Increasing a freshwater lens below a creek ridge using a controlled artificial recharge and drainage system: a case study in the Netherlands
- First Online:
- Cite this article as:
- Pauw, P.S., van Baaren, E.S., Visser, M. et al. Hydrogeol J (2015) 23: 1415. doi:10.1007/s10040-015-1264-z
- 1.1k Downloads
A controlled artificial recharge and drainage (CARD) system was used to increase freshwater lenses below creek ridges to increase freshwater supply. Creek ridges are typical geomorphological features that lie up to 2 m higher than the surroundings in the reclaimed tidal flat landscape of the southwestern Netherlands. The 5–30-m thick freshwater lenses below the creek ridges are a vital source for irrigation, as the groundwater and surface waters are predominantly saline. However, freshwater supply from these lenses is commonly not sufficient to meet the irrigation demand, which leads to crop damage. The CARD system was tested in the field and the development of the freshwater lens was monitored during the period May 2013 to May 2014. Numerical models, which were used to investigate a long-term effect of the CARD system, predicted that below the center of the creek ridge, the 13–15-m thick freshwater lens increased 6–8 m within 10 years. The total volumetric increase of the freshwater lens was about 190,000 m3 after 10 years, which was about 40 % of the total recharge (natural and artificial recharge). From this increased freshwater lens, up to three times more water can be extracted using horizontal wells, compared to the initial size of the freshwater lens. A higher water table in the CARD system leads to a thicker freshwater lens but a lower storage efficiency. A lower water table has the opposite effect.
KeywordsArtificial recharge Coastal aquifers Creek ridge Salt-water/fresh-water relations The Netherlands
Augmentation de la taille d’une lentille d’eau douce sous une butte de crique à l’aide d’un système contrôlé de recharge artificielle et de drainage: un cas d’étude en Hollande
Un système contrôlé de recharge artificielle et de drainage (CARD) a été employé pour augmenter la taille de lentilles d’eau douce sous les buttes de crique, afin d’augmenter l’approvisionnement en eau douce. Les buttes de criques sont des structures géomorphologiques typiques, qui peuvent atteindre 2 m au-dessus du sol environnant, dans les paysages de bas-fonds intertidaux récupérés sur la mer, en Hollande du sud-ouest. Les lentilles d’eau douce, d’épaisseur de 5–30 m sous les buttes constituent une ressource vitale pour l’irrigation, car l’eau souterraine et les eaux de surface sont généralement salées. Néanmoins, l’approvisionnement en eau douce à partir de ces lentilles n'est généralement pas suffisant pour satisfaire la demande d'irrigation, ce qui est dommageable aux récoltes. Le système CARD a été testé sur le terrain et le développement de la lentille d’eau douce a été suivi pendant la période de mai 2013 à mai 2014. Les modèles numériques, qui ont été utilisés pour rechercher les effets à long terme du système CARD prédisent que, sous le centre des buttes, les lentilles d’eau douce de 13–15 m d’épaisseur augmenteraient de 6–8 m en 10 ans. L’augmentation du volume de la lentille d’eau douce serait de 190 000 m3 au bout de 10 ans, soit environ 40 % de la recharge totale (naturelle et artificielle). A partir de cette lentille d’eau douce dont la taille est augmentée, il serait possible d’extraire par des puits horizontaux jusqu’à trois fois plus d’eau par rapport à leur taille initiale. Un niveau de nappe élevé à l’intérieur du système CARD conduit à une lentille d’eau douce plus épaisse mais avec une efficacité de stockage plus faible. Un niveau de nappe plus bas produit l’effet inverse.
