Combining unsaturated and saturated hydraulic observations to understand and estimate groundwater recharge through glacial till

Journal of Hydrology 391 (2010) 263–276

Contents lists available at ScienceDirect

Journal of Hydrology journal homepage: www.elsevier.com/locate/jhydrol

Combining unsaturated and saturated hydraulic observations to understand and estimate groundwater recharge through glacial till M.O. Cuthbert *, R. Mackay, J.H. Tellam, K.E. Thatcher Hydrogeology Research Group, School of Geography, Earth and Environmental Sciences, University of Birmingham, Birmingham B15 2TT, UK

a r t i c l e

i n f o

Article history: Received 19 June 2009 Received in revised form 13 May 2010 Accepted 21 July 2010 This manuscript was handled by P. Baveye, Editor-n-Chief Keywords: Groundwater recharge Till Fracture Unsaturated zone Superficial deposits

s u m m a r y Although there has been much previous research into various aspects of the flow mechanisms through glacial till, an integrated analysis of the flow system from the ground surface to the aquifer is lacking. This paper describes such an approach with reference to a detailed field study of the hydraulic processes controlling groundwater recharge through lodgement till in Shropshire, UK. A fieldsite was instrumented with tensiometers and piezometers at a range of depths through the profile, and the geology investigated in detail through field and laboratory testing. The median matrix hydraulic conductivity of the 6 m thick till is found to be around 2  1010 m/s on the basis of laboratory measurements. Using the barometric efficiency of the till derived from on-site pressure responses, the specific storage for the till is found to be in the range 2  106–6  106 m1 and approximately 3  106 m1 for the underlying Permo-Triassic sandstone, the regional aquifer. The hydraulic data indicate that till water table responses to rainfall occur during the summer period even when large tensions are present higher in the profile. This is thought to be due to preferential flow through hydraulically active fractures in the till, which were observed in a test pit dug on-site. The field evidence indicates that the fractures are usually infilled with a variety of materials derived and transported from clasts within the till. The bulk hydraulic conductivity of the till seems to be greatly enhanced by these features and it is shown on the basis of hydraulic testing and numerical modelling that the bulk hydraulic conductivity of the till is orders of magnitude greater than that of the till matrix and reduces with depth below ground surface. The paper furthers understanding of the hydraulic processes contributing to recharge through till and makes the link between the detail of these processes and simplified models of recharge estimation, which may be needed for larger scale water resource studies. The results are relevant also to contaminant migration studies and aquifer vulnerability assessments. Ó 2010 Elsevier B.V. All rights reserved.

1. Introduction Although understanding the processes involved in groundwater recharge is crucial to both water resources studies and aquifer vulnerability assessments, the influence of superficial deposits on groundwater recharge is poorly understood (e.g. Lloyd, 1980; Lerner et al., 1990; Rushton, 2005). This is particularly true of areas covered by glacial till. Preferential flow is a widely recognised hydrological phenomenon and in the context of till deposits, early research quickly identified fracturing as an important feature. Many studies report field measurements of hydraulic conductivity which are several orders of magnitude greater than core-scale laboratory values measured at equivalent locations (Keller et al., 1986; Fredericia, 1990; McKay et al., 1993; McKay and Fredericia, 1995; Klinck et al., 1996, 1997; Gerber et al., 2001). Further evidence for preferential flow path* Corresponding author. Tel.: +44 121 414 6167; fax: +44 121 414 5528. E-mail address: [email protected] (M.O. Cuthbert). 0022-1694/$ - see front matter Ó 2010 Elsevier B.V. All rights reserved. doi:10.1016/j.jhydrol.2010.07.025

ways has come from the use of applied and environmental tracers, in particular the presence of tritiated water at much greater depths than would be expected due to flow through the low permeability matrix alone (Keller et al., 1986, 1988; Hendry, 1988; D’Astous et al., 1989; Ruland et al., 1991; Gerber et al., 2001). An early publication giving evidence for the influence of fractures on the movement of groundwater through glacial till was by Williams (1967). The presence of fractures, which provide pathways for preferential flow, has been used to explain this scale dependence of hydraulic conductivity measurement ever since. Since drilling is an inefficient method of investigating till fractures due to the small volume sampled, test pits of a variety of sizes and designs have been excavated by many researchers to allow a more detailed fracture mapping to be carried out (Christy et al., 2000; Jakobsen and Klint, 1999; Klint and Gravesen, 1999). Although the most intense fracturing is often associated with the oxidised weathered zone of the upper few meters, fractures have been found to extend significantly deeper than this zone to depths <10 m (Ruland et al., 1991) and as deep as 21 m (Grisak et al., 1976).

