Convective Meteorology (Mesoscale Dynamics)

Deciphering the significance of serial observation assimilation order

Chris Kerr

School of Meteorology

04 December 2015, 3:00 PM

National Weather Center, Room 5600
120 David L. Boren Blvd.
University of Oklahoma
Norman, OK

Serial observation assimilation using an ensemble Kalman filter (EnKF) is commonly used over simultaneous assimilation. A serial assimilation analysis is assumed identical to the corresponding simultaneous assimilation analysis. This assumption is truly valid if the observation error distribution is Gaussian. Also, the forward operator(s) must be linear without covariance localization applied to reach an identical analysis. When forward operators are nonlinear and covariance localization is utilized, analyses vary slightly when observations are assimilated in different orders. There is potential for an optimal order based on observation type and/or localization length. This exciting subject matter will be presented using an idealized supercell simulated by the Weather Research and Forecasting (WRF) model. Synthetic radar and satellite observations are assimilated using a variant of the ensemble square root filter available through the Data Assimilation Research Testbed (DART). While an astronomical number of assimilation orders can be tested, this study simply switches the observation type order. Through numerous experiments, the hope is to create some generalizations on observation assimilation order and its relationship to analysis accuracy.

Convective Meteorology (Mesoscale Dynamics) Seminar Series website