Weather May Be Theoretically Predictable Up to 129 Days Ahead, Study Suggests

Meteorologists have debated how far into the future weather could ever be predicted, even with perfect knowledge of the atmosphere, since numerical weather prediction began in the late 1950s. Past efforts to find that limit generally examined how small errors in current forecasts grow over time, but researchers writing in Advances in Atmospheric Sciences say that approach leaves a gap, because no one knows how errors smaller than today's would behave.

Lead author Dr. Wei Zhang, a climate scientist at the University of Miami and the NOAA Cooperative Institute for Marine and Atmospheric Studies, and co-author Dr. Zoltan Toth, formerly of NOAA, instead asked what would stop a perfect forecast from continuing indefinitely if the atmosphere's initial state and physical laws were known exactly. They found that one source of uncertainty would remain: the unknown quantum-scale phase of photons arriving in sunlight. As solar energy moves through the atmosphere and eventually reaches every molecule, they argue, this uncertainty spreads until the atmosphere loses the information from its precisely known starting state, a point they call the "energy turnover point."

Using estimates of the atmosphere's total energy, incoming solar energy and the uncertainty in those measurements, the researchers calculated a theoretical predictability limit of 129 plus or minus 7 days. Today's useful forecasts extend to roughly 14 days. The researchers estimate that about half of the additional time could offer a meaningful extension of current forecast skill, while the rest would provide only low-confidence guidance; under ideal conditions, they estimate, today's 5-day forecast skill might theoretically extend to about 62 days.

Zhang and colleagues are developing independent methods to test the proposed 129-day limit, according to the study, which was published on July 29 in Advances in Atmospheric Sciences.