Aumento de lentes de agua dulce por debajo cordones utilizando un sistema controlado de recarga artificial y drenaje: un caso de estudio en los Países Bajos
Se usa un sistema controlado de recarga artificial y drenaje (CARD) para incrementar los lentes de agua dulce por debajo de cordones para incrementar el abastecimiento de agua dulce. Las cordones son características geomorfológicas típicas que se encuentran hasta 2 m por encima del entorno en un paisaje de planicie de marea recuperada del suroeste de los Países Bajos. El espesor de 5–30-m de lentes de agua dulce debajo de los cordones es una fuente vital para la irrigación, dado que el agua subterránea y el agua superficial son predominantemente salinas. Comúnmente, sin embargo, el abastecimiento de agua dulce a partir de estas lentes no es suficiente como para cubrir la demanda de la irrigación, lo cual conduce a daños en los cultivos. El sistema CARD fue probado en el campo y se monitoreó el desarrollo de lentes de agua dulce durante el período mayo de 2013 a mayo de 2014. Los modelos numéricos, que fueron usados para investigar el efecto a largo plazo del sistema CARD, predijeron que debajo del centro del cordón, los espesores de 13–15-m de las lentes de agua dulce se incrementaron entre 6–8 m durante 10 años. El aumento volumétrico total de las lentes de agua dulce fue alrededor de 190,000 m3 después de 10 años, lo cual fue cerca al 40 % del total de la recarga (natural y artificial). A partir de este incremento de las lentes de agua dulce, se puede extraer hasta tres veces usando pozos horizontales, comparada con el tamaño inicial de las lentes de agua dulce. Un nivel freático alto en el sistema CARD conduce a un mayor espesor de la lente de agua dulce pero a una menor eficiencia en el almacenamiento. Un nivel freático bajo tiene el efecto opuesto.
Ampliação da lente de água doce subjacente a crista de drenagem utilizando um sistema artificial de recarga e dreno: estudo de caso na Holanda
Para aumentar os recursos de água doce disponíveis, um sistema artificial de recarga e dreno controlado (SARDC) foi utilizado para ampliar as lentes de água doce subjacentes as cristas de drenagem. Cristas de drenagem são feições geomorfológicas que tipicamente se elevam 2 m acima em relação às áreas de entorno das planícies recuperadas do sudoeste holandês. As lentes de água doce de 5–30 m de espessura subjacente aos bancos são fontes essenciais para a irrigação, uma vez que tanto a água subterrânea como as águas superficiais são predominantemente salinas. Entretanto, os recursos hídricos disponíveis nessas lentes de água doce são comumente insuficientes para atender a demanda de irrigação, consequentemente prejudicando as plantações. O SARDC foi testado em campo e o desenvolvimento da lente de água doce foi monitorado durante o período entre Maio de 2013 e Maio de 2014. Modelos numéricos desenvolvidos para investigar os efeitos do SARDC a longo prazo indicaram que a porção central da crista de drenagem (banco), correspondente a lente de 13–15 m de espessura, aumentou em 6–8 m em 10 anos. O aumento volumétrico da lente de água doce foi de 190,000 m3 em 10 anos, correspondente a aproximadamente 40 % da recarga total (recarga natural e artificial). Em consequência à ampliação da lente de água doce, até três vezes mais água pode ser bombeada utilizando poços horizontais em comparação aos recursos disponíveis com o tamanho original da lente de água doce. O aumento do nível piezométrico no SARDC gera uma lente de água doce mais espessa, porém diminui a eficiência de armazenamento. Níveis piezométricos mais baixos apresentam o efeito contrário.
In the southwestern delta of the Netherlands large areas lie at or below mean sea level (MSL). Natural dunes and man-made dikes prevent these areas from flooding, and the surface-water level in the ditches is artificially controlled using weirs, dams, and mechanical pumps. In this landscape, sandy tidal creek deposits lie up to 2 m higher than the surroundings. These geomorphological features are known as ‘creek ridges’. Below the creek ridges, 5–30-m-thick freshwater lenses are present. These lenses are an important source for irrigation, because the surrounding saline groundwater is predominantly saline and because in summer most of the surface waters are brackish to saline and there is no supply of fresh surface water from other areas (De Louw et al. 2011). The amount and rate of groundwater extraction from freshwater lenses is regulated by the local water board to prevent excessive saltwater upconing and drawdown of the water table (Province of Zeeland 2002). Currently, the irrigation demand already exceeds the permissible groundwater extraction, which leads to crop damage. As the anticipated sea-level rise, climate change, and land subsidence will further jeopardize fresh groundwater resources in the future (Oude Essink et al. 2010; Pauw et al. 2012), measures are necessary in this delta to assure a sustainable freshwater supply.
