dc.description.abstract | During summer and autumn seasons, tropical cyclone has been the main source of precipitation in Taiwan, but it also leads to serious disasters. If we could predict the formation of tropical cyclones, that will be helpful to the disaster mitigations. There were some researches employed satellite data to estimate the total heat energy and relative vorticity during typhoon formation period, and tried to find possible signals or thresholds as whether typhoons will occur or not. However, they did not consider the possible bias induced by using a fixed computing coverage. Therefore, this study will use a dynamic computing coverage by considering the cluster size variation of the tropical cyclone system, and employ SSM/I and QuikSCAT satellite data to estimate the total heat energy and relative vorticity, respectively, in finding better thresholds for these two physical values as whether typhoons will occur or not in the Northwest Pacific.
This study selects 106 tropical cyclone cases during May to November, 2000-2007 in the Northwest Pacific, two-thirds of these cases are used to establish the formation thresholds, one-thirds of these cases, which total to 35 cases, are regarded as dependant cases for verification. The result shows that if we used a fixed tropical cyclone system size, there were 31 cases can be announced almost two days earlier before the official JTWC warnings were issued, and the prediction accuracy reaches 88.6%. If we considered the cluster size variation of these tropical cyclone systems, the prediction accuracy could be raised to 91.4%. It reveals that the dynamic tropical cyclone system size could improve the prediction accuracy. Furthermore, the verification result of 24 independent cases during 2008-2009 shows that 18 cases could be predicted before JTWC. It could improve tropical cyclone prediction accuracy by considering the cluster size variation of tropical cyclone systems.
| en_US |