Estimations of Hydrogeological Properties in Vulnerability and Risk Assessments

Doctoral thesis, 1995

The main objective of this thesis was to evaluate and develop approaches for hydrogeological vulnerability and risk assessments, starting with scant data and updating the results as new information becomes available. This work was performed within three separate projects between 1989 and 1995. The first project was aimed at application and evaluation of properties of standardized groundwater vulnerability classification systems. The work was mainly directed at the DRASTIC rating methodology. It was shown that (1) DRASTIC has some favorable statistical properties from the use of a fairly large number of statistically correlated key parameters; (2) the selection of parameters is relevant for general groundwater vulnerability assessments in Sweden; (3) the concept of hydrogeological settings descriptions is very useful; (4) the implications of the vulnerability index are not clear and the results may therefore be less distinctive and less useful than desired; and (5) the system tends to overestimate the vulnerability of porous media aquifers compared to aquifers in fractured media. The second project was aimed at development of a guidance framework for monetary risk assessments before collection of new data. The framework has a dual-site approach, assuming a situation of a contaminant source site and a receptor site. A two-step procedure leads to the monetary risk assessment with respect to existing compliance levels: (1) conceptual model development, and (2) quantitative model realization, addressing the engineering design of the contamination source, the hydraulic connection between the source and the receptor, and the contaminant transport conditions. The third project was aimed at the development of approaches and methods for preparing geological and hydrogeological decision bases in the siting and construction phases of a nuclear waste repository. The concept of a Positioning Index (PI) was developed for describing the conditions for fuel canister positioning with respect to specific compliance levels. A Bayesian Markov Geostatistical Model (BayMar) was developed for probability estimations with respect to set compliance levels. Application of the methodology showed that BayMar is capable of predicting conditions with respect to lithology, hydraulic conductivity, and rock quality designation index (RQD) reasonably well for both site and repository tunnel scales. A comparison between BayMar and indicator kriging showed that the two methods are analogous in describing the spatial correlation structure since the autocovariance function of BayMar is the reciprocal of the variogram. Main advantages of BayMar are that (1) it is capable of handling all mutually exclusive states simultaneously; (2) no fit to any standard variogram is required; and (3) it provides a formal way for handling both previously existing information and professional judgements. When comparing predictions at Äspö, BayMar gave results closer to observed conditions than did indicator kriging.