Â鶹ӰԺ

Recent applications of the hybrid particle-ensemble Kalman filter in Lagrangian data assimilation

Lagrangian data assimilation consists of directly assimilating the positions of (semi-) passive drifters in order to estimate the underlying flow field. It is characterized by a high-dimensional fluid flow state governing the movement of a low-dimensional, highly nonlinear drifter position state. The hybrid particle-ensemble Kalman filter is a recently-developed method that takes advantage of the benefits of two ensemble methods, the ensemble Kalman filter (EnKF) and the particle filter (PF), in the context of Lagrangian data assimilation. This method has already been shown to improve on the EnKF in low-dimensional idealized fluid flow systems (Slivinski et al 2015), with regards to both the mean estimates and quantification of uncertainty, and has recently been investigated in higher dimensions. Results will be shown with the hybrid filter and EnKF in a quasi-geostrophic fluid flow system, when drifters encounter different coherent structures (such as a stable center and unstable saddle.) Issues that have arisen, such as generating a reasonable ensemble of flow states in a higher-dimensional system and correctly assimilating multiple drifters, will be discussed, along with ideas for future applications of the hybrid filter.