Introduction

Accurate and reliable characterization of aquifer heterogeneity remains one of the foremost problems in hydrogeology. Failure to characterize aquifer heterogeneity can have a number of undesired consequences for contaminant investigations (Boulding and Ginn 2004), particularly when trying to predict the transport pathways for various groundwater contaminants. Traditional hydrogeologic measurements used to characterize heterogeneity, such as grain size analysis and hydraulic testing, yield sparse data as they are limited to the vicinity of boreholes, and thus tend to sample small aquifer volumes. Near-surface geophysical methods used for hydrogeological studies, namely hydrogeophysics, can provide information about the subsurface structure of aquifers and the various confining units or aquitards, and can identify potential permeable pathways offered by faults, fracture zones or coarse-grained sediments (e.g., van Overmeeren 1998; Lunt et al. 2004; Weissmann et al. 2004; Ismail et al. 2005; Rayner et al. 2007; Hirsch et al. 2008; Perozzi and Holliger 2008; Rabeh et al. 2009). Hydrogeophysics can also provide some information on the hydrogeologic properties between or away from boreholes (e.g., Day-Lewis et al. 2002); however, the relationship between geophysical properties and hydraulic conductivity is complex (Day-Lewis et al. 2002), non-unique and scale dependent (e.g., Hyndman et al. 1994, 2000).

This study uses a combination of borehole geophysical logging and ground penetrating radar (GPR) surveying in an attempt to interpret aquifer structure within coarse-grained glacial outwash sediments. The study site is located within the Abbotsford–Sumas aquifer, situated in the central Fraser Valley of southern British Columbia (BC), Canada and northern Washington (WA), USA (Fig. 1). The aquifer is the largest unconfined aquifer in the region, covering an area of approximately 161 km2 (62 sq. miles), and is roughly bisected by the Canada–USA border. The motivation for the study was to characterize subsurface heterogeneity at different scales, which may influence the flow of groundwater and the transport of groundwater contaminants. The study does not propose to investigate “scale effects” related to the hydraulic and transport properties of aquifers, which can only be accomplished through hydraulic and tracer tests. Specifically, the study aimed to identify potential preferential pathways based on the continuity of geologic units and inferred grain size distribution. The presence of continuous high permeability gravel beds within otherwise sand-dominated aquifer could offer such conduits. Similarly, lower permeability confining units would tend to restrict groundwater flow and, if present near ground surface, protect underlying units from surface contamination.

Fig. 1
figure 1

Location of the Abbotsford–Sumas aquifer in the Lower Fraser Valley. The PARC study site and gravel pit are situated within 1 km of the international border between Canada and the United States. The grey zone indicates the extent of the aquifer; representative regional groundwater flow directions are indicated by arrows

Full size image

Several GPR surveys were conducted at Agriculture and Agri-Food Canada’s Pacific Agri-Food Research Centre (PARC) substation in Abbotsford, BC (Figs. 1, 2). Borehole geophysical logs, including natural-gamma and electromagnetic induction (conductivity), were acquired in nine PVC-cased monitoring wells around the perimeter of the site. These wells provide not only water table elevation data, but also historical dissolved ion concentrations (notably nitrate) at specific monitoring depths (Hii et al. 1999, 2005). Geologic constraints are provided from borehole lithologic (driller’s) logs as well as from an abandoned gravel pit (cross-sectional pit face) situated approximately 200 m south of the PARC site.

Fig. 2
figure 2

Site map showing the locations of the GPR lines and the nested monitoring wells. The black square on Line 1 indicates the location of the CMP survey midpoint

Full size image

Background

Hydrogeology of the Abbotsford–Sumas aquifer

The Abbotsford–Sumas aquifer is located on a broad outwash plain, which is elevated above the adjacent river floodplains (Fig. 1); topographic relief is roughly 150 m. The uplands are centered on the City of Abbotsford, BC, and extend westward through Langley, BC, and south to Lynden, WA. The largest valley is Sumas Valley, which runs northeast to southwest from the City of Sumas to the City of Chilliwack (Fig. 1), and contains the lower drainage of the Sumas River. The Sumas River flows to the northeast and picks up a significant baseflow component from aquifer discharge on its eastern side. To the south is the Nooksack River, which flows west and then south. Most of the surface and groundwater flow from the Abbotsford–Sumas aquifer ends up in the Nooksack River. The climate in this coastal region is humid and temperate, with a mean annual precipitation of 1,000–2,100 mm, falling mostly as rain from October to May. Recharge to the aquifer (900–1,100 mm) is primarily from direct precipitation (Scibek and Allen 2005).

