TROPICAL CYCLONE TRACK PREDICTION USING HIGH-RESOLUTION SATELLITE DATA WITH A NEW METHODOLOGY

Tropical cyclone track forecasting remains a difficult problem. Here, we summarise eleven forecast cases, almost all in the Australian Regio n (AR), each of which was regarded as a difficult forecast case. There are three new aspects in these forecasts. First, the modelling and da ta assimilation have been performed at very high (15 km) resolution. S econd, a newly developed data source, namely, high spatial and tempora l resolution cloud-drift winds (CDWs), has been used to augment the of ten quite poor observational database upon which operational forecasts currently are based. Finally, a range of continuous assimilation sche mes, including recently developed four-dimensional (4-D) variational a ssimilation and hourly nudging have been tested. In these cases, conve ntional forecast guidance and CLIPER (a statistical model (Neumann 197 2) based on CLImatology and PERsistance) usually were poor. In contras t, the continuous assimilation procedures, namely, one-hourly nudging and 4-D variational assimilation, took advantage of the high spatial a nd temporal resolution wind data and provided much improved forecasts, especially beyond 12 hours. Overall, the initialisation provided by c ontinuous assimilation, combined with a substantial high spa tial and temporal resolution database and high resolution modelling. has shown a capacity to improve greatly the accuracy of tropical cyclone track f orecasts over the data-sparse oceans.