基于ARIMA模型的台风频次预测

Typhoon generation frequency prediction based on ARIMA model

基于日本东京台风中心的逐月台风生成频次资料,为最近30 a的台风逐月生成频次时间序列建立了合适的ARIMA乘积季节模型,并用该模型对2010—2014年的台风逐月生成频次进行了预测。结果表明:时间序列中的ARIMA乘积季节模型在预测台风频次方面具有一定的预测技巧,除2013年外,其余4 a模型提前1 a预测的全年台风生成总数的异常情况均与观测一致。

In this paper, an ARIMA product seasonal model is established for the monthly typhoon generation frequency time series during the last 30 years based on the monthly typhoon generation frequency data from Tokyo typhoon center of Japan. The monthly typhoon generation frequency prediction for the years 2010-2014 is completed through this model. The results show that, the ARIMA product seasonal model performs quite well in predicting the typhoon generation frequency. Except for 2013, the anomalies of the total annual typhoon generation frequency of the remaining four years predicted one year in advance by this model are all consistent with the observations.