The Abbotsford–Sumas aquifer is highly productive, and provides water supply for nearly 10,000 people in the USA (towns of Sumas, Lynden, Ferndale, Everson and scattered agricultural establishments) and 100,000 in Canada, mostly in the City of Abbotsford, but also in the township of Langley (Mitchell et al. 2003). The aquifer is mostly unconfined and comprised of coarse-grained sediments of glaciofluvial drift origin, referred to as the Sumas Drift (11,000–10,000 BP) (Armstrong et al. 1965). The Sumas Drift consists of diamictons (lodgement and flow tills), thick and well-sorted glaciofluvial outwash (uncompacted) sands and gravels (advance and recessional), glaciolacustrine sediments, and ice-contact sediments deposited during the Sumas Stade (Armstrong et al. 1965). It also contains lenses of till. The thickness of Sumas Drift can be up to 65 m, and it is thickest in the northeast where glacial terminal moraine deposits are found. The aquifer is underlain by an extensive glaciomarine deposit, the Fort Langley Formation, which outcrops in the uplands to the west. Groundwater flows regionally from north to south, as illustrated with representative regional flow lines in Fig. 1, although there are some local variations (Liebscher et al. 1992), particularly in the vicinity of the streams that drain the aquifer. There is minimal vertical flow within the aquifer, except in the near surface through direct recharge, which is of particular importance in respect of water and nitrate transport through the unsaturated zone.

This outwash deposit was situated close to the ice margin; a high energy glaciofluvial environment, with highly variable braided streams that changed course frequently (Easterbrook 1969). Consequently, the deposit is highly heterogeneous and hosts variable mixtures of sand and gravel with minor fines. Scibek and Allen (2005) summarize ranges of hydraulic conductivity, K, for the Sumas Drift. K ranges from 2 to 2,377 m/day, with a mean value of 60 ± 4 m/day based on the results of 136 well tests. The highest values represent highly permeable portions of the aquifer and are not typical of most locations. This heterogeneity can be expected to result in complex groundwater paths at both regional and local scales. For example, Scibek and Allen (2005) confirmed a north to south regional flow through numerical modeling, but noted that groundwater was redirected locally, particularly around streams, as a result of heterogeneity of the surficial sediments.

Aquifer heterogeneity appears to have some bearing on the distribution of nitrate contamination within this agricultural setting. Groundwater nitrate concentrations have been monitored historically across the Abbotsford–Sumas aquifer, and are typically observed to decrease with increasing depth (Hii et al. 1999). McArthur and Allen (2005) showed that in the shallowest wells, with depths less than 10 m, most nitrate-nitrogen (NO3-N) concentrations have been historically (1988–2004) below 30 mg NO3-N/L, with an average nitrate concentration of 13 ± 7 mg NO3-N/L. However, there were several wells with higher concentrations, approaching 60 mg NO3-N/L. The average nitrate concentration in moderate depth wells between 10 and 40 m depth was 14 ± 5 mg NO3-N/L. All wells deeper than 50 m had nitrate concentrations below 30 mg NO3-N/L and most were below the maximum allowable concentration (MAC) of 10 mg NO3-N/L. However, at several locations within the aquifer, such as the PARC site, long-term monitoring in nested monitoring wells indicates moderate NO3-N concentrations at shallow depths, and higher concentrations at intermediate and deeper depths. For example, at the southeast corner of the PARC site, NO3-N concentrations at 91-7 (22.1 m) are historically lower (ave. 10.4 ± 2.7 mg NO3-N/L) than those measured in the deep monitoring well, 91-4 (46.0 m).(ave. 14.7 ± 3.8 mg NO3-N/L). In all monitoring wells, the screen is located across the bottom 1 m of each monitoring well, and the sample is drawn from mid-screen depth. Thus, nitrate concentrations are generally representative of the monitoring well’s depth. Such concentration inversions, as found at the PARC site, are suggested to be the result of preferential pathways within the subsurface, through which contaminated groundwater travels rapidly away from source areas (here associated with raspberry crops) and reaches deeper portions of the aquifer.

Geophysical approaches for aquifer characterization

GPR has been used successfully in a number of studies to improve aquifer characterization, including mapping of clay–sand boundaries (Young 1995), imaging sedimentary units (Greenhouse et al. 1987; Beres and Haeni 1991; Jol and Smith 1991; Aiken 1993; Beres et al. 1995, 1999; Olsen and Andreasen 1995; Asprion and Aigner 1997; van Overmeeren 1998) and characterizing aquifer heterogeneity (Gawthorpe et al. 1993; Lunt et al. 2004; Day-Lewis et al. 2002; Close et al. 2004). Locally (within the Fraser Valley), an extensive GPR survey was conducted in the similarly heterogeneous Brookswood aquifer (Rea 1996; Rea and Knight 1998; Rea et al. 1994; Rea and Knight 2000). Aquifer characterization was conducted by identifying and mapping radar architectural elements, which were identified as hydrogeological units, and then combined with drillers’ logs to reconstruct the depositional environment (Rea 1996; Rea and Knight 2000). The depth of penetration on the GPR profiles ranged from about 5 m in the eastern part of the aquifer to about 15 m in the western part due to changes in attenuation of the GPR signal. In that study, electrically conductive clay bodies were suggested to correspond to low-permeability boundaries.

Much of the work done using borehole geophysics for aquifer characterization has been to examine aquifer lithology, aid stratigraphic correlations, or map water quality. Of the various tools available, aquifer lithology is commonly examined using both natural-gamma and conductivity logging (Cromwell 1992; Nobes and Schneider 1996; Barrash and Morin 1997; Pullan et al. 2002; Siron and Segall 1997; Paillet and Reese 2000). Because natural-gamma is sensitive to clay content and grain size, it is often the only geophysical log used to examine aquifer lithology (West 2002; Norris 1972; Baldwin and Miller 1979; Dixon-Warren and Stohr 2003).

In 1993, as part of the Fraser Lowland Hydrogeology Project, two monitoring wells at the PARC site (94-2 and 94-3) were logged using natural-gamma, conductivity and magnetic susceptibility sondes (Geological Survey of Canada 2003). These data were used for comparison with the logs collected during this study and, as noted later, were used to adjust conductivity logs affected by a minor calibration error. Irving and Knight (2003) conducted cross-hole GPR logging between different pairs of monitoring wells at the PARC site in order to investigate saturation-dependent radar wave velocity anisotropy in both the saturated and unsaturated zones. Their findings indicated that in the unsaturated zone, anisotropy is greater than in the saturated zone. This was thought to be because the coarse-grained units drain more readily than the finer-grained units, hence the finer-grained units would retain moisture and result in higher anisotropy. They determined an average vertical radar velocity of 0.12 m/ns in the unsaturated zone, and 0.07 m/ns in the saturated zone. Their survey was conducted in late July/early August when ground saturation levels are near their lowest in the aquifer; the water table was situated at a depth of 17.5 m.

Methodology

Site description

All of the geophysical surveying for this current investigation was completed at the PARC site in Abbotsford, BC. The PARC site is approximately 200 by 400 m, and is used for test crops. As part of its monitoring program, Environment Canada installed ten monitoring wells at this facility (Fig. 2); these wells range in depth from 19.4 to 46.4 m, with an average depth of 27.9 m. The site was chosen due to availability, existence of a number of monitoring wells on-site, and because there is increased nitrate contamination within the aquifer in this area (Hii et al. 1999). The field work was completed over 5 days during May and August of 2005. Water levels in the monitoring wells were measured during field work excursions. The water table ranges in depth from approximately 17 to 21 m across the site.

An abandoned gravel pit, located 200 m to the south of the PARC site (Fig. 1), was also visited in order to investigate the local geological framework. A series of photographs were taken in a north–south direction along a wall within the gravel pit (Figs. 3, 4).

Fig. 3
figure 3

Gravel pit excavation photo. Unit 1 is horizontally bedded coarse sand with gravel units. Unit 2 is medium to coarse sand. Unit 3 is bedded gravel with sand and large cobbles that is dipping slightly to the south. Unit 4 is medium sand with some gravel and cobbles. Unit 5 is talus that has fallen off the excavation face. The white box at the north end indicates the location of Fig. 4

Full size image
Fig. 4
figure 4

Close-up of the area indicated in Fig. 3. The interbedded nature of the coarse sand and gravel deposits is apparent. The pen circled in red is 17 cm long

Full size image

Borehole logging

The borehole logs were collected for all monitoring wells at the site, except CDA2 (see Fig. 2 for monitoring well location). The system was a Mount Sopris MGX-II portable digital logger, including a 305 m winch. Two tools were used: a 2PGA-1000 natural-gamma tool and a 2PIA-1000 (Geonics EM-39) electromagnetic induction (apparent conductivity) tool. Conductivity was logged first, followed immediately by gamma, and all logs were collected in an upward direction for a consistent run of the tool. Disturbance of the water column is not an issue because formation apparent conductivity is measured (about 1 m from the borehole), not fluid conductivity. Fluid conductivity itself is low within the borehole and in the surrounding aquifer, with specific conductance values less than 200 μS/cm based on water samples collected in each monitoring well. As these boreholes are cased with solid PVC, except at the screen, the fluid conductance of standing water column reflects that of the screen interval and will be uniform with depth. Thus, fluid conductance is not likely to influence formation conductance.

