Ensemble-based data assimilation in tropical cyclone forecasting

A key problem in tropical cyclone track forecasting is the assimilation of observed data, to provide initial conditions for the model. The provision of reliable estimates of flow dependent error covariances is crucial. We use an ensemble of idealized MM5 tropical cyclone forecasts to test the efficacy of an ensemble-based Kalman Filtering technique (the Ensemble Transform Kalman Filter or ETKF). Using this technique, the computational demand is greatly reduced compared with traditional Kalman Filter-based assimilation schemes in which processing time and memory requirements are prohibitive. The quality of the ETKF scheme depends on the ability of the ensemble perturbations to capture the leading directions of analysis error. The ETKF also provides a unified framework for targeting observations in the vicinity of the cyclone.