Townley, L.R., Barr, A.D., Braumiller, S., Kawanashi, M., Lever, D.A., Miyakawa, K., Morris, S.T., Raffensperger, J.P., Smoot, J.L., Tanaka, Y., and Trefry, M.G. (1992), Alligator Rivers Analogue Project Final Report, Volume 6, Hydrogeological Modelling, ANSTO, Lucas Heights, N.S.W., Australia, 133pp.

The secondary dispersion fan of the No.1 Orebody at Koongarra is believed to be the result of groundwater transport and chemical processes over long periods of time. Understanding the hydrogeology of the site has therefore been a significant goal of ARAP, both directly, through hydrogeological field studies, and indirectly, through the use of geophysics, petrophysics, water chemistry and rock chemistry as indicators of past and present hydrogeological processes. At the same time, there has been a major effort to use existing hydrogeological flow models to interpret available data and to predict rates and directions of groundwater flow.

This Volume describes a number of attempts to predict groundwater movement near the orebody. The goal was to use available data, both physical and chemical, quantitative and qualitative, to predict rates and directions of movement under present-day hydrogeological and climatic conditions. Paleohydrology was specifically excluded, mainly because it was perceived that there would not be enough data to make significant progress in that direction. Volume 14 of this series focuses on radionuclide transport, thus coupled flow and transport modelling is also specifically excluded from this volume. Any use of chemical data in this Volume is qualitative only, in that spatial distributions of dissolved or sorbed species may provide evidence for past or present directions of flow.

This Volume presents a hierarchy of models applied by a number of individuals working in various locations worldwide. The hydrogeology of the Koongarra site is complex and difficult to model because (i) subsurface flow is a minuscule component of the overall water balance of the region, (ii) significant flow occurs as fracture flow rather than as porous media flow, (iii) the region is extremely heterogeneous, with a wide range of material types and fracturing on several scales, (iv) many of the available hydrogeological measurements were collected from exploration holes, which were constructed to determine the extent of uranium orebodies rather than specifically to understand groundwater flow, and (v) observations are concentrated near the No.1 Orebody, hence there is limited knowledge about the regional geology and about boundary conditions for regional groundwater flow.

All but one of the models described in this Volume are porous media models, which assume that the fractured medium acts as a continuum and that flow is governed by Darcy's Law with some effective or average hydraulic conductivity. One model is a fracture flow model, based on generating a finite number of discrete fractures with random orientations. Another model is an inverse model, which attempts to utilise data from aquifer tests to estimate heterogeneous transmissivities and aquifer storativities in a number of zones.

The hierarchy of models includes models with varying dimensionality and models which are both steady and transient. Each type of model is able to describe at least some features of the Koongarra site.

The first models actually applied were two-dimensional models in vertical section. These models showed that if there exists a water table beneath the Mount Brockman Massif, flow from Mount Brockman can indeed occur towards the Cahill Schist with flows moving generally upwards from the Koongarra Fault through the primary orebody towards the land surface. Details of the flow pattern and velocities near the orebody are quite sensitive to the assumed hydraulic conductivity of the fault zone.

The early cross-sectional models assumed that a groundwater divide exists beneath Mount Brockman and that there is also a no-flow boundary beneath Koongarra Creek, which acts as the final discharge point for the flow system. In order to check these assumptions, another two-dimensional cross-sectional model was set up on a regional scale. This model assumed boundaries on the Arnhem Land Plateau and on the far side of Mount Brockman, and also included the Sawcut Fault as another major fault system. Sensitivity studies showed that it is indeed very likely that a local flow system exists as originally hypothesised.

The next models developed were three-dimensional. Two separate models were created for a region 3 km square. The purpose of these models was to demonstrate, with similar assumptions to those used in two-dimensional vertical sections, that flow directions could deviate significantly from a plane. By including a highly conductive zone to represent Koongarra Fault, it was found that significant flows could occur horizontally along the Fault. A third three-dimensional model was developed independently at a much more local scale. The purpose of this model was to demonstrate the significant effect of anisotropy, which was determined based on Borehole TV measurements of the orientations of planes of schistosity.

Aquifer testing described in Volume 5 of this series uses the concept of transmissivity in its interpretation of aquifer response to pumping. The concept of an aquifer, a layer transmitting significant quantities of water in a mainly horizontal direction, seems hard to accept in an environment as heterogeneous as that at Koongarra. But modelling of aquifers both in one dimension and two-dimensionally in plan has contributed significantly to our understanding of the site. A one-dimensional model with three layers (often described as a quasi two- dimensional model) was applied to flow between the Fault and Koongarra Creek. Being a transient model, this model was able to show that reverse flows can indeed occur back towards the Fault, but only if there is distributed recharge over the orebody as well as a mechanism for the Fault, or a region near the Fault, to remove water from the simulated cross-section. The model also showed clearly that the response of the three-layered system, consisting of a highly weathered zone, a fractured transmissive zone and a less conductive lower schist zone, is governed mainly by the transmissivity and storage coefficient of the middle layer. The storage coefficient of the higher layer has little effect. A two-dimensional model in plan used a description of anisotropy to show that reverse flows can also occur even without a conducting Fault.

Modelling of a three-dimensional region using discrete fractures showed that it is certainly possible to simulate systems like that observed at Koongarra, but that large amounts of data are probably needed to obtain realistic descriptions of the fracture networks. Inverse modelling of aquifer test data has shown that inverse procedures may have some advantages over fitting of analytical type curves, especially in highly heterogeneous environments.

The objective of determining rates and directions of groundwater movement has not been achieved with any degree of confidence. The most that can be said is that there are many combinations of hydrogeological and climatic variables which could produce groundwater velocities in the range apparently needed by coupled flow and transport models. In the final analysis, hydrogeological data do not contain as much information as the chemical data, thus fitting of coupled models is more likely in the long term to be able to identify the best combination of model parameters.

On the other hand, in order to interpret present-day spatial distributions of dissolved chemical species, it is in principle necessary to run models for millions of years. Since it is yet not possible to run three-dimensional models for these time scales, there may be results that can be achieved from three-dimensional hydrogeological modelling alone that can not yet not be achieved from coupled modelling. There is clearly a role for hydrogeological modelling, but modelling is not a substitute for careful design of field experiments and careful analysis of data.


Copyright © 2005 by Lloyd Townley
Last revised: 6 May 2005