Groundwater flow in the vicinity of two artificial recharge ponds in the Belgian coastal dunes A. Vandenbohede & Emmanuel Van Houtte & Luc Lebbe Abstract Since July 2002, tertiary treated wastewater has been artificially recharged through two infiltration ponds in the dunes of the Belgian western coastal plain. This has formed a lens of artificially recharged water in the dunes’ fresh water lens. Recharged water is recovered by extraction wells located around the ponds. Hydraulic aspects of the artificial recharge and extraction are described using field observations such as geophysical borehole loggings and a tracer test. Borehole logs indicate recharged water up to 20m below surface, whereas the tracer test gives field data about the residence times of the recharged water. Furthermore, a detailed solute transport model was made of the area surrounding the ponds. Groundwater flow, capture zone, residence times and volume of recharged water in the aquifer are calculated. This shows that the residence time varies between 30days and 5years due to the complex flow pattern. The extracted water is a mix of waters with different residence times and natural groundwater, assuring a relatively stable water quality of the extracted water. Keywords Groundwater recharge/water budget . Capture zone . Residence times . Numerical modeling . Belgium
Introduction The phreatic aquifer of the dunes in the Belgian western coastal plain is important as a source of fresh water such as that used for the local water supply. In the surrounding
Received: 5 November 2007 / Accepted: 27 May 2008 Published online: 26 June 2008 © Springer-Verlag 2008 A. Vandenbohede ()) : L. Lebbe Research Unit Groundwater Modelling, Department of Geology and Soil Science, Ghent University, Krijgslaan 281 (S8), 9000, Gent, Belgium e-mail:
[email protected] Tel.: +32-(0)9-2644652 Fax: +32-(0)9-2644653 E. Van Houtte Intermunicipal Water Company of the Veurne Region (IWVA), Doornpannestraat 1, 8670, Koksijde, Belgium Hydrogeology Journal (2008) 16: 1669–1681
areas (sea, shore and polder), mainly brackish to salt water is found. Otherwise, fresh-water lenses present in the phreatic polder aquifer contain very limited fresh-water resources (Vandenbohede and Lebbe 2002). Tertiary sediments below the Quaternary phreatic aquifer consist mainly of low permeable sediments and/or contain brackish water or salt water. Consequently the fresh-water lens under the dunes forms the main groundwater reserve which can be exploited. Groundwater exploitation for the production of drinking water started in 1947 at the water extraction St-André by the Intermunicipal Water Company of the Veurne Region (IWVA; Fig. 1). The IWVA is responsible for the distribution of drinking water in the western part of the Belgian coastal plain. To remediate decreasing water levels and also to guarantee current and future water extraction possibilities, alternative exploitation methods were studied during the 1990s (Van Houtte and Verbauwhede 2005; Vandenbohede et al. 2008). As in many other water extractions (e.g. Asano 1992; Van Breukelen 1998; Bouwer 2002; Greskowiak et al. 2005; Massmann et al. 2006), it was opted to artificially recharge the aquifer using superficial ponds. In July 2002, the IWVA started with artificial recharge of the phreatic dune aquifer using treated wastewater effluent. This effluent undergoes an extra (tertiary) treatment using ultra filtration and reverse osmosis prior to recharge in the dune aquifer. Recharge is realised by means of two shallow interconnected ponds. The water is extracted using a well battery (well battery two on Fig. 1) located around the ponds. A mixture of artificially recharged and native dune water is extracted. In this way, it is possible to reduce the net groundwater extraction, whereas the total volume of extracted water has increased. Consequently, the result is a sustainable water extraction in an area with limited freshwater reserves (Vandenbohede et al. 2008). The purpose of this paper is to discuss hydraulic aspects of the interaction between recharge ponds and extraction wells. This is done by integrating two approaches. Firstly, results of a number of field measurements and tests are available. Geophysical borehole loggings (electromagnetic conduction) give information of the extension of the recharged water whereas a tracer test makes it possible to derive residence times. Secondly, a groundwater flow model, using the MOCDENS3D computer code (Oude Essink 1998), is developed in which all available information (recharge and extraction rates, geological data etc.) is DOI 10.1007/s10040-008-0326-x
1670
Fig. 1 The St-André water extraction is located in the Belgian western coastal dunes near the French-Belgian border (square in index map). The water extraction consists of two well batteries and the artificial recharge ponds are located between the wells of well battery 2
summarized. This results in a more detailed and threedimensional insight into the groundwater flow in the vicinity of the recharge ponds. Calculation of residence times of the recharged water and capture zone of well battery 2 are important tools necessitating the use of the MOCDENS3D computer code (Oude Essink 1998).
