Monday, 13 January 2020: 8:45 AM
258A (Boston Convention and Exhibition Center)
The VORTEX-SE program aims to reduce loss of life and property from tornado damage. Among the challenges of building a probabilistic convection-allowing forecasting system, there is evidence that increased resolution of a limited-area numerical weather prediction model may better represent the updrafts associated with tornadogenesis in the southeastern US. These may occur in both cellular and linear convective modes. However, a threefold increase in grid resolution requires over a thirty-fold increase in computational demand; as such, given a finite computational resource, the number of ensemble members must be balanced with grid spacing.
This presentation details results comparing 3-km and 1-km ensemble forecasts of severe-thunderstorm outbreaks to assess the impact of increased resolution on storm characteristics. Evaluation is performed with a range of scoring techniques, including object-based and scale-aware methods, reflecting the multi-faceted challenge of storm-scale verification. The object-based verification discriminates between linear and cellular systems, and reveals common differences between 3-km and 1-km storm-object characteristics (e.g., size, updraft speed).
Verification was grouped as (1) deterministic, traditional methods based on point-wise contingency tables; (2) a scale- and time-aware probabilistic metric, and (3) a novel evaluation of thunderstorms via object identification and information theory. Results are mixed; 1-km forecasts perform better in detecting low- and midlevel rotation, but at cost to weak/moderate reflectivity forecasts. The nature of improvement was sensitive to the case, variable, forecast lead-time, and magnitude/severity, precluding a straightforward aggregation of results. However, the distribution of object-specific information gain over all cases shows greater average gain evaluating the 1-km storm objects. Further to the mixed results, we reiterate the importance of verification methodology appropriate for the problem of interest.
This presentation details results comparing 3-km and 1-km ensemble forecasts of severe-thunderstorm outbreaks to assess the impact of increased resolution on storm characteristics. Evaluation is performed with a range of scoring techniques, including object-based and scale-aware methods, reflecting the multi-faceted challenge of storm-scale verification. The object-based verification discriminates between linear and cellular systems, and reveals common differences between 3-km and 1-km storm-object characteristics (e.g., size, updraft speed).
Verification was grouped as (1) deterministic, traditional methods based on point-wise contingency tables; (2) a scale- and time-aware probabilistic metric, and (3) a novel evaluation of thunderstorms via object identification and information theory. Results are mixed; 1-km forecasts perform better in detecting low- and midlevel rotation, but at cost to weak/moderate reflectivity forecasts. The nature of improvement was sensitive to the case, variable, forecast lead-time, and magnitude/severity, precluding a straightforward aggregation of results. However, the distribution of object-specific information gain over all cases shows greater average gain evaluating the 1-km storm objects. Further to the mixed results, we reiterate the importance of verification methodology appropriate for the problem of interest.
- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner