Tidal influence on self-potential measurements
Abstract
Long-term surface and borehole self-potential (SP) monitoring was conducted in the UK Chalk aquifer at two sites. The coastal site is ~1.7 km from the coast, and the inland site is ~80 km from the coast. At both sites, power spectral density analysis revealed that SP data contain the main ocean tidal periodic components. However, the principal lunar component (M2), the dominant ocean tidal component, was most significant at the coastal site. The M2 signal in surface-referenced SP data at the inland site was partly due to telluric currents caused by the geomagnetic ocean dynamo. Earth and/or atmospheric tides also contributed, as the SP power spectrum was not typical of a telluric electric field. The M2 component in borehole-referenced data at the inland site was below the significance level of the analysis method and was 2 orders of magnitude smaller than the M2 signal in borehole-referenced SP data at the coastal site. The tidal response of the SP data in the coastal borehole is, therefore, primarily driven by ocean tides. These cause changes in fluid pressure and chemical concentration gradients within the coastal aquifer, leading to time varying electrokinetic and exclusion-diffusion potentials. Borehole-referenced SP measurements could be used to characterize and monitor tidal processes in coastal aquifers such as the intrusion of seawater.
Key Points
- SP from coastal and inland aquifers in the UK display tidal power spectra, suggesting distinct SP contributions at both sites
- The tidal SP is 2 orders of magnitude larger at the coast and is dominated by electrokinetic and/or exclusion-diffusion potentials
- Borehole-referenced SP data in the coastal aquifer are dominated by ocean tidal processes and could be used to monitor seawater intrusion
1 Introduction
Coastal aquifers are an important water resource but are threatened by the intrusion of seawater. Reduced recharge and/or groundwater abstraction decreases freshwater hydraulic head and can cause the movement of a saline front toward abstraction boreholes [van Weert et al., 2009]. The recovery of the aquifer in the area adjacent to a contaminated borehole can take considerable time [Zhou et al., 2005]. Geophysical monitoring may assist in avoiding contamination of coastal water supplies.
The most common method for monitoring seawater intrusion is borehole measurements of fluid electrical conductivity (FEC) [Food and Agricultural Organisation, 1997; Werner et al., 2009]. However, FEC monitoring does not facilitate proactive management of abstraction. Active geophysical methods, including electrical resistivity tomography (ERT) and time domain electromagnetic methods (TDEM), can offer a means of remotely detecting the saline front. However, ERT commonly relies on surface monitoring and it can be logistically difficult to lay out long-term surface electrode arrays in densely populated areas. Furthermore, and despite recent advances in such techniques [de Franco et al., 2009; Falgàs et al., 2009; Rosas-Carbajal et al., 2013], ERT and TDEM provide information only about the saturated rock resistivity rather than directly about fluid dynamics within the aquifer.
An alternative, passive geophysical method is self-potential (SP) monitoring. SP arises in order to maintain electrical neutrality when a separation of charge occurs due to gradients in pressure (electrokinetic potential) and/or concentration (exclusion-diffusion potential) [Revil, 1999; Maineult et al., 2004]. In coastal aquifers, tides alter the pressure and concentration gradients [Kuan et al., 2012] and are expected to give rise to SP signals with a tidal signature. An electrokinetic response to pumping has previously been detected in the UK Chalk aquifer [Jackson et al., 2012a]. Furthermore, modeling suggests that SP may indicate the movement of a saline front some distance from a monitoring borehole [Gulamali et al., 2011; Bolève et al., 2011; Jackson et al., 2012b; Jougnot et al., 2015]. Thus, SP could be used to monitor, remotely, seawater intrusion in real time using a permanent array of borehole electrodes with a reference electrode placed at the surface or within the monitoring borehole.
