Rapid, Short-Term Forecast Adjustment Through Offline Ensemble Data Assimilation

Rapid updates of short-term numerical forecasts remain limited by the time it takes to frequently assimilate observations and run the dynamic model to produce new forecasts.   Statistical methods exist to rapidly adjust forecasts based on long-term, fixed covariances with observations as they become available (e.g., the LAPS method), but these methods do not sample the flow-depended covariances unique to the current weather situation. Ensemble forecasts and data assimilation techniques are able to provide and exploit these flow-dependent covariances and can describe the sensitivity of forecast fields to the current atmospheric state.  An ensemble forecast adjustment technique is explored that uses the ensemble square-root Kalman filter to assimilate observations as they become available and adjust not only the current analysis, but also subsequent forecast fields based on covariances between the ensemble estimate of the observation and future forecast states.  This allows rapid adjustment of forecasts "offline" without the expense of running the full dynamical model.  Furthermore, by utilizing ensemble techniques for the data assimilation, short-term forecast uncertainty from the ensemble is also adjusted by this method.