St-Andre infiltration project The dunes of the Belgian western coastal plain are very well suited for the extraction of water. The phreatic aquifer consists of fine medium sands in which lenses of silty or clayey fine sand can occur. A shallow silty layer exists, for instance, under the western pond and this layer acts as a semi-permeable layer. The influence of this layer will be discussed further in the report. The aquifer has a thickness of about 30 m; the width of the dunes (from high water line to polder) is 1.75 km. The substratum of the aquifer is formed by the clay of the Kortrijk Formation, Ieper Group. This clay is of Eocene age and is considered here as an impermeable boundary. Landward from the dunes, a polder (artificially drained low-lying area) is located. A fresh-water lens is found in the dune aquifer. In the polder, brackish water to salt water is found, whereas a saltwater lens occurs above fresh water under the shore. More Hydrogeology Journal (2008) 16: 1669–1681
details on the origin and evolution of the fresh-water–saltwater distribution under the shore and in the dunes and polder can be found in Lebbe (1978), Vandenbohede and Lebbe (2006) and Vandenbohede et al. (2008). This fresh-water– salt-water distribution means that the only relatively important fresh-water reserves are situated in the dune aquifer. Extraction of water from the dunes for the production of drinking water started in St-André in 1947. The extraction rate steadily increased from about 0.5 million m3/year during the 1950s to 1.75 million m3/year during the 1960s. From then on, extraction rates remained 2 million m3/years up to the second millennium. Initially, the water extraction started with one well battery with 109 wells (well battery 1 in Fig. 1). In 1968, a second well battery was put in use with 54 wells. Extraction rates of both well batteries were more or less the same. Extraction wells had screens between 6 and 10 m below surface; in the 1980s new wells were drilled with screens between 12 to 16 m below surface. In 2002, 70 wells were active in conjunction with the first well battery and 112 wells in conjunction with the second well battery. To partially restore hydraulic heads and natural groundwater flow in the dunes around the St-André extraction and to ensure water production which meets the demands, it was decided in the mid 1990s to artificially recharge the dune aquifer. After considering different options, the artificial recharge project started in DOI 10.1007/s10040-008-0326-x
1671
July 2002. The system which is used now consists of two recharge ponds surrounded by wells (well battery 2). The ponds are interconnected through a subterranean pipe. Effluent, from a nearby waste-water treatment plant, is used for recharge after additional treatment using ultra filtration (UF) and reverse osmosis (RO) (Van Houtte and Verbauwhede 2005). The recharged water is thereafter extracted via the wells of well battery 2 situated north and south of the ponds (Fig. 2). Figure 2 shows, instead of the 112 wells, only 110 wells. This is because Fig. 2 shows the well positions in the finite difference grid of the model whereby 2×2 wells are located in the same finite difference cell in two occasions. Wells 1–10 have submersed pumps, whereas wells 11–110 are connected via a siphon to a central collector from which water is extracted. The artificial recharge ponds are located in a dune slack. The recharge ponds are in direct connection with the saturated zone, there is no unsaturated zone between the bottom of the ponds and the water table. The artificial recharge project loops the water cycle: extracted water goes to the users and their waste water is purified. This water, after extra treatment, is then again reused to recharge the dune aquifer. Up to 2.5 million m3/year can be recharged and the same amount of water can be extracted. An additional amount of 1.7 million m3 natural dune water can be extracted (1 million by well battery 2 and 0.7 million by well battery 1). This means that the capacity of the water extraction is more than doubled (from 2 million to 4.2 million m3/year), whereas the net amount of water extracted from the dune aquifer is reduced from 2 million to 1.7 million m3/year. Figure 3 shows the total dissolved solids content (TDS, mg/l) of the recharge and extracted water as a function of time. From the start of the artificial recharge project until May 2004, TDS in the recharged water was approximately 100 mg/l. This water consisted of 90% RO and 10% UF water. The latter is added to the former to mineralise the recharged water to a certain extent. This mineralisation is required before recharging the dune aquifer. From May 2004, NaOH is used to mineralise and adjust the pH of the recharged water instead of UF water and TDS of the recharged water became from then on smaller, approxi-
Fig. 3 Total dissolved solids (TDS, mg/l) of the recharge and extraction water as function of time
mately 50 mg/l. The extracted water is a mixture of recharged water (80%) and native dune water (20%). The recharged water becomes also mineralised by for instance chalk dissolution in the aquifer (unpublished data). This means that the TDS of the extracted water is substantially higher than that of the recharged water. The TDS in the extracted water before the start of the artificial recharge project was approximately 800–900 mg/l. This is larger than the TDS of shallow dune water which is approximately 550 mg/l due to the fact that water originating from deeper parts of the dune aquifer is also flowing to the extraction wells. This deeper water has a slightly higher TDS. TDS of the extracted water decreased steadily in the first year of the project to about 325 mg/l or slightly more than 3 times the TDS of the recharged water. Change in composition of the recharged water in May 2004 is also visible in the extracted water since the TDS decreased to about 225 mg/l or 4.5 times the TDS of the recharged water. Notice that during the second part of 2002, TDS of the extracted water increased to 800 mg/l. During this period, recharge rates were very low (70,000 m3/month instead of 175,000 m3/month) due to start-up problems with the production of recharged water. The extraction rate remained, however, unaltered, meaning that more native dune water was extracted. Therefore, TDS of the extracted water increased temporally.
