摘要
综合分析了四种空间插值方法,即距离权重倒数法、局部多项式法、普通克里金法和考虑海拔的协克里金法的特点,并以深圳市36个雨量站2006年雨季27天的日累计降水量数据为实例进行了研究和验证,采用交叉检验的方法对插值结果进行比较。研究表明,四种方法均能反映降水总体情况,但插值曲面相对真实曲面较平滑,距离权重倒数法的插值曲面最为平滑;多项指标比较显示,克里金法优于距离权重倒数法和局部多项式法。与普通克里金法相比,考虑海拔的协克里金法对插值精度没有明显提高。依照雨量站海拔、日均降水量分别对插值结果数据分组统计比较表明,海拔高的雨量站插值结果普遍大于实测结果,海拔低的雨量站则相反;日均降水量较大时(>50mm),插值误差明显增大。
Discrete or continuous rainfall data are required to run many GIS models for environment and planning. The paper attempts to make a general comparison on different spatial interpolation methods. It carried out a study on four spatial interpolation methods： Inverse Distance Weighing （IDW）, Local Polynomial, Ordinary Kriging and co-Kriging with respect to elevation. Daily rainfall data,obtained from 36 rainfall stations during 27 inconsecutive days in the rainy season of 2006 in Shenzhen was employed in this study. Cross validation of the results shows that the four methods could to some extent reflect the rainfall situation of the region, yet the four interpolated surfaces are more smoothing than the practical circumstance. Particularly, IDW is the most smoothing one of the four. Besides, all the criteria （Mean Error, Mean Absolute Error, Root Mean Square Error and Percentage Error） that brought up in our study to measure accuracy of the four methods demonstrated that Ordinary Kriging and co-Kriging methods are superior to Local Polynomial and IDW. Inclusion of elevation in the co-Kriging method does not lead to improvement of result compared with Ordinary Kriging method. Furthermore, the interpolation data was grouped according to elevation and average daily rainfall and the same criteria above was bought to the statistics of those groups. Comparison on the criteria reveals that, the interpolated result on rain gauge stations with high elevation tend to larger than the observed data, and those rain gauge stations with low elevation are on the contrary. The interpolation error increases sharply while the average daily rainfall is bigger than 50mm.
出处
《地球信息科学》
CSCD
2008年第5期566-572,共7页
Geo-information Science
基金
国家自然科学基金项目(40701134)
国家863计划项目(2007AA12Z216)