264

M.O. Cuthbert et al. / Journal of Hydrology 391 (2010) 263–276

In addition to fractures, the literature points towards other factors in controlling the hydraulic conductivity of tills such as source materials, the depositional mode and post-depositional processes (Haldorsen and Kruger, 1990). Burrowing and root-hole development may be other contributing factors (Klint and Gravesen, 1999). Several studies have found that weathered till may be several orders of magnitude more permeable than equivalent unweathered till (Bonell, 1972; Cravens and Ruedisili, 1987; D’Astous et al., 1989; Klinck et al., 1996, 1997). This more permeable weathered zone may extend as far as 18 m below ground surface in the Prairie region of Canada (Hendry, 1988) although depths of <8 m are more commonly reported in other areas (Cravens and Ruedisili, 1987; Ruland et al., 1991; McKay et al., 1993; Klinck et al., 1996, 1997). Despite intensive investigation of till geology as described in the brief overview of existing literature above, previous research into recharge through till has tended to fall into two groups. First, studies have focussed on saturated zone flows through superficial materials, with relatively little attention being paid to unsaturated zone processes. The popularity of flooding tracer experiments is an example of this (e.g. Nilsson et al., 2001; McKay et al., 1999) but from this type of experiment it is impossible to discern the likely flow pathways and timescales under natural gradients in the unsaturated zone. Second, unsaturated zone processes are often studied and modelled in detail by soil scientists but with little attention given to the fate of water once it has left the shallow soil zone (Villholth et al., 1998a; Villholth and Jensen, 1998b; Jansson and Jansson, 2003; Gerke and Kohne, 2004). Hence, although there has been much research into the flow mechanisms through tills, very few studies have given consideration to the whole system from the ground surface to the aquifer. The research presented in this paper presents such an integrated approach, by developing a detailed understanding of the hydraulic processes for a fieldsite in Shropshire, UK. Results of hydraulic monitoring, experimentation and sampling of both the saturated and unsaturated zones are presented, used to derive conceptual models, and then tested using numerical modelling. 2. Fieldsite 2.1. Locality The experimental site is located in a rural area of Shropshire, UK (Fig. 1) within the catchment of the Potford Brook (catchment area 22.5 km2), a tributary of the River Tern (catchment area 880 km2), itself a tributary of the River Severn. It is a lowland area of gently

Fig. 1. Location and layout of the fieldsite (C – boreholes, T – tensiometers, P – piezometers).

undulating terrain underlain predominantly by a Permo-Triassic sandstone aquifer of regional significance. Variable superficial deposits cover much of the catchment comprising glacial till, glaciofluvial deposits and valley alluvium. Soils generally reflect the nature of the underlying superficial geology. Long term (1970– 1999) annual average rainfall and potential evapotranspiration for the Potford Brook catchment are around 670 mm and 597 mm respectively (based on UK Meteorological Office data as described by Cuthbert (2006)). Based on regional groundwater modelling studies the recharge to the aquifer in outcrop/outwash sand covered areas and till covered areas is estimated to be around 240 mm/a and 90 mm/a respectively in the Potford Brook catchment between 1980 and 2000 (Streetly and Shepley, 2005). However there is significant uncertainty in recharge distribution and hydraulic processes at the field scale which, in part, has driven the research on which this paper is based (Cuthbert et al., 2009; Thatcher, 2009). The site is located near Wood Farm (national grid reference (NGR) SJ 6215 2302) and is underlain predominantly by till overlying the sandstone at around 6 m below ground level (bgl). The site lies in the corner of a field used for grazing but until 2001 had a mixed arable use. The layout of the site instrumentation is shown in Fig. 1. A hedgerow forms the western boundary of the site, to the west of which is a ditch and then a tarmac road. The northern boundary is lined with oak trees, the branches of which overhang the site by around 10 m in places. 2.2. Instrumentation 2.2.1. Boreholes Two boreholes (C1 and C2) and two piezometers (P1 and P2) were continuously cored to depths up to 6.6 m below ground level at 90 mm ID (internal diameter) using a sonic drilling rig. Core from all four holes was logged according to British Standard (BS5930) and sub-sampled for hydraulic testing. 2.2.2. Piezometers Data from an existing piezometer at the site (P3) were available: this piezometer is screened between 7.0 and 8.0 m bgl within the Permo-Triassic aquifer. Boreholes C1 and C2 were backfilled with bentonite after coring but piezometer installations were carried out in P1 and P2 in June 2004 with 40 cm long monitoring sections comprising a pre-fabricated filter pack around 25 mm ID plastic casing with 0.3 mm slots. The installations were sealed above the filter pack with bentonite, the top 30 cm being filled with a cement grout, and then fitted with covers at ground level. Piezometers P1 and P2, screened within the till at 3.3–3.7 m bgl and 4.6–5.0 m bgl respectively, were fitted with custom-made electronics including differential pressure transducers (vented to atmosphere) to enable data to be logged automatically at regular intervals. Water levels were recorded manually every few days from June to October 2004 and were then automatically logged from November 2004 to September 2005 at 5 or 10 min intervals. 2.2.3. Tensiometers and thermocouples Tensiometers were built in-house at the University of Birmingham, incorporating the use of pressure transducers to allow the option of datalogging as described by Greswell et al. (2009). The tensiometer design comprised a 21 mm OD unglazed porous ceramic cup with a bubbling pressure of 1.5 bar and a pore size of around 2 lm, attached to a length of 20 mm outside diameter (OD), 17 mm inside diameter (ID) ABS and fitted with a circuit board including a differential pressure transducer in order to allow data logging. Small Luer valves were fitted at the top end of the tube just above ground level and used to fill the instrument with de-aired water via a syringe. The short length of tensiometer