In this paper a new system to increase the thickness of freshwater lenses below creek ridges is presented. This controlled artificial recharge and drainage (CARD) system aims at increasing the freshwater lens by raising the water table during winter, when there is a precipitation excess. In view of the classical Badon Ghijben-Herzberg relationship, a higher water table leads to a thicker freshwater lens. Evidently, from thicker freshwater lenses, more fresh groundwater can be extracted (Dagan and Bear 1968; Volker et al. 1985; White and Falkland 2009). The CARD system differs from other artificial recharge techniques (Bouwer 2002) that are aimed at increasing freshwater supply in coastal areas such as spreading basins (Stuyfzand 1993; Tredoux et al. 2003; Pyne 2005) and aquifer storage and recovery (ASR) using wells (Miotlinski et al. 2013; Zuurbier et al. 2014a, 2014b), in that agricultural drain tiles are used to raise the water table.
Material and methods
Theory and definitions
In this paper, the classification of the salinity of surface water and groundwater into ‘fresh’, ‘brackish’, or ‘saline’ is based on the electrical conductivity of the water referenced to a temperature of 20 °C (ECw20). The water is fresh if the ECw20 ≤ 2.0 mS/cm, the water is brackish if the ECw20 is between 2.0 and 5.0 mS/cm, and the water is saline water if the ECw20 ≥ 5.0 mS/cm. This classification is based on guidelines that freshwater users in the area use for irrigation; freshwater can be used safely for irrigation, brackish water is only used occasionally in times of severe water scarcity, and saline water is unfit for irrigation.
The thickness of the freshwater lens is defined as the depth at which the ECw20 or ECbulk (the electrical conductivity of the water and porous medium) is equal to 50 % of the difference between their maximum and minimum value below the creek ridge. This value is referred to as Dmix (modified from De Louw et al. 2011). Dmix approximately corresponds to the centre of the mixing zone of the freshwater lens. The reason for using Dmix rather than depth of the ‘fresh–brackish’ interface for the thickness of the freshwater lens is that, from the different geophysical measurement techniques that were used, it was difficult to determine the fresh–brackish interface. The main reason for this was that data about the formation factor for the conversion of the ECbulk to ECw20 using Archie’s law (Archie 1942) were not collected.
Study area and design of the CARD system
The study area is located on Walcheren (Fig. 1). In the low-lying areas of Walcheren, the upper hydrogeological unit is a semi-confining layer up to 10 m thick, which consists of Holocene peat, clay, and silt deposits. Below this unit, sandy deposits from the Quarternary and Neogene periods form a 10–80-m-thick semi-confined aquifer. In the semi-confined aquifer, low-permeability layers are present, but they do not form a coherent confining unit. The semi-confined aquifer is bounded below by Oligocene marine clay, which is considered as the hydrogeological base (TNO 1997; Goes et al. 2009; De Louw et al. 2011; Stafleu et al. 2011).
In April 2013, the CARD system was installed on the two fields and a dam in the small ditch between the fields was placed to prevent drainage of the fresh groundwater (Fig. 3). The drain tiles of the CARD system are composed of plastic perforated tubes of 60 mm with an outer envelope of polypropylene (PP450). The depth of the drain tiles varies between 1.1 m (higher parts of the creek ridge) and 0.8 m (lower parts of the creek ridge) below ground level (BGL). At the higher parts of the creek ridge the distance between the drain tiles is 7 m. Artificial recharge using fresh surface water takes place here. The fresh surface water is collected downstream of a weir just north of the CARD system (Fig. 3). The weir controls the discharge of fresh surface water into a brackish/saline ditch. Discharge mainly occurs in late autumn, winter, and early spring, as a result of the precipitation excess, so that, in this period, water is available for artificial recharge. The collected freshwater is passed through a 0.2-mm filter (Meeuwse Handelsonderneming), is transported over the brackish/saline ditch, and is then pumped into the CARD system at an intake point (Fig. 3). When the water table is higher than the highest desired drainage level (about 1 m BGL at the highest parts of the creek ridge), the pump automatically turns off. During the other period of the year, artificial recharge is generally not possible, as the weir prevents the discharge of freshwater into the brackish/saline ditch.
At the lower parts of the creek ridge, artificial recharge does not take place, but the drainage level is kept as high as possible (between 0.6 and 0.8 m BGL) to reduce the groundwater flow from the higher parts of the creek ridge. A water table higher than 0.6 m BGL at the lower parts of the creek ridge is undesired as it leads to crop damage. In Fig. 3 it is shown how the drainage level is controlled. Individual drain tiles at 6-m spacing are oriented perpendicular to the plane of Fig. 3. These drain tiles are connected to a collector drain. The water in this collector drain discharges into a collector well, where the drainage level can be controlled at the desired level. Additional illustrative pictures of the CARD system are given in the electronic supplementary material (ESM) of this paper.