However, a significant challenge for SP monitoring in coastal aquifers is that the signals of interest, i.e., changes in concentration and pressure gradients induced by ocean tides, may coincide with diurnal and semidiurnal SP fluctuations caused by other sources such as telluric currents, Earth and atmospheric tides, and changes in temperature (Table 1). A number of studies have examined SP responses to induced periodic processes, including harmonic pumping tests [Maineult et al., 2008; Revil et al., 2008] and drainage-imbibition cycles [Allègre et al., 2014], but few have examined the SP response to natural periodic processes. Diurnal or semidiurnal SP signals have been observed with source mechanisms other than ocean tides, including diurnal spring discharge [Perrier et al., 1999] and flow under glaciers driven by Earth and atmospheric tides [Kulessa, 2003]. Perrier et al. [1997] and Bumpus and Kruse [2014] attributed diurnal SP signals to evapotranspiration causing vertical flow within the unsaturated zone. Other studies attribute the presence of diurnal and semidiurnal SP signals to telluric currents caused by the geomagnetic field [Perrier et al., 1998; Trique et al., 2002].
| Period (h) | Cycles per Day (cpd) | Notation | Tidal Mechanism | Amplitude Relative to M2 | Possible Nonoceanic Sources |
|---|---|---|---|---|---|
| Diurnal | |||||
| 23.934 | 1.003 | K1 | Principal lunar-solar | 0.5842 | Earth tides |
| 24.066 | 0.997 | S1 | Principal solar | 0.1933 | Earth tides/telluric/temperature/atmospheric tides |
| 25.819 | 0.929 | O1 | Principal lunar | 0.4148 | Earth tides/atmospheric tides |
| 26.868 | 0.893 | Q1 | Elliptic lunar | 0.0795 | Earth tides |
| Semidiurnal | |||||
| 11.967 | 2.005 | K2 | Lunar-solar | 0.1267 | Earth tides |
| 12.000 | 2.000 | S2 | Principal solar | 0.4656 | Earth tides/telluric/temperature/atmospheric tides |
| 12.421 | 1.932 | M2 | Principal lunar | 1.0000 | Earth tides/atmospheric tides |
| 12.658 | 1.896 | N2 | Lunar elliptic | 0.1915 | Earth tides |
| Short Period | |||||
| 6.301 | 3.809 | MN4 | Shallow water components found in the English Channel | 0.0297 | NA |
| 6.209 | 3.865 | M4 | 0.1000 | NA | |
| 6.101 | 3.934 | MS4 | 0.0553 | NA | |
| Long Period | |||||
| 13.661 days | 0.070 | Mf | Fortnightly | 0.1722 | Earth tides |
| Nonoceanic Components | |||||
| 8.000 | 3.000 | M3 | Principal solar | NA | Telluric/temperature |
| 6.000 | 4.000 | M4 | |||
The few studies that report SP measurements at coastal sites have obtained data using only surface electrode arrays. Gokhberg et al. [2009] observed a tidal power spectrum in SP data obtained using 2 km long cables in coastal Japan. They were unable to explain the magnitude of the observed SP solely by invoking Earth tides and suggested that ocean tides contributed. Kang et al. [2014] monitored SP at a coastal sea dyke. They observed a semidiurnal SP signal, which they attributed to changes in pressure gradients across the tide, but they did not use frequency analysis to assess the SP data. Furthermore, they used a 6 h sampling rate so aliasing will have occurred and installed steel electrodes, which are unstable and of inadequate quality for SP monitoring [Perrier et al., 1997]. If changes in salinity occurred at their electrodes as seawater moved through the dyke, then redox potentials will have arisen [Castermant et al., 2008]. Both Gokhberg et al. [2009] and Kang et al. [2014] neglected the impact of the geomagnetic ocean dynamo, and Kang et al. [2014] neglected Earth tides, yet these processes may have led to the SP responses they observed.
The aim of this study was to establish whether SP responds to changes in pressure and concentration gradients in a coastal aquifer driven by ocean tidal processes. The objectives were to (i) measure SP in the UK Chalk aquifer, (ii) assess the contribution of various tidal processes by comparing inland and coastal, and surface- and borehole-referenced SP data, and (iii) determine whether ocean tides induce a measureable tidal SP. Table 1 shows the dominant tidal components in the English Channel along with hypothesized SP sources.
