基于地统计的区域气象要素空间插值与分布研究

Spatial Interpolation Method and Spatial Distribution Characteristics of Temperature and Precipitation in South Loess Plateau

利用地统计学的理论与方法,以黄土高原南部地区为例,分别运用径向基函数法、普通克里格法和反距离权重法对研究区内139个气象台站2010年的年平均气温和年降水量进行了空间插值,并利用交叉检验方法对插值精度进行了评估。评估结果表明,3种空间插值方法对研究区域内的气象要素进行统计内插的效果都较好。对气温来说,径向基函数法最好,其均方根预测误差值仅为2.263,其次是反距离权重法和普通克里格最好,均方根预测误差值分别仅为2.377和2.38;对降水来说,则是普通克里格法最优,均方根预测误差最小,且其标准均方根预测误差值为0.8205,接近于1。由此可见,对年均降水量采用普通克里格插值法的精度较高。

Temperature and precipitation data from 139 meteorological stations in the south loess plateau of China were interpolated by using ra- dial basis function, Ordinary Kriging（OK）and Inverse Distance Weighing （IDW）with arcGIS. Cross- validation were applied to evaluate the three interpolation method, from which the Mean Absolute Error（MAE） and Root Mean Squared error（RMSE） were calculated. The results showed that RBF gave the best simulation of temperature. RMSE by RBF was 2.263, followed by IDW and then by OK. For precipitation, OK was the best simulation. RMSs by OK was 0. 825, closely to 1 respectively. It revealed OK gave the best simulation.