An Ensemble Filter Assimilation System for Global and Regional Models

Jeffrey Anderson

NCAR Data Assimilation Initiative

Ensemble filter methods for data assimilation have been under development for atmospheric and oceanic prediction for about a decade. Recent advances appear to have made these
methods competitive with existing variational assimilation methods. In addition to providing high quality state estimates, ensemble filters also provide a sample of the probability distribution
function of the state and are extremely easy to apply. In fact, it is possible to add an atmospheric GCM to a carefully crafted ensemble assimilation system with less than a week's worth of
effort. This ease of development has led to a surge in applications making use of ensemble filters.

The National Center for Atmospheric Research (NCAR) is developing a community facility for ensemble data assimilation called the Data Assimilation Research Testbed (DART). An overview of the key
algorithms in DART including a novel algorithm for dealing with model error will be presented in a tutorial form. The credibility of the system will be supported by results from numerical weather prediction experiments performed using NCAR's community atmospheric model (CAM). Results from assimilations with the Weather Research and Forecasting (WRF) regional model will also be discussed briefly.