2 Method
Two sites in the Upper Chalk aquifer [Bristow et al., 1997] of the UK were used to investigate the source of tidal SP. Figure 1 shows the layout of the surface and borehole arrays at the inland site near Trumpletts Farm in Berkshire and the coastal site at Saltdean in East Sussex.
2.1 Site Descriptions
The inland site has been extensively investigated as part of a research program on lowland catchments [Williams et al., 2006; Wheater et al., 2007; Mathias et al., 2007; Butler et al., 2009]. The site consists of six 100 m boreholes close to an abstraction borehole which is infrequently used. The boreholes penetrate New Pit Chalk, ~3 m of Chalk Rock and then ~20 m of Lewes Nodular Chalk [Schürch and Buckley, 2002; Woods, 2006]. The Seaford Chalk formation then extends to the surface. Borehole PL10B was instrumented (Figures 1b and 1c).
The borehole at the coastal site is used to monitor FEC (as a proxy for salinity) seaward of the Balsdean abstraction borehole (Figures 1d–1f). FEC in the borehole can reach 15,000 μS/cm as regular breakthrough of saline water occurs [Jones and Robins, 1999]. The borehole penetrates the fine-grained Upper Cretaceous Seaford Chalk and the coarser grained Lewes Nodular Chalk in the lower 12 m of the borehole. Ocean tides are recorded at Newhaven (Figure 1d) [British Oceanographic Data Centre, 2015].
2.2 Data Acquisition
At both sites at least one surface reference electrode was available. The shallowest borehole electrode was used as the borehole reference. The borehole arrays comprised nonpolarizing Silvion Ag/AgCl WE300 potable water electrodes [Silvion, 2015]. Surface electrodes were Silvion Ag/AgCl WE200 soil electrodes [Silvion, 2013]. In both electrode types, the Ag/AgCl rod is housed inside a plastic casing containing KCl gel. The casing is separated from its surroundings by a low permeability ceramic disk. The WE300 electrodes have a 0.05 M internal KCl concentration, and the WE200 electrodes have a 0.5 M internal KCl concentration. Surface temperature was measured using Campbell Scientific 107-L thermistors. AquaTroll 200 FEC/T/P probes were used to measure FEC, temperature (T), and pressure (P) in the borehole and were placed next to selected electrodes (Figures 1c and 1f). At both sites, data were logged at a 5 min time interval with a Campbell Scientific CR3000 data logger. This device has a 20 GΩ internal impedance, a resolution of 0.67 μV, and an accuracy of 0.04%. The electrodes were connected to the logger using 5 mm diameter coaxial cable. Data at the coastal site were collected from May 2013 to January 2015 and at the inland site from May 2015 to August 2015, although not all data types were recorded throughout.
At the inland site, four surface monitoring electrodes were installed around the monitoring borehole (Figure 1b) and were referenced against a surface electrode 90 m north of the monitoring borehole. The electrodes were installed at a depth of ~0.5 m, in a bucket filled with saturated bentonite, ensuring a good electrical connection with the ground [Corwin, 1990; Perrier et al., 1997; Trique et al., 2002]. Temperature was measured at Ref-IL, Surf2-S-IL, and Surf3-E-IL. Four electrodes spaced with respect to the shallowest electrode, at ~24 m, ~12 m, ~8 m, and ~4 m, were installed in the monitoring borehole (Figure 1c). The maximum spacing between electrodes matched the maximum electrode spacing at the coastal site. An FEC/T/P probe was installed at the top of the borehole.
At the coastal site, only a reference electrode was installed at the surface and 14 electrodes spaced at 2 m intervals were installed in the borehole (Figure 1f). Three FEC/T/P probes were installed at the top, middle, and bottom of the water column. The spacing between the shallowest and deepest electrodes was ~24 m. In order to minimize exposure to diurnal and seasonal temperature variations, the surface reference electrode and a thermistor were buried at ~1.5 m in a plastic bucket filled with saturated bentonite (Figure 1e). It was not possible to install a surface array at the coastal site.
