地质统计学是用于模拟土体固有空间变异性的方法之一，以变差函数为工具，采用Kriging插值提供未采样点处土工参数值的最优线性无偏估计。将地质统计学方法应用于宿-新（宿迁至新沂）高速公路某试验段内孔压静力触探（piezocone penetration test，CPTU）锥尖阻力qt空间变异性研究中，采用回归分析移除数据中的趋势项，从而获得具有弱平稳性的残差数据。指数型理论变差函数能够准确描述试验段内土体的连续空间变异性特征。根据估计结果，试验段内锥尖阻力qt残差的变程具有显著各向异性，在水平方向和竖直方向分别为4.05 m和1.2 m。采用普通Kriging插值结合趋势分析，绘制了qt在试验段的空间分布图和平面投影图，用于指导工程实践。结果表明，普通Kriging插值的估计结果能够与试验段内实测资料形成较好的对比，仅仅在部分极值变化和远离采样点的位置处估计值可靠性会降低。

Geostatistics is one of the most powerful techniques for modeling the inherent spatial variability of soil properties. In geostatistics, the semivariogram is utilized to describe the continuous variation characteristics and then the Kriging interpolation is applied to provide the best linear unbiased estimates of soil properties. Geostatistics is used to model the inherent spatial variability of piezocone penetration test(CPTU) cone tip resistance qt in a testing section of Suqian-Xinyi Expressway for site characterization. First of all, theoretical background of the geostatistics involving constructing experimental semivariogram, theoretical semivariogram fit and Kriging prediction is reviewed briefly. Least square regression analysis is performed to remove the linear trend function within the qt data to generate the weak stationary residuals. It is found that exponential theoretical semivariogram can accurately describe the correlation structure and variability of the residuals at this site. Significant anisotropy is observed in the correlation structure of qt residuals. Results show that the vertical range is about 1.2 m while the horizontal range is 4.05 m for qt at this site. The distribution map of qt data at the target site is obtained using ordinary Kriging interpolation and compared with in-situ measurement data. It is concluded that the Kriging estimates are comparable to the measurements at this site. However, the reliability of estimates may decrease at the locations corresponding to peak values and those far away from observations. This research can improve the reliability of geotechnical design in practice since the inherent spatial variability of qt data is accurately evaluated.