ABSTRACT
The region of northwestern Pacific is often attacked by typhoons. During a typhoon, the sea water level rises rapidly, and storm surge disasters appear under the effect of strong wind and low pressure, which may cause serious economic losses and causalities. In order to reduce the typhooninduced damages, it is essential to provide a high-precision forecast service of storm surges in coastal areas, which needs a reliable forecasting of typhoon tracks. In recent years, many weather forecast centers have been using numerical models for the typhoon forecast, which, however, cause some errors to arise from the inaccurate initial boundary conditions, physical parameters, numerical schemes, etc. In order to reduce such kinds of errors, an ensemble of forecast methods based on the initial disturbance or the mode perturbation are usually employed (Zhang and Krishnamurti, 1997; Zhu and Dai, 2000; Zhou et al, 2003; Wang and Liang, 2007). These methods can generate multiple groups of forecast results through the combination of different initial conditions, physical parameters, or even different meteorological models (e.g. Duan and Wang, 2004; Rao and Srinivas, 2014). On the basis of statistical analysis, the methods have been proved to be able
2 METHODOLOGY
2.1 General introduction
This super ensemble method is proposed based on the weighted average of forecast results from different weather centers. Without any loss of representativeness, the 24-hour forecast data from four operational weather forecast centers-China Meteorological Administration (CMA), Japan Meteorological Agency (JMA), Joint Typhoon Warning Center (JTWC) of USA, and Taiwan Meteorological Center (TMC)—is used for the study. The forecast procedure is divided into two periods: training period and forecasting period. The weighted coefficients for each center are determined by the center’s performance of typhoon forecast in the training period. A higher weighted coefficient will be assigned to the center that performed more accurately in the training period.