2.3 Spectral Analysis Method
A number of different spectral analysis techniques have been used in previous studies to analyze periodic SP signals [Allègre et al., 2014; Kulessa et al., 2003; Maineult et al., 2008]. In this study, the Lomb-Scargle power spectral density (PSD) periodogram [Lomb, 1976; Scargle, 1982] was used to analyze SP. The Lomb-Scargle PSD, a least squares spectral analysis technique, was chosen because it can be used with time series data that have gaps or that are not evenly spaced in time [Lomb, 1976; Trauth, 2010]. The Lomb-Scargle PSD offers an alternative approach to dealing with data gaps to that proposed by Mariethoz et al. [2015]. It also allowed determination of the probability of a certain frequency peak being related to a real periodic signal and not the result of random fluctuations; i.e., the probability of detection [Trauth, 2010]. The significance limits were set at 50, 90, 99 and 99.99% and are shown in all PSD plots using dashed lines. No data smoothing or filtering was conducted prior to spectral analysis.
3 Results
3.1 Inland Site
Surface SP data from three north-south dipoles at the inland site are shown in Figures 2e–2g. All dipoles displayed the diurnal principal solar component (S1). The north-south dipoles also displayed the semidiurnal principal lunar component (M2). The magnitude of the M2 component decreased with the length of the dipole. Temperature measured at the electrodes, and temperature differences across the dipoles, also displayed S1 but did not show any other spectral components (Figures 2b–2d; the raw data are shown in Figure 2a).
In the monitoring borehole at the inland site, the temperature was stable and showed no specific frequency response (Figure 3a) but the head contained the diurnal (S1) and semidiurnal (S2) principal solar components, although both were below 50% significance (Figures 3b–3d).
Figure 4 shows the PSD of the SP recorded at the deepest borehole electrode, referenced sequentially against different electrodes in the surface array. All dipoles displayed S1 and M2 and, as the distance of the reference electrode from the borehole increased, the M2 component also increased.
In contrast to the surface-referenced borehole data, there were no significant periodic components recorded in the borehole electrodes when referenced against the shallowest borehole electrode (BH4-IL) (Figure 5). The origin of the long-term trend observed in Figure 5a is unclear. Each of the other borehole electrodes (not shown) displayed a slightly different onset and nature of trend, suggesting an electrode artifact. However, further analysis of the trend is beyond the scope of this paper (see Jackson et al. [2012a] for further discussion of SP monitoring at the inland site).
3.2 Coastal Site
In order to compare the PSD analyses from the inland and coastal sites it was necessary to use a similar time series length of ~3 months. Comparing 1.5 years of data from the coastal site (Appendix A) with 3 months of data from the inland site would have produced a bias in favor of the coastal site [Karl, 1989; Trauth, 2010]. It was not possible to gather a longer time series from the inland site. Data from the coastal site collected from May to August 2013 were analyzed, which corresponds to the same time of year as the inland data acquisition providing confidence in the comparison of the data sets. It was not possible to collect simultaneous data sets.
Figure 6 shows the results of PSD analysis for the tidal gauge data at Newhaven. Data from the shorter time period are shown; the full data are presented in Appendix A. The main frequency components identified in the PSD of the data from the shorter time period are the same as those identified over the long time period for all data types (tidal gauge, borehole head, temperature, conductivity, and SP). The tidal gauge data were dominated by the diurnal principal lunar component (M2). There was also evidence of S2 and N2 and an indication of K1, MN4, M4, and MS4 (Figures 6b–6d and Appendix A).
Head fluctuations measured in the borehole at the coastal site contained the main tidal components M2, K1, S2, and N2, similar to the tidal gauge data (Figures 7a and 7b). The borehole FEC and temperature data were dominated by the M2 component (Figures 7c–7f). FEC increased slightly in mid-August 2013 due to decreased head. There were no significant temperature fluctuations at the surface electrode (Figures 7g–7h).
