Geostatistical analysis of groundwater chemistry in Japan. Evaluation of the base case groundwater data set

The groundwater chemistry is one of important geological environment for performance assessment of high level radioactive disposal system. This report describes the results of geostatistical analysis of groundwater chemistry in Japan. Over 15,000 separate groundwater analyses have been collected of deep Japanese groundwaters for the purpose of evaluating the range of geochemical conditions for geological radioactive waste repositories in Japan. The significance to issues such as radioelement solubility limits, sorption, corrosion of overpack, behavior of compacted clay buffers, and many other factors involved in safety assessment. It is important therefore, that a small, but representative set of groundwater types be identified so that defensible models and data for generic repository performance assessment can be established. Principal component analysis (PCA) is used to categorize representative deep groundwater types from this extensive data set. PCA is a multi-variate statistical analysis technique, similar to factor analysis or eigenvector analysis, designed to provide the best possible resolution of the variability within multi-variate data sets. PCA allows the graphical inspection of the most important similarities (clustering) and differences among samples, based on simultaneous consideration of all variables in the dataset, in a low dimensionality plot. It also allows the analyst to determine the reasons behind any pattern that is observed. In this study, PCA has been aided by hierarchical cluster analysis (HCA), in which statistical indices of similarity among multiple samples are used to distinguish distinct clusters of samples. HCA allows the natural, a priori, grouping of data into clusters showing similar attributes and is graphically represented in a dendrogram Pirouette is the multivariate statistical software package used to conduct the PCA and HCA for the Japanese groundwater dataset. An audit of the initial 15,000 sample dataset on the basis of information on depth and temperature reduces the number of samples to 7,140. These waters are then screened on the basis of thermal imprints effects (recorded temperature > predicted temperature from geothermal gradient and depth information). These 880 thermally altered waters are used in volcanism scenario analyses. The remaining samples are divided into two groups, groundwaters at < 200m depth and groundwaters > 200m depth. The former group is identified for use in the uplift and erosion scenario (in which the repository may approach closer to the ground surface in the far future due to these two processes). The remaining 950 samples define the deep groundwaters for a potential repository. A subset (560 samples) of the 950 deep samples based on maximum sample depth information is selected by JNC considering rock formation (excluded Quaternary unconsolidate sediments) as the Deep Repository Groundwater Dataset for the PCA/HCA. Combined PCA and HCA of the Deep Repository Groundwater Dataset substantiates three reference groundwater categories: 1) a dilute, Na-HCO3 type groundwater possibly divided into two subtypes based on pH [one with high pH (∼8) and one with low pH (∼6), 2) a higher temperature (∼40degC) moderate ionic strength Na-HCO3 type groundwater with pH between 7 and 9; and 3) a saline (near seawater ionic strength), high pH (∼8) brine. Comparison of these results with the proposed hypothetical reference groundwater samples indicates that some of the proposed reference groundwaters do not appear to closely represent any of the categories derived by PCA/HCA for the deep Japanese groundwater data sets used. There may be several reasons for this. First, the initial data set of 15,000 samples does not necessarily represent a complete, uniform coverage of groundwater have been eliminated before PCA/HCA because the samples lacked one of key major chemical variables selected for the analyses (e.g., Na+). Thus, in order to assure a broad scope of potential repository waters, a low-pH saline water (SRLP) and a high-carbonate water (MRNP) are included as possible reference groundwaters. (author)