地理科学进展

地理科学进展 ›› 2004, Vol. 23 ›› Issue (2): 34-42.doi: 10.11820/dlkxjz.2004.02.005

• 水文与信息技术 • 上一篇    下一篇

降雨信息空间插值的不确定性分析

朱会义, 贾绍凤   

  1. 中国科学院地理科学与资源研究所,北京 100101
  • 收稿日期:2004-01-01 修回日期:2004-02-01 出版日期:2004-03-25 发布日期:2004-03-25
  • 作者简介:朱会义(1966-),男,副研究员,博士。主要从事土地科学、地理信息系统应用研究。E-mail:zhuhy@igsnrr.ac.cn
  • 基金资助:

    国家自然科学基金资助项目(40271008)及中科院地理科学与资源研究所知识创新工程领域前沿项目(CXIOG-E01-08-01)

Uncertainty in the Spatial Interpolation of Rainfall Data

ZHU Huiyi, JIA Shaofeng   

  1. Institute of Geographic Sciences and Natural Resources Research, CAS, Beijing 100101,China
  • Received:2004-01-01 Revised:2004-02-01 Online:2004-03-25 Published:2004-03-25

PDF (PC)

赞

被引次数

196 | 67

摘要：

文章以潮白河流域为样区，根据58个雨量站1990年的降雨观测数据，采用反距离权重法、克立格法、样条函数法、趋势面法等插值方法，分析了站点数量变化、时间尺度变化、栅格像元的尺度变化、插值方法的差异对降雨数据空间插值结果的影响，剖析降雨插值中的不确定性。结果表明：（1）插值站点数量越大，区域降雨插值的不确定性越小；（2）像元尺度在50m~1000m间变化对降雨插值的不确定性只有微弱的影响；（3）对应于时间尺度由年到月到日的变化，降雨插值的不确定性随时间尺度的减小而显著增大；（4）不同插值方法影响到降雨空间插值的不确定性水平。为了减少降雨信息空间插值的不确定性，根本途径是要引入第三方相关变量，并将其整合到现有的插值算法中。高相关性变量的选取及其与插值模型的整合方式将成为降雨插值研究的主导方向。

关键词: 不确定性, 降雨信息, 空间插值

Abstract:

Taking Chaobaihe Basin as a study area, and using the data from 58 stations in 1990, this paper analyzes the uncertainty in the spatial interpolation of rainfall data caused mainly by the number of stations, temporal scale, cell size of interpolation grid and different interpolation methods. IDW, Kriging, Spline and Trend methods are all adopted in the paper work. The results imply that:(1) the more the number of stations in the interpolation,the less the uncertainty reflected by MAE in rainfall data interpolation; but for certain point, adding some more stations will not absolutely increase its accuracy because of their spatial distribution;(2) the variations of cell size from 50m, 100m, 200m to 1000m does not affect the accuracy remarkably; (3) when temporal scale is shortened from year to month and day, the uncertainty of interpolation results based on the same number of stations increases greatly; (4) different interpolation methods bring different levels of uncertainty. According to the analysis above, the basic way to reduce the uncertainty in rainfall data interpolation is to introduce other relative variations with high sample density, and to integrate them in present interpolation methods. So the choice of those relative variations and their integration with interpolation methods should be the core of the future research in rainfall interpolation.

Key words: rainfall data, spatial interpolation, uncertainty