抄録
Several numerical tests in this paper can be performed to check the validity of cokriging estimation of the groundwater levels in space and to show its cross-correlative effects in using the insufficient data set. Cokriging, as a type of kriging, seeks to estimate values of a certain variable at the unsampled locations. The cokriging technique makes use of the cross-correlative dependence between several variables. A groundwater level to be cokriged is evaluated as a linear combination of the available sampling data on groundwater levels and on bed elevations, which are spatially correlated with each other. Three steps are needed in the cokriging application. First, several geostatistical models, which adequately describes the spatial structure of the variables of interest, must be proposed. The second step involves using the available data to identify the unknown parameters of the models described in the first step and to select the best fitting model among proposed ones on the basis of the four information criteria. Third and finally, cokriging is employed to estimate a spatial distribution to be desired. Numerical tests are presented in this work to demonstrate the cokriging effects subject to the insufficient data. Cokriging yields reasonable and valid results because it can interpolate less sampling data due to its ability to adequately pick up the available cross-correlative data. The high applicability and reliability of the cokriging method can be reinforced with the above results.
