基于克里金插值估算区域降水量的抽样方法对比分析——以甘肃省为例  

Comparison analysis of sampling methods to estimate the regional precipitation based on Kriging interpolation method: a case study in Gansu Province

作　　者：刘世伟[1,2] 吴锦奎[1,3] 张文春 周嘉欣[3] 杨俊华[1] 

机构地区：[1]中国科学院寒区旱区环境与工程研究所冰冻圈科学国家重点实验室,甘肃兰州730000 [2]中国科学院大学,北京100049 [3]中国科学院寒区旱区环境与工程研究所流域水文及应用生态实验室,甘肃兰州730000 [4]甘肃省酒泉水文水资源管理局,甘肃嘉峪关735200

出　　处：《冰川冻土》2015年第3期650-657,共8页Journal of Glaciology and Geocryology

基　　金：国家重大科学研究计划项目(2013CBA01806);国家自然科学基金项目(41130638;41271085);冰冻圈国家重点实验室开放基金项目(SKLCS-OP-2013-05)资助

摘　　要：降水数据的空间插值精度由降水的空间变异特征、插值方法和降水观测站分布决定的,其中,降水观测站及插值方法的选择尤为重要.利用空间随机抽样、空间分层抽样和空间三明治抽样等空间抽样方案,以甘肃省境内的气象站点为研究对象进行抽样分析,在抽样结果的基础上进行区域降水量克里金插值,并对比了插值结果精确度.结果表明:在较为丰富的先验知识的前提下,空间三明治抽样所得的区域降水量插值结果的误差项ME(1.97)、MSE(0.0066)和RMSSE(1.0184)都是最优的,空间分层抽样次之,空间随机抽样最差.与另外两种抽样方案相比,空间三明治抽样是一种使区域误差最小,适用范围更广,精确度更高的抽样方法.The accuracy of spatial interpolation of precipitation data is determined by the spatial variability of precipitation,the interpolation method and the distribution of observatories w hose selections are particularly important. In this paper,the spatial sampling programs including space random sampling,space stratified sampling and space sandw ich sampling w ere used to analyze the meteorological stations of Gansu Province,and the accuracy of Kriging interpolation method conducted on the basis of the sampling results w ere compared. It is found that the errors of regional precipitation interpolation based on space sandw ich sampling including ME（ 1. 97）,MSE（ 0. 0066） and RMSSE（ 1. 0184） are minimum under the premise of abundant prior knowledge.The result of space stratified sampling is poor,and that of space random sampling is the w orst. Space sandw ich sampling is the best sampling method w hich minimizes the regional error,and has a w ider scope of application w ith a higher degree of accuracy as compared w ith the other two.

关 键 词：克里金插值 空间抽样 三明治抽样 降水量 甘肃 

分 类 号：P426.613[天文地球—大气科学及气象学]