Surface-referenced borehole SP data were dominated by the M2 component, but K1, S2, and N2 components were also present (Figure 8), similar to the tidal gauge and borehole head data. It is notable that the M2 component in surface-referenced borehole SP data at the coastal site was an order of magnitude larger than in comparable data at the inland site (Figure 4f). There was also an indication of S3, MN4, M4, MS4, and S4 (Appendix A).
Borehole-referenced SP was dominated by the M2 component, although a small K1 component was also present (Figure 9). The magnitude of the M2 and K1 responses increased as the separation of the electrodes increased. The smallest separation at which M2 appeared was ~8 m (Figure 9h). It is again notable that these data show a strong M2 component, unlike the comparable data at the inland site (Figure 5).
4 Discussion
Table 2 shows a summary of the periodic components in the SP data at both sites and the hypothesized source mechanisms. The components in borehole-referenced SP at the coastal site almost exactly match those of ocean tides. Furthermore, for the reasons outlined below, the absence of significant periodic tidal components in the borehole-referenced SP at the inland site suggests that Earth, atmospheric or geomagnetic tides, are not primarily responsible for the tidal SP response observed at the coastal site. Figure 10 shows the M2 component for surface and borehole-referenced SP at both sites for two dipoles of comparable length. The M2 component was orders of magnitude larger in the coastal SP data, suggesting that distinct SP sources are present in the coastal aquifer and are driven by ocean tides. However, because the M2 component was present in surface-referenced SP data at the inland site, caution was required before attributing the M2 response at the coastal site to ocean tides. We discuss the various possible sources of the observed SP response at each site in more detail in the following sections.
| Data Set | Semidiurnal | Diurnal | Largest Periodic Component | Hypothesized Primary SP Sources | ||||
|---|---|---|---|---|---|---|---|---|
| M2 | S2 | N2 | K1 | O1 | S1 | |||
| SP Data | ||||||||
| Coastal BH surface reference | 3 | 1 | 2 | 2 | 2 | 0 | M2 | Oceanic tide (EED/EK) |
| Coastal BH referenced | 3 | 2 | 1 | 2 | 2 | 0 | M2 | Oceanic tide (EED/EK) |
| Inland surface array | 2 | 1 | 0 | 0 | 0 | 3 | S1 | Temperature (electrode) |
| Inland BH surface referenced | 2 | 1 | 0 | 0 | 0 | 3 | S1 | Temperature (electrode) |
| Inland BH referenced | 0 | 0 | 0 | 0 | 0 | 0 | NA | NA |
| Possible SP Source Mechanisms | ||||||||
| Oceanic tide | 3 | 2 | 1 | 2 | 1 | 0 | M2 | |
| Geomagnetic field | 1 | 2 | 0 | 0 | 0 | 3 | S1 | |
| Temperature | 0 | 1 | 0 | 0 | 0 | 3 | S1 | |
| Earth tideb | 3 | 2 | 1 | 2 | 1 | 0 | M2 | |
| Atmospheric tideb | 0 | 3 | 0 | 0 | 0 | 1 | S2 | |
- a Dark grey shading indicates borehole-referenced SP. Light grey shading highlights surface-referenced SP. Periodic magnitudes are ranked: 3—largest, 2—intermediate, 1—smallest, 0—below 50% significance. NA, not applicable.
- b Earth tides were estimated from Agnew [2007] and atmospheric tides from Lindzen and Chapman [1969] and Dai and Wang [1999].