Field observations Geophysical borehole measurements
Fig. 2 Model grid centred over the two ponds and position of the extraction wells of well battery 2. Well numbers of individual wells are indicated as are the observation wells (WP) Hydrogeology Journal (2008) 16: 1669–1681
A focussed electromagnetic induction tool (EM39, Geonics) is used for geophysical borehole logging. The EM39 is specially designed for use in wells encased with electrical non-conductive materials. This is a major advantage in comparison with older electrical methods (long normal and short normal) which could only be performed in open boreholes or in fully screened wells. EM39 employs a small internal transmitter coil energised with an audio-frequency current to induce eddy currents in the soil surrounding the well. These eddy currents generate an alternating secondary magnetic field which can be observed by small receiver coils located at some distance from the transmitter. The small secondary magnetic field is linearly proportional to the DOI 10.1007/s10040-008-0326-x
1672
electrical conductivity of the surrounding material and the device can be calibrated to read the terrain conductivity directly (McNeill 1986). The arrangement of coils provides a relatively large lateral range and a high degree of vertical resolution which makes it very suitable for hydrogeological research. EM39 measures the electrical conductivity of the surrounding sediments within a distance range from 20 to 100 cm from the well axis while being insensitive to conductivity of the borehole fluid and disturbed material situated near the well axis. The vertical resolution is a few decimetres. EM39 measurements were made in wells WP6.1, WP7.1, WP24, WP23, WP21 and WP22 (see Fig. 2 for locations) in August 2006. Earlier measurements (August 2002) were also performed in all but WP7.1. Vertical scale is expressed in mTAW, whereby 0 mTAW is the Belgian reference level (2.36 m below mean sea level). WP7.1 is located south of well battery 2, which means that native dune water is expected in this location. Mean electrical conductivity of the dune aquifer at WP7.1 (Fig. 4) is 20 mS/m corresponding with a TDS of the pore water of 550 mg/l. There are two zones with a slightly higher conductivity representing layers or horizons with slightly increased silt or clay content. Conductivity increases also below –20 mTAW, due to the higher TDS values of the pore water at the bottom of the phreatic aquifer and due to the transition of the Quaternary sediments to the clay of the Kortrijk Formation. WP6.1 is located close to the ponds and the EM39 profile can be subdivided in two parts. A low conductivity of 6 mS/m is seen in the first 15 m of the August 2006 measurement. This corresponds with a TDS of 175 mg/l and is the recharged water which has already undergone mineralisation because of the passage through the upper part of the dune sediments. A lens or horizon with higher conductivity is present around –10 mTAW, which is due to an increase in silt or clay content of the sediments. Below –13 mTAW, conductivity increases to 27 mS/m, which is due to the fact of the native dune water becoming more mineralised deeper in the dune aquifer. Notice that there is an important difference between the 2002 and 2006 measurement. The August 2002 measurement was made only 1 month after the start of the
artificial recharge project. Also, water with a higher TDS was recharged than is currently the case (Fig. 3). Although the distinction between recharge and native dune water can be seen on the 2002 measurement, conductivity of the former water is still larger than in the 2006 measurement. WP24 is located between the western pond and the extraction wells but has only a casing in the upper part of the aquifer. Mean conductivity is 6 mS/m in 2006, indicating the presence of recharged water. A higher conductivity was again measured in 2002. Notice also that the recharged water is observed from the top of the profile onwards. EM39 measurements in WP23, WP21 and WP22 show the same results as WP24 and are not shown here.
Tracer test A tracer test was performed to derive the residence time of the artificially recharged water in the dune aquifer. Therefore, NaF was dissolved in the recharged water whereby fluoride was used as a tracer. After addition of the tracer, fluoride concentration in the eastern pond measured 0.95 mg/l, whereas this was 1.33 mg/l in the western pond. Fluoride concentrations were measured in wells WP21, WP22, WP23, WP24 and in the extracted water. The wells WP are nested wells with three screens, respectively at a depth of 3(XX.4), 7 (XX.3) and 11.5 (XX.2) m below surface level. Distance of WP21 and WP24 to the ponds is 10 m, whereas this is 30 m for WP22 and WP23. Fluoride was measured every two or three days in these wells. Figure 5 shows the fluoride measurements in the observation wells and in the extracted water. In the wells located closest to the ponds (WP24 and WP21), breakthrough of fluoride was observed after 11 days since the start of the test. Increased fluoride concentrations were measured in WP23 and WP22 after 19 days. However, no samples where taken before day 19, so time of breakthrough is uncertain in these cases. There is an important difference between wells WP23–WP24 and wells WP21–WP22. Notice that the deepest screen of WP24 (WP24.2) shows larger concentrations than the shallower screens. This is also the case for WP23, albeit that the difference in concentration
Fig. 4 EM39 measurements giving the electrical conductivity (EC) in function of depth in WP7.1, WP6.1 and WP24 Hydrogeology Journal (2008) 16: 1669–1681
DOI 10.1007/s10040-008-0326-x
1673
Fig. 5 Fluoride concentration in function of time for the different observation wells WP21, WP22, WP23, WP24 (a–d) and of the extracted water (e). The time scale is in the number of days since start of the test
between the different screens of WP23 is much less than for WP24. In WP21, the largest concentrations are observed in the shallowest screen. The reason for this difference is aquifer heterogeneity. As indicated before, a shallow semipermeable layer is present below the western pond but not below the eastern pond. Consequently, there is an important lateral flow component above the semi-permeable layer near the western pond. Therefore shallow observation wells of WP21–WP22 have a larger fluoride concentration than deeper wells. Near the eastern pond, however, recharged water can more easily flow deeper in the groundwater reservoir and the largest fluoride concentrations are observed in the deepest observation wells.
