【作者】 金志凤； 李波； 黄敬峰； 李仁忠；

【Author】 Jin Zhifeng1) Li Bo2) Huang Jingfeng3) Li Renzhong1) 1)(Zhejiang Climatic Center, Hangzhou 310017) 2) (Institute of Public Administration,Zhejiang University, Hangzhou 310029) 3) (Institute of Agricultural Remote Sensing & Information Technology, Zhejiang University, Hangzhou 310029)

【机构】 浙江省气候中心； 浙江大学公共管理学院； 浙江大学农业遥感与信息技术应用研究所；

【摘要】 高精度的空间化蒸发量信息对于区域水资源管理、生态环境治理等方面都具有重要意义。本文选取浙江省的72个气象站点1971～2008年38年的年平均蒸发量资料,并分析其与经度、纬度、海拔高度的内在关系,提出了一种利用BP神经网络进行年平均蒸发量空间插值的新方法。用没有参与建模的10个站点进行验证,结果显示,BP神经网络的插值方法的绝对平均误差为79.57mm,相对平均误差也仅为5.97%,相对误差的分布范围有80%都集中在0%～5%和5%～10%两个区间内,取得了较好的模拟效果。研究表明,该方法不仅能够用于年平均蒸发量的空间插值,而且依此获取的高精度、高分辨率蒸发量数据,对于与其相关领域的研究具有重要意义,还对降雨、平均气温等的空间插值有借鉴意义。更多还原

【Abstract】 High precision spatial evaporation information is greatly significant in many fields such as regional water resource management, ecological environment protection. This paper adopted mean annual evaporation of 72 meteorological stations from 1971 to 2008, and analyzed the relationship among evaporation, longitude, latitude and altitude, and provided a new method based on BP neural network spatial interpolation for mean annual evaporation. 10 meteorological stations were used to validate the result, the result indicated that mean absolute error of BP neural network spatial interpolation method was 79.57mm, mean relative error was 5.97%, 80% relative errors were concentrated in the 0% ~ 5% and 5% ~ 10%, a better simulating result was obtained. The research demonstrated that BP neural network spatial interpolation method not only could be applied to interpolate for mean annual evaporation, but also could be used to obtain high spatial resolution evaporation datasets. BP neural network spatial interpolation method was greatly significant for relative research fields, and had great significance for spatial interpolation of precipitation, mean temperature.更多还原