M.O. Cuthbert et al. / Journal of Hydrology 391 (2010) 263–276

265

protruding from the ground surface (<30 cm) was covered by a white PVC cover. To install the tensiometers, a 20 mm screw auger was used to auger to the desired depth. The complete tensiometer was then gently pushed into the slightly undersized hole to ensure a snug fit between the ceramic cup and the soil. The installations were carried out in August 2004. The depth of the ceramic cup of each tensiometer is given in Table 1. Each tensiometer had a thermocouple attached to its ABS tube just above the porous cup for monitoring the ground temperature. 2.2.4. Data quality Owing to the potential for noise in pressure transducer readings produced by variations in solar radiation (Cain et al., 2004) all the cables running from the monitoring devices to the loggers were buried 10 cm beneath the ground. The piezometers and tensiometers were recorded manually every few days from August to October 2004 and were then automatically logged using four-channel, HOBO U12 dataloggers from November 2004 to September 2005 at 5 or 10 min intervals. The piezometer transducer/data logger equipment gave results in excellent agreement with manual dips taken sporadically through the monitored period. Tensiometers T2 and T4 unfortunately malfunctioned within 1 month of installation due to electronic failure. The cause of this problem has not been discovered. The other tensiometers generally performed well although pressure effects due to freezing of the water column occurred at times correlated with sub-zero ground temperatures measured by onsite thermocouples. 3. Results 3.1. Geology The geology is known from the four cored holes and the preexisting piezometer (P3), from shallow augering and the test pit described below and, indirectly, from resistivity surveys (Cuthbert, 2006; Cuthbert et al., 2009). Topsoil is encountered from the surface to between 0.3 and 0.4 m bgl, and comprises loose soft dark-brown variably-sandy clayey silt with rootlets and organic material. Below the topsoil reddish-brown poorly laminated silty sandy clay, interpreted as a solifluxion deposit, is present in several of the boreholes at the site down to a maximum depth of 0.7 m bgl. In one location (C2), a 0.15 m thick layer of fine sand was found between the topsoil and the solifluxion deposit. These deposits overlie a layer of till to a depth of around 6.3 m bgl thought to be associated with Late Devensian glaciation. Underlying the till are red to red-brown, fine to medium grained, pebble-free, dune-bedded sandstones of aeolian origin and Permo-Triassic age that form the major regional aquifer. A schematic cross section is shown in Fig. 2 to indicate the likely geometry of the different lithologies. Also shown are the piezometer screen intervals (P1–P3) and tensiometers (T1– T5). The depth of each installation is accurate but the horizontal position is only approximately projected onto the section.

Table 1 Tensiometer installation depths. Tensiometer

Depth of base of ceramic cup (cm bgl)

Monitored unit

T1 T2 T3 T4 T5

29 57 88 124 140

Soil Till Till Till Till

Fig. 2. Schematic cross section of the fieldsite.

The till encountered in the field investigations was often highly over-consolidated and is interpreted as being laid down during ice advance by sub-glacial lodgement processes. Typically in the field it was described as a firm to very stiff reddish brown slightly gravely slightly sandy clay with matrix supported clasts up to 20 mm and, more rarely, 50 mm in diameter. A range of laboratory tests was carried out on core materials from the site as described by Cuthbert (2006). The particle size distribution (PSD) data derived from sieving indicate that the till comprises 30–50 w.% sand and 0–10% gravel, with the remainder made up of silt and clay. PSD data (derived using an automated method with MasterSizer 200, Malvern Instruments) for the <2 mm fraction of 12 samples confirm that the till is poorly sorted and the size distribution is generally unimodal. It was found that the proportion of clay size particles within the till ranges between 11% and 15% of the fine fraction (silt and clay). Thin sections of till samples further confirm that the samples are poorly sorted and furthermore have angular to sub-rounded clasts dominated by quartz grains (feldspar and rock fragments are less common) set within a matrix of fines. No preferential grainsize orientation or layering is evident. The results of limited semi-quantitative X-ray diffraction analysis (Cuthbert, 2006) are consistent with the thin section evidence. Quartz and clay minerals make up over 80% of the rock mass with the clay mineral assemblage being dominated by illite but also including a significant proportion of the swelling clay montmorillonite. The till has approximate values of porosity in the range 0.23– 0.32 with dry bulk density ranging from 1600 to 2100 kg/m3 (derived using ratios of saturated and oven dried moisture contents and volumes as described by Cuthbert (2006)). Fracturing was not evident from any undisturbed samples or cores with the exception of one large semi-vertical fracture found in borehole P1 between 1.5 and 2.0 m bgl. The fracture seems to be planar and showed blue discolouration (in contrast to the red-brown host rock) and the remains of organic matter. The original fracture may have been exploited by the root network of later surface vegetation. The root activity of large trees is likely to have affected the upper few meters of much of the catchment in the past before the clearance of trees for agriculture. However no palaeoroot holes were evident in core samples (White et al., 2008).

