A study estimates that the maximum duration during which the atmosphere could retain a trace of its initial state is approximately 129 days. This figure describes a theoretical physical limit.
Today, a forecast is generally considered to retain useful information about the atmosphere’s overall evolution for around 14 days. The farther ahead we look, the more the small errors present at the start grow.
Pixabay illustration
The new study approaches the problem differently. It does not only seek to simulate the growth of tiny errors. The authors examine how much energy the atmosphere contains and how quickly that energy is renewed through radiative exchanges with space.
Even with perfect knowledge of the atmosphere, incoming solar radiation contains microscopic variations that are impossible to know exactly. According to the authors, this noise first appears at the smallest scales. Air movements then gradually transmit it to larger-scale phenomena.
After approximately 129 days, all the energy initially present would therefore have been replaced. Within the proposed framework, the atmosphere would then have lost all usable memory of its initial state. This would be the absolute limit, even with a perfect model, perfect observations, and perfectly known external conditions.
This does not mean that conventional weather forecasts could become reliable up to day 129. The researchers estimate that the truly useful horizon could, at most, gain about 57 days compared with the current 14 days. In their ideal scenario, a forecast quality achieved today at five days could, for example, be reached again around day 62.
The remainder corresponds to much weaker information. Between approximately 72 and 129 days, it would mainly concern very large atmospheric movements, with little value for forecasting rain or the temperature in a city. This part could nevertheless be of interest for certain specialized scientific applications.
This estimate is based on several assumptions, particularly how noise propagates between the different scales of atmospheric movements. The authors themselves indicate that their result will need to be confirmed or reconsidered by independent studies.
It also remains to be determined what fraction of the additional 57 days could actually become accessible through better observations and improved models.