What is the relevance of the butterfly effect for practical weather prediction?

• Applicant: Dr. Tobias Selz
• Subject Area: Atmospheric Science
• Term: since 2024
• Project identifier: Deutsche Forschungsgemeinschaft (DFG) - Project number 548044620

Even if perfect observations and models were available, the time interval for which weather forecasts can be accurate is limited. This intrinsic limit is linked to fundamental physical characteristics of the earth's atmosphere, wherein minor initial errors rapidly amplify on small scales and subsequently disperse – a phenomenon known as the “butterfly effect”. Since current errors in the initial condition estimation are significantly larger than “butterflies”, it is in principle still possible to substantially improve weather forecasts by reducing these errors. Previous research suggests a potential improvement of approximately 5 days, implying that the quality of a typical recent 7-day forecast could extend to a 12-day forecast, if an optimal observational and data assimilation system became available. These estimates are, however, based on averages over large regions and derived from small samples, primarily focused on midlatitudes. The goal of this proposal is to determine how the intrinsic limit and the improvement potential varies in space and time. Special attention will be given to cases that show unusually low predictability in current operational weather prediction systems (sometimes referred to as “forecast busts”). It is possible that in such cases, the intrinsic limit is transiently so short that there is essentially no improvement potential left. This outcome would have significant implications for developers of weather prediction systems. Furthermore, my plan involves expanding the analysis of intrinsic predictability beyond midlatitudes to include other regions of the world such as the tropics. Prior research suggests a longer intrinsic limit and a greater improvement potential in tropical areas. These quantitative predictability estimates will be complemented by process-based analyses to deepen our understanding of the butterfly effect and error growth and propagation in weather forecasting. Regarding intrinsic predictability, I am proposing to explore the significance of the kinetic energy spectrum, the role of equatorial waves in the tropics, and the role of slantwise ascent in warm conveyor belts in the midlatitudes. To achieve these goals, a large sample of ensemble simulations is required. To address computational feasibility, these simulations will be conducted at a relatively low resolution, but supplemented by a stochastic convection scheme to account for the missing convective motions and variability. In a second set of numerical experiments, a small, carefully selected sample of the most interesting cases will be revisited through global high-resolution experiments employing explicit convection, which are considered the most accurate numerical setup currently attainable.

• DFG Programme: Research Grants