Improving the Assimilation Ability for the Extreme Eventsby Proposing a Nonlinear Machine Learning DataAssimilation Approach

Extreme weather and climate events impose substantial societal and economic costs under continued global warming, but their prediction remains a challenge in meteorology and other geosciences. Their emergence and development are results of nonlinear multivariate interactions within the earth system at a wide range of spatial and temporal scales. A favorable initial state is essential for triggering the evolution of extreme weather and climate events, except for large‐scale drivers, positive feedbacks and stochastic processes. Thus, accurate initialization of coupled systems poses a fundamental challenge in extreme event prediction. Several operational weather forecasting centers have successfully established their own covariance-based data assimilation (DA) systems to address this challenge. However, the conventional assimilation approaches, such as the ensemble Kalman filter (EnKF), tend to underestimate extreme events due to their inability to capture these nonlinear coupling features, given their reliance on linear background error covariance estimation. Thus, in this study, we aim to consider the complex coupling features in state estimation by leveraging the capabilities of machine learning (ML) algorithms in nonlinear representation. The novel ML-based assimilation method effectively and nonlinearly projects the observational information to the prior predictions, generating reliable analysis for extreme phenomena. This data driven approach effectively characterizes the time-variant and complex multivariate relationships, thereby nonlinearly projecting the innovation onto the ensemble subspace. This significant improvement enables the ML-based approach to increase the analysis accuracy for extreme phenomena by up to 66% over EnKF, and its ensemble increment distribution is well aligned with that of the target increments, showing the potential of data driven assimilation approach for advancing the capabilities of capturing and triggering the extreme events.