatives of the electron. The denser the air, the more muons die on their way to the ground. A surface detector catches this flux and, from it, reconstructs air density not at a single point but across a large volume around itself.

The authors built a numerical experiment: the reference weather model WRF reproduced cyclone Freddy, and from it they generated synthetic observations of the muon flux. These were assimilated using an ensemble Kalman filter, a method that corrects forecasts based on an ensemble of 50 simulations with a localization radius of 640 km. It turned out that muon data reduce the error in surface pressure by up to 10% or more compared to a free forecast. This is stronger than the effect of a point pressure measurement at the same location. The reason is simple: muons sense density over a large volume around the detector, so correlations with pressure away from the measurement point are stronger. For a noticeable advantage, a detector with an effective area of just 0.27 m² and hourly accumulation would suffice.

It was known that the atmospheric muon flux depends on air density, but no one had shown that this dependence could be used to improve weather forecasts in an operational data assimilation system. Now it is clear: assimilating the muon flux in an ensemble Kalman filter improves estimates of surface pressure, wind, and temperature, and the effect from muons is stronger than from a point pressure measurement at the same location. Already today, inexpensive muon counters could be deployed to improve local forecasts, especially in regions with sparse weather station networks, such as over oceans or in polar areas.

Astrophysical observatories like IceCube or Pierre Auger could become incidental suppliers of data for weather models. A natural next step is to use not just the total integrated flux but also its angular distribution: the anisotropy of muons carries information about the direction of atmospheric inhomogeneities. Another direction is optimizing localization and assimilation frequency for specific meteorological tasks. It is promising to deploy networks of inexpensive scintillation detectors in regions where traditional observations are scarce. The muon thermometer, applied to the atmosphere, has already shown it can catch the heat of a cyclone.