Deep learning-enhanced ensemble-based data assimilation for high-dimensional nonlinear dynamical systems

Ashesh Chattopadhyay, Ebrahim Nabizadeh, Eviatar Bach, Pedram Hassanzadeh

Abstract: Highlights•EnKF-based DA requires large ensembles and are computationally expensive.•Smaller ensemble size leads to sampling error and inaccurate state estimation.•ML-based H-EnKF uses large data-driven ensembles at low computational cost.•H-EnKF outperforms EnKF in terms of accuracy of estimated states at the same cost.