摘要
为了实现离散降水数据到连续降水数据的插值,以内蒙古自治区多年(1950-2000年)平均降水量为例,讨论了不同空间插值方法的优劣,发现基于地统计方法的Kriging插值具有最佳的插值效果。进一步对比Ordinary Kriging和CoKriging插值方法的4种不同的半变异函数模型,结果表明利用指数半变异函数的CoKriging插值方法具有最小的预测误差。地统计方法对于预测降水具有重要的价值,而不同的地区具有各自适合的模型,理解该方法的数学原理和操作过程,对相关研究具有借鉴意义。
Spatial interpolation of precipitation data is very important to those areas where the observation stations are few and at random. As no single method suits to precipitation interpolation for all regions, this paper discussed the differences between geostatistical analysis and other several interpolation methods by using annual average precipitation in Inner Mongolia of China. The results indicate that the geostatistical analyst methods have the advantage of other methods in interpolation precision. The comparison between two kinds of geostatistical analyst methods, Ordinary Kriging and CoKriging respectively shows that the latter is more effective and has a higher precision, and this owes to its inclusion of elevation which has an influence on precipitation processes. Comparing four different semi-variogram models of Kriging interpolation methods, the exponential model has the minimum RMSE, and the spherical model in the next place. This work could potentially enforce the related researches in interpolations of precipitation, and even other spatial climatic data.
出处
《水文》
CSCD
北大核心
2010年第1期14-17,58,共5页
Journal of China Hydrology
基金
国家"十一五"重大科技项目(2008ZX07526-002-02)
城市与区域生态国家重点实验室自主项目(SKLURE2008-1-02)
关键词
地统计学
空间插值
克里格插值
多年平均降水量
geostatistical analyst
spatial interpolation
Kriging interpolation
mean annual precipitation