Overview of the different types of field measurements and their measurement location (ML)
1, 2, 4, 6
Sampling using piezometer nest
1, 5, 6
1, 2, 3, 4, 5
Lithology and ECbulk
A, B, C
Continuous vertical electrical sounding (CVES) measurements were carried out at three transects to estimate the distribution of ECbulk in a vertical cross section (Fig. 4b). The ABEM Lund Imaging System, comprising the ABEM SAS 4000 terrameter and the ES10-64 electrode selector, was used for this purpose. The Wenner method was used with an electrode distance of 2 m to measure the ‘apparent’ ECbulk distribution (Reynolds 1997). The RES2DINV inversion program (Loke 2006) was used to obtain a model of ECbulk.
After the installation of the CARD system, the water table, ECbulk and ECw20 were monitored at MLs 1, 2, 4, 5, and 6 (Fig. 4b), for the period May 1, 2013–May 1, 2014. Measuring location 6 served as a reference measurement, as the CARD system was not installed here.
At ML 2, a vertical electrical resistivity monitoring system (subsurface monitoring device; SMD) was installed. The borehole for this system was made using pulse drilling with a diameter of about 15 cm. The SMD was used to measure the vertical distribution of ECbulk at a daily interval, using multiple electrodes located at various depths. The electrode spacing was 0.1 m from −6.5 to −10.3 m NAP. Above and below this section the electrode distance gradually increased to 0.7 m. The Wenner configuration was used to measure ECbulk.
At MLs 1 and 4, six 2.5-cm-diameter piezometers with a 40-cm-long screen were installed at depths where the mixing zone was present before the installation of the CARD system. The borehole was made using rotary flush drilling, with a diameter of about 20 cm. Bentonite plugs in between the screen were used to prevent short-circuiting of the groundwater (Stuyfzand 1993). About two times the volume of the piezometer was pumped to determine the ECw20, which was sufficient to obtain a stable value. During the period September 2013–May 2014, the ECw20 was sampled approximately bi-monthly.
In December 2013 and February 2014, electromagnetic (EM) borehole logging using a thin probe (SLIMFLEX) was conducted at MLs 1, 5, and 6. In the probe, an alternating electrical current is generated in a ‘transmitting’ coil to produce a primary EM field. This primary EM field produces electrical currents in the subsurface which lead to a secondary EM field. Both fields are measured in another coil at 50-cm distance from the transmitting coil. From the difference in amplitude and phase lag of the EM fields, ECbulk can be determined (McNeill 1980). The coils are situated such that the influence of the water in the borehole is negligible.
At MLs 1, 2, 4, and 6, groundwater levels were recorded at a 10-min interval using pressure transducers installed in piezometers. The piezometers had a 40-cm screen at the bottom, and the screens were located at 4 m from the surface. The borehole for the piezometer was made using hand pulse drilling equipment. The pressure transducers were read out, compared with manual measurements and re-installed approximately bi-monthly. The maximum difference between the manual measurement and the recorded pressure was 4 cm.
Set-up of the initialization model
SEAWAT version 4 (Langevin et al. 2007) was used to simulate variable density groundwater flow and solute (salt) transport in the study area (Fig. 4a). The horizontal extent of the model is 1,600 × 1,600 m and discretized using cells of 10 × 10 m. The vertical extent of the model is from +1.5 m NAP to −87.0 m NAP. Up to a depth of −25 m NAP, the model layers are 0.5 m thick. The hydraulic conductivity distribution in this section was based on the hydrogeological model GEOTOP (Stafleu et al. 2011). Below −25 m NAP, the thickness of the model layers increases to up to 5 m. The hydraulic conductivity distribution in this section was based on the regional hydrogeological model REGIS II (REGIS 2005). The vertical anisotropy ratio—the horizontal hydraulic conductivity (Kh) divided by the vertical hydraulic conductivity (Kv)— in the numerical model is 1.4. In Fig. 4c, a cross section of Kh in the numerical groundwater model is shown, which indicates the incision of the semi-confining layer by the more permeable deposits below the creek ridge. This is in line with the concepts shown in Figs. 2 and 3. The effective porosity (0.3), confined storage coefficient (1E-5 1/m), specific yield (0.15), longitudinal dispersivity (0.1 m), transverse dispersivity (0.01 m), and molecular diffusion coefficient (8.64E-5 m2/d) were assumed to be constant in the model. These parameters were estimated from general literature (Domenico and Schwarz 1997) and from groundwater flow studies in the coastal plain of the Netherlands (Oude Essink et al. 2010; De Louw et al. 2011; Zuurbier et al. 2014b) and Belgium (Lebbe 1999; Vandenbohede and Lebbe 2006).
