【作者】 赵旭寰；

【导师】 王振会；

【作者基本信息】 南京信息工程大学 ， 大气遥感科学与技术， 2008， 硕士

【摘要】 掌握雷电活动规律可以为雷电防护工作提供相应的参考,而对雷电活动做出准确的预报则可以趋利避害,将雷电造成的灾害尽量减小。但是由于雷电活动在一定时间尺度上发生、发展,具有瞬时性,随机性与非线性的特点,很难对其规律进行准确的把握,并对其做出预报。在江苏省气象灾害重点实验室重点项目(KLME050101)的支持下,本研究利用江苏省气象局新建成的闪电定位系统,首先对发生在全省境内的地闪资料进行了统计分析,在此基础上,利用神经网络方法对雷暴的发生做了模拟预报研究,并进一步提出了基于该方法的闪电预报系统模型。首先利用2005年8月至2006年8月一年来江苏省气象局闪电定位系统的地闪监测资料,对江苏地区闪电活动的规律进行了统计分析。统计结果显示:(1)江苏地区闪电活动主要集中于夏季,7月的闪电总数占全年的52.9%,主单峰特征十分明显,春秋两季闪电活动相对较少,而冬季则很少发生。(2)正,负地闪电流强度谱分布一致,均呈单峰分布形式,正地闪平均电流强度较大,为35.21KA,说明正闪对人身财产威胁更大。(3)江苏地区闪电主要发生在洪泽湖地区以及溧水山区,闪电活动与下垫面性质有很大关系。(4)全年四季中地闪密度分布不均匀,地闪的发生随季节也存在纬度上的变化。然后利用人工神经网络的方法对南京地区的雷暴天气进行模拟预报。南京站的探空资料(每天08时和20时各一次)可以反映雷暴发生前大气层结的初始状态,而江苏省闪电定位网资料则可以作为预报量加以使用。本研究利用探空资料,计算了一些与雷暴发生相关的预报因子,用以建立南京地区雷暴的神经网络预报模型。并应用独立样本进行了初步的模拟预报检验。初步结果表明,预报模型取得了比较令人满意的效果,应用神经网络的方法预报雷暴的发生是可行的。最后提出了基于该方法的预报系统模型,并给出了预报流程说明。更多还原

【Abstract】 The knowledge of lightning activities is important in lightning protection work and also in the fields of lightning forecast. But the spatial and temporal distributions of lightning activities are difficult to be measured for its randomicity and nonlinearity. Supported by the Natural Science Foundation of Jiangsu Province (KLME050101) in 2005, the Cloud to Ground lightning flash (CG) data from Lightning Location System (LLS) in Jiangsu Province has been analyzed to find out the rule of regional CG flashes. And a neural network based on the LLS data has been developed for lightning forecast.Since August in 2005, the Lightning Location System in Jiangsu province has been functioning for more than one year. The lightning data in the last year partly reveals the lightning activities in and around Jiangsu province, although the detection efficiency remains to be improved. The present study statistically analyses the data and find: (1) the most cloud to ground lightning fleshes(CG) occur during April to July, and the peak value of lightning activity was find to be around 15:00(LMT)(2) The spectra distribution of return stroke current intensions exhibits normal distribution (3) the prominent maxima of lightning density was find to be around Hongze lake and Yisu mountain area, which reflects that thunderstorm activities were well related to the surface characteristics. (4) the lightning density was not equable in different seasons.A neural network-based scheme to do a multivariate analysis of the occurrence of thunderstorm is presented. Many sounding-derived indices are combined together to build a short-term forecast of thunderstorm in the city of Nanjing. For thunderstorm forecasting , sounding and lightning strike data from June to September have been used to train and validate the network. Output form network show relatively satisfied performance in this preliminary study.更多还原