2017 Volume 59 Issue 2 Pages 105-123
The purpose of this study is to develop a kriging interpolation method to estimate soil/rock types in the deep zones from the water well data for compensation of limited core-sampling data. The system provides the probability value of each soil/rock type by using the indicator kriging. The soil/rock types of the maximum probability are specified in a regular grid. The probabilistic-based system determines the data search radius and the anisotropic ratio of variogram range at each point according to the categories of topography, geology and data amounts with search radiuses. Re-estimation Score (ReS) in the leave-one-out cross validation is compared to determine the local parameter settings. A case study of the proposed method in the Toyohira River alluvial fan, Sapporo, Japan, shows that the kriging estimation covers almost the entire region. In addition, the averaged ReS is close to the highest value as the smallest search radius is applied.
