Weather Predictability (Core Science Keynote)

Soon after the beginning of numerical weather prediction, the following question presented itself: “What degree of improvement in the prediction can be expected from a given improvement of the initial condition?” If only small improvements were to be obtained for a much more accurate initial condition, then there would be an “…intrinsic finite range of predictability, which cannot be lengthened by bettering the observations.” (Lorenz 1969). A consensus has formed behind Lorenz's idea that predictability is limited for flows with many scales of motion in which errors in small scales grow faster than, and spread to, errors in larger scales. This idea is supported by numerical experiments using fluid-flow models under varying degrees of idealization and approximation. However, as complicated as the latter models are, they are considerably simpler than numerical weather prediction models as they do not take into account moist convection, cloud microphysics, complex orography/physiography, flow regimes, etc. In this talk, I will briefly recount the historical context of Lorenz's radical concept of limited predictability, the consequences of which are still being grappled with by mathematicians, meteorologists and social scientists. For mathematicians limited predictability raises questions on the well-posedness of the fluid-flow equations. For meteorologists the existence of a scale-dependent, finite-time barrier helps in deciding which forecast problems are potentially tractable. For social scientists, limited predictability means there is a time beyond which the marginal (economic/social) benefits of forecast-error reduction are small relative to the marginal cost of forecast-system improvement; this work is essential for knowing how scarce resources are to be allocated in the collective weather-prediction enterprise. I will attempt to illustrate some of these ideas with some examples from current-day, experimental, real-time numerical prediction of small-scale convective weather systems. Finally I will describe what brought me to this subject, the relevant parts of my background and examples of the sorts of jobs where knowledge of weather predictability is an important asset.

Reference

Lorenz, E. N., 1969: The predictability of a flow which possesses many scales of motion. Tellus, v. 21, 289–307.