摘要
使用人工神经网络方法建立了湖北省汛期 (6～ 8月 )总降水量的短期气候预测模型 ,该神经网络模型的输入是汛期前期 (2～ 4月 )的北半球月平均 5 0 0 h Pa高度场、海平面气压场和太平洋海温场的扩展自然正交展开 (EEOF)的前几个主要模态的时间系数 ,输出了湖北汛期降水场的自然正交展开 (EOF)的前 2个主要模态的时间系数。41 a历史资料的交叉检验表明 :样本试验的预报技巧评分平均为 0 .2 4 6 ,虽然该模型对各年的预报效果仍存在一定的不稳定性 。
The authors constructed artificial neural network (ANN) models of short-term climate forecasting to predict summer (June-August) precipitation in Hubei Province. The inputs of the model were the extended empirical orthogonal functions (EEOF) of the 500 hPa height field,the sea level pressure field in the Northern Hemisphere and sea surface temperature field over the Pacific before summer flood season (February-April), and the outputs were the empirical orthogonal functions of the summer precipitation totals of representative stations. The cross-validation over 41 years has shown that the forecasting skill score of ANN models is 0.246. Though the forecasting skills year by year are still unstable, positive skills exist obviously statistically for summer precipitation forecasting in Hubei Province.
出处
《气象学报》
CSCD
北大核心
2001年第6期776-783,共8页
Acta Meteorologica Sinica
基金
中国科学院大气物理研究所 LASG实验室提供资助
关键词
湖北
汛期
降水量
人工神经网络
短期气候预测
交叉检验
自然正交展开
Artificial neural network,Short-term climate forecasting,Cross-validation,Empirical orthogonal functions.
