Improving EnKF spin-up for typhoon assimilation and prediction

To initialize the mesoscale EnKF for a regional model, it is common to use initial conditions from the global (re)analysis products and initial ensemble perturbations constructed based on the 3D-Var background covariance. Such initial conditions don't have enough mesoscale information and the perturbations are less than optimal due to the lack of mesoscale flow-dependency. Therefore, mesoscale EnKF requires a spin-up period to reach its asymptotic level of accuracy. In addition, for the case of typhoon assimilation, such spin-up usually corresponds to the developing stage of typhoons, when satellite data supposed to be valuable on the open sea. Thus, the importance of the satellite data, e.g. the satellite surface wind, may not be best explored with an EnKF system affected by spin-up.

To improve the analysis quality during the spin-up, the “running in place (RIP)” method proposed by Kalnay and Yang (2010) is implemented based on the framework of Local Ensemble Transform Kalman Filter (LETKF) with the Weather Research and Forecasting (WRF) model. Results from OSSE experiments show positive impact from the RIP method on adjusting the dynamical structures of the typhoon during the spin-up and improving the forecast track and intensity as well. With the RIP, satellite surface wind data have more influence on typhoon assimilation and prediction.