A Dynamical Initialization Scheme for Real-time Forecasts of Tropical Cyclones Using the WRF Model

To improve the skill of real-time forecasts for tropical cyclones (TCs), a dynamical initialization scheme is developed and implemented into a real-time forecast system for Northwest Pacific TCs based on the Weather Research and Forecasting (WRF) model. In this dynamical initialization scheme, each cycle run is initialized at 6 h before the initial forecast time (t0-6) and integrated for 6 h to the initial forecast time (t0) using the NCEP GFS analysis fields as the initial and lateral boundary conditions. The axisymmetric component of the model vortex after the 6 h spinup is then subtracted to replace the original axisymmetric vortex in the analysis field at t0-6 for the next cycle run. The cycle run will be terminated once the intensity of the simulated TC is comparable to the observed. If for any case, the storm does not deepen during the first two cycles, a bogus vortex, which makes the storm intensity about 80% of the observed, will be embedded into the initial cycle time for the subsequent cycles. To reduce the model bias in simulating the large-scale environmental flow, a large-scale spectral nudging approach is applied to the integration of each cycle run. This spectral nudging considerably improves the track forecast. In addition to the more realistic TC structure and intensity at the initial forecast time, the new dynamical initialization scheme also spins up all model physics, such as the planetary boundary layer and cloud microphysics, an advantage for a warm startup of the model forecast. To demonstrate the skill of the proposed dynamical initialization scheme, forecast experiments with and without the dynamical initialization are conducted for TCs over the Northwest Pacific in 2010. The results show that the dynamical initialization has an overall positive effect on track forecasts while improves the intensity forecasts substantially. In particular, the initial shock in storm structure and intensity is largely reduced with the use of the dynamical initialization. This is because the dynamical initialization improves the 3-dimensional TC structure and the spin-up of model physics at the initial forecast time.