人工神经网络方法在大气降水氚浓度恢复中的应用  

APPLICATION OF ARTIFICIAL NEURAL NETWORK IN THE RESTORATION OF TRITIUM CONCENTRATION IN PRECIPITATION

作　　者：龙文华[1,2] 陈鸿汉[1] 段青梅[3] 李志[2] 潘洪捷[2] 刘荣益 

机构地区：[1]中国地质大学水资源与环境学院,北京100083 [2]内蒙古自治区地质调查院,内蒙古呼和浩特010020 [3]内蒙古国土资源勘查开发院,内蒙古呼和浩特010020 [4]内蒙古自治区地质环境监测院,呼和浩特010020

出　　处：《地质与资源》2008年第3期208-212,共5页Geology and Resources

摘　　要：根据人工神经网络能识别输入输出数据间复杂的非线性关系等特性,选用北半球(纬度22～74°)70个站点的IAEA/WMO大气降水氚浓度观测数据,建立了大气降水氚年平均浓度的恢复模型.通过对比认为:人工神经网络恢复的降水氚浓度能客观地反映其真实值,为无资料地区恢复1953年以来大气降水氚浓度提供了一种更好的思路.The artificial neural networks are able to distinguish the complex nonlinear relations between the input/ output data. Based on such characteristics, this article selects IAEA/WMO observation data of tritium concentration in atmospheric precipitation from 70 gauging stations in Northern Hemisphere （latitude 22 - 74° ） to establish the restoration model for the annual average concentration of tritium in atmospheric precipitation. With comparison, it is concluded that, the tritium concentration restored by the artificial neural networks can objectively reflect its true value, which provides a new thought for the datum-free areas to restore the tritium concentration in atmospheric precipitation from 1953.

关 键 词：人工神经网络 氚浓度 影响因素 大气降水 

分 类 号：P426.6[天文地球—大气科学及气象学]