| 摘要: |
| [目的]介绍ANUSPLIN在黄土高原多沙粗沙区时间序列气象要素的空间插值过程,为相关人员在ANUSPLIN的参数设置、误差分析和协变量要素选择等方面提供参考.[方法]以薄盘样条函数为插值理论,以专用气象数据插值软件ANUSPLIN为实现工具,并引入一个或多个协变量线性子模型,来实现多个气象要素的空间插值.[结果]完成了黄土高原多沙粗沙区时间序列(1980-2000年)多个气象要素月平均数据的栅格化,计算了气象要素随其影响因子变化的关系.[结论]ANUSPLIN以薄盘光滑样条函数为理论基础,引入协变量线性子模型,能较好地提高气象要素空间插值精度,且能反映气象要素随其影响因子变化的比率关系.在大多数情况下,该区模型选择以样条次数为3次的局部薄盘光滑样条函数模型为最佳.温度的空间插值相对比较容易,且误差较小,1995-07平均相对误差为1%;风速、水汽压的误差中等,日照时数和降雨量的误差较大,个别情况相对误差可超过50%. |
| 关键词: ANUSPLIN 时间序列 气象要素 空间插值 |
| DOI： |
| 分类号: |
| 基金项目:中澳合作ACIAR PROJECT(LWR1/2002/018)；中国科学院西部之光项目（B184/2004） |
|
| Interpolation for time series of meteorological variables using ANUSPLIN |
|
|
| Abstract: |
| 【Objective】 The interpolation processes of the meteorological variables using the ANUSPLN in coarse sandy hilly catchments of the loess plateau were introduced in this paper.It should be a useful reference for setting parameters,analyzing the errors and selecting the correct covariates.【Method】 In the interpolation of different meteorological variables,a professional interpolation package ANUSPLIN was used in which one or more influenced factors were introduced as covariate sub-models.【Result】 Time Series of monthly meteorological data from 1980 to 2000 on the geomorphologically complex Coarse Sandy Hilly Region in Loess Plateau were interpolated to surfaces，and the lapes rate of the meteorological variable changing with its influence factors were calculated.【Conclusion】 Based on the thin plate smoothing spline function,using multiple covariates as linear sub-models in addition to the independent spline variables, ANUSPLIN can develop the interpolation accuracy and reflect the rates between the meteorology variables and their influenced factors,and especially adapt to time series of data.For the research area,partial thin plate smoothing spline model using one or more linear sub-models with 3 spline order is the best model for most meteorological variables.Temperature interpolation is easier with less error,for 1995-07 month,the mean relative error is about 1%,the interpolation errors for wind and vapor pressure are moderate,and higher for sunshine hour and rain in which max relative error reach 50% in some way. |
| Key words: ANUSPLIN time series meteorological variable spatial interpolation |
