摘要
运用反距离加权插值法(IDW)和梯度反距离加权插值法(GIDW)对全国183个气象站的2001年、2002年平均气温进行了内插,并在此基础上进行了幂指数优化和邻近点选择优化。交叉验证结果表明,对于IDW方法,幂指数为3、邻近点选择采用三角网法的插值结果最优;对于GIDW方法,幂指数为2、邻近点选择采用固定数目法的插值结果最优。在幂指数和邻近点选择优化的基础上,比较了IDW方法与GIDW方法的插值结果,考虑经纬度和海拔高程的GIDW方法明显优于IDW方法。在此基础上,提出了基于K-means聚类的空间插值优化方法,实践证明聚类后再插值比直接插值效果更佳,聚类为插值前的数据预处理提供了一种新的思路。
This paper used the inverse distance weighting(IDW) method and the gradient plus inverse distance weighting(GIDW) method to interpolate the yearly mean temperature data from 183 meteorological stations of China in 2001 and 2002. On the basis of this, the paper made the power exponent and the nearest points selecting optimization. Cross-validation tests showed that the power exponent being setting to 3 and using triangle net to select the nearest points reached the beast results for IDW, and the power exponent being setting to 2 and using fixed number points reached the beast results for GIDW. After the power exponent and the nearest points selecting optimizing, this paper compared the results of IDW and GIDW, which showed that the result of GIDW was obviously superior to the result of IDW. Finally, this paper put forward an optimization method for spatial interpolation based on K-means clustering. The experiment results show that interpolating after clustering is superior to directly interpolating, and clustering can provide a new thought for data pretreatment before interpolating.
出处
《地理空间信息》
2017年第7期86-89,共4页
Geospatial Information