3.2. Hydraulic properties Laboratory hydraulic conductivity values, known from the results of falling head permeameter tests for 19 vertically-oriented till samples taken from cores, range from 1  1010 m/s–2  106 m/s

266

M.O. Cuthbert et al. / Journal of Hydrology 391 (2010) 263–276

(1  105 m/d–0.2 m/d). Four samples also tested for horizontal hydraulic conductivity spanned the same range. However, most values plot in the range 1  1010–1  109 m/s with only two samples having significantly higher vertical hydraulic conductivity. No discontinuities were apparent from visual inspection to explain the increased permeability of the two more permeable samples. A limited number of hydraulic tests were also carried out at the site as described by Cuthbert (2006). These tests were kept to a minimum so as to limit the disturbance of the natural state of the site for hydraulic observation during the period of the study. The results of double ring infiltrometer tests suggest that, under ponded conditions, infiltration rates as high as 4.2  105 m/s (3.6 m/d) are possible and that the saturated hydraulic conductivity of the soil may be in the range 8.1  107–3.5  106 m/s (0.07– 0.3 m/d). Results of falling head piezometer tests carried out on both piezometers screened within the till yielded estimated hydraulic conductivity values of 3  1010 m/s and 2  107 m/s for P1 and P2 respectively. The result for P1 is very close to the median value of till hydraulic conductivity measured in the laboratory described above. The result for the test on P2 is three orders of magnitude greater than that for P1 but lies within the range of outlying values measured in the laboratory. 3.3. Hydraulic responses and evidence of preferential flow 3.3.1. Short term temporal variations in pressure, barometric efficiencies and specific storage Time series of pressure head fluctuations recorded by tensiometers T1, T3, and T5 (under saturated conditions) and in piezometers P1–P3 are shown in Fig. 3 for the period 28/12/04–22/1/05. Also shown are relative atmospheric pressure fluctuations as recorded at Bowling Green climate station situated approximately 2.5 km north of the fieldsite (data supplied by the UK Natural Environment Research Council’s Lowland Catchment Research (LOCAR) Data Centre). Short timescale variations in pressure head are evident which correlate strongly with changes in atmospheric pressure. An inverse relationship between atmospheric pressure and pressure head, and vice versa, is to be expected for horizons monitored within confined or low permeability formations (Jacob,

1940; Todd, 1959). Owing to the low permeability of the till matrix, and the relatively large diameter of the piezometer tubes P1 and P2, the instantaneous barometric response seen at these locations is unexpected, given Hvorslev (1951) calculations carried out using data from the falling head experiments. These responses can only reasonably be explained by the movement of water into, and out of, storage within the gravel pack surrounding the piezometer screen and within any fractures which intersect the borehole. This may provide the necessary volume of sufficiently permeable and compressible material to account for the rapid volume change needed to cause the observed change in water level within the piezometers. The barometric efficiencies for till (T1, T3, and T5) and for the sandstone aquifer (P3) have been calculated using the graphical method of Gonthier (2007) and are shown in Table 2. Values of specific storage were also calculated, assuming the porosity values given in Table 2, using well known formulae (for example described by Domenico and Schwartz (1998), p. 129). Since the barometric effects in P1 and P2 do not reflect the nature of the formation in which they are situated but the installation configuration, they have been omitted from the analysis. The values are relatively large implying reasonably low values of compressibility for the till, consistent with the slow drilling rates achieved by the sonic rig used at the site and the interpretation of it being formed by sub-glacial lodgement. Derived values for specific storage are at the high end of those reported in the literature for till (Ostendorf et al., 2004; Keller et al., 1989). The derived specific storage value for the Permo-Triassic sandstone is consistent

Table 2 Barometric efficiency and derived values of specific storage. Location

Formation

Barometric efficiency

Assumed porosity

Specific storage (m1)