MOCDENS3D model A three-dimensional density-dependent groundwater flow model was needed to obtain a more detailed view of groundwater flow around the artificial recharge ponds, determine capture zones, travel and residence times and the geometry of the volume of recharged water. For this simulation MOCDENS3D (Oude Essink 1998), which is based on the three-dimensional solute transport code MOC3D (Konikow et al. 1996), was chosen and adapted for density differences and Visual Hydrogeology Journal (2008) 16: 1669–1681
MOCDENS3D (Vandenbohede 2007) was selected for the visualisation of model results and the calculation of captures zones and residence times. The MOCDENS3D model of the recharge process is based on an existing regional scale model simulating the groundwater flow and fresh-water–salt-water distribution in the dunes, polders and shore of the area surrounding the St-André water extraction (Vandenbohede et al. 2008). This regional scale model spans an area of 4,200×4,500 m and includes the evolution of the water extraction from its start in 1947 until the end of 2006. First step was, however, the simulation of the current fresh-water–saltwater distribution. Therefore, the replacement of salt by fresh water because of land reclamation and the formation of the dune belt was simulated. Subsequently, the evolution of the water extrication from 1947 to June 2002 was simulated using stress periods of 2 or 3 years. The model was calibrated with hydraulic head observations and water quality data. Stress periods of 2 months were used from the start of the artificial recharge project (July 2002) until the end of 2006 to calculate the effect of the artificial recharge on the groundwater flow and freshwater–salt-water distribution in the whole dune aquifer. The model described here zooms in on the 975×525-m area surrounding the ponds (Fig. 6) and consists of 105 rows and 195 columns. Dimension of every finiteDOI 10.1007/s10040-008-0326-x
1674
Fig. 6 a Horizontal cross-section through the regional flow model at the level of the screens of the extractions wells (Vandenbohede et al. 2008). The rectangular box shows the detailed model around the ponds. The grey scale represents TDS (mg/l) of the pore water, black lines are lines of equal fresh-water heads and arrows indicate the groundwater flow. b Cross-section at the same level through the detailed model showing the chloride concentration (mg/l) and the groundwater flow by means of the flow vectors and lines of equal fresh-water head. Grey scale in b is the reverse from grey scale in a
difference cell is 5×5 m and 12 layers are considered, each with a thickness of 2.5 m, which means that the complete phreatic aquifer is considered in the model and that the clay of the Kortrijk Formation determines the impermeable lower boundary. The north, south, east and west boundaries of the model are constant head boundaries. The values for these constant heads are derived from the regional scale model and are altered after every stress period. Stress periods of 2 months are used, coinciding with the stress periods of the regional scale model. After Hydrogeology Journal (2008) 16: 1669–1681
every stress period, the recharge and extraction rates change as does the concentration of the recharged water if necessary. The solute concentration in the model is that of the chloride concentration of the pore water. Horizontal hydraulic conductivity of the dune sediments is 12 m/d for every layer. Ratio of the horizontal to the vertical conductivity is 20. This anisotropy is 40 between layer 1 and 2 under the western pond, simulating the increased hydraulic resistance of the semi-permeable layer present at this location. This layer is identified in a number DOI 10.1007/s10040-008-0326-x
1675
of drilling descriptions located in the western part of the model domain, but the exact lateral extension is not known. Therefore, it is conceptualised that this layer is present in the western part (from column 1−118) of the model. Anisotropy is also 40 for layer 10−12 because the lower part of the aquifer is less permeable. Horizontal hydraulic conductivity is also smaller (8 m/d). Natural recharge is 280 mm/year, whereas the longitudinal, horizontal transverse and vertical transverse dispersivity are respectively 0.1, 0.01 and 0.001 m. Effective porosity is 0.38. Influences of boundaries are minimised in the model in two ways. First issue is the distance of the boundaries from the wells. From the larger model, the zone of influence of well battery two is known. Based on this, position of boundaries is chosen so that the drawdowns are minimal. Secondly, a combination of constant head and no flow boundary is applied. A constant head boundary assumes the presence of imaginary wells mirroring the pumping wells or injection wells in a number of injection or pumping wells respectively. At the position of the boundaries, the superposition of the pumping or injection (in the model) and the imaginary injection or pumping wells cancel each other resulting in a constant head. However, between the pumping or injection wells and the boundary, the drawdown is influenced because of the imaginary wells. Because of the presence of four constant head boundaries, the model wells result in four sets of imaginary injection and pumping wells. These are also mirrored by the boundaries and so on, resulting in a number of imaginary
injection and pumping wells of first order, second order, etc. The higher the order, the farther these wells are located from the boundaries and the smaller their influence is. These effects can be partially cancelled out by halving the hydraulic conductivities of the first and last rows and the first and last columns of each layer. These values must be regarded as the mean of the expected conductivity values and a conductivity of zero, the latter being an impermeable boundary. Impermeable boundaries are realised in the model by imaginary pumping or injection wells, respectively for pumping or injection wells in the model. Consequently, by the combination of constant head boundaries and the adjustments of the hydraulic heads, the odd numbered imaginary wells (and most important the first order wells) are deleted.