The PCG package was used for solving the variable density groundwater flow equation. The hydraulic head closure criterion was 10−5 m and the flux closure criterion was 0.1 m3. The advection term of the solute transport equation was solved using the MOC solver, with a minimum number of nine particles per cell. A Courant number of 1.0 was used. The sink and source and the hydrodynamic dispersion terms of the solute transport equation were solved using the GCG solver, with a relative concentration closure criterion of 10−6.
Stress periods of 6 months were used, representing winter and summer conditions. The recharge rate in winter (1.5 mm/d) and the evapotranspiration rate in summer (−0.02 mm/d) were based on average data (period 1971–2000) from a nearby weather station—station Vlissingen, ~11 km from the study site (KNMI 2015). For the evapotranspiration, crop factors based on the crops on the fields of the farmers were taken from Hooghart and Lablans (1988), to correct for the reference Makkink evapotranspiration (Makkink 1957). At the vertical sides of the model, a head-dependent flux boundary condition was applied with the GHB package, using hydraulic heads and concentrations taken from a larger-scale model (Van Baaren et al. 2011). The conductance value was set to 1 m2/d. Testing showed that an order of magnitude higher or lower value for the GHB conductance did not significantly change the model results at the study site. This was in line with the expectations, as the boundaries were located relatively far from the CARD system. The bottom of the model is a no-flow boundary condition as it corresponds with the hydrogeological base. Ditches were simulated in the model using the RIVER package. Tile drainage was modelled using the DRAIN package. The conductance for the cells where the RIVER or DRAIN packages were used was based on the methods of De Lange (1999) and Pauw et al. (2014). More information on the input parameters of the DRAIN and RIVER conductance is given in the ESM.
Calibration of the initialization model and sensitivity analysis
The initialization model was ended at the moment the thickness of the freshwater lens below the creek ridge (Dmix) reached a virtually stable value. The simulated end-thickness of the freshwater lens was compared with the thickness of the freshwater lens measured by the CPTs. Using the parameters that were described in the previous section, the thickness of the freshwater lens was overestimated in the model. Therefore, the model was calibrated to obtain a better match with the field observations. For the calibration, the groundwater recharge rate in winter was reduced, as this was considered the most uncertain parameter. Overestimation of the groundwater recharge is likely, as ponding and subsequent surface runoff frequently occurs at the field site due to the low permeability of the deposits close to the surface of the creek ridge, due to the fining-up sequence. As a result of ponding and surface runoff, the groundwater recharge is lower than the calculated value as described in the previous section (Healy 2010; Voortman 2010).
In addition to the model calibration, the sensitivity of other model parameters was investigated. However, due to the long runtime of the initialization model (approximately 1 week on a modern desktop PC), an extensive sensitivity analysis was not conducted. The effect of a higher evapotranspiration rate in summer was investigated by increasing it from 0.2 to 0.4 mm/d. This value is comparable with the anticipated increase of the evapotranspiration rate in summer to climate change (De Louw 2013; KNMI 2015). In addition, the hydraulic conductivity in the upper 15 m below the higher parts of the creek was multiplied by a factor of 2, for investigating the effect of the hydraulic conductivity in decreasing the thickness of the freshwater lens. Furthermore, a two times higher and two times lower value of the conductance of the RIVER and DRAIN cells was investigated, as these values were based on various assumptions (see ESM).
Scenario simulations to investigate the effect of the CARD system
The simulated hydraulic heads and concentrations at the end of the initialization simulation with the calibrated groundwater recharge in winter were used as initial condition for three scenario simulations where the effect of the CARD system on the thickness of the freshwater lens was simulated. In addition, a ‘reference’ simulation where the CARD system was not simulated was used for comparison with the three scenario simulations. In all four simulations, the yearly period May 1, 2013–May 1, 2014 was simulated using weekly stress periods. The recharge and evapotranspiration rates were based on data from a nearby weather station—station Vlissingen, ~ 11 km from the study site (KNMI 2015). The yearly period was repeated to cover a period of 10 years to study a long-term effect of the CARD system.
