Ultra Rapid Data Assimilation for Real Time Weather
The growing network connectivity of high quality environmental sensors placed in road weather stations, aircrafts and automobiles makes now available meteorological observation streams with measurement rates of the order of minutes. This observational frequencies challenge currently operational forecasting systems as the assimilation of observations at this rate is often prohibitively expensive and typically aggravates the problem of introducing imbalance into the analysis during the assimilation step. In order to address these problems, an Ultra-Rapid Data Assimilation (URDA) algorithm is being developed at the German Weather Service (DWD) for the regional prediction system COSMO-KENDA. The URDA methodology is based on the preemptive forecasting concept, which allows the generation of observationally constrained ensemble forecast while avoiding additional model reinitializations. Here the ensemble transformation matrix, as given for example by an ensemble square root filter, is employed to update a reduced set of the state variables of an existing model forecast. The feasibility of this new method has been successfully tested in both linear and nonlinear control systems. For the COSMO-KENDA system the first URDA experiments show very promising outcomes in particular for leadtimes of the order of minutes, where forecasts generated by traditional sequential DA cycles are heavily polluted by spinup waves.
Dr. Walter Acevedo German Meteorological Service (DWD)