Model results Groundwater flow and water-quality distribution Figure 6b shows a horizontal cross-section through the aquifer at the level of the extraction wells for the end of 2006. The grey scale represents chloride concentration, whereas the arrows show direction and magnitude of groundwater flow. Figure 7 shows two vertical crosssections through the aquifer. Figure 7a, c are cross-sections according to column 80, which is through the western pond; whereas b and d are cross-sections according to column 145, which is through the eastern pond. Distribu-
Fig. 7 Vertical cross-sections through the dune aquifer. a and c Cross-sections according to column 80 which is through the western pond; b and d Cross-sections according to column 140, which is through the eastern pond. Chloride concentration is shown at the end of 2004 (a and b) and at the end of 2006 (c and d). Arrows show the direction and magnitude of the groundwater flow and lines represent the freshwater heads. The position of the extraction wells is indicated with black vertical bars Hydrogeology Journal (2008) 16: 1669–1681
DOI 10.1007/s10040-008-0326-x
1676
tion of chloride is shown at the end of 2004 (a and b) and at the end of 2006 (c and d). Arrows show the direction and magnitude of the groundwater flow and lines represent the fresh-water heads. In these figures, differences between the two ponds are visible. The area with recharged water in the cross-section is smaller for the eastern than for the western pond. Additionally, the presence of recharged water is confined to the area between the pond and the extraction wells for the eastern pond. This is not the case for the western pond. Here, water is present beyond the limits of the extraction wells. This is also visible in the horizontal cross-section in Fig. 6b. There are two reasons for these differences. First of all, the eastern pond is smaller than the western pond. This means that less water recharges and that the volume of recharged water present under the eastern pond will be smaller than the volume of recharged water present under the western pond. Secondly, aquifer heterogeneity is of importance. The presence of a semipermeable layer under the western pond means that recharged water encounters a larger hydraulic resistance for vertical flow. The result is that recharged water will spread more laterally from the western pond. Of note is, for instance, the flow towards the wells located south of the pond. Recharged water flows at first mainly laterally away from the pond before flowing towards the wells; thus, the area where recharged water is present extends south of the extraction wells. This means that the recharged water first flows over the position of the extraction wells before flowing back towards them. The influence of the semi-permeable layer in the simulation agrees very well with the conclusions of the tracer tests. Further, recharged water is found to a depth of approximately 25 m under the western pond and to approximately 18 m under the eastern pond. This agrees with the EM39 measurements in WP6.1, which is situated
between both ponds. In WP6.1, recharged water is observed to a depth of approximately 20 m.
Capture zones Figure 8 shows a horizontal cross-section through the capture zone of the extraction wells. This cross-section is made at the top level of the model giving the time recharged water (natural and artificial) needs to travel from the recharge point to the extraction wells; or, conversely, it gives the residence time of the recharged water (natural and artificial) in the aquifer. Figure 9 shows two vertical cross-sections through the capture zone according to columns 80 and 140. Capture zones are calculated by means of a particle-tracking algorithm within Visual MOCDENS3D. A number of particles are placed in the model grid and their paths are calculated using the x, y and z velocity components of the nodes of the finite difference grid by means of three-dimensional interpolation. The time after which particles are removed from the model grid through the wells is registered. A contour plot of these travel times gives the capture zones. A number of different conclusions can be made from Figs. 8 and 9. For water recharging in the pond, it takes in general less than 50 days to reach the extraction wells. There are, however, a number of exceptions. For instance, a water divide exists in the centre of both ponds. Water recharging south of the divide flows to the southern wells of well battery 2, whereas water recharging north of the divide flows towards the northern wells. This divide is clearly visible on the vertical cross-sections through the capture zone in Fig. 9 as well as in Figs. 7 and 8. Water recharging close to the water divide needs a longer time to reach the extraction wells because it is involved in a longer flow cycle towards these wells. Whereas water
Fig. 8 Horizontal cross-section through the capture zone of the extraction well. Grey scale represents the time (in days) needed for water recharging in the first layer of the model to reach the extraction wells. Times larger than 350 days are not shown here. Position of ponds and extraction wells are indicated by black lines Hydrogeology Journal (2008) 16: 1669–1681
DOI 10.1007/s10040-008-0326-x
1677
Fig. 9 Vertical cross-sections along column 80 (a) and column 140 (b) of the capture zone of the extraction well. Time is in days; times larger than 350 days are not shown. The position of the extraction wells are indicated by black vertical bars
recharging close to the edges of the ponds flows more or less directly to the wells, water recharging in the centre of the ponds flows downwards in the aquifer. When this latter recharged water reaches the bottom of the recharged water lens, it flows upwards and towards the wells. The flow pattern results in the characteristic lobate crosssection through the capture zone with the ponds in the centre as can be seen for the eastern pond (Fig. 9b). The western pond shows a more complex situation because of the location of wells 1–10. These wells are located farther from the pond. Figure 8 shows that wells 1–10 receive recharged water from the north-western part of the western pond but this takes significantly longer than 50 days. The capture zone of these wells is also clearly delineated from the rest of the northern wells (11–40) or from the nearest southern wells (41–50). Figure 9a shows a third lobate zone which is a cross-section through the capture zone of wells 1–10. Notice, for example, that water recharging just north of the water divide does not flow towards the closest wells but towards the wells 1–10. The time to reach the extraction wells is larger than the 350 days shown on the Hydrogeology Journal (2008) 16: 1669–1681
