Analysis component of the Kalman filter
Gain matrix (or statistical weights): K
Innovations (or observation departures): y-HXb
- Kalman filter adds a temporal dimension to the the modelisation of the background-error covariances which now becomes time and flow dependent