4.1 Temperature and Conductivity
At the inland site, the surface-referenced SP data showed a significant S1 component (Figures 2 and 4), similar to the temperature data at the surface electrodes (Figures 2b–2d). Temperature fluctuations at, and between, the surface electrodes, and the temperature sensitivity of the electrodes, likely contributed to the S1 response in the surface-referenced SP data. Differences in temperature fluctuations between electrode locations (Figures 2c–2d) explain the differing magnitude of S1 across the site (Figures 2e–2g). The electrodes have a reported temperature coefficient of −0.65 mV/°C [Silvion, 2015]. However, at the coastal site, the surface reference electrode was buried at 1.5 m, below the penetration depth of diurnal temperature fluctuations (Figure 7g) [Perrier et al., 1997]. Temperature fluctuations in both the inland and coastal boreholes were small, which partly explains why there was no S1 component in SP data from the coastal site, or in borehole-referenced SP data at either site (Table 2).
Borehole FEC data at the inland site showed no detectable frequency response. However, at the coastal site, both temperature and FEC fluctuated with a predominantly M2 component, although a K1 component was also present, most clearly in the temperature data (Figures 7d and 7f). Thus, the borehole temperature and FEC share the same key frequency components as the borehole SP data at the coastal site. However, the magnitudes of the fluctuations in temperature and FEC (~0.08°C and 50 μS/cm, respectively) are too small to be responsible for the SP response. The temperature fluctuations yield a maximum temperature contribution of 52 μV for the electrode sensitivity reported above, which is an order of magnitude less than the observed fluctuation of ~600 μV in the SP data (Figure 12a).
Figure 12c shows a crossplot of borehole FEC versus SP and clearly illustrates that there is no direct relationship. Furthermore, the SP fluctuations do not correspond to the FEC fluctuations (Figure 12(b)); moreover, when the FEC increased in late July 2013, there was no corresponding change in SP (compare Figure 7e and Figures 8a and 9a). Thus, SP was not directly caused by the FEC fluctuations observed in the borehole.
4.2 Geomagnetism
Trique et al. [2002] interpreted the M2 component in SP data obtained from a surface electrode array in the Alps as ocean tidal modulation of the geomagnetic field. The observation of M2 in the geomagnetic field is a global phenomenon [Cochrane and Srivastava, 1974; Lilley and Parker, 1976; Cueto et al., 2003; Maus and Kuvshinov, 2004; Love and Rigler, 2014]. The movement of ocean water through the geomagnetic field induces an electric current [Kendall and Chapman, 1970; Hewson-Browne, 1973], a phenomenon known as the ocean dynamo [Rosser and Schlapp, 1990; Palshin et al., 1996]. The ocean dynamo is known to generate telluric currents in the UK [Chapman and Kendall, 1970; Osgood et al., 1970; Brown and Woods, 1971; Hart et al., 1983b; Hart et al., 1983a]. The strength of the telluric contribution to measured SP depends on the length and orientation of the electrode dipoles [Orihara et al., 2012]. Thus, the presence of the M2 component in surface-referenced SP at both sites (Figures 2, 4, and 8) suggests that geomagnetism could have contributed (Table 2). Figure 11 shows the PSD of the geomagnetic field declination and horizontal and vertical intensity from three stations around the SP monitoring sites (Figure 1a).
The geomagnetic field is dominated by the principal solar component (S1) and contains S2, S3, and S4 harmonics. The presence of S3 and S4 in the long time series surface-referenced SP at the coastal site (Appendix A) provides an indication of geomagnetic influence, because S3 and S4 can only be explained by telluric currents.
The relative magnitude of each component at the three geomagnetic stations was approximately the same (Figure 11). Most significantly, the M2 component was consistently smaller than S1 and S2. Globally, the S2 component of the geomagnetic field is larger or comparable to M2 [Cochrane and Srivastava, 1974; Maus and Kuvshinov, 2004; Love and Rigler, 2014; Bindoff et al., 1988]. If the geomagnetic field were dominantly responsible for the SP fluctuations at both sites, then the SP data would be expected to show a similar power spectrum to the geomagnetic data. This is not the case, although the relationship between dipole length and SP response at the inland site (Figure 2) indicates that the geomagnetic field does contribute to the S1 and M2 components observed in surface SP data, as suggested by Trique et al. [2002].
