Monday, 14 January 2002: 11:14 AM
Using Verification Techniques to Evaluate Differences Among Convective Forecasts
Although convective forecasts reside at different spatial and temporal scales and often attempt to capture different aspects of convective activity, it is important that these forecasts be compared and evaluated so that the underlying differences can be determined and used by aviation decision-makers. Understanding the differences among the forecasts can lead to improvements in the forecasts and can provide additional information to decision-makers so that the correct forecast (or combination of forecasts) is applied to the appropriate weather problem. The challenge, however, is to develop verification techniques that not only address the particularities of each forecast, but can be applied similarly to each of the forecasts. For instance, difficulties arise when trying to modify the verifying observation dataset to fit the forecast resolution, identifying appropriate datasets that represent what the forecasts are trying to capture, and making informative comparisons between the forecasts so that improvements can be made. In this paper, we will present the statistical verification approaches that were developed for a convective forecast intercomparison exercise during summer 2001, which were used to evaluate very different types of convective forecasts, and will demonstrate how these methods can be used to understand the underlying differences among the forecasts.