| 本文已被：浏览 1541次 下载 1000次 |
码上扫一扫！ |
| RBF神经网络的汛期旱涝预报方法研究 |
|
王艳姣1, 张鹰1, 邓自旺2, 宋德众3
|
|
1.南京师范大学地理科学学院, 江苏, 南京, 210097;2.南京信息工程大学KLME实验室, 江苏, 南京, 210044;3.福建省专业气象台, 福建, 福州, 350001
|
|
| 摘要: |
| 运用福建省25个代表站汛期降水量资料,得到了能够反映全省旱涝状况指标,以此指标为预报量,运用相关分析和逐步回归分析方法,从前期海温场、大气环流场中选取了预报因子,并据此建立了福建汛期旱涝的多元线性回归和RBF神经网络预测模型。结果表明,RBF神经网络模型在历史样本拟合精度上、样本交叉检验和模型的实际预测能力上都明显优于传统的线性回归方法,该模型在实际预测中具有良好的应用能力和推广价值。 |
| 关键词: RBF神经网络 汛期旱涝 预测模型 |
| DOI： |
| 分类号: |
| 基金项目:国家科技部项目《福建省气候灾害短期气候预测业务服务系统》(2001DIB20116)南京信息工程大学KLME开放课题(KJS02108)资助 |
|
| THE APPLICATION OF RBF NEURAL NETWORK IN FORECASTING THE RAINY SEASON DROUGHT/FLOOD IN FUJIAN |
|
WANG Yan-jiao1, ZHANG Ying1, DENG Zi-wang2, SONG De-zhong3
|
|
1.Geography College of Nanjing, Normal University, Nanjing 210097, China;2.KLME of Nanjing University of Information Science & Technology, Nanjing 210044, China;3.Special Observatory of Fujian Province, Fuzhou 350001, China
|
| Abstract: |
| A provincial drought/flood index is constructed on the basis of precipitation data of rainy seasons from 25 representative stations in Fujian province. For the prediction of the index, the multi-line regression model and RBF neural network model (RNNM) are introduced and the series of predictors are selected from previous monthly SST and 500hPa height field data by means of correlation and stepwise regression. The results show that the RNNM is much better than multi - line regression model in terms of the precision of historical sample fittings, the value of sample intercrossing test and actual prediction ability. The model proves to be widely applicable. |
| Key words: RBF neural network drought and flood in rainy season forecast model |