Natural-gamma logs were conducted at a logging speed of 2.75 m/min, with a sampling interval of 0.005 m. The data were smoothed using a 61-point moving average filter to reduce noise and to allow for comparison. This filter was selected through experimentation to provide some smoothing to the data, but not so much as to lose the character of the logs. As there is some inherent error associated with the gamma logging, monitoring well 91-4 was logged in both the upward and downward directions; the latter at a logging speed of 1.5 m/min and with a sampling interval of 0.005 m. Gamma counts were consistent in both logging directions (McArthur 2006), reducing the uncertainty associated with the interpretations of the logs. As well, the correlation of units across several boreholes (as discussed later) would suggest that there are minimal effects from the boreholes themselves.

Electrical conductivity logs were conducted at a logging speed of 2.5 m/min, with a sampling interval of 0.01 m. Most of the logs exhibit negative conductivities, which is not uncommon for Geonics EM-39 conductivity meters. Negative values are primarily noticeable at low conductivity (approx. <5−10 mS/m) values because the probes are not perfectly calibrated by Geonics for these low conductivities. In this study, we had the benefit of being able to directly compare these logs to those acquired by the Geological Survey of Canada a few years previously, in which more realistic (+) minimum conductivities were measured. A small uniform positive shift of 4 mS/m was thus applied to the data to compensate for the calibration error. The variability with depth, however, was consistent between the two sets of logs; only the magnitude differed. No additional processing was applied to the conductivity logs.

Ground penetrating radar

GPR data were collected using a Sensors & Software pulseEKKO 100 GPR system, with 50 and 100 MHz center frequency antennas. This GPR unit consists of separate transmitting and receiving antennas, a control console and a laptop. Antenna positioning was maintained using a tape measure. GPR lines were oriented such that they passed by the monitoring wells as closely as possible, but due to rows of raspberry bushes, they were generally restricted to the grass roadways between the test plots. Unfortunately, a GPR line could not be run adjacent to the nest of monitoring wells in the northwest corner of the site.

Both common-offset and common-midpoint data were collected. A common-midpoint (CMP) sounding was conducted first along Line 1 at both 50 and 100 MHz antenna frequencies (see Fig. 2 for location). For the CMP sounding the initial antenna separation was 0.2 m, with an increase in separation of 0.2 m for each subsequent measurement, for a total offset of 38 m. A prominent reflection was visible in the subsurface near 400 ns in the CMP data; however, the reflection was discontinuous with interference from other events thus hindering a reasonable CMP velocity depth analysis. Therefore, near-surface velocity was initially determined from the first arrival of the ground wave. For both the 50 and the 100 MHz antenna frequencies, the direct radar wave velocity was 0.11 m/ns.

Common-offset surveys were conducted at both 50 and 100 MHz on 7 lines (Fig. 2). Profile surveys at 50 MHz used a constant offset of 2 m (based on the results of the CMP survey) with a station spacing of 0.4 m. Profile surveys at 100 MHz used a constant antenna offset of 1 m with a station spacing of 0.2 m.

GPR data were processed using the software ReflexW (Sandmeier 2005). After applying a constant time-zero shift, a moving average dewow filter with a window length of 30 ns for the 50 MHz antenna data and 15 ns for the 100 MHz antenna data was applied. This was followed by energy decay gaining. The energy decay gain is based on the envelope of amplitude decay of the average trace calculated for a profile, where the gain is the inverse of the decay curve. Energy decay gain is preferred over automatic gain control (AGC) for interpretation since the same gain function is applied to all traces in a profile, thus better preserving amplitude variation laterally along the profile and with depth. All of the profiles exhibited prominent near surface (top 2 m) diffractions (from boulders) that interfered with the stratigraphic image. Velocities determined from the diffraction hyperbolas ranged from 0.095 to 0.105 m/ns (mid-value of 0.100 m/ns). These diffractions were therefore filtered by applying a constant velocity migration (using the mid-value velocity of 0.100 m/ns) to the data.