Borehole-referenced SP data at both sites showed no S1 component (Figures 5 and 9), and only the coastal site showed a significant M2 component (Figure 9). Thus, the geomagnetic field did not make a significant contribution to the large M2 component observed in the SP data at the coastal site.
4.3 Earth and Atmospheric Tides
There was some indication of the principal solar component (S1) and its harmonic (S2) in the head fluctuations at the inland borehole (Figure 3d). Such head fluctuations are most likely caused by atmospheric pressure, which fluctuates with S1 and S2 period [Dai and Wang, 1999]. Rojstaczer and Riley [1990] and Kumpel [1997] showed that diurnal (O1, S1) and semidiurnal (M2, S2) water level fluctuations in boreholes can be caused by atmospheric and Earth tides and can occur in confined and unconfined aquifers (Table 2). Thus, Earth and atmospheric tides can cause groundwater flow and electrokinetic potentials, via periodic compression and release of the Earth's crust.
Earth tides share the same relative amplitude for each of the periodic components as ocean tides [Agnew, 2007]; thus, they may contribute to the M2 component observed in borehole or surface-referenced SP. Kulessa et al. [2003] invoked Earth tides as the primary driving mechanism for subglacial flow leading to a significant K1 component in their SP. Gokhberg et al. [2009] modeled observations of tidal SP on kilometer long surface cables in Japan and attributed the presence of the main tidal components (O1, P1, K1, N2, M2, and S2) to the electrokinetic potential caused by Earth tides. However, assuming that Earth tides are of the same order of magnitude at the inland and coastal sites [Agnew, 2007], then Earth tide effects are too small to explain the SP components at the coastal site, primarily because of the absence of any significant M2 component within the inland borehole-referenced SP (Figures 5 and 10).
SP processes in the unsaturated zone have been investigated by previous authors, who found that the source mechanisms are nontrivial. For example, Maineult et al. [2008] and Revil et al. [2008] found that SP measured during harmonic pumping tests and referenced using an electrode in the unsaturated zone produced nonlinear responses to the periodic change in pressure induced by the pumping tests. Maineult et al. [2008] attribute these nonlinear effects to saturation and desaturation processes occurring in the unsaturated zone. Unsaturated zone processes are briefly discussed here because they may also explain why M2 was observed in surface-referenced data at both sites. Adler et al. [1997] attributed SP fluctuations, from a few hours to 1 day, to periodic variations in saturation driven by atmospheric fluctuations. The M2 component is a constituent of atmospheric tides [Geller, 1970; Vial and Forbes, 1994; Forbes et al., 1994], albeit not as prominent as S1 or S2 [Lindzen and Chapman, 1969]. Kulessa et al. [2003] invoked atmospheric tides to explain the S2 component in their SP, suggesting that semidiurnal atmospheric fluctuations contributed to subglacial flow. Earth or atmospheric tides may not generate any significant change in pressure gradients within the aquifer but may drive flow or changes in saturation, within the unsaturated zone and in turn generate a measureable electrokinetic potential at electrodes installed at, or close to, the surface [Linde et al., 2011; Jougnot et al., 2012; Allègre et al., 2014]. Thus, Earth and/or atmospheric tides may explain the M2 signal recorded in our surface-referenced data. It is clear that surface SP monitoring is fraught with interpretation difficulties, including multiple overlapping source mechanisms.
4.4 Ocean Tides
The borehole-referenced SP data at the inland site contained no significant frequency components. In contrast, the borehole-referenced SP data at the coastal site contained significant M2 and K1 components, consistent with the borehole head and tidal gauge data. The above analysis, by ruling out alternative processes, suggests that these tidal components in the coastal borehole SP data must be caused by changes in pressure and/or concentration gradients (i.e., the electrokinetic (EK) or exclusion-diffusion (EED) potential) in the aquifer, driven by tides in the English Channel. Figure 12 shows coastal SP, head, and conductivity in July and August 2013. The SP, here smoothed with a 2 h window, is anticorrelated with head, which is consistent with a negative EK coupling coefficient as measured in chalk [Jackson et al., 2012a].
