Focus: To explore advanced assimilation techniques, including Four-Dimensional Variational (4D-Var) and Ensemble Kalman Filter (EnKF) approaches, aimed at developing a unified WRF data assimilation system and with a particular focus on assimilating remotely sensed radar data at convective/mesoscales.

MMM's advanced data assimilation program researches a number of aspects of mesoscale data assimilation, including 4D-Var and EnKF approaches. As this work progresses, we make the advances available to the university community and operational centers.

The EnKF is a promising approach to developing a combined ensemble forecast and data-assimilation system. In essence, one begins with an ensemble of analyses and makes an ensemble of short-range forecasts (using the full nonlinear forecast model) to the time of the next available observations. This ensemble of forecasts is used to estimate the forecast covariances required to assimilate the new observations via the standard Gaussian formalism of the Kalman filter. Each ensemble member is then updated given the new observations by assimilating a set of perturbed observations (that is, the actual observations plus noise consistent with the observational uncertainty). This approach shares with 4D-Var the benefit of flow-dependent forecast covariances, but, unlike 4D-Var, it does not require the linearized or adjoint versions of the forecast model, nor does it require the off-line estimation of first guess forecast error covariances. The ensemble Kalman filter also has the attractive feature of providing a short-range ensemble forecast and of ?initializing'' the ensemble members for longer-range ensemble forecasts.

The optimal use of existing observation networks is also a key aspect of data assimilation research. With regard to the prediction of severe convection, Doppler radar data is a primary source of information. Efforts focus on the optimal preprocessing, quality control, and thinning of radar radial velocity and reflectivity observations.

Analysis and 3-hour forecast of the rain water mixing ratio at z = 2.25 km for the June 12, 2002 squall line observed during IHOP (lower panels). The reflectivity observations are shown in the upper panels for comparison. The forecast was performed using a cloud model initialized by data from four NEXRADs and a 4D-Var radar data assimilation system. ...more...

Related Links

The Data Assimilation Testbed Center at MMM

The Data Assimilation Testbed Center at RAL

WRF-Var User Guide

Want to learn more about what “Data Assimilation” is all about? View UCAR's Met Ed tutorial, “Understanding Data Assimilation: How Models Create Their Initial Conditions”.