There was also a strong reflection from the water table at the PARC site, between 235 and 320 ns. Based on measured water table elevations at the site at the time of the survey, a velocity of 0.12 m/ns is estimated. This involved a two step process. To first provide additional radar-based evidence that the reflection we expected to be the water table was indeed at the approximate depth, we first used a depth scale determined independently from velocity hyperbola fitting and the direct wave. Then the depth scale was re-calibrated for the water table depths measured in the wells at the time of the survey. The velocity was slightly higher (by 0.01 m/ns) than that obtained from the CMP survey (ground wave arrival) and diffraction hyperbolas, but is the same as the average vertical unsaturated zone velocity of 0.12 m/ns determined by Irving and Knight (2003) using cross-borehole GPR. All the values are typical of partially saturated sands where the observed velocity variation can reasonably be explained by expected variations in porosity and related variations in volumetric water saturation. Also, the site is used for agricultural research; therefore, the top soil layer may be enriched with organics, and the site is irrigated, thus, the surface layer may have a higher moisture content: consistent with the observed slower surface velocity. Because the water table was measured in all of the monitoring wells, and these wells provide an accurate estimate of the water table depth at the site, the depth scale for the common-offset data was determined using a velocity of 0.12 m/ns, thus providing a reasonable average velocity to the water table depth.

Results

Borehole logging

Figure 5 shows the apparent conductivity and natural-gamma logs for well 91-1, situated at the northwest corner of the site (see Fig. 2). These logs are typical of those collected at the site (McArthur 2006). Below the surface organic soil layer, gamma counts range between 20 and 50 cps (see also Fig. 6), indicating predominantly sands and gravels around the monitoring wells (Keys 1997). These cps results are consistent with the limited amount of lithologic data collected during drilling and rudimentary grain size analyses conducted at the PARC site, which yield dominantly coarse sand and gravel. Apparent conductivity values for this and other boreholes at the site range from near zero to approximately 7 mS/m, but average approximately 2.5 ± 3.6 mS/m.

Fig. 5
figure 5

Natural-gamma log (black) (61-point moving average) and conductivity log (grey) from monitoring well 91-1. See Fig. 2 for location. The water table is indicated by the horizontal dashed line with inverted triangle at approximately 17 m depth

Full size image
Fig. 6
figure 6

Natural-gamma logs (61-point moving average) from monitoring wells 91-4, 91-5 and 91-7. See Fig. 2 for well location. Well separation is only a few meters as indicated by the scale. A fining upward sequence repeats on a scale of 3–10 m (representative arrows shown) and some units may be laterally correlated at a scale of approximately 10 m or more (as illustrated by the grey shaded area)

Full size image

What is immediately apparent in Fig. 5 is the abrupt shift in conductivity at the water table, and the apparent negative correlation between gamma cps and conductivity in the saturated zone. Within the unsaturated zone (0–17 m), conductivity ranges from slightly above 0–1.5 mS/m over a range of gamma (~32–45 cps). In the saturated zone (17–46 m), conductivity ranges from 1.6 to 6.0 mS/m over a range of gamma (~26–40 cps). Overall, gamma cps is lower below the water table, probably because of higher attenuation due to the higher density of water with respect to air as in the unsaturated zone. The relatively sharp rise in conductivity at the water table suggests a relatively thin capillary zone (i.e., an abrupt contact between the unsaturated and saturated zones). This suggests that the water table should produce a prominent radar reflection because contrasts in electrical permittivity are mainly controlled by water content.

In the natural-gamma logs, fining upward (and limited coarsening upward) sequences can be seen in the sand and gravel (Fig. 5). Fining upward sequences were identified on the basis of a noticeable increase in gamma count (upward), which persist vertically over 3–10 m. There are also sudden shifts in the natural-gamma counts (either up or down) that reflect the abrupt changes in grain size interpreted to be associated with erosional unconformities. Due to the high energy depositional environment in which these sands and gravels were deposited, and the proximity to their source, there was frequent erosion and deposition of subsequent layers (Easterbrook 1969) leading to such unconformities. A rudimentary grain size analysis was completed by Environment Canada on the “91” series monitoring wells at some of the depth intervals. In each interval, the average grain size was calculated, and plotted against the mean gamma count for the same depth interval. The depth intervals of the grain size samples were 1.2 m. There was no correlation (results not shown), which is likely due largely to the large depth intervals over which the grain size analysis was completed and, therefore, the calculation of mean gamma. Thus, the resolution was too low to detect thin units of finer or coarser material. Had the grain size sampling interval been smaller, a better relationship may have been established from the gamma data.

