1: \begin{abstract}

2: We present a new strategy for filtering high-dimensional

3: multiscale systems characterized by high-order non-Gaussian statistics

4: using observations from leading-order moments. A closed stochastic-statistical

5: modeling framework suitable for systematic theoretical analysis and

6: efficient numerical simulations is designed. Optimal filtering

7: solutions are derived based on the explicit coupling structures of

8: stochastic and statistical equations subject to linear operators,

9: which satisfy an infinite-dimensional Kalman-Bucy filter with conditional Gaussian dynamics.

10: To facilitate practical implementation, we develop a finite-dimensional stochastic filter model that approximates the optimal filter solution. 

11: We prove that this approximating filter effectively captures key

12: non-Gaussian features, demonstrating consistent statistics with the optimal filter first in

13: its analysis step update, then at the long-time limit guaranteeing stable convergence

14: to the optimal filter. Finally, we build a practical ensemble filter

15: algorithm based on the approximating filtering model, which enables accurate recovery of the true model statistics. 

16: The proposed modeling and filtering strategies are applicable to a wide range challenging problems in science and engineering, particularly for statistical

17: prediction and uncertainty quantification of multiscale turbulent

18: states.

19: \end{abstract}

20: