基于Pair Copula函数的空间预测模型及其应用

对于空间数据的插值预测,大多采用传统的空间插值方法如反距离加权插值法和克里金插值法,这2种方法在边缘分布或存在异常值的情况下会导致预测精度相对较低;采用基于Copula理论的方法克服了这一问题。通过Pair-Copula函数描述了空间相依结构并利用MCMC方法(贝叶斯估计法)估计参数,讨论基于空间数据对未观测位置相关数据进行了空间插值预测;结合重庆市雾霾数据对该方法与反距离加权插值法、普通克里金和泛克里金插值法进行比较,结果发现基于Pair-Copula函数的空间预测模型具有更高的精度。

Spatial Prediction Model and Its Application Based on Pair Copula Functions

The interpolation prediction of spatial data usually uses traditional spatial interpolation methods such as inverse distance weighted interpolation and Kriging interpolation, whose prediction accuracy is relatively low under the impact of marginal distribution or outlier, as a result, the method based on copula overcomes the problem. Spatial correlation structures are described by Pair Copula function and the parameters are estimated, and spatial interpolation prediction method is discussed in corresponding values of none observation stations based on spatial data. This model is compared with inverse distance weighted interpolation, original Kriging interpolation and universal Kriging interpolation based on the data of fog in Chongqing, and the results show that the spatial prediction model based on Pair Copula function posses the higher accuracy.