An Efficient Dual-Resolution Ensemble Data Assimilation Approach and Tests with the Assimilation of Doppler Radar Data

A new efficient multi-resolution data assimilation algorithm is developed based on ensemble Kalman filter method (EnKF) and tested with simulated radar data for a supercell storm. In it, radar observations are assimilated on single high-resolution grid by using flow-dependent background error covariance estimated from low-resolution ensemble forecasts. The single high-resolution analysis is used to adjust the ensemble mean of the low-resolution analysis. It is shown that flow-dependent and dynamically consistent background error covariances estimated from the low-resolution forecast ensemble can be used successfully for the assimilation of radar observation into single high-resolution model. In general, the system is able to reestablish the model storm pretty well after a number of assimilation cycles. This method has advantage of reducing the computational cost of ensemble Kalman filter assimilation method. For the ensemble forecast, the single very resolution model run may provide more small-scale details, therefore reduce the impact of model error; while the use of low-resolution ensemble allow for a better estimate of the forecast error covariance by increasing the ensemble size. Forecast of the supercell storm starting from the high-resolution analysis is found to remain very close to the model-generated truth at least one and half hours.