Predictability Assessment and Improving Ensemble Forecasts
ARIZONA UNIV TUCSON DEPT OF ATMOSPHERIC SCIENCES
Pagination or Media Count:
PROJECT GOALS AND OBJECTIVES. The PI continues to examine atmospheric predictability with the goal of improving ensemble forecasts at ranges of 12 hours to 10 days. The research is addressing several issues, including 1. Documentation of analysis uncertainty from mesoscale and global analyses. 2. Design of optimal ensemble forecast systems EFS, with an emphasis on precipitation forecasts. 3. Calibration of EFS output by artificial neural networks and other statistical techniques. 4. Design of stochastic physics parameterizations that improve under-dispersion in EFS s. Most progress during the past year involved topics 1 and 2, with limited progress on topics 3 and 4, so descriptions of new results will primarily emphasize topics 1 and 2. The PI also served as Co-Chief Scientist to Dr. M. Steven Tracton for ONR initiative on Predictability in the Atmosphere and Ocean, and presumably will continue to play some role on any follow-up ONR predictability initiative. DOCUMENTATION OF ANALYSIS UNCERTAINTY. As noted in last year s report, we are estimating the statistics of analysis errors Eo from differences between different analysis-forecast systems. The method gives a component of the analysis uncertainty that, although difficult to relate with high precision to Eo, quite likely denotes a reasonable lower bound of its magnitude. Although this methodology is not as comprehensive or precise as statistics from emerging ensemble data assimilations or from well-constructed observing system simulation experiments OSSE s, it is currently tractable, very economical, and useful guidance can be quickly obtained.
- Operations Research