The 129-Day Limit: How Far Can We Predict Weather?

Imagine waking up to a weather report that remains meaningful four months from now. It sounds like science fiction. Yet a new study argues there may be a hard physical ceiling to how long the atmosphere can hold usable information — about 129 days.

The question of long-range predictability is as old as numerical weather prediction itself. Since the 1950s, forecasters and theoreticians have probed how tiny errors in initial conditions balloon into large forecast mistakes. That method—watching error growth—has taught us a lot, but it leaves a blind spot: what happens if those errors were far smaller than anything we can currently measure?

Two atmospheric scientists approached the puzzle from a different angle. Instead of tracking errors, they followed energy. If you knew the atmosphere's state perfectly and understood the governing physics, what could possibly destroy that knowledge? The surprising culprit: sunlight. Not its warmth, but the quantum-scale uncertainty carried by incoming photons.

Photons arriving from the Sun have phases that are fundamentally unknowable at the quantum level. Those tiny unknowns get folded into the atmosphere through radiation, then mixed and amplified by winds, convection, and chemistry. Over time, the uncertainty spreads, replacing the memory of the original, perfectly known state. The team calls the moment this happens the energy turnover point — the instant when internal atmospheric predictability peters out.

Crunching estimates of total atmospheric energy, solar input, and realistic observational uncertainties, the researchers arrived at a likely internal predictability limit of 129 ± 7 days. That number sits far beyond the roughly two-week horizon of modern high-skill forecasts. But it is not a promise of equal clarity across those extra months. According to their analysis, only about half of that extension would offer genuinely skillful guidance. For example, the kind of certainty we now expect for a 5-day forecast might, in principle, persist out to roughly 62 days, after which confidence fades.

The work, published in Advances in Atmospheric Sciences, comes from Dr. Wei Zhang of the University of Miami and the NOAA Cooperative Institute for Marine and Atmospheric Studies, together with Dr. Zoltan Toth, formerly of NOAA. The result is a theoretical ceiling for internal predictability — external drivers, like volcanic eruptions or sudden ocean changes, would still impose their own limits.

Theory like this reshapes ambition. It offers a benchmark: a distant horizon that separates what physics permits from what our instruments and models must achieve. Now the authors and others are building independent estimates to test the idea. If the 129-day figure holds up, it will mark both an invitation and a challenge for forecasting science — how far will we push toward that horizon?