Nature
PUBLISHED
How far in advance can meteorologists forecast the weather without getting it so wrong that they seriously upset umbrella manufacturers and ice cream shops? It turns out there is a theoretical limit beyond which it’s practically impossible to predict the weather with any real certainty.
Today, it’s possible to make “skillful forecasts” within 14 days. Beyond that, it’s still possible to get a reasonable stab, but things start to get increasingly hazy and less predictable as the days go on.
However, a new study suggests a higher ceiling for accurate weather forecasts than previously expected. That limit is apparently around 129 days, beyond which quantum-scale uncertainty takes over and makes a forecast too noisy to use.
How do we predict the weather?
Modern weather forecasting is all about gathering huge quantities of data from every corner of the globe – measuring everything from air temperature and surface temperature to air pressure, humidity, wind speed, and ocean currents – then plugging it into a computer model that can fathom all the different underlying dynamics at play.
The challenges come with the wealth of data points and the many variables within the atmosphere that dictate changes in the weather. Within this, there are also complex feedback loops that can magnify small errors over time.
This is the hallmark of a chaotic system, as illustrated in the famous adage of a butterfly flapping its wings and causing a tornado the following week. One slight miscalculation, or missing data point, and a prognosis can easily go to pot.
But if we knew all the right data points in the atmosphere to a sufficient level of accuracy and understood its underlying dynamics perfectly, couldn't forecasts theoretically stay accurate indefinitely? According to this new research, the answer is still no.
Quantum-scale uncertainty enters the equation
To find a theoretical limit to weather forecasts, a team of scientists led by Dr. Wei Zhang, a climate scientist at the University of Miami and the NOAA Cooperative Institute for Marine and Atmospheric Studies, looked deeply into the atmosphere's energy cycle.
They found that long-term weather forecasts will inevitably be tripped up by one unavoidable renegade: the quantum-level uncertainty of incoming solar photons.
Light from the sun arrives as photons, and quantum mechanics predicts that a photon's phase has no fixed value until it interacts with something. Until that point, this value is essentially unknowable, and so the behavior of a photon isn’t just difficult to predict; it is fundamentally beyond our ken.
Over relatively short timeframes, this uncertainty introduced by photons is manageable. But because solar energy eventually reaches every corner of the atmosphere, influencing virtually every molecule, the uncertainty adds up, overtaking and blurring the predictability of any model.
“The state of a system perfectly known at one time remains known, preserving information about the state forever. Noise entering into the atmosphere from the unknown quantum properties of photons, however, changes this situation,” the study reads.
The moment where this reaches its breaking point — the "energy turnover point" — works out to be roughly 129 days, give or take 7 days either side.
That might seem a relatively short time in a world where artificial general intelligence and god-like processing power are, apparently, just around the corner. However, the researchers were keen to reiterate that the current window of skillful forecasts is just 14 days, so getting to this limit would require an amost 10-fold improvement.
So, while the next few years could, theoretically, lift us closer to that 129-day limit, make no mistake, getting beyond that point would, they conclude, actually be impossible.
The new study is published in the journal Advances in Atmospheric Sciences.