The head fluctuations observed in the coastal borehole are ~0.5 m (Figure 12a), and the maximum EK contribution is estimated using the coupling coefficient (−0.575 ± 0.080 mV/mH2O in Jackson et al. [2012a]) resulting in a ~290 μV EK contribution, around half the ~600 μV fluctuation observed in the borehole-referenced SP data. This estimate is consistent with observations made by Jackson et al. [2012a] who measured a ~60 μV response to a ~0.1 m drawdown during pumping of the abstraction borehole at the inland site (Figure 1b). Thus, it appears that the EED potential, caused by the remote movement of the saline front, must contribute to the observed tidal SP in the coastal borehole.
The arguments outlined above provide compelling evidence that a combination of EK and EED potential mechanisms are primarily responsible for the tidal SP observed in the borehole-referenced SP at the coastal site. However, the relative contribution of EK and EED potentials to the observed tidal SP needs to be investigated separately using numerical modeling.
4.5 Implications for SP Monitoring in Coastal Aquifers
Since tidal SP is caused by EK and EED potentials, borehole-referenced SP monitoring could aid aquifer management. SP could be used to characterize coastal aquifers through joint inversion with hydraulic data [Soueid Ahmed et al., 2014; Straface et al., 2007; Chidichimo et al., 2015; Straface et al., 2011], including determining aquifer transmissivity using tidal boreholes [Carr and Van Der Kamp, 1969; Millham and Howes, 1995; Rotzoll and El-Kadi, 2008; Rotzoll et al., 2013; Trefry and Bekele, 2004]. SP could also be used to determine dispersivity [Revil and Jardani, 2010; Straface and De Biase, 2013], allowing characterization of the saline transition zone [Abarca et al., 2007; Kerrou and Renard, 2010]. Finally, modeling suggests that the movement of the saline front may contribute to SP measured in remote boreholes [Gulamali et al., 2011; Jackson et al., 2012b]. Thus, SP could provide early warning of saline breakthrough. However, further research is required.
5 Conclusions
The main tidal constituents (M2, the semidiurnal, principal lunar cycle and K1, the diurnal, and principal lunar-solar cycle) were present in borehole-referenced SP data acquired at a coastal site, suggesting that SP responds to ocean tidal processes in the aquifer. No frequency components were found to be significant in borehole-referenced SP data acquired at an inland site. The M2 component was observed in surface-referenced SP data but was less significant than the S1 component which was driven by the diurnal temperature cycle. The M2 component in surface-referenced SP is likely to be influenced by a combination of Earth, atmospheric and geomagnetic tides. These processes must be accounted for in surface measurements or surface-referenced borehole measurements of SP if diurnal or semidiurnal processes are of interest, even in studies conducted far from the coast. The analysis presented here demonstrates that borehole-referenced SP in coastal aquifers is primarily driven by ocean tidal process in the aquifer, implying that EK and EED potentials are the primary SP source mechanisms. Further work is required to understand the competing contributions of EK and EED potentials. However, an initial analysis suggests that EK is unlikely to be solely responsible for the observed tidal SP. The results suggest that SP may have multiple applications in coastal aquifers, including characterization and monitoring, and could form part of a borehole contamination early warning system.
Acknowledgments
MacAllister was supported by NERC CASE studentship NE/I018417/1. The authors would also like to thank Southern Water for access to the borehole at Saltdean. Atkins Global and Southern Water are thanked for funding installation of the equipment and for additional funding under the NERC studentship. The laboratory components of this work were carried out in the TOTAL Reservoir Physics Laboratory at Imperial College London, and their support is gratefully acknowledged. Jackson acknowledges partial support from TOTAL under the TOTAL Chairs program. The data supporting the conclusions of this work are available through the corresponding author.
Appendix A:
The results from the PSD analysis of the hydrodynamic and SP data from the long-term monitoring experiment at the coastal site are presented here (Figures A1-A6).
