A PNN prediction scheme for local tropical cyclone intensity over the South China Sea

作者：Huang, Xiaoyan; Guan, Zhaoyong^*; He, Li; Huang, Ying; Zhao, Huasheng

来源：Natural Hazards, 2016, 81(2): 1249-1267.

DOI：10.1007/s11069-015-2132-9

• Probabilistic neural network (PNN)
• South China Sea local tropical cyclone (SLTC)
• TC intensity
• Climatology and persistence (CLIPER)
• Multiple linear regression (MLR)

A nonlinear tropical cyclone (TC) intensity scheme is introduced with probabilistic neural network (PNN) approach and is based on climatology and persistence (CLIPER) factors to predict local TC intensity in the South China Sea from May to October in 1949 to 2012. In this study, we investigate the local TCs that generate over the South China Sea and the maximum wind speeds near the center of TCs. The performance of the new prediction model is assessed with mean absolute error and forecast trend consistency rate. Results indicate the followings: (1) the new model is effective because of its low error rate. The absolute forecast error proportion of samples, which is less than or equal to 5 m/s, is more than 80 % in 24 h, 60 % in 48 h, and 50 % in 72 h. (2) The prediction capacity of the PNN model is more powerful than that of the CLIPER model and the multiple linear regression model based on the same samples and predictors. (3) The maximum skill level based on 17 forecasts is over 0.6, which is suitable for operational TC intensity application.

• 出版日期2016-3
• 单位广西壮族自治区气象减灾研究所; 南京信息工程大学