作　　者： ; ; ; ;

机构地区： 中国林业科学研究院森林生态环境与保护研究所

出　　处： 《Meteorological and Environmental Research》 2010年第5期7-11,57,共6页

摘　　要： 以获取三江源地区1971～2000年的各月降水及平均气温的栅格数据为目的,选取青海省及其周边共58个基准气象站点的气候数据,结合100m×100m分辨率的数字高程模型(DEM)数据,采用协同克里格(Co-Kriging,COK)和薄盘光滑样条函数(Thin plate smoothing splines,TPS)2种方法进行气温及降水的空间插值。用广义交叉验证(General cross validation,GCV)的最小统计误差筛选最佳模型,用均方根预测误差(RMSE)和平均绝对误差(MAE)比较2种方法的精度。结果表明,①不论是COK还是TPS,引入协变量后模型插值结果的精度更优;②TPS的插值效果显著优于COK:对月均温模型,TPS的RMSE值比COK提高了69.48%,MAE验证指标提高了70.56%;而对于月降水模型,TPS的RMSE值比COK提高了28.07%,MAE值提高了29.06%。 In order to get the spatial grid data of monthly precipitation and monthly average temperature of Sanjiangyuan area, the Co-Kriging (COK) and thin plate smoothing splines(TPS) interpolation methods were applied by using the climate data during 1971-2000 of 58 meteorological stations around Qinghai Province and the 3 arc-second digital elevation model (DEM) data. The performance was evaluated by the smallest statistical errors by general cross validation (GCV). Root-mean-squared predicted errors (RMSE) and mean absolute errors (MAE) were used to compare the performance of the two methods. The results showed that: 1) After combing covariates into the models, both methods performed better; 2) The performance of TPS was significantly better than COK: for monthly average temperature, the RMSE derived from TPS was 69.48% higher than COK, as MAE increased by 70.56%. And for monthly precipitation, the RMSE derived from TPS was 28.07% higher than COK, as MAE increased by 29.06%.

关 键 词： 三江源地区 插值

领　　域： [天文地球]