Forecast Verification Progress, Challenges, and Prospects (Invited Presentation)

Over the last 15 years, the science and practice of forecast verification has undergone significant development and been extended into areas that couldn't be imagined 30 years ago. Some examples include the evaluation of very high resolution precipitation and cloud predictions, the application of uncertainty measures to examine the statistical significance of verification results, the development of new approaches for the evaluation of extremes, the development of methods to evaluate the performance of ensembles, and the application of image analysis approaches to investigate spatial attributes of forecast performance. In addition, numerous software packages and tools have been implemented that make the job of verification more feasible. With these advances, new challenges have also become evident, including the need to appropriately account for observation uncertainty in verification studies and the need to apply methods that are appropriate for specific applications (e.g., to evaluate predictions from integrated precipitation, hydrologic and land surface models). Developing methods that are appropriate for such specific applications requires an understanding of the processes connecting the different components of the forecasting and application system, identification and interpretation of new observation sources (e.g., from satellite measures), and understanding the links between the various components of the systems. This talk will consider the current status of forecast verification science and practice and discuss some of the challenges that still remain. In addition, prospects for coping with some of these challenges will be considered.