Current issues...
Impact of model error
- 4D-Var implicitly assumes that the model is perfect and that any difference between a forecast and the observations is related to an error in the intial conditions
Simplifications are introduced in the model used to reduce the computational cost
- What is the impact of using a simplified physics on top of reducing the resolution
- Consistency between the simplified physical parameterizations and those used in the high resolution operational model with sophisticated parameterizations
Background term used in 4D-Var requires flow dependent forecast error covariances
- Use of a simplified form of Kalman filter is proposed to properly cycle 4D-Var (e.g., reduced rank extended Kalman filter, ensemble Kalman filter, etc.)