| This thesis describes new methods for the characterization of fractured bedrock aquifers and, via their application to a set of field data collected at Smithville, Ontario, Canada, provides new insight into the nature of fracture versus matrix flow in rock.;Two complementary Maximum Likelihood methods are developed that provide log-normal parameter estimates from borehole measurements of interval transmissivities. The methods are tested using the Monte Carlo method based on interval transmissivities simulated from the properties of specified fracture populations. The first method was shown to estimate the transmissivity distribution of the most transmissive fractures from among the population at a user-specified scale (the inverse of the density of the Poisson Point Process assumed to describe their locations along the borehole), while the second method was shown to estimate sums of the transmissivities of the remaining fractures. The properties of the first method were shown to resolve certain difficulties resulting from the use of existing methods for both the analysis and design of fixed-interval length hydraulic testing programs, while the properties of the two complementary methods together provide tractable definitions for fracture versus matrix flow.;Expressions for aquifer hydraulic conductivity and average linear groundwater velocity per unit hydraulic gradient are developed in terms of the fracture set scale parameter, the log-normal parameter estimates for the fractures transmissivity, and log-normal parameter estimates for the matrix interval transmissivity. Estimation accuracy for these important quantities is shown to be optimal for a moderately-heavy-tailed distribution for the fracture population transmissivities. It is pointed out that the job of drilling-geologists during core logging and analysts of hydraulic test data is one and the same. That is, to determine at what scale and with what physical properties such a fracture set exists. |
