The Use of Weather Scenarios from Convective-Allowing Ensembles to Convey Forecast Uncertainty

Often, actions of ATM personnel are binary: the decision to set a plan or to move traffic is usually based on the present best forecast of the weather. If an ensemble forecast of convective weather is used, often it is simply the mean that is presented and used, even though the ensemble spread, indicating uncertainty, is also available. Ensembles contain far more information than simply their mean and their spread, and offer a set of unique forecast solutions that can be clustered into a most likely outcome, as well as a second most likely outcome. Using ensemble-based scenarios, traffic planners could seek strategies to mitigate weather-related impact associated with a small number of future atmosphere states. For example, decision makers could use this information to determine a NAS ‘plan B’ that mitigates multiple weather scenarios.

An object-oriented framework for determination of weather scenarios using the uncertainty information from forecast ensembles is presented in this work. Hierarchical/agglomerative clustering was first applied to the centers of mass of the precipitation objects to partition the regions according to the precipitation activity, followed by the application of principal component analysis (PCA) clustering to the precipitation field within each geographical cluster to form the scenarios within each region.

The results suggest the utility of this combined geographical and PCA clustering. PCA clustering shows promise in the formation of sensible scenarios that cluster similar regional patterns together. Separating the regions by geographical clustering prior to PCA clustering is demonstrated to increase the information content of the resulting scenarios by accounting for the regional aspect of the precipitation patterns.

The approach for the identification of forecast scenarios was developed using SREF output, but the goal is to apply this approach to an ensemble of convective-allowing members. Preliminary results using a prototype of the High-Resolution Rapid Refresh Ensemble (HRRR-E) will be included.