利用最小二乘回归支持向量机LS-SVMR(least square support vectors machines for regression)对2个不同长度的时间序列资料,国家气候中心1982年1月~2005年12月Nino3区逐月海温距平指数(短序列),及1950年1月~2006年12月Nino3区逐月海温...利用最小二乘回归支持向量机LS-SVMR(least square support vectors machines for regression)对2个不同长度的时间序列资料,国家气候中心1982年1月~2005年12月Nino3区逐月海温距平指数(短序列),及1950年1月~2006年12月Nino3区逐月海温距平指数(长序列)资料进行了预测试验,以验证支持向量机对气候变化中非线性时间序列的预测效果。结果表明:通过训练建立的最小二乘回归支持向量机模型,较好地反映了Nino3区海温距平指数的变化规律,36个月的预报效果较好,具有一定的可信度。资料的长度越长,预测结果与实测值的变化趋势越接近,但资料长度对均方根预报误差不敏感。展开更多
Large-scale and mesoscale analyses are made for extremely heavy rain (EHR) and meso-β scale low vortex (MSLV) in Jiading District of Shanghai Municipality during 6-7 July 2001.It is shown that the EHR forms in the si...Large-scale and mesoscale analyses are made for extremely heavy rain (EHR) and meso-β scale low vortex (MSLV) in Jiading District of Shanghai Municipality during 6-7 July 2001.It is shown that the EHR forms in the situation of northern westerly trough linking together with southern inverted typhoon trough at northwest side of the West Pacific Ocean subtropical high. Numerical simulation is made using a 21-layer improved REM (regional η coordinate model) for this course.The results show that the precipitation forms earlier than MSLV.and the strong convergence in wind velocity mate (WVM) triggers the strong precipitation.The formative reasons of WVM.especially the weak wind velocity center are discussed,and the formative mechanisms of the MSLV and EHR are discussed using high spatial and temporal resolution model- output physical fields.The results show that the heavy rain releases latent heat and warms the air column,and enhances the low level positive vorticity that existed before.Then it causes the formation of MSLV.There is a positive feedback mechanism between low vortex and precipitation,so CISK must be an important mechanism.展开更多
文摘利用最小二乘回归支持向量机LS-SVMR(least square support vectors machines for regression)对2个不同长度的时间序列资料,国家气候中心1982年1月~2005年12月Nino3区逐月海温距平指数(短序列),及1950年1月~2006年12月Nino3区逐月海温距平指数(长序列)资料进行了预测试验,以验证支持向量机对气候变化中非线性时间序列的预测效果。结果表明:通过训练建立的最小二乘回归支持向量机模型,较好地反映了Nino3区海温距平指数的变化规律,36个月的预报效果较好,具有一定的可信度。资料的长度越长,预测结果与实测值的变化趋势越接近,但资料长度对均方根预报误差不敏感。
基金supported by the Air Force Foundation under Grant No.KJ99099
文摘Large-scale and mesoscale analyses are made for extremely heavy rain (EHR) and meso-β scale low vortex (MSLV) in Jiading District of Shanghai Municipality during 6-7 July 2001.It is shown that the EHR forms in the situation of northern westerly trough linking together with southern inverted typhoon trough at northwest side of the West Pacific Ocean subtropical high. Numerical simulation is made using a 21-layer improved REM (regional η coordinate model) for this course.The results show that the precipitation forms earlier than MSLV.and the strong convergence in wind velocity mate (WVM) triggers the strong precipitation.The formative reasons of WVM.especially the weak wind velocity center are discussed,and the formative mechanisms of the MSLV and EHR are discussed using high spatial and temporal resolution model- output physical fields.The results show that the heavy rain releases latent heat and warms the air column,and enhances the low level positive vorticity that existed before.Then it causes the formation of MSLV.There is a positive feedback mechanism between low vortex and precipitation,so CISK must be an important mechanism.
