Maintaining Realistic Uncertainty in Model and Forecast
Abstract
LONG-TERM GOALS. My long term goal is to better understand ensemble forecasting in general and the dynamics of initial uncertainty given only imperfect models when predicting any physical system. Of particular interest to me are methods which aim to evaluate predictions of high dimensional systems like the atmosphere, methods to best combine different models, and determination of aims which are both achievable by the forecaster and useful for the user. OBJECTIVES. I wish to establish exactly how uncertainty in the initial condition and imperfections in the model limit the prediction of nonlinear dynamical systems (NDS) like the atmosphere and ocean. Of particular interest is the evaluation and interpretation of ensemble forecasts, now routinely made at the European Centre for Medium-Range Weather Forecasting (ECMWF) and the National Center for Environmental Prediction (NCEP). Particular questions include (i) how to best distribute computational resources between using larger ensembles versus higher resolution models, (ii) when to select a best model versus when to use an ensemble over models, (iii) how to evaluate current ensemble formation schemes and (iv) whether the answers to these questions change when regional forecasts are of interest. While the focus is on geophysical fluid dynamics, the insights gained are applicable across a wide range of forecasting and control systems, including mechanical and industrial processes and general questions of data analysis and assimilation..
Document Details
- Document Type
- Technical Report
- Publication Date
- Sep 30, 1999
- Accession Number
- ADA630144
Entities
People
- Leonard Smith
Organizations
- University of Oxford