figures. Wells 41–50 which are not located exactly around the pond receive only a small amount of recharged water. Moreover, the most distant wells from these extract mainly naturally recharged water. The flow vectors in Figs. 6 and 7 and the cross sections through the capture zones indicate that most of the artificially recharged water flows towards the extraction wells. This is evidently due to the fact that the ponds are firmly surrounded by extraction wells. The fact that no extraction wells are located west and east of the ponds has no effect on this. At the western side of the ponds, recharged water flows ultimately towards wells 1–10 or 41–50. The largest zone where no wells are placed is situated northeast of the eastern pond. Water which recharges in the eastern part of the eastern pond, however, flows towards wells 30–40 or 100–110.
Residence times Residence times of the artificially recharged water in the dune aquifer are discussed more in detail here also using a DOI 10.1007/s10040-008-0326-x
1678
particle-tracking algorithm of Visual MOCDENS3D. A large number of particles (13 per finite difference cell) are placed in all cells of the model representing the ponds. The movement of these particles is calculated and followed until they reach an extraction well. The time to travel the distance from the ponds to an extraction well is the residence time of the particle in the aquifer. Thereafter, a statistic analysis of the residence times of all particles is made. Figure 10a shows a time-frequency distribution of the residence times of water recharged in the ponds, whereas Fig. 10b shows the cumulative distribution. These figures can be regarded in different ways. First, it indicates when water recharging at the same point in time reaches the extraction well. Eight percent of the recharged water reaches the extraction wells after 27 days; after 36 days, this percentage increases to 25%. Half of the amount of water which was recharged at the same point in time is recovered after 55 days, whereas this is 117 days for 75%. Total recovery of the artificially recharged water is achieved after 1,813 days or almost 5 years. To ensure clarity of the figures, only residence times up to 350 days for the time-frequency plot and up to 500 days for the
cumulative distribution are shown in Fig. 10. Notice that the time-frequency distribution of Fig. 10a is highly asymmetric, which is due to the complex nature of the capture zone of the extraction wells and the high variety of flow paths as indicated in the previous section. About 50% of the water flows relatively quickly to the extraction wells, which it reaches after less than 60 days. Water recharging, for example, in the vicinity of the water divide in the ponds flows in a deep and much longer flow cycle towards the extraction wells and is spread over a longer time span. This contributes to the tailing of the distribution in Fig. 10a. Another issue is the position of the extraction wells. In all, 75% of the wells are located at a distance which is less than 60 m from the ponds—other wells are between 60 and 100 m. Water flowing from the western pond towards wells 1–10 will have in general a larger residence time because of the larger distance. The same accounts for water flowing from the eastern part of the eastern pond towards wells 100–110. Consequently, water recharging in the ponds at the same point in time reaches the wells at different times. However, 50% of the water reaches the extraction wells within 60 days, which refers to water flowing to the extraction wells via shallow flow
Fig. 10 Frequency–residence time plot (a) and cumulative frequency–residence time plot (b) of the residence time of the artificially recharged water in the dune aquifer Hydrogeology Journal (2008) 16: 1669–1681
DOI 10.1007/s10040-008-0326-x
1679
paths or water flowing to the nearest extraction wells. The other 50% of the water flows in deeper (and slower) flow paths to the extraction wells or to more distant extraction wells. In the latter case, the result is an increase of frequency on the general decreasing frequency-residence time plot of Fig. 10. Notice for instance the small increase in frequency around a residence time of 90 days and between 150 and 200 days. Figure 10a can also be considered to be the equivalent of a breakthrough curve of a tracer and can thus be compared with the results of the tracer test, more particularly with Fig. 5e. This shows enlarged fluoride concentrations between 30 and 45 days, thus coinciding with the highest frequencies of the time-frequency diagram of Fig. 10a. Consequently, calculations are in good agreement with the results of the tracer test. Figure 9 can also be seen as a distribution of the residence times of the water which enters the extraction wells at one time. Thus, it is derived that 50% of the extracted water simultaneously has a residence time of 55 days and so on, whereas 50% of the water has a longer residence time of up to 5 years. This means that the extracted water is a mix of water with different residence times from about 30 days up to 5 years. A number of particles will have flow paths ending in the same well. A statistical analysis was made for each individual well. This is visualised by means of box plots (Fig. 11) for each well of well battery 2. Box plots show lower quartile, median and upper quartile values. The whiskers are lines extending from each end of the box to show the extent of the rest of the data. Outliers (+) are data with values beyond the ends of the whiskers. Some interesting trends can be seen from Fig. 11. About 40% of the wells (wells 10–20, 50–75, 90–100) have a lower quartile between 25 and 30 days and a median between 30 and 40 days. This means that, as shown in Fig. 10, about 50% of the recharged water is recovered in less than 60 days. Residence time of water
