摘要:以西南地区1996~2000年93个气象台站观测的月均降雨量为基础，对各月降雨量进行空间自相关性，变异特征等空间分析后，采用反距离加权法（IDW）和以不同变异函数模型（指数模型、球面模型、高斯模型）为基础的普通克里金（O-Kriging）两种方法进行空间插值，通过交叉验证结果对两种方法进行分析比对。结果表明：（1）西南地区月均降雨量存在明显的空间集聚现象，并具有显著的空间自相关性和变异特征，可对该研究区域降雨量进行空间插值研究。（2）在O-Kriging插值时，变异函数选用指数模型的效果最好，球面模型次之，高斯模型最差。（3）两种方法对月均降雨量及其极大值和极小值插值时，O-Kriging的插值误差均小于IDW，插值误差整体上与降雨量呈正相关关系。在剔除各月降雨量极大值较为集中的两个站点后进行插值，插值结果的误差均明显降低。（4）对研究区域整体来说，O-Kriging的插值效果优于IDW，但就单个站点来看，结果并非如此。在降雨量的空间插值中，由于研究区域和时间尺度的不同，并不存在绝对的最优方法，应根据实际应用效果选择最适方法。
关键词:降雨量/
空间插值/
空间自相关/
交叉验证
Abstract:Based on monthly precipitation observations collected at 93 meteorological stations in Southwest China from 1996 to 2000, this study investigates the spatial interpolation results with the Inverse Distance Weighting (IDW) and O-Kriging interpolation methods. Firstly, we analyze the spatial autocorrelation and spatial variability character of monthly average precipitation data. Secondly, the IDW and O-Kriging based on three semi-variograms (exponential, spherical and, Gaussian model) are used to spatially interpolate monthly precipitation. Finally, the interpolation results are compared and discussed using the cross-validation method. The conclusions are:(1) Monthly precipitation distribution in Southwest China shows a spatial aggregation feature with high spatial autocorrelation and variation, which favors for the spatial interpolation. (2) Compared to the three semi-variograms used in the O-Kriging interpolation method, the best performance is from the exponential model, while the worst is from the Gaussian model. (3) When the O-Kriging and IDW are used in spatial interpolation of monthly average and maximum and minimum precipitation, the former one perform better than the latter one. The errors between interpolated data and observations overall increase with monthly precipitation magnitude, and the errors from both interpolation methods are obviously reduced after removing the maximum monthly precipitation points. (4) For the study area as a whole, the interpolation effect of the O-Kriging is better than that of IDW, however, this is not true at single sites. There is no absolute optimal method in the spatial interpolation of precipitation for every study area and on all time scales. The optimal interpolation method depends on the actual demands and applications.
Key words:Precipitation/
Spatial interpolation/
Spatial autocorrelation/
Cross validation
PDF全文下载地址:
http://www.iapjournals.ac.cn/qhhj/article/exportPdf?id=20190104