T1 T3 T5 P3

Lodgement till Lodgement till Lodgement till Permo-Triassic sandstone

0.21 0.34 0.64 0.43

0.28 0.28 0.28 0.24

6.3E-06 3.9E-06 2.1E-06 2.7E-06

-10 -20

T1

Pressure (cm H2O)

-30

T3

-40 T5

-50 -60

Atmospheric

-70 -80

P2

-90

P1

-100

P3

22/01/05

21/01/05

20/01/05

19/01/05

18/01/05

17/01/05

16/01/05

15/01/05

14/01/05

13/01/05

12/01/05

11/01/05

10/01/05

09/01/05

08/01/05

07/01/05

06/01/05

05/01/05

04/01/05

03/01/05

02/01/05

01/01/05

31/12/04

30/12/04

29/12/04

28/12/04

-110

Fig. 3. Averaged hourly pressure variations monitored by the fieldsite instruments in comparison to relative atmospheric pressure fluctuations measured at Bowling Green climate station (LOCAR Data Centre, 2005).

267

M.O. Cuthbert et al. / Journal of Hydrology 391 (2010) 263–276

an infinite homogeneous medium (e.g. Keller et al., 1989), suggests that pressure responses through the sandstone confined by till should take less than a day to propagate to the site for realistic values of transmissivity and storage coefficient. Thus, it is likely that the water level changes seen in P3 reflect an aggregated response of the sandstone aquifer across the catchment in addition to any direct effects of recharge through the till at this location.

with results from other studies in the area (Streetly and Shepley, 2005). 3.3.2. Saturated zone temporal head variations Time series of groundwater levels in piezometers P1–P3 are shown in Fig. 4 for the entire monitored period. Manual readings are plotted as points and logger data presented as continuous lines. As can be seen, the small timescale and relatively small magnitude variations in water level due to barometric effects are superimposed on a larger seasonal trend. Groundwater levels for all three piezometers are at a minimum during September and begin to recover during October. Water levels in P2 and P3 have a maximum in January/February with a subsidiary peak in April in contrast to P1 in which a plateau of high water levels persists between January and May before summer recession begins. The magnitude of the seasonal fluctuations in water level varies greatly between the piezometers with P2 and P3 having a variation of just 0.4 m in comparison to 1.4 m for P1. The greater seasonal fluctuations seen in P1 are likely to be due to the proximity of the monitored interval to the water table. The pressure response is then attenuated with depth as would be expected within low permeability till, with smaller fluctuations recorded by P2 situated deeper within the till. It is noted that after the falling head test had been carried out on P2 the responses in this piezometer became nearly indistinguishable from those recorded in P3, i.e. the head difference recorded between these two piezometers fell significantly. These observations imply that the falling head test caused an increase in the hydraulic connectivity between the two monitored intervals. This is discussed further below. Also shown in Fig. 4 is the groundwater levels recorded within the Permo-Triassic sandstone at Drakely Heath, approximately 1.4 km away from the site, adjusted to a different datum for comparison with P3. This indicates that the fluctuation within the sandstone below the till at the fieldsite is very similar to seasonal variations in a nearby area of outcrop sandstone. This is unsurprising given that the site is only 500–1000 m from areas of sandstone outcrop. Using an analytical model which simulates the propagation of a sinusoidally varying head boundary condition through

3.3.3. Temporal variations in hydraulic gradients Time series of average daily total heads (relative to ground level) for the fieldsite piezometers and tensiometers are shown in Fig. 5 for the whole monitored period. A persistent downward hydraulic gradient is evident throughout the year between the water table and the groundwater level in the Permo-Triassic sandstone of approximately 0.4 during the summer and nearly 0.5 during the winter. There is some uncertainty as to the exact position of the summer water table but it can be inferred from the data from T5 to P1 to be at approximately 1.7 m bgl. As noted above, water levels in P2 closely follow those in P3 suggesting a good degree of hydraulic connection between the two piezometers. The absence of a significant time lag between these piezometers suggests that the till bulk hydraulic conductivity is significantly greater than that of the till matrix in this location. Fig. 6 shows daily average values of tensiometer pressure head for the whole monitored period. During the winter period between early November and early May the pressure head at each tensiometer was always positive with the exception of periods of several days at a time when tensions of up to 30 cm were recorded in T1. A downward hydraulic gradient was present throughout the upper 1.4 m of the profile during this time with values of around 0.2–0.6 present between T1 and T3 and T3 and T5. Following large rainfall events T1 took as much as 1 day to respond with initial pressure increases at T3 and T5 at around 1.5 and 2 days respectively. After the initial response the pressure increase was relatively sudden for T1 becoming more and more gradual for T3 and T5. In contrast to the saturated pressure responses of the tensiometers to recharge, the hydrograph for P1 is very smooth. Although some attenuation of the recharge signal will occur due to the

