节点文献
浙江省汛期降水定量预测技术
The Study for Quantitative Precipitation Prediction of Flood Season in ZheJiang
【Author】 Hu Bo1 Yu Shan Xian2 Teng Wei Ping2 Mao Yan Jun3 (1 Zhejiang Meteorological Observatory,ZheJiang,HangZhou,310017 2 Zhejiang institute of Meteorology,ZheJiang,HangZhou,310017 3 Zhejiang climate center,ZheJiang,HangZhou,310017)
【机构】 浙江省气象台; 浙江省气象科学研究所; 浙江省气候中心;
【摘要】 首先,采用旋转EOF方法,将浙江梅汛期和汛期的降水量空间分布分别划分为浙北北部区、浙东南区、浙西南区、浙中东区和浙北区、浙东南区、浙西南区,将分区内站点的雨量数据进行平均,得到各个分区的平均雨量时间序列。然后,对NCEP月平均常规再分析场和导出场分别进行2~6个月累加距平处理,得到3000个初始因子场。利用REOF中得到的各个分区平均雨量时间序列对初始因子场进行全球相关普查(要求初始因子场的时间不能早于预报对象前一年6月份),然后对高相关区取4×6或3×8网格区;将这些高相关区与各个预报目标场分区进行典型相关分析,取1~2对显著相关的典型变量,共得到近100000对典型相关变量,将所选的典型相关变量与分区内单站的雨量时间序列进行单点相关,相关系数在0.5以上的进入单站预报因子库备用。最后,以各个分区得到典型相关变量为预报因子,梅汛期和汛期38站雨量的时间序列为预报目标,选取其中12个典型变量作为预报方程的初选因子,在这12个因子范围之内根据因子差异法任意选取7个因子,构建浙江省梅汛期和汛期雨量BP人工神经网络预报模型,模型对2003~2005年进行试报,要求这3年的预报正确,单站一般有数十个通过此条件的预报模型,将这些预报进行简单平均,形成集合预报。对2006~2009年梅汛期和汛期的降水进行试报,显示最佳的预报方案4年平均ps评分分别达到69和84分,比较好的预报出降水趋势。随后对预报因子空间进行分析表明(1)NCEP导出资料在汛期降水预测中具有重要作用;(2)梅汛期和汛期因子在全球范围内均有五个集中区,其中四个区分别与ENSO、冰岛低压、阿留申低压等大气活动中心相对应;(3)较长时段的汛期雨量预报需要考虑南半球和北半球的大气环流因子影响。(4)大气能量收支因子对短期气候预测具有十分重要意义。
【Abstract】 Based on REOF and CCA methods,the rainfall between May and July and between May and September at 38 stations in flood season from 1961 to 2005 are analyzesd and the canonical variables are taken as prediction factors by use of NCEP monthly data.Then based on three schemes the BP artificial neural network prediction models are set up.The results of prediction tests from 2006 to 2009 show that the ps scores for the best scheme reach 69 and 84 corresponding to May to July and May to September respectively.The spatial analyses of prediction factors show that(1) The derived NCEP reanalysis data play an important role.(2) The factors of May to July and May to September have five concentration zones.The four zones are corresponding to atmospheric active centers of ENSO、Iceland depression and Aleutian depression.(3) The precipitation prediction of longer flood season needs to consider both the factors of circulation of north and south hemishere.(4) The budget factors of atmospheric energy are very important to climate prediction.
【Key words】 rotated EOF; canonical correlation analysis; flood season; rainfall prediction; concentration zone of factors;
- 【会议录名称】 第28届中国气象学会年会——S5气候预测新方法和新技术
- 【会议名称】第28届中国气象学会年会
- 【会议时间】2011-11-01
- 【会议地点】中国福建厦门
- 【分类号】P457.6
- 【主办单位】中国气象学会