Data Assimilation and Predictability: Ensemble-Based Data Assimilation

This article introduces the algorithm of ensemble-based data assimilation (EDA) and the main issues in its application to atmospheric sciences. EDA is drawing increasing attentions in data assimilation community mainly due to its flow-dependent background error covariance determined using a short-range ensemble forecast and ease of implementation. Many types of EDA have been applied with different models at different scales in both research and operational or quasi-operational communities. Various aspects involved in EDA are discussed including observations, ensemble initialization, sampling error, covariance inflation and localization, model error, verification, nonlinearity and non-Gaussian errors, intercomparison, and hybrid with variational schemes.