Correlation of the fining upward sequences between wells was challenging due to the coarse-grained nature of the sediments, as shown by the cross-section constructed from the natural-gamma logs from wells 91-4, 91-5 and 91-7 (Fig. 6). Correlation from one well to the next was done on the basis of repeating patterns of one or more fining upward sequences, and arguably, such correlations are questionable based on the borehole data alone. Therefore, GPR data were used to constrain the interpretation. Figure 6 shows unit contacts which are similarly defined by prominent GPR reflections in Fig. 7a (red sub-horizontal lines). The grey shaded area on Fig. 6 (representing a package of units) is also shown on Figs. 7a, b for comparison.

Fig. 7
figure 7

GPR lines collected along Line 2 (a and b) and Line 5 (c and d). Shown in a and c are the 50 MHz profiles, and shown in b and d are the 100 MHz profiles. The 50 MHz profiles are annotated with arrows identifying prominent reflections and the water table. Corresponding reflectors are interpreted on the 100 MHz profiles. Natural-gamma logs for monitoring wells 91-4, 91-5 and 91-7 (Line 2) and CDA-1 (Line 5), which are adjacent to the GPR lines, are shown. The blue shaded zone on the 100 MHz line coincides with that shown in Fig. 6

Full size image

The difficulty in correlating units attests to the heterogeneity of the subsurface, which occurs both vertically and horizontally. Layers appear to be continuous laterally for about 10 m (i.e., between boreholes). The photographs from the gravel pit (Figs. 3, 4) also show distinct units of coarser and finer material. The thickness of the sand and gravel units range from 20 cm to approximately 3 m in this exposure. Most of these units are laterally continuous, from several meters up to greater than the extent of the photo. There are also smaller units that are only laterally continuous over a scale of tens of centimeters to a few meters, which are smaller than those indicated by the natural-gamma logs.

At this site, interpretation of the borehole data is limited by existing wells and their spacing. Notwithstanding, the borehole geophysics provided valuable information on vertical and lateral heterogeneity at small spatial scales. To investigate whether this heterogeneity can be extended beyond the nested monitoring wells and across the site, GPR surveys were conducted.

Ground penetrating radar

GPR data for three of the lines are shown: Line 2 and Line 5 are approximately 165 and 180 m long, respectively (Fig. 7a, b), and Line 3 is 385 m long (Fig. 8). The top profiles show the 50 MHz data, and the bottom profiles show the 100 MHz data. Also shown on two of the 50 MHz profiles are the gamma logs collected in the adjacent monitoring wells. In Fig. 7a, prominent reflections that correlate with the gamma log contacts in Fig. 6 are also shown. The 50 MHz profiles show a deeper penetration than the 100 MHz surveys; however, the 100 MHz surveys provide a much higher resolution above the water table. Below the water table reflection, especially in the 100 MHz profiles, the signal is often lost due to increased attenuation within the more conductive water-saturated zone. The depth scale on the right side of the images is based on a velocity of 0.12 m/ns.

Fig. 8
figure 8

GPR lines collected along Line 3. a 50 MHz profile, and b 100 MHz profile. The 50 MHz profiles are annotated with arrows identifying prominent reflections and the water table. Corresponding reflectors are interpreted on the 100 MHz profiles

Full size image

The water table is represented as a strong reflection between 360 and 400 ns on most of the sections; shown as blue line in the 100 MHz sections only. The slope of the water table is consistent with the water table gradient observed in the monitoring wells at the PARC site; groundwater at the PARC site flows northwest to southeast. The fact that this reflection is so strong is indicative of the abrupt nature of the water table at this site. That is, due to the relatively coarse-grained sediments, the capillary fringe is thin to absent, as was suggested by the borehole conductivity data. Water table reflectivity is frequency-dependent, with lower frequencies sensing a more abrupt “interface” compared to higher frequencies sensing more of a gradation; this effect can be seen in comparing the 50 and 100 MHz profiles. In other study areas where finer-grained materials, such as silt or clay are present, the water table reflection may be less continuous and less abrupt due to variations in soil moisture (e.g., Galagedara et al. 2003). As indicated earlier, Rea and Knight (2000) identified electrically conductive clay bodies in the nearby Brookswood aquifer. In that particular area, glaciomarine clays of the Fort Langley Formation outcrop near surface and occur more frequently as lenses in the subsurface as compared to the Abbotsford aquifer.

Hydrostratigraphic interpretation of the GPR data was based on reflection configurations described in Beres and Haeni (1991). The GPR profiles reveal a variety of reflections that are interpreted as primary sedimentary structures, and that can be defined as radar facies. Radar facies are mappable, 3D units composed of reflections whose internal reflection configuration (e.g., shape, orientation), continuity, amplitude, polarity, spacing, and external 3D geometry differs from adjacent units (Labey et al. 2009). Although not mapped in 3D, at least two radar facies are apparent in the GPR profiles. All of the profiles exhibit similar reflections and an overall similar character.