0 -50 -100 Falling head tests

-350

-150 -200

-370 -250 -300 -390 Pumping for water quality sampling

-350 -400

-410 Pumping test in sandstone

Groundwater Level for P1 (cm asd)

Groundwater Level for Drakely Heath, P2 & P3 (cm asd)

-330

-450 -500

Drakley Heath

P2

P2

P3

P1

Sep-05

Aug-05

Jul-05

Jun-05

May-05

Apr-05

Mar-05

Feb-05

Jan-05

Dec-04

Nov-04

Oct-04

Sep-04

Aug-04

Jul-04

Jun-04

May-04

Apr-04

Mar-04

Feb-04

Jan-04

-430

P1

Fig. 4. Time series of groundwater levels within the till at the fieldsite and within the Permo-Triassic sandstone aquifer at Drakely Heath, around 1.4 km from the fieldsite.

268

M.O. Cuthbert et al. / Journal of Hydrology 391 (2010) 263–276

40

0 T1 -100

35 T3

P1

-200

T1

T5 30

-300 P2 25

P3

-400 -500

20

-600

T3

15

Rainfall (mm/d)

Total Head (cm asd)

T5

-700 10 -800 Rainfall

-900

5

0

T1

T2

T3

T5

P1

P2

T1

T3

T5

P1

P2

Rainfall

Sep-05

Aug-05

Jul-05

Jun-05

May-05

Apr-05

Mar-05

Feb-05

Jan-05

Dec-04

Nov-04

Oct-04

Sep-04

Jul-04

Aug-04

-1000

P3

Fig. 5. Time series of average daily hydraulic head and rainfall (recorded at Bowling Green).

1000

40

35 800

600 25

400

20

15

Rainfall (mm)

Tension (cm)

30

200 10 T1 T3

5

T5

NO DATA AVAILABLE

T2

T3

T5

T1

T3

T5

Sep-05

Jul-05

Jun-05

May-05

Apr-05

0 Mar-05

Feb-05

Dec-04

Nov-04

Oct-04

Sep-04

Aug-04

Jul-04

T1

Jan-05

Rainfall

-200

Aug-05

0

Rainfall

Fig. 6. Time series of average daily tension, cm above site datum (asd).

specific storage of the till, the smoothing is likely to be due mainly to the piezometer configuration and is consistent with the results of the falling head test carried out on this piezometer described above indicating slow equilibration between the piezometer and the surrounding formation. 3.3.4. The Hydraulic argument for preferential flow During May 2005, as evapotranspiration rates began to become more significant, the tension in the topsoil zone as recorded by T1 increased very quickly and was out of range at 8.8 m by mid June.

Tensions in T3 and T5 also began to increase during this time but at a slower rate reaching a maximum of 600 and 400 cm respectively by mid-August. The point at which transition is made to an upward hydraulic gradient between T1 and T3 coincides with the start of the summer recession in piezometer P1. During the late summer in 2004 an upward hydraulic gradient was in place between P1 and T5 and between T5 and T3. However summer rainfall had been sufficient to keep the topsoil relatively moist and a downward gradient was present between T1 and T3. By early October pressure heads in T3, T5, and P1 increased quickly

