基于地统计方法的气候要素空间插值研究

A Study on Spatial Interpolation Methods for Climate Variables Based on Geostatistics

在回顾了地统计学的产生、发展及其基本原理的基础上,对于目前众多可以提供计算网格的气候要素的空间插值方法中,具体探讨了普通克里格法和协同克里格法。将甘肃省1961-1990年30年平均降水量和蒸发量作为区域化变量,根据不同的半变异函数理论模型,采用普通克里格法和双变量协同克里格法,通过对比分析得到:(1)不论是多年均降水量还是多年平均蒸发量在空间上都呈现明显的梯度变化,二者的空间变程都很大,而降水量变化幅度更大。降水量从东南部向西北部逐渐减少,蒸发量则相反,从东南向西北逐渐增加。(2)基于地统计的插值方法,根据半变异函数云图和试验方差最小的原理,选择合适的半变异函数理论模型进行变量的空间插值,能够较好地模拟区域化变量的空间连续分布格局,并取得较好的效果。对比普通克里格法和协同克里格法,后者增加了高度对降水量和蒸发量的影响,在空间分布上更为合理,插值的精度也要明显好于普通克里格法。(3)采用地统计方法虽然在总体上能够较好地反映气候要素的空间分布格局,但检验显示,两种方法空间插值的精度都还不是很高,插值的精度还有待进一步提高。

Based on reviewing the origin, development and basic principles of Geostatistics, two kinds of interpolation methods concretely ordinary Kriging and Cokriging are discussed. As no single method among so many available ones to spatial interpolation of climate variables is optimal for all regions and all variables, the interpolation methods are discussed based on Geostatistics by using annual average precipitation and evaporation in Gansu province from 1961 to 1990. Based on different semivariogram theory models we adopt ordinary Kriging and Bivariate Cokriging, by comparing of them the obtained conclusions are as follows： （1） No matter multi-year average precipitation or evaporation all presented obvious gradient change on space, the change ranges of both were great, the former was larger than the latter. Multiyear Annual average precipitation decreased gradually from southeast to northwestward, but evaporation was opposite, increased gradually from southeast to northwest. （2） According to semivariogram cloud plots and experiment variance minimum principle selected suitable semivariogram theory models based on Geostatistics interpolation method to interpolate, which could simulate space pattern spreading continuously of regionalized variable well, then get better interpolation effect. Comparing ordinary Kriging with Cokriging, as the latter input altitude which had an influence on precipitation and evaporation, it was more rational on space distribution and had a higher interpolation precision. （3） With Geostatistics methods the spatial interpolations could reflect the general space pattern of climate variables better in general, but the spatial interpolations precision of two methods were still not high, which still remained further improving.