NUMERICAL MODELLING OF SALTWATER PLUME MOVEMENT IN A FRACTURED AQUIFER ON SUB-REGIONAL SCALE

Transcription

1 NUMERICAL MODELLING OF SALTWATER PLUME MOVEMENT IN A FRACTURED AQUIFER ON SUB-REGIONAL SCALE Sascha E. Oswald (Groundwater Protection and Restoration Group, University of Sheffield, United Kingdom) INTRODUCTION In many countries saltwater plumes in freshwater aquifers pose a concern in groundwater protection matters. Of main interest often are the catchment zones of groundwater abstractions and predictions of salt concentration in wells, because high salt concentrations are a hazard to the use as drinking water. These problems arise mostly in the vicinity of coastlines, but there are also cases of aquifers exposed to the intrusion of ancient saltwater from other aquifers. The latter case is subject of this study, modelling a fractured sandstone aquifer located in Southern Africa based on field data. The semi-arid climate with low recharge values implies the catchment zone of the water wells to be large in extent. These large horizontal dimensions make modelling as a fracture network very demanding, and additionally the sandstone matrix of the aquifer is assumed to contribute to the flow of water. Therefore, an equivalent porous medium approach was chosen to model the movement of the saltwater plume. The effect of increased pumping are studied including density effects via calculating the time development of the salt concentration distribution, starting from an assumed steady state, and the breakthrough curve in a well field. INTRUSION OF ANCIENT SALTWATER INTO A FRESHWATER AQUIFER The Palla Road aquifer is part of a Graben structure and is located in the semi-arid southeastern part of Botswana. It consists of the Ntane fractured sandstone with a thickness varying between approximately 60 m and 120 m. The major part is covered with a basalt layer with a thickness of up to 100 m. However, in a significant part of the area the sandstone aquifer is only covered by Kalahari sands. The main withdrawal of groundwater started in 1988 in two well fields in the southeastern part from wells fully screened in the aquifer. The withdrawal of water was increased steadily, and the rate was planned to be increased further during the time of this study. The study area covers about 50 x 20 km, and thus includes the main part of the current catchment zone of the well fields. FIGURE 1. Two-dimensional regional model, simulated heads (all units in meters). Area of three-dimensional model is located in southeastern part as indicated by bright line. 1

2 Inflow of freshwater takes place mainly at the western boundary of the aquifer and outflow at the eastern boundary (see also figure 1). Infiltration from the top usually is minor due to the overlying basalt unit, but in the northeastern part without basalt cover the recharge rates are relatively high. This recharge delivers a major component of the water extracted in the well fields, which are not far away. Along the central part of the southern boundary of the aquifer, which corresponds with a fault zone, there is a lateral intrusion of saline water of about 1 % salt mass fraction from a deeper fossil aquifer. MODEL AND RESULTS An existing, calibrated, two-dimensional, flow and transport model covering the whole, regional study area (Siegfried and Kinzelbach, 1997) was used as a basis. Then a three-dimensional aquifer model on sub-regional scale was telescoped in. It covered the lateral saltwater intrusion as well as the two well fields. This model area was delineated by the hydraulic heads calculated with the twodimensional model and the location of intruding saline water identified via according transport calculations. The horizontal dimensions are approximately 30 x 8 km and the thickness of the aquifer is between 80 and 100 m. The aquifer dips there by about one percent from north to south. The recently developed three-dimensional variable-density flow code d 3 f (Fein, 1999) was applied, because the saltwater plume may exhibit density effects which could change the movement of the saltwater plume and thus impact the salt concentration in the wells. This could be triggered by vertical differences inside the aquifer as well as by its dip. The new model consisted of four homogeneous hydrostratigraphic units of varying permeability and porosity (see table 1). TABLE 1. Parameters used in the simulations Parameter Upper unit 1 Upper unit 2 Lower unit 1 Lower unit 2 Horizontal permeability k h [m 2 ] 6.83 x x 10 -l x x Vertical permeability k v [m 2 ] 1.37 x x x x Porosity n [-] Longitud. dispersivity α L [m] 600 Transverse dispersivity α T [m] 60 Effect. diffusion coeff. D m [m 2 /s] 8.7 x 10 -l0 The three-dimensional geometric representation, defined by the linear interpolation of borehole data, was transferred into a model grid consisting of tetrahedrons and triangular prisms. Longitudinal and transversal dispersivities were estimated to be 600 m and 60 m, respectively. Due to the resulting large vertical dispersion the vertical concentration changes are small and therefore the three-dimensional aquifer model behaves more like a dipping two-dimensional aquifer. 2

