基于Kriging模型的地面气温空间插值研究  

Spatial Interpolation of Surface Air Temperature by Kriging Models

作　　者：李静思[1] 潘润秋[1] 范馥麟 

机构地区：[1]武汉大学资源与环境科学学院,武汉430079 [2]斯特拉斯克莱德大学电子与电力工程学院

出　　处：《西南师范大学学报（自然科学版）》2016年第5期21-27,共7页Journal of Southwest China Normal University(Natural Science Edition)

基　　金：青年科学基金项目(41401446)

摘　　要：基于中国气象局提供的全国192个气象观测站的日均地面气温(SAT)数据,利用Kriging模型对研究区域内任意位置SAT进行插值评估.为了使应用于Kriging模型的试验数据满足弱平稳性,首先使用傅立叶级数拟合并去除原数据中的年趋势;并根据各位置的地理坐标,包括海拔和经纬度,拟合出研究区域的空间趋势.反距离加权(IDW)插值法作为比较对象也被应用于试验中.采用交叉验证的方式,将192个观测站依次作为目标点进行插值并与该站的实际数据比较,计算各站插值评估的平均绝对误差(MAE)与均方根误差(RMSE),从而比较两种方法的精确度.结果表明Kriging模型在174个观测站得到的插值结果优于IDW插值法.并且,在Kriging过程中,去除时间趋势和空间趋势对准确获取研究区域内的空间相关性起到了重要作用.This paper aims to describe spatial interpolation methods to estimate surface air temperatures（SATs）.The SAT at a particular location where SAT observations are not available is estimated through a Kriging interpolation between SAT measurements from 192 meteorological sites at which daily SAT observations have been obtained.A temporal de-trending method based on a Fourier series is used to model and remove the annual trend in original data in order to ensure the stationarity of de-trended data from which Kriging parameters are determined.Furthermore,a spatial or surface de-trending in terms of geographic coordinates including altitude,latitude and longitude of each location is adopted in a Kriging model.Besides a Kriging model,an inverse distance weighting（IDW）interpolation method is tested as a comparison.The accuracies of both spatial interpolation approaches are assessed by calculating and comparing their mean absolute error（MAE）and root mean square error（RMSE）when taking each meteorological site as the target location in a cross-validation procedure.The results show that the Kriging model performs better than the IDW method at 174 sites.In addition,the temporal and spatial de-trending methods make the main contribution to the accurate capture of spatial correlations in the study area in a Kriging process.

关 键 词：地面气温 KRIGING模型 反距离加权插值法 傅立叶级数 去除时间趋势 去除空间趋势 

分 类 号：P413[天文地球—大气科学及气象学]