9:00 AM - 9:15 AM
[HCG21-01] Evaluation of borehole siting for assessing the 3D distribution of fossil seawater using collocated cokriging
Keywords:Collocated cokriging, Borehole survey, Electromagnetic survey, Paleo-saline water, Geological disposal
Introduction
The hydraulic environment with slow groundwater flow rates is considered to be a suitable geological environment for the site selection of the geological disposal of high-level radioactive waste. Seawater trapped during sedimentation is a source of paleo-saline water. The presence of such saline water indicates that groundwater flow in the area is sufficiently slow. The distribution of saline water can be assessed based on geochemical properties, such as chloride ion concentration (Cl−) and stable isotope (δ18O, δD) compositions of groundwater. Therefore, the site selection process requires a technique to accurately estimate the three-dimensional (3D) distribution of the aforementioned parameters.
Collocated cokriging (COK), which combines borehole and electromagnetic (EM) data, is an efficient method for estimating hydrologic conditions. In terms of the time and cost constraints of borehole survey, achieving reasonable precision with the small number of boreholes is necessary. In this study, we aim to evaluate borehole siting based on COK results for estimating the 3D distribution of saline water.
Methods
Cl− and δ18O data obtained from 10 boreholes (HDB-1–HDB-11, excluding HDB-2) in the Horonobe area and 3D resistivity distribution from the EM survey were used for COK. First, the 3D distribution of Cl− and δ18O were estimated by COK using the results from all boreholes (full model). Next, fifteen partial datasets were created from one to three boreholes, which were subsampled by considering resistivity and geological features. COK was performed using the datasets in the same manner. The root mean square error (RMSE) of the entire model was calculated for each dataset to evaluate the estimation accuracy.
Results
Of the fifteen partial datasets, three sets (HDB-1, HDB-3+5, and HDB-3+5+7) produced the 3D distribution of Cl− and δ18O, which were well congruent with that of the full model, and resulted in a comparably low RMSE. In HDB-1, where a single borehole was used, the corresponding EM data were generally low, and the distributions of Cl− and δ18O in the subset were similar to those of the full model. In HDB-3+5, where two boreholes were combined, each borehole showed a skewed distribution of Cl− and δ18O. However, the distribution was continuous after integrating both boreholes. HDB-3+5+7, a set adding HDB-7 to HDB-3+5, produced the RMSE that was comparable to HDB-3+5, indicating that the additional data did not improve the accuracy.
The remaining twelve datasets had higher RMSE, but the estimated 3D distributions differed from that of the full model. In these cases, Cl−, δ18O, and the corresponding EM data showed discontinuous and skewed distribution.
RMSE was strongly correlated with the square of the difference in means of the estimated value obtained from the partial dataset and full data for Cl− and δ18O.
Discussion
In this study, we suggested that the number of borehole surveys can be reduced by adequate siting without a significant decrease in estimation accuracy. The partial dataset can produce comparable results with the full model when the Cl−, δ18O, and EM data can be considered as a representative of the full model. COK uses the mean, correlation coefficient, and covariance of the primary (borehole) and secondary (resistivity) data. Therefore, in the Horonobe area, where Cl− and δ18O are correlated with resistivity, borehole survey at a site covering the entire range of resistivity obtained by the EM survey may reflect representative geochemical features and result in adequate coefficients in COK.
This study was conducted as part of the “FY2021 Technical Development Project on Geological Disposal of High-Level Radioactive Waste (Research and development on groundwater flow evaluation technology in bedrock)” commissioned by the Agency for Natural Resources and Energy, Ministry of Economy, Trade, and Industry.
The hydraulic environment with slow groundwater flow rates is considered to be a suitable geological environment for the site selection of the geological disposal of high-level radioactive waste. Seawater trapped during sedimentation is a source of paleo-saline water. The presence of such saline water indicates that groundwater flow in the area is sufficiently slow. The distribution of saline water can be assessed based on geochemical properties, such as chloride ion concentration (Cl−) and stable isotope (δ18O, δD) compositions of groundwater. Therefore, the site selection process requires a technique to accurately estimate the three-dimensional (3D) distribution of the aforementioned parameters.
Collocated cokriging (COK), which combines borehole and electromagnetic (EM) data, is an efficient method for estimating hydrologic conditions. In terms of the time and cost constraints of borehole survey, achieving reasonable precision with the small number of boreholes is necessary. In this study, we aim to evaluate borehole siting based on COK results for estimating the 3D distribution of saline water.
Methods
Cl− and δ18O data obtained from 10 boreholes (HDB-1–HDB-11, excluding HDB-2) in the Horonobe area and 3D resistivity distribution from the EM survey were used for COK. First, the 3D distribution of Cl− and δ18O were estimated by COK using the results from all boreholes (full model). Next, fifteen partial datasets were created from one to three boreholes, which were subsampled by considering resistivity and geological features. COK was performed using the datasets in the same manner. The root mean square error (RMSE) of the entire model was calculated for each dataset to evaluate the estimation accuracy.
Results
Of the fifteen partial datasets, three sets (HDB-1, HDB-3+5, and HDB-3+5+7) produced the 3D distribution of Cl− and δ18O, which were well congruent with that of the full model, and resulted in a comparably low RMSE. In HDB-1, where a single borehole was used, the corresponding EM data were generally low, and the distributions of Cl− and δ18O in the subset were similar to those of the full model. In HDB-3+5, where two boreholes were combined, each borehole showed a skewed distribution of Cl− and δ18O. However, the distribution was continuous after integrating both boreholes. HDB-3+5+7, a set adding HDB-7 to HDB-3+5, produced the RMSE that was comparable to HDB-3+5, indicating that the additional data did not improve the accuracy.
The remaining twelve datasets had higher RMSE, but the estimated 3D distributions differed from that of the full model. In these cases, Cl−, δ18O, and the corresponding EM data showed discontinuous and skewed distribution.
RMSE was strongly correlated with the square of the difference in means of the estimated value obtained from the partial dataset and full data for Cl− and δ18O.
Discussion
In this study, we suggested that the number of borehole surveys can be reduced by adequate siting without a significant decrease in estimation accuracy. The partial dataset can produce comparable results with the full model when the Cl−, δ18O, and EM data can be considered as a representative of the full model. COK uses the mean, correlation coefficient, and covariance of the primary (borehole) and secondary (resistivity) data. Therefore, in the Horonobe area, where Cl− and δ18O are correlated with resistivity, borehole survey at a site covering the entire range of resistivity obtained by the EM survey may reflect representative geochemical features and result in adequate coefficients in COK.
This study was conducted as part of the “FY2021 Technical Development Project on Geological Disposal of High-Level Radioactive Waste (Research and development on groundwater flow evaluation technology in bedrock)” commissioned by the Agency for Natural Resources and Energy, Ministry of Economy, Trade, and Industry.