Kernel Density Estimation Applied to Tropical Cyclones Genesis in Northwestern Pacific

摘要

It is recognized that although the conditions necessary for generating cyclones are actually present in the tropical latitudes for much of the year, the storms themselves are comparatively rare. In addition, the skill of forecasting the occurrence of individual tropical cyclones remains poor, largely because of our continuing ignorance o the physics underpinning tropical cyclone genesis. This situation gives an obstacle to predicting future changes in the origin of tropical cyclones. In this paper, authors will give a projection of changes origin of tropical cyclone in the future. First, the estimation of tropical cyclones in northwestern pacific genesis probability function is obtained by using the nonparametric kernel smoothing method which makes no assumption on the prior distribution form of observations and is different from parametric methods used traditionally. Second, by choosing the different bandwidth, the fitting probability density estimation is different. Lastly, from the comparison, the conclusion obtained is that the smoothing parameter of kernel is very critical. The optimal bandwidth value can be perfectly fitting the true density of tropical cyclones genesis, and then the future changes of occurrence will be modeled.