269

M.O. Cuthbert et al. / Journal of Hydrology 391 (2010) 263–276

until a downward hydraulic gradient persisted throughout the profile by mid November. The recovery in water level in P1 occurred while tensions of several hundred cm were present in T3 and T5, i.e. over much of the upper 1.4 m of the profile. As shown by Cuthbert (2006) using a variably saturated 1-dimensional model, in which bulk permeability was set equal to the till matrix permeability, it is not possible for significant amounts of moisture to flow to depth through the till while these tensions exist in the upper profile. The wetting of the profile between the topsoil down to a depth of 1.4 m bgl (T5) in 1 month is also problematic if vertical matrixonly flow is invoked as the flow mechanism. Furthermore, the pressure increase in P1 occurs earlier and is steeper than that observed in P3 suggesting that the recovery higher up in the profile cannot simply be due to a rise in the piezometric level of the sandstone. These observations strongly suggest the presence of preferential pathways enabling the relatively rapid wetting of the profile. Although this paper focuses on evidence from the hydraulic measurements, Cuthbert (2006) has also shown that porewater chemistry data for this fieldsite are also suggestive of preferential flow through the till. In particular, on the basis of nitrate and tritium concentrations, it was shown that modern water is present in the profile as deep as 4.4 m and 5.9 m respectively which would not be possible under natural hydraulic gradients assuming flow through the low permeability till matrix alone. 3.4. Geological evidence for fracturing In order to look for the presence of possible features contributing to such preferential flow through the till at the fieldsite a test pit, approximately 2  2  2 m in size, was excavated using a 1.5 tonne excavator. A slightly undersized timber frame was built to sit inside it to provide support in case of collapse. Once completed, three sides of the pit were worked by hand with a spade and trowel to remove the smeared surface created by the digger. Since it was vertical features that were primarily expected to be found, material was removed using the tools with a horizontal motion in order to avoid creating artificial vertical features. Since the excavator was working at the limit of its reach the fourth side became very uneven and thus required too much work by hand to create a mappable surface. Once prepared, a 3 by 3 grid of 60 cm squares was constructed on each of the three surfaces (to the north, south and west of the pit) using string and small wooden stakes. Then each face was photographed and the main features were sketched on a scale drawing. A diagram of the west face is shown in Fig. 7 which has features representative of the other faces of the pit. A number of significant Features (1–5) are highlighted and corresponding photographs of each feature are shown in Figs. 8–10. A layer of topsoil was present to a depth of 30–40 cm bgl with a very sharp transition to the underlying till or sand deposits. Plant roots (such as shown in Feature 2, Figs. 7 and 8) were observed to around 1 m bgl and decaying root material was present within fractures to as deep as 1.9 m bgl. In most places clayey till or solifluxion deposits were present below the topsoil and near vertical open fractures were seen at a spacing of 10–20 cm to a depth of around 70 cm bgl (Feature 3, Figs. 7 and 9). In the upper part of the north face lumps of till had fallen away from the face revealing fracture faces. These faces were often coated with topsoil indicating the downward movement of material into the fractures from above. In the corner of the west and south faces a large weathered boulder of sandstone had been incorporated into the till just underlying the topsoil layer (Feature 1, Figs. 7 and 8). This comprised fine to medium grained sand varying in colour from grey to orange brown indicating different states of weathering. At the

Fig. 7. Sketch of the west face of the test pit.

2

1 10 cm

Fig. 8. Photograph of Features 1 and 2 shown in Fig. 7.

boundary between the sand and the finer grained clayey till beneath, a thin rim of rust coloured material was sometimes present in the till. This iron oxide precipitation is most likely due to a rise in pH as infiltrating water comes into contact with the till which contains carbonate (the dominant clast type is mudstone which shows a vigorous reaction with dilute HCl). In three locations on the south and west faces, fractures up to 2 cm wide were present to around 1.7 m bgl and infilled with red sandy material (Feature 4, Figs. 7 and 10). The source of the infill appeared to be the weathered sandstone clast incorporated into the till higher in the profile. Below 60 cm depth, near vertical fractures approximately 1–2 mm wide infilled with very fine grained light grey material were common in all the faces of the test pit (Feature 5, Figs. 7 and 10). These fractures had an average spacing of around 30 cm to approximately 1.4 m bgl and those persisting to at least 2 m bgl had a spacing of approximately 60 cm. The material

270

M.O. Cuthbert et al. / Journal of Hydrology 391 (2010) 263–276

3

10 cm

Fig. 9. Photograph of Feature 3 shown in Fig. 7.

ing of the fractures with carbonate are possible but, given the spatial relationships between infill and clast lithology, a mechanism of fracture infilling due to the downward transport of weathered material seems most likely in this case. The base of the pit was excavated on its north side to enable the 3-dimensional pattern of fracturing to be observed. A perspective drawing of the fractures in this location is shown in Fig. 11 with a photograph for comparison shown in Fig. 12. It is apparent that the fractures form a network around columnar blocks of till. The pattern of decreasing intensity of fracturing with depth is consistent with the fractures being formed in response to stress changes at the ground surface. In other areas, such stresses have been shown to have been caused by desiccation in response to long dry periods in the past perhaps combined with lower regional groundwater levels, due to freeze thaw cycles and/or due to stress relief due to the removal of ice loading as the glacier retreated (Grisak et al., 1976; McKay and Fredericia, 1995; Klint and Gravesen, 1999).

60 cm

N9 N8 4

5

Fig. 11. Perspective sketch of fractures in north face and base of test pit.

60 cm

30 cm Fig. 10. Photograph of Features 4 and 5 shown in Fig. 7.

filling these fractures was, in the field, indistinguishable from weathered calcareous mudstone which is the dominant clast found in the till within the test pit. This suggests that the fractures have at one time been hydraulically active, transporting this material through the fractures from above. This explanatory mechanism for the infilling is consistent with the observation that the type of fracture infill is closely associated with the types of clasts that the fractures intersect as illustrated by the sand filled fracture shown as Feature 4 in Figs. 7 and 10. This fracture extends from a large weathered sandstone clast situated higher in the profile and then intersects a soft mudstone clast at around 1.6 m bgl. The fracture then continues vertically downwards from this clast but is infilled not with sand but with material indistinguishable (to the eye) from the mudstone clast. Other mechanisms for infill-

N9

Fig. 12. Photograph showing the connectivity of fractures in north face and base of test pit – fractures lie between the dashed lines.