Radar facies 1 (RF-1) is defined by reflections that are slightly undulating, generally horizontal, and parallel to sub-parallel. These reflections are generally short and most appear not to be laterally continuous beyond 10–15 m along the profile. For example, the same blue shaded area that was illustrated on the borehole cross-section (Fig. 6) is also shown on Fig. 7a, b. Within this zone, several GPR reflections correlate with the tops and bottoms of units identified in the borehole logs (unit contacts from the gamma logs are delineated by red lines). Similar contacts are apparent above and below this shaded zone in both the GPR and gamma data. At monitoring well CDA1, several dominant shifts in gamma occur, notably at 8 and 18 m; both of these are also clearly associated with reflections, but there is no clear indication of unit contacts in the gamma log.

These small-scale reflections are interpreted to result from interbedded sands and gravels, typical of a high energy fluvial environment, and correspond to abrupt changes in grain size over small distances (<10 m) as evidenced in the borehole logs. In areas that appear more discontinuous in nature, smaller scale (not fully resolved) sand and gravel cross-bedding and/or smaller lenses are present, with larger diffractions, occurring predominantly close to the ground surface, being caused by boulders. Such features, including boulders, are observed in the gravel pit (Figs. 3, 4).

Radar Facies 2 (RF-2) is defined by reflections that are much longer, extending up to a few hundred meters (dominant reflections, other than the water table) (Figs. 7, 8). These are interpreted as beds. A few reflections appear to be continuous across the site; the more prominent are colored in orange on the 100 MHz lines and are indicated by white arrows on the 50 MHz lines. Lines 2 and 5 (Fig. 7) show evidence of a deep continuous reflector at 30 m depth. Shallower continuous reflectors are clearly evident on Line 5 at 6 m (near monitoring well CDA2) and 8 m (Fig. 7). These latter two reflectors appear to be present on Line 3, which extends across the center line of the site (see Fig. 2). The same reflectors are evident on Line 6 (data not shown, see McArthur 2006). These results suggest that while the internal structure of horizons at the site may not be laterally continuous, larger scale beds are continuous (up to at least 360 m). At the gravel pit, such continuous beds are clearly evident (see Fig. 3), and are associated with unconformities originating from erosion of bed material.

Discussion

The outwash sediments of the Abbotsford aquifer were deposited within a high energy glaciofluvial environment, characterized by frequent erosion, and deposition of subsequent units. The high energy of this depositional environment likely resulted in little to no fine-grained (e.g., clay/silt size) material remaining in the deposits, and to the heterogeneous nature of the sand and gravel deposits. Both the geophysics and the gravel pit exposure support this interpretation. For example, the gamma count does not exceed 50 cps. The relatively low and consistent gamma counts combined with driller’s logs are consistent, with an absence of significant units of finer-grained material and especially a lack of clay. Also, grain size analysis conducted on sediments collected during drilling of a well situated close to the gravel pit showed a high percentage of coarse sand and gravel, and yielded a range of saturated hydraulic conductivity of 52–3,450 m/day based on the maximum and minimum grain size curves. These relatively high values suggest that fines are largely absent.

On the GPR profiles, the undulating to discontinuous nature of RF-1 at scales of 10-15 m (Figs. 7, 8) is interpreted to result from interbedded and cross-bedded sands and gravels, and likely represents contacts between these discontinuous fining upward sequences; although, the GPR cannot resolve the fining upward grain size distribution. This depositional variability extends across the PARC site, resulting in an overall heterogeneous character to the subsurface at scales less than 500 m. However, GPR profiles also point to a few laterally continuous horizontal to sub-horizontal reflectors, defined as RF-2, which extend at least up to 360 m. These are interpreted as beds, possibly bounded by unconformities, based on evidence of gravel bars, truncation of underlying units, as well as scour and fill features in the gravel pit exposure (Figs. 3, 4).

Overall, there appears to be two scales of heterogeneity: (1) small scale (<15 m) heterogeneity reflected in the units comprised of fining upward (and occasional coarsening upwards) sediments observed in the borehole logs, which correspond to RF-1. These sequences may be laterally continuous at scales of 10–15 m, based on gamma log correlations between monitoring wells and GPR reflectors that can be traced between the wells. The continuity of the sequences beyond 10–15 m in the GPR profiles is less evident as GPR reflections are more discontinuous at this scale; and (2) larger scale (up to at least 360 m) heterogeneity resulting in continuous GPR reflections along beds that may be associated with unconformities (RF-2) as discussed above.

