局部回归克里格在气温栅格化中的应用  

Application of Local Regression Kriging in Mean Annual Air Temperature Estimation

作　　者：郭春霞[1,2] 栗忠魁[3] 诸云强[1,4] 孙伟[5] Guo Chunxia Li Zhongkui Zhu Yunqiang Sun Wei(Institute of Geographic Sciences and Natural Resources Research, CAS, Beijing 100101 University of Chinese Academy of Sciences, Beijing 100049 Gollege of Resource Environment and Tourism, Capital Normal University, Beijing 100048 Jiangsu Center for Collaborative Innovation in Geographical Information Resource Development and Application, Nanjing Jiangsu 210023 Agricultural Information Institute of CAAS, Beijing 100081)

机构地区：[1]中国科学院地理科学与资源研究所,北京100101 [2]中国科学院大学,北京100049 [3]首都师范大学资源环境与旅游学院,北京100048 [4]江苏省地理信息资源开发与利用协同创新中心,江苏南京210023 [5]中国农业科学院农业信息研究所,北京100081

出　　处：《首都师范大学学报（自然科学版）》2016年第6期85-92,共8页Journal of Capital Normal University：Natural Science Edition

基　　金："十二五"国家科技支撑计划项目(2012BAB11BOO);科技基础性工作专项重点项目(2013FY110900)

摘　　要：气温是全球环境变化、农业、生态等研究领域的重要输入变量,气温栅格数据能有效地与其他空间数据进行叠加分析,目前主要通过空间插值获得.本文采用局部回归克里格(Local Regression Kriging)构建全国1981—2010年30年累年年均气温空间分布数据集,并与常规插值方法进行精度对比.插值结果与我国整体的气温分布趋势一致,同时也反映了典型区域气温的局部特征.西北地区受海拔高低起伏的影响,温度呈现出以青藏高原地区为中心,向外围逐渐升高的特点.交叉验证结果表明,充分考虑了气温空间异质性以及残差空间自相关性的邻域为50的局部回归克里格法插值精度相对最高,其均方根误差(RMSE)为1.788℃,平均绝对误差(MAE)为1.127℃,拟合优度R2为0.916 9.Temperature is an important input variable in many research fields, such as global environmental change, agriculture, ecology and so on. Temperature grid can be overlaid to other spatial data efficiently. At present, air temperature raster can be obtained mainly by interpolation techniques. In this paper, mean annual air temperature raster of 30 years from 1981 to 2010 was obtained by using local regression kriging, which was also compared with the other six methods. Interpolation results indicate： temperature interpolation agrees with distribution of Chinese air temperature overall, at the same time reflects local features of air temperature in typical regions. Influenced by topography of the Tibetan Plateau, temperature in the Tibetan Plateau is lowest, which rises continually to the periphery. Cross validation results indicate： local regression kriging with 50 neighbors considering spatial heterogeneity and spatial autocorrelation presents relatively highest accuracy, whose RMSE and MAE are 1. 788℃ and 1. 127% respectively. The goodness of fit R2 is 0. 9169.

关 键 词：局部回归克里格 累年年均气温 气温栅格 局部建模 精度分析 

分 类 号：P208[天文地球—地图制图学与地理信息工程]