Improving Air Quality Data Assimilation by Covariance Regularization
*伟 林 (北京大学)
Data assimilation in air quality models is very challenging due to the nonlinearity of the dynamical system and non-Gaussianity of the state distribution. Borrowing ideas and techniques from recent developments in large covariance estimation, we review the existing covariance regularization methods for ensemble-based data assimilation and suggest some improvements. Numerical comparisons and illustrations will be reported.