Evaluating Numerical Predictions of Meteorological Features
As with any forecast, it is important to understand the quality of the predictions. Methods of evaluating “feature-specific” predictions are actively being developed by the research community. In this study, we apply subjective feature-based evaluation methods using an Euclidean distance approach to a series of numerical weather prediction forecasts. These results will be compared to “traditional” forecast verification statistics that are computed as a function of the difference between observed and predicted values. We will also compare the results of the subjective evaluation methods to automated techniques for evaluating feature-specific predictions. The goal is to gain insight on the quality and usefulness of the various forecast evaluation methods and to determine whether new objective verification methods are providing information that is consistent with subjectively-determined forecast evaluations.
This study was conducted as part of a new sophomore-level, research-oriented laboratory at Purdue University in the Atmospheric Science program.