NG24A-08
The Differences between Ensemble and Variational Methods and Implications for their Fusion
Tuesday, 15 December 2015: 17:45
300 (Moscone South)
Daniel Hodyss1, Craig H Bishop1,2 and Matthias Morzfeld3, (1)US Naval Research Laboratory, Monterey, CA, United States, (2)Naval Research Lab, Monterey, United States, (3)University of California Berkeley, Berkeley, CA, United States
Abstract:
Recently there has been a surge in interest in coupling ensemble-based data assimilation methods with variational methods (commonly referred to as 4DVar). Here we discuss a number of important differences between ensemble-based and variational methods that must be accounted for if one attempts to fuse these methods. The main difference between ensemble-based and variational methods is the interpretation of what the forecast covariance matrix is that is required by a data assimilation system whose state estimate is to be the posterior mean (ensemble method) or the posterior mode (4DVar). This difference has important implications to both the interpretation of the resulting state estimate and the design of data assimilation algorithms that fuse aspects of ensemble and variational techniques.
