海南岛秋季旱涝的模糊神经网络预测模型研究.pdfVIP

, Journal ofAnhuiAgr.i Sci. 2010, 38( 26): 14549- 14551 黄小燕 况玲玲 1, 2 2, 3 吴 慧 , 林 熙 ( 1. , 570203; 2. , 570203; 3. , 570203) 利用 1961~ 2008年海南岛 18 个市县观测站的秋季降水量和NCEP 再分析资料, 通过合 分析和相关分析方法, 选取前期海温 场位势高度场风场相对湿度场等环境场中影响海南岛秋季旱涝关键区作为因子, 经过因子的降维处理, 建立海南岛秋季旱涝的模 型结果表明, 综合考虑多个环境场因子构建的模型优于单 个或少数环境场因子构建的模糊神经网络预测模型, 其对历史样本的拟合 精度高, 独立样本对趋势预测的准确率高而稳定, 对特旱和特涝年也有较好的预测能力, 在新评分标准下 2004~ 2008年的独立样本预测 检验评分达 86, 可以投入业务使用 秋季旱涝; 模糊神经网络; 预测; 海南岛 P429 A 0517- 6611( 2010) 26- 14549- 03 Research on FuzzyNeuralNetwork PredictionModel forAutumn Droughtflood in Hainan WU Huiet al (Hainan C lmi ateCenter, H ai ou, H ainan 570203) Abstract W ith the data of autumn rainfall in 18 m eteorological observatories in Hainan and reanalyzed data of NCEP of 1961- 2008, using comprehensive and correlation analysis, after dmi ension reduction, a fuzzy neural networ prediction model for autumn droughtflood in H ainan was developed ta ing previous SST, geopotential height, w ind and relative hum idity that effect the autumn droughtflood as factors. The resu lts show ed that the fuzzy neuralnewt or prediction modelw ith multi factors is superior to thosew ith single or several factors, and it show s advanta ges such as high fit precision and stability to h istory samp les, especially for the independent samples, and w ith good pred iction ability for severe drought and severe flood years. The tests for independent sam ples in 2004- 2008w ith new criterion reaches 86 and it is of great help foropera tional prediction. eywords Autumn droughtflood; Fuzzy neural newt or ; Prediction; H ainan island , ,, , , , 2000 10, , 1 , , 1. 1 18 1961~ 2008 , 26. 5 ;NCEP/NCAR 2004,, 0 0 , 2. 5 2. 5 500 hPa850 hPa( 18. 3 u v) 850 hPa