In order to understand and predict groundwater flow and contaminant transport in this, and similarly coarse-grained heterogeneous aquifers, consideration must be given to both the nature of the heterogeneity and its scale, as well as to the dominant flow direction within the unsaturated and saturated zones.

In the unsaturated zone, water infiltrates and moves predominantly downward. The rate at which the groundwater infiltrates is dependent largely on the hydraulic properties of the sediments, which are dependent upon the soil moisture (Galagedara et al. 2003). Layering will result in variations in the hydraulic conductivity due to variable saturation of the coarse and finer-grained materials. For example, Chesnaux and Allen (2008) undertook simulations of nitrate transport through the unsaturated zone using grain size data from this aquifer. They showed that even with a material comprised of sand and gravel, the finer-grained sediment has a greater water retention capacity relative to the coarser-grained sediment, which results in a longer time for a dissolved contaminant (nitrate in that study) to reach the water table. Therefore, characterizing the nature of vertical heterogeneity of the sediments is important in hydrogeological studies where infiltration and contaminant transport though the unsaturated zone is of concern. The results of this study suggest that gamma logs can provide insight into vertical layering within coarse-grained aquifers and grain size distribution (fining upward) within individual coarse-grained units. Thus, borehole logging in combination with higher frequency sediment sampling for grain size analysis would be beneficial for such studies.

In the saturated zone, groundwater flow is predominantly horizontal in this, and many, aquifers (low vertical gradient). Thus, the lateral continuity of geologic units is of fundamental importance for understanding groundwater flow at a range of scales; groundwater can be anticipated to move quickly through coarser-grained horizons, with these possibly acting as permeable pathways, and slower in finer-grained units, with these possibly acting as confining units. The GPR data suggest that some units are continuous laterally over distances of at least 500 m. This continuity of units has important implications for conceptualizing the hydrostratigraphy across the site. While it was not possible to identify which units might act as permeable pathways, the GPR data nonetheless suggested that this coarse-grained aquifer has structure, and that elements of this structure can be detected at smaller scale (i.e., within the borehole logs). Unfortunately, a lack of accurate grain size data did not permit constraint of the borehole log interpretation. This lack of detailed information on grain size is not uncommon in hydrogeological investigations, particularly those conducted at regional scales that rely on domestic well records. Typically, water well drillers only casually record major changes in grain size, and infrequently are samples collected for grain size analysis.

In summary, most hydrogeological studies would benefit from the use of geophysics (borehole and other surface methods, including GPR) to aid in characterizing the subsurface hydrostratigraphy and its structure. Ideally, a combination of methods should be used, as one can be used to aid in the interpretation of the other, as demonstrated here in using GPR reflections to constrain unit tops and bottoms in borehole logs. This combined approach offered a means to interpret the data in this coarse-grained aquifer; something that would have been challenging to do otherwise.

Conclusions

Geophysical techniques were used successfully to characterize the nature of subsurface heterogeneity within a coarse-grained aquifer, over a range of scales. Borehole gamma logs show evidence of fining upward sequences, occasional coarsening upward sequences, and abrupt changes in grain sizes. The fining upward sequences, which persist vertically for 3–10 m, appear to be laterally continuous from one monitoring well to the next, over a distance of at least 10 m. Unit contacts also appear to correlate with GPR reflectors, which have an undulating to discontinuous nature at lateral scales of 10–15 m (radar facies 1). RF-1 is interpreted to result from interbedded and cross-bedded sands and gravels. GPR profiles also point to a few laterally continuous horizontal to sub-horizontal reflectors, which extend at least up to 360 m, which are interpreted as unconformities (RF-2), based on evidence of gravel bars, truncation of underlying units, as well as scour and fill features in the gravel pit exposure. GPR profiles also show a strong reflection that is interpreted as the water table. The fact that this reflection is so strong is indicative of the abrupt nature of the water table at this site. That is, due to the relatively coarse-grained sediments, the capillary fringe is thin to absent.

The nature of the heterogeneity at this and other similar sites will influence groundwater flow and the transport of contaminants both within the unsaturated zone and in the saturated zone, and such heterogeneity should be taken into account in hydrogeological investigations. Representing the vertical heterogeneity is particularly important for characterizing flow and contaminant transport within unsaturated zone, as there will be an impact on transport times and concentrations due to variably saturated conditions. In the saturated zone, lateral continuity of coarser-grained materials may contribute to permeable pathways, which may be important for predicting flow and transport at larger scales.