In scenario A the influence of the CARD system on the thickness of the freshwater lens was simulated. The RIVER package was used to represent infiltration and drainage by the drain tiles of the CARD system. The hydraulic head of each RIVER cell was based on the weekly average hydraulic heads that were measured at MLs 1, 2, and 4. For verification of this approach, the simulated artificial recharge was compared with the measured artificial recharge at the water collection point of the CARD system. Further information on the simulation of the CARD system is available as ESM.
The effect of a lower availability of fresh surface water for artificial recharge was investigated in scenario B by simulating a 0.1-m-lower water table during the winter period (i.e., November 1, 2013–May 1, 2014) compared to scenario A. In scenario C a 0.1-m-higher water table compared to scenario A was simulated to investigate the effect of a higher water table.
Estimated Dmix from the CPTs and simulated Dmix in the numerical model at MLs 1–5
Surface elevation (m NAP)
Dmix (m NAP)
Simulated Dmix (m NAP)
The CPT at ML 1 indicates low-permeability deposits between −12 and −12.5 m NAP. This is also observed in the CPTs at MLs 2, 3, and 4, but at different depths between −8.0 and −15.0 m NAP. In the CPT at ML 5, low-permeability deposits at depth seem to be absent. These sections of low-permeability deposits are interpreted as bank deposits of the former creek system. This type of heterogeneity is not present in the GEOTOP model. In the ESM the Rf is shown in combination with the vertical hydraulic conductivity at MLs 1–5.
ECbulk prior to the installation of the CARD system
Monitoring of ECbulk and ECw20
The SLIMFLEX measurement at ML 1 showed that Dmix lowered by about 0.8 m in the period between December 10, 2013 and February 7, 2014. This agrees well with the sampling results for this ML (Fig. 8). The SLIMFLEX measurements at ML 5 show a lowering of Dmix of about 0.5 m. At ML 6 (the reference ML) the SLIMFLEX measurements show no significant changes in ECbulk. This was expected, as the CARD system was not installed here.
Calibrated initialization model
The recharge rate in winter was reduced (calibrated) from 1.5 to 1.2 mm/d to obtain a reasonable match between the simulated Dmix and the measured Dmix by the CPTs, at MLs 1–5 (Table 2). At ML 1, the simulated Dmix is about 1 m deeper than the measured Dmix. This is attributed to the presence of low-permeability deposits at ML 1, indicated by the increase of the friction ratio Rf in the CPT, and the absence of these low-permeability deposits in the numerical model. At MLs 2 and 3, the CPTs indicate more sections of low-permeability deposits than at ML 1. These sections are also not present in the numerical model; therefore, at these MLs, the simulated Dmix is even deeper than the measured Dmix, compared to ML 1. At ML 4, the difference between the simulated and measured Dmix is comparable to the difference at ML 1. At this ML, the CPT indicated low-permeability deposits up to about −8.0 m NAP, i.e., the semi-confining layer. In the numerical model the semi-confining layer is also present. At ML 5 the simulated Dmix is within the estimated uncertainty (0.5 m) of the measured Dmix. At this ML, the CPT indicated no low-permeability deposits (Fig. 5). In the ESM, further information on the comparison between the CPTs and the numerical simulations is given.
Effect of the CARD system
At ML 5, the measured 0.5 m lowering of Dmix between December 10, 2013 and February 7, 2014 using SLIMFLEX (Fig. 10) agrees well with the simulated lowering of Dmix (0.3 m) in the equivalent period during the first year (0.3 m). At ML 1, the difference between the measured and simulated lowering of Dmix is larger. The sampling results (Fig. 8) indicate that between September 20, 2013 and March 28, 2014, Dmix lowered about 2.0 m, whereas in scenario A, Dmix lowered about 1.4 m. This difference is attributed to the initialization model, where the simulated Dmix at ML 1 is about 1 m lower than the measured Dmix (Table 2). The thicker the freshwater lens, the less the freshwater lens will increase in thickness upon a given increase of the water table, which can be deduced from the Badon Ghijben-Herzberg principle.
