Friday, 30 June 2017: 3:15 PM
Salon F (Marriott Portland Downtown Waterfront)
Through high-resolution deterministic and ensemble sensitivity experiments with both regional and global models, and with both realsitically large and minute idealized initial perturbation uncertainties, we seek to answer what is the ultimate predictability limit of multi-scale mid-latitude day-to-day weather phenomena. Highlights will be given to severe weather events such as midlatitude winter storms, tropical cyclones and tornadic thunderstorms. These experiments suggests such a limit may exist both in terms of overall global error energy at different wavelengths and in terms of feature-based verifications of individual events. Such a limit is intrinsic to the underlying dynamical system and instabilities even if the forecast model and the initial conditions are nearly perfect. Minute uncontrollable initial conditions originated from small-scale instabilities can grow upscale that will eventually limit the predictability of various weather at increasingly larger scales. More specifically, from a global perspective, on average, the practical predictability limit of the midlatitude weather by the current state-of-the-science global models from lead numerical weather prediction centers is about 10 days or so while the intrinsic limit is likely to be around 2 weeks, consistent with earlier estimates by Ed Lorenz in the 1960s. In other words, even with a perfect model, reducing the initial condition uncertainties to an order of magnitude smaller than the realistic current level of uncertainty will at most extend the deterministic lead times by 3-4 days for midlatitude day-to-day synoptic weather; much smaller room in forecast lead times will be for smaller scale phenomena. Nevertheless, this 3-4 more days’ limit remains an encouraging and impactful goal to be achieved that will bring significant socioeconomic benefits. Achieving this additional predictability limit needs coordinated efforts by the entire community through designing better numerical models performing at higher resolutions with better observations and through better use of observations with advanced data assimilation and computing techniques.
- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner