返回
The importance ensemble information for an ensemble Kalman filter
雷荔傈
Nanjing University
Ensemble-based data assimilation has been widely used for geophysical applications. Ensemble-based data assimilation relies on ensemble forecasts to provide flow-dependent error statistics. The flow-dependent error statistics may not be able to fully present the errors of geophysical systems that often contain features cross scales. Before a unified data assimilation paradigm for a multi-scale geophysical system, the role of ensemble information that is essential for ensemble-based data assimilation need be understood. Traditionally, multi-scale data assimilation either uses an iterative assimilation procedure or divides ensemble information from small to large scales. Here, the ensemble information is examined from an aspect of ensemble mean and ensemble perturbations separately. For large scale features, ensemble mean is more prominent than ensemble perturbations, and vice versa for small scale features. Therefore, an analog offline ensemble Kalman filter (AOEnKF) is proposed for large scale data assimilation, which constructs ensemble priors from a control climate simulation for each assimilation time based on an analog criterion using proxy observations. Even though AOEnKF is an offline scheme, it has the ability to capture “flow-dependent” background error covariances that help spread observation information through climate fields. An integrated hybrid EnKF is proposed for small scale data assimilation, which updates both the ensemble mean and ensemble perturbations by a hybrid background error covariance in the framework of EnKF. The integrated hybrid EnKF approximates the static background error covariance by use of climatological perturbations through augmentation or additive approaches.