Referee comment

Overall I feel positive about the manuscript. The analytical work provides some interesting interpretations on the estimation of model error covariances. The numerical demonstrations are also interesting. However,