In Fig. 13b it is shown that at ML 1 the decrease of the thickness of the freshwater lens due to buoyancy during the summer months is much less than the increase of the thickness of the freshwater lens during the artificial recharge period. This effect also holds for the whole freshwater lens below the CARD system, which indicates a net increase of the thickness of the freshwater lens over a year. This freshening continues for at least 10 years (Fig. 13b). After 10 years, the freshwater lens has not reached its end-thickness, but the freshening rate has considerably decreased. In Fig. 13c it is shown that Dmix has increased about 6 m in depth at ML 1 and about 8 m at ML 5 in scenario A after 10 years of simulation.
For brevity, the results of scenarios B and C (analogous to Fig. 13) are given in the ESM, and only the most important results of these scenarios are described here. As expected, in scenario B, in which the water table in the CARD system is 0.1 m lower than in scenario A, the increase of Dmix in depth is less than in scenario A. After 10 years, Dmix at MLs 1 and 5 has increased about 5 and 6 m compared to the reference simulation, respectively, and the freshwater lens has almost reached a new dynamic equilibrium. As expected, in scenario C, in which the water table in the CARD system is 0.1 m higher than in scenario A, the increase of Dmix in depth compared to the reference simulation (8 m at ML 1 and 10 m at ML 5) after 10 years is larger than in scenario A. The differences in increase of the freshwater lens between the scenarios A, B, and C indicate the significant effect of raising the water table below the creek ridge in winter on the thickness of the freshwater lens.
As expected, the yearly VAR and dVlens are smaller in scenario B compared to scenario A. However, the yearly Seff is higher; in the first 7 years the yearly dVlens is larger than the yearly VAR and in the last 3 years the yearly Seff has decreased from 94 to 80 %. In scenario C the yearly VAR and dVlens is larger than in scenario A, but the yearly Seff is smaller. Only in the first 2 years is the dVlens larger than VAR. In the last 8 years the yearly Seff has decreased from 94 to 48 %.
In Fig. 14d–f, the cumulative VAR and dVlens are shown for the three scenarios. The total volume of the freshwater lens below the CARD system at the end of the initialization model is 373,088 m3. In scenario A the cumulative VAR (191,186 m3) and dVlens (186,945 m3) after 10 years are about equal. The total simulated natural recharge volume is 274,440 m3; hence, after 10 years, the cumulative dVlens is about 40 % of the sum of the total natural recharge and the cumulative VAR. As expected, after 10 years, the largest cumulative VAR and dVlens are present in scenario C, whereas the smallest VAR and dVlens are present in scenario B. The cumulative Seff in the three scenarios is higher than the yearly Seff. This is indicated by the difference in intersection of the lines of VAR and dVlens between Fig. 14a–c and Fig. 14d–f. In Fig. 14e, the lines of VAR and dVlens do not intersect. Hence, cumulatively, the total volume of the increase of the freshwater lens below the CARD system in scenario B is larger than the total volume of artificial recharge over a period of 10 years. Summarizing, the results of Figs. 13 and 14 indicate that the higher the water table is maintained using the CARD system, the larger the increase of the freshwater lens, the more water is needed for artificial recharge, and the lower the storage efficiency.
Increase of extraction due to increase of the freshwater lens
As an approximation, Dmix is taken here as the depth of the sharp fresh–saline interface. At ML 1, Dmix was about −13 m NAP at the end of the initialization model (Fig. 11). Furthermore, the following parameters are chosen based on the initialization model: a = 13 m, b = 37 m, δ = 0.0223, d = 7 m (so the horizontal well is located at a depth of −6 m NAP), Kh = 7.0 m/d, Kv = 5.0 m/d, and n = 0.3. Below the well, at x = 0, the maximum pumping rate Qmax can be calculated over a period (t) of 180 d in case ζ = d/4. Using these values, application of Eq. (5) yields a Qmax of 0.33 m3/d, per meter length of the horizontal well.
Limitations of this study and recommendations for further research
Although the results of this study have indicated the potential of the CARD system to increase a freshwater lens below a creek ridge and to increase freshwater supply, the study has limitations. First of all, the effects of chemical, biological, and mechanical clogging of the CARD system were not investigated. Belcher and D’Itri (1995) described the ambiguous experiences regarding the influence of clogging on controlled drainage in their summary of a collection of field experiments on controlled drainage and sub-irrigation systems. Clogging will depend on the local conditions such as soil type, artificial recharge regime and the amount of suspended solids in the water. Further research is needed to determine the effect of clogging of the CARD system. Furthermore, although the CARD system is intended to be a relatively cheap method for artificial recharge, the economic aspects of the CARD system need to be investigated in detail.
