11th Symposium on Integrated Observing and Assimilation Systems for the Atmosphere, Oceans, and Land Surface (IOAS-AOLS)

7.3

Adaptive sampling with the Local Ensemble Transform Kalman Fitler (LETKF): implementation on a global model

Junjie Liu, University of Maryland, College Park, MD; and E. Kalnay

Adaptive observation strategies (AOS) aim to improve forecasts by adding additional observations at a few locations that have no standard observations. There are two basic kinds of adaptive strategies. One is adjoint-based technique, such as sensitivity to initial conditions and singular vectors to identify the sensitive regions in which additional observations will be taken (Palmer et al, 1998; Montani et al, 1999). This kind of method does not consider the effects of the changing background uncertainties and observation uncertainties. The other is ensemble-based technique such as the ensemble spread technique (Lorenz and Emanuel 1998; Morss 1998) and the adaptive methods directly derived from the output of the ensemble Kalman Filter (Bishop et al, 1999. Liu et al, 2005).

We discuss a simple ensemble spread method, a local Pa method that includes different observational errors in the selection, and a method combining both, all of which belong to the ensemble based techniques. They are based on output from the Local Ensemble Transform Kalman Filter (LETKF, Hunt, 2005), which is an accurate and efficient type of ensemble Kalman filter. These methods have already been proven to be accurate and efficient in Lorenz-40 variable model (Liu et al., 2006). With 15 ensemble members, they reach smaller analysis error than that of singular vector method implemented with a 1024 member ensemble Kalman filter method (Hansen et al, 2000).

Here, we further explore these different adaptive strategies on the SPEEDY model (Molteni, 2003), which is a primitive equation global model with a simple but complete set of physical parameterizations. We will compare these strategies with the system used at NCEP for operational adaptive observations (ETKF), evaluating these strategies in selecting the optimal flight routes rather than single targets. With this realistic model we will explore the use of interactive adaptive observation strategies for satellite observing systems including “observing 10% of the area to obtain 90% of the impact” as required for future lidar wind instruments. The best strategies to follow for tropical and extratropical adaptive observations, which may be quite different, will be also explored.

.

Session 7, Advanced methods for data assimilation (Part II)
Wednesday, 17 January 2007, 1:30 PM-2:30 PM, 210B

Previous paper  Next paper

Browse or search entire meeting

AMS Home Page