extracted in the other 60% of the wells is larger and spread over a larger time span, which is, as already explained, due to the combination of longer flow cycles of the recharged water and position of extraction wells. Water extracted from wells 1–10, for instance, has a longer residence time which is partly due to the larger distance to the western pond. Also, the amount of water reaching these wells follows deep and longer flow cycles as already shown in Fig. 9. Water reaches wells 42–49 also after a longer time—note that there is a clear trend from well 42 to well 49, which is mainly due to the position of the wells. Moreover, no recharged water is extracted from well 41, which is located furthest from the pond. A similar observation is made for wells 101–110 which are located southeast of the eastern pond. Wells 107–110 are located too far away from the pond to extract recharged water. For wells 25–40, there is a general trend of residence time increasing from wells 25 to 31 and decreasing from 32 to 40. The same can be said for wells 75–90. The increase of residence time, however, is not as large as for instance wells 1–10, 41–50 or 100–110. The reason for this is the position of these wells, which are located between the two ponds or north or south of the smallest part of the eastern pond. The combination of the facts that less water infiltrates here with the larger distance between pond and extraction wells (mainly to the southern wells) makes the residence times of water entering these wells larger. Of note is also that, in general, the larger the residence time is of water entering a particular well, the larger the difference between lower and upper quartile values is. This difference is smallest for the wells 10–20, 50–75, 90– 100, which have the shortest residence times. The increase of the difference with increasing residence time is very well illustrated by, for instance, wells 100–110, wells 25– 40 or wells 75–90. Longer residence times are found for wells which are positioned relatively far from the infiltration ponds. These wells also receive recharged water taking part in the long and deep flow cycle, which
Fig. 11 Box plots of the residence time of infiltration water reaching the different wells of well battery 2. The whiskers are lines extending from each end of the box to show the extent of the rest of the data. Outliers (+) are data with values beyond the ends of the whiskers Hydrogeology Journal (2008) 16: 1669–1681
DOI 10.1007/s10040-008-0326-x
1680
means that the spectrum of residence times will be much larger than for wells located close to the ponds. Therefore the difference between the lower and upper quartile values will be larger for these former wells.
Volume of infiltration water Figure 12 shows the evolution of the volume of recharged water present in the dune aquifer. This is subdivided into total volume of recharged water and volume of recharged water present under the eastern and western pond. During the first half year of the artificial recharge project (until the end of 2002), the volume of recharged water is increasing relatively quickly. Afterwards, this volume is still increasing but the rate of increase is decreasing. At the end of 2007, the volume of the recharged water continues to increase. This means that, although the recharge and extraction rates are more or less fixed during the last 3 years, the volume of recharged water present in the aquifer was still increasing. At the end of 2006, a volume of 1.1502 106 m3 of water is calculated as being present in the aquifer. Taking into account a porosity of 0.38, this means a volume of 3.0268 106 m3 of dune aquifer. About 70% of the recharged water is present under the western pond. To study how these volumes will evolve in the near future, the groundwater model was extended for 5 years, whereby recharge and extraction rates varied as in 2006. Between 2007 and 2012 there is still an increase in volume of recharged water, although this increase becomes very small. The fact that this volume of recharged water will not be at a dynamical equilibrium, even after 10 years of recharge is not unexpected. This is the same as, for instance, the formation of fresh-water lenses where a dynamical equilibrium is reached after more than 100 years (Vandenbohede and Lebbe 2002). Obviously, the increase in volume of recharged water will become very small within the next 10 years.
Conclusions Since July 2002, tertiary treated wastewater has been artificially recharged in the phreatic dune aquifer of the Belgian western coastal plain. This was done to realise a sustainable water extraction in an area where fresh-water resources are limited. Hydraulic aspects and storage and recovery of the recharged water were studied using borehole measurements, results of a tracer test and a three-dimensional solute groundwater flow model. Capture zones and residence times were studied to identify characteristics of the system. Because of the artificial recharge, a lens of recharged water has formed in the native fresh dune water lens. Dimensions of this lens were derived from borehole measurements and the modelling. Borehole measurements show a clear distinction between recharged and native dune water, the former having a lower electrical conductivity because of the lower TDS. Recharged water is found up to a depth of 20 m below surface at WP6.1, located almost between the two ponds. Modelling shows that the depth of the recharged water lens under the eastern pond is slightly less than 20 m, whereas this is slightly more than 20 m for the western pond. Because of its larger surface, more recharged water is present under the western than under the eastern pond. The volume of recharged water in the aquifer is, 5 years after the start of the recharge project, still increasing, although this increase will become very small in the next 5–10 years. Groundwater flow and flow paths show that the occurrence of recharged water is not restricted to the zone between the ponds and the extraction wells, especially around the western pond. This is due to a shallow semi-permeable layer present below the western pond and illustrates the effects of aquifer heterogeneity. The importance of the semi-pervious layer was also indicated by the tracer test. All recharged water is, however, recovered as is indicated by the capture zones.