3 FIGURE 2. Horizontal cross-section of steady state salt concentration distribution (red indicating maximum salt concentrations, blue freshwater). The boundary conditions matched to the ones of the two-dimensional model where boundaries were geometrical identical. The inner flow boundaries of the three-dimensional model were chosen mainly as fixed pressure according to the appropriate head equipotential lines and salt concentrations were set to be close to zero there. The initial freshwater distribution transforms into a saltwater plume due to the saltwater intrusion at the central southern boundary with fixed salt concentrations. After several thousand years the plume becomes stationary (see figure 2). The water abstraction of the past years and the future abstraction was put together into an abstraction scenario. This assumed the pumping rates to increase for 12 years, then to remain constant, and the outflow from the aquifer to vary accordingly. The intrusion of saline water and the recharge rates were set to be constant in time. After the start of pumping the saltwater plume moves slowly towards the eastern part and exhibits an upconing towards the main well field (see figure 3). There the salt concentration in the extracted water increases significantly, reaching a maximum after about one hundred years, then decreases while tending to a new steady state after several hundred years (see figure 4). However, the concentrations do not exceed critical levels. 3

4 FIGURE 3. Horizontal cross-section, ca. 110 years after the start of pumping: a) Saltwater plume b) Horizontal flow field (towards the main well field). For comparison purposes the simulations were repeated in the same manner without accounting for density effects. The results show a very similar behavior, especially the salt concentrations breakthrough curves. It can be concluded that density effects probably influence the resulting salt concentration in the wells only weakly, because of the high degree of mixing in this particular situation, though the aquifer has a noticeable dip. DISCUSSION AND CONCLUSIONS The simulation results show that the water quality in the wells is not a critical factor for water abstraction under conditions similar to the assumed scenario. This is the case for simulations accounting for density effects as well as without density effects. However, the results illustrate that for water resources management, which aims to achieve a sustainable use of the freshwater resource, time scales of hundreds of years have to be taken into account. 4

5 FIGURE 4. Calculated breakthrough curve in main well field (line : with density effects; crosses : without density effects; concentration in salt mass fractions). Modelling of salt transport in a large, but relatively thin three-dimensional fractured aquifer like the one studied here is a challenging task, especially for variable density flow conditions. The large dimensions and the lack of sufficient data make a fracture network approach very problematical. Additionally, the aspect ratio of more than 100:1 imposes difficulties in simulating vertical differences in salt concentrations. However fractures as preferential flow paths, as well as incomplete vertical mixing in the aquifer, could affect the plume movement significantly, especially in the vicinity of the pumping wells. Thus, in cases of strong upconing below the wells, a modelling approach including fractures and high vertical grid resolution would be necessary to assess the real behavior in such an aquifer. REFERENCES Fein, E. (editor) d 3 f A Simulator for Density-Driven Flow. User's Manual. Gesellschaft für Anlagen- und Reaktorsicherheit (mbh), Braunschweig, Germany. Siegfried, T., and W. Kinzelbach Modelling of groundwater withdrawal and its consequences to the aquifer: A case study in semi-arid Botswana. Report of the Institute of Hydromechanics and Water Resources Management, ETH Zurich, Zurich, Switzerland. TOC 5