Investigating uncertainty in predictions

The NCAR ensemble project has demonstrated the feasibility of a real-time, conterminous U.S. coverage, ensemble analysis and forecast system for high-impact, short-range weather predictions.  The extended period of continuous forecast operations has evolved into a one-of-a-kind multi-year dataset of high-resolution ensemble forecasts that has garnered considerable interest within the research community.  Community research activities with the NCAR ensemble dataset span investigations of snowfall prediction in the Inter-Mountain West, power interruptions in the Northeast, and predictions of severe thunderstorms and flash floods among others. Recent advances in applying object-based methods enable extraction of richer data from simulations (e.g., Sobash et al. 2016) that can be used to relate metrics of simulated extreme storms against characteristics of the near-storm environment.

Uncertainty (Predictability) figure

The ensemble storm viewer was recently added to the NCAR ensemble website suite of probabilistic forecast guidance. This new tool enables interactive investigation of forecast severe storm objects. For example, simulated storms that produce strong mid-level rotation along with hail forecasts of one inch or larger (shown), analogous to forecasts of supercell thunderstorms capable of producing high-impact weather hazards, are plotted along with characteristics of the object such as how rotation intensity varies in time and space (shown) for a 48-hr ensemble forecast initialized from May 27, 2017.

Understanding the role of turbulent processes in mesoscale weather systems is a forefront area of research linking mesoscale and microscale meteorology, and is becoming practically important as the resolution of prediction models begins approaching scales where the largest turbulent eddies can be resolved.  MMM scientists have been analyzing turbulence-resolving simulations of tropical cyclones to determine the effects of turbulence on the wind field with particular attention to low-level extreme values.  Examining turbulence in simulated hurricanes, a coherent structure (“misovortex”) associated with peak wind gusts in the simulations has been identified, which develops when a perturbation in the strongly sheared surface layer below ~50 m is tilted and then amplified at the inner edge of the eyewall.   Maximum wind gusts in eddy-resolving turbulence simulations of tornadoes have shown that present-day surface observing systems in field projects probably have a systematic low bias in wind speed, as large as 45 m/s.  

Uncertainty figure

The average properties of a "misovortex," which is associated with the strongest wind gusts, in a large-eddy simulation of a hurricane.  Color shading shows wind speed (m/s) at 8 m ASL, black vectors illustrate the near-surface flow relative to the misovortex, and white contours denote surface pressure.

The NCAR ensemble project has demonstrated the feasibility of a real-time, conterminous U.S. coverage, ensemble analysis and forecast system for high-impact, short-range weather predictions.  The extended period of continuous forecast operations has evolved into a one-of-a-kind multi-year dataset of high-resolution ensemble forecasts that has garnered considerable interest within the research community.  Community research activities with the NCAR ensemble dataset span investigations of snowfall prediction in the Inter-Mountain West, power interruptions in the Northeast, and predictions of severe thunderstorms and flash floods among others. Recent advances in applying object-based methods enable extraction of richer data from simulations (e.g., Sobash et al. 2016) that can be used to relate metrics of simulated extreme storms against characteristics of the near-storm environment.