Fig. 12 Volume of recharged water present in the dune aquifer in function of time. The total volume of both ponds is given, as is the volumes of water present under the eastern and western ponds Hydrogeology Journal (2008) 16: 1669–1681
DOI 10.1007/s10040-008-0326-x
1681
Analysis of residence times has indicated that 50% of the water recharged at the same point in time reaches extraction wells in less than 60 days. It takes up to almost 5 years for the last water recharged at the same point in time to reach the extraction wells. These calculations are confirmed by the results of the tracer test. The distribution of residence times indicates that the extracted water is a mix of water with residence times ranging from slightly less than 30 days up to almost 5 years. Large residence times are the result of water flowing in a longer and deeper flow cycle towards the extraction wells. Extraction of this recharged water is therefore spread over a longer time span. There are distinct patterns in the distribution of residence times of water reaching each individual well. This distribution is clearly influenced by the position of the well and the distance of this well to the recharge ponds. A larger distance for instance means that the mean residence time is longer. It also means that the spectrum of residence times of the extracted water is larger. The spread of the residence times, together with the fact that also native dune water is extracted (80% recharge and 20% native dune water) makes the water quality of the total extraction water stable throughout the year. This is, for example, confirmed by the constant chloride concentration of the extracted water (Fig. 3). Acknowledgements This research was done as part of a research project funded by Institute for the Promotion of Innovation by Science and Technology in Flanders (IWT-Flanders), grant IWT/ OZM/050342. The first author was also supported by funding from IWT-Flanders. The authors thank two anonymous reviewers and the editors for their constructive comments.
References Asano T (1992) Artificial recharge of groundwater with reclaimed municipal wastewater: current status and proposed criteria. Water Sci Technol 25:87–92
Hydrogeology Journal (2008) 16: 1669–1681
Bouwer H (2002) Artificial recharge of groundwater: hydrogeology and engineering. Hydrogeol J 10:121–142 Greskowiak J, Prommer H, Massmann G, Johnston C, Nützmann G, Pekdeger A (2005) The impact of variable saturated conditions on hydrochemical changes during artificial recharge of groundwater. Appl Geochem 20:1409–1426 Konikow LF, Goode DJ, Hornberger GZ (1996) A three-dimensional method-of-characteristics solute-transport model (MOC3D). US Geol Surv Water Resour Invest Rep 96–4267 Lebbe L (1978) Hydrogeologie van het duingebied ten westen van De Panne [Hydrogeology of the dune area west of De Panne]. PhD Thesis, Geological Institute, Ghent University, The Netherlands Massmann G, Greskowiak J, Dünnbier U, Zuehlke S, Knappe A, Pekdeger A (2006) The impact of variable temperatures on the redox conditions and the behaviour of pharmaceutical residues during artificial recharge. J Hydrol 328:141–156 McNeill JD (1986) Geonics EM39 Borehole conductivity meter: theory of operation. Technical Note TN-20, Geonics Ltd., Mississauga, ON, Canada Oude Essink GP (1998) MOC3D adapted to simulate 3D densitydependent groundwater flow. In: Proc. of MODFLOW’98 Conference, vol 1, 4–8 October 1998, Golden, CO, USA, pp 291–303 Van Breukelen BM, Appelo CAJ, Olsthoorn TN (1998) Hydrogeochemical transport modeling of 24years of Rhine water infiltration in the dunes of the Amsterdam Water Supply. J Hydrol 209:281–296 Vandenbohede A (2007) Visual MOCDENS3D: visualisation and processing software for MOCDENS3D, a 3D density dependent groundwater flow and solute transport model. User manual, Research Unit Groundwater Modelling, Ghent University, The Netherlands Vandenbohede A, Lebbe L (2002) Numerical modelling and hydrochemical characterisation of a fresh water lens in the Belgian coastal plain. Hydrogeol J 10(5):576–586 Vandenbohede A, Lebbe L (2006) Occurrence of salt water above fresh water in dynamic equilibrium in coastal groundwater flow systems. Hydrogeol J 14(4):462–472 Vandenbohede A, Van Houtte E, Lebbe L (2008) Sustainable groundwater extraction in coastal areas: a Belgian example. Environ Geol. doi:10.1007/s00254-008-1351-8 Van Houtte E, Verbauwhede J (2005) Artificial recharge of treated wastewater effluent enables sustainable groundwater management of a dune aquifer in Flanders, Belgium. Proc. Ismar 5, 5th International Symposium on Management of Aquifer Recharge Systems for Protecting and Enhancing Groundwater Resources, 10–16 June 2005, Berlin
DOI 10.1007/s10040-008-0326-x
