Recently high-resolution Global Climate Model (GCM) results show that future Tropical Cyclone (TC) characteristics, such as arrival frequency, developing process and intensity, will be changed by global climate changes. However, there are major two difficulties for assessment of future TC disaster. One is uncertainty of future projection of GCM, and another is shortage of sample TC data for statistical analysis. In this paper, we generated many synthetic TC data with Global Stochastic Tropical Cyclone Model (GSTCM) and estimated future changes of TC properties and reduced uncertainty by ensemble averaging of multi-GCM projection results. In addition, GSTCM which have empirical temporal correlation algorithm was improved for the reproducibility of arrival TC statistics by cluster analysis of TC data. This upgrade could pave the way to local future projection of TC disaster.