The hydrological analysis that was presented also has limitations. An important reason for this is the long simulation time of the numerical models that were used. For this reason, the amount of simulations for the calibration, the sensitivity analyses and the scenarios were limited. Future work should focus on constructing a computationally efficient model which can be used to better estimate the uncertainty and sensitivity of the model parameters. Such a model can also be used to more extensively investigate the influence of the availability of surface water for artificial recharge on the thickness of the freshwater lens, and for a more thorough analysis of the increase of the maximum (sustainable) freshwater extraction from freshwater lenses. Although the aspects that control the thickness of the freshwater lens below a creek ridge were not investigated quantitatively to a large extent, some hydrogeological aspects are qualitatively discussed here.
It is expected that low-permeability layers, in particular their spatial coherence and permeability, play an important role in the thickness of the freshwater lens and, hence, the effect of the CARD system on the increase of the freshwater lens. Low-permeability layers can induce significant vertical hydraulic head gradients, such that the vertical pressure distribution in the freshwater lens is not hydrostatic. This has a negative effect on the thickness of the freshwater lens. Furthermore, low-permeability layers can retard or even prevent the increase of the freshwater lens. The field observations at ML 2 suggest an influence of thin low-permeability layers on the freshening process. For applying the CARD system elsewhere, it is important to realize that in hydrogeological models these thin layers may not always be present, as was the case in this study. Cone penetration tests offer a relatively cheap and effective technique with very high vertical resolution, for detecting (thin) low-permeability layers.
The increase of the thickness of the freshwater lens is furthermore influenced by the permeability of the sandy creek deposits and the permeability of the semi-confining layer. The lower the permeability of the sandy creek deposits and the confining layer, the less water (and energy) is needed to maintain an elevated water table to increase the thickness of the freshwater lens. A low permeability of the sandy creek deposits has a positive effect on the amount of water that is needed to maintain an elevated water table, but has a negative effect on the rate of the freshening. The rate and extent of the freshening furthermore depend on how much the water table can be raised during the year, as was indicated by Fig. 14. It is expected that an increase of the period where artificial recharge is applied increases the thickness of the freshwater lens, as this influences the time-averaged water table below the creek ridge, but this was not further investigated. Furthermore, for applying the CARD system elsewhere it is important that the unsaturated zone is thick enough to accommodate the rise of the water table.
The results of this case study indicate the potential of a controlled artificial recharge and drainage (CARD) system to increase a freshwater lens below a creek ridge. On the field site where the CARD system was tested, in the southwestern part of the Netherlands, the 13–15-m-thick freshwater lens is expected to increase 6–8 m within 10 years, as a result of the 0.5 m rise of the water table in winter in the CARD system, based on the results of numerical model simulations. A lower water table leads to a lower increase of the volume of the freshwater lens, but a higher storage efficiency (the volume of the increase of the freshwater lens relative to the volume of artificial recharge). A higher water table has the opposite effect. Based on an analytical solution, the increase of the freshwater lens after 10 years allows for a 2–3 times higher pumping rate using a horizontal well, which indicates the potential for the CARD system to increase freshwater supply.
It is expected that the most important factors that influence the effect of the CARD system on the increase of the freshwater lens are low-permeability layers, the availability of surface water for artificial recharge, the permeability of the sandy creek sediments, the permeability of the semi-confining layer, and the thickness of the unsaturated zone to accommodate the rise of the water table.
The subsurface monitoring device (SMD) was designed, financed and operated by Imageau. The CARD system was installed by H. Rutten. Fugro conducted and financed the CPTs. Meeuwse Handelsonderneming designed the pump and transportation system. The farmers of De Waterhouderij, in particular Werner Louwerse and Johan Sanderse, are warmly thanked for their support. Mike van der Werff, Rein Lantman, Pieter Doornenbal, Thomas Boerman, Maitri Fischer, and Marjan Sommeijer are thanked for their assistance with the field measurements. Numerical model input and output was processed using the Python project ‘Flopy’; code.google.com/p/flopy/. This work was carried out within the Dutch ‘Knowledge for Climate’ program.
Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made.