M.O. Cuthbert et al. / Journal of Hydrology 391 (2010) 263–276

McKay and Fredericia (1995) suggest that fractures started due to desiccation in the upper part of a clay at a fieldsite in Ontario, Canada, could extend downwards well below the depth of the water table. This assertion is based on a linear elastic fracture formation model which simulates the formation of vertical fractures in clay (Mase et al., 1990). It may be possible therefore that, if due to desiccation, such fractures penetrate the entire 6 m deep till unit at the fieldsite described in the current study. However, based on the observed doubling of fracture spacing with a corresponding doubling of depth, by 6 m bgl the spacing of the fractures is likely to be 1–2 m. Stress relief due to glacial unloading may have also led to the formation of deeper fractures. A schematic 3-dimensional sketch of the fracture distribution with depth is shown in Fig. 13. The infill of material derived from clasts within the till matrix is evidence of historic flow within these features. However since the fractures are now predominantly infilled with fine-grained material it is not clear how permeable they remain to the present day. During the excavation of the base of the pit it was noticed that several areas became wet with time. It is noted that the groundwater level recorded by piezometer P1 was above the base of the pit at the time of excavation at 1.4 m bgl. When a fracture in the base of the pit on its north side was freshly excavated seepage was evident

0.3 m 3m

Fig. 13. Sketch of fracture distribution in 3-dimensions.

10 cm

Seepage

Fig. 14. Photograph showing seepage from fracture in base of test pit.

271

from the fracture as shown in Fig. 14. This is strong evidence for the hydraulic significance of these fractures in the present day. With further investigation it was apparent that seepage from the fractures was discontinuous and localised in certain areas indicating that flow within the fracture network may be channelised. As was noted above, the occurrence of the piezometer test carried out on P2 seemed to increase the hydraulic connectivity between this piezometer and the underlying sandstone. Given the relatively small volumes of rain water used for this test, this is more likely to have been due to a physical particle flushing effect rather than enhancement of the permeability of a connecting fracture by dissolution of the carbonate infill. It is also evidence that fracturing occurs within the till to a depth of at least 5 m. Assuming an unfilled fracture width of 2 mm and range of fracture spacings from 0.3 m to 2 m, for a hexagonal arrangement of matrix blocks the total fracture porosity would be 0.2–1.3%. Since the fractures are infilled and flow is channelised, the effective porosity of the till will be much lower than this range in reality. 3.5. Conceptual model of hydraulic processes The data and analysis presented so far enable a picture of the hydraulic processes contributing to recharge at the fieldsite to be proposed. Infiltration into the shallow soil zone seems to keep pace with all but the heaviest rainfall events. Owing to the uneven ground surface some local temporary ponding does occur after the heaviest rainfall events in several places across the site and may take a few days to evaporate or infiltrate. Unless the till profile has become completely saturated, the generation of significant amounts of runoff at the site is unlikely due to the flat topography, the moderate infiltration capacity of the soil and the storage provided by the uneven ground surface. Preferential infiltration through closely spaced fractures is likely to be the primary mechanism during dry periods but the amount of water reaching the water table will depend on how much water is taken up by the matrix during such fracture flow. During the summer months, transpirative demand for grass growth leads to tensions of several hundred cm building up over the upper part of the profile to depths of more than 1.4 m bgl. This drying occurs progressively from the top of the profile downwards and coincides with a recession in piezometric pressure deeper in the till. It is likely that this recession is due to a combination of evaporative demand from above, slow vertical drainage to the sandstone below, and the falling piezometric level in the underlying sandstone aquifer. Pressure increases in the deeper till imply that recharge to the till water table can occur during the summer period in response to persistent heavy rainfall. It is hypothesised that this response is made possible by fracture flow through the till unsaturated zone. The relatively large variation in pressures in the saturated zone of the till is consistent with a low specific yield for the till. During late summer the wetting of the profile begins with the initial re-saturation of the topsoil and underlying sandy horizons. From this moderately permeable wet upper layer it is thought that water flows through partially saturated fractures in the till enabling relatively rapid recharge to the water table with a much slower wetting of the till matrix blocks. As time progresses greater equilibrium between the till fractures and matrix is achieved as the entire till profile becomes saturated. It is thought that limited lateral flow occurs through the till since the relatively flat topography limits the build up of significant lateral head gradients and, as a result, flow to a field boundary ditch (which primarily collects runoff from the adjacent tarmac road) adjacent to the western site boundary is minimal. Thus water infiltrating into the till is either drawn back to the atmosphere by

272

M.O. Cuthbert et al. / Journal of Hydrology 391 (2010) 263–276

evapotranspiration or flows vertically to recharge the underlying sandstone. However, it is noted that these observations were made over a drier than average year and it may be that runoff and lateral flow to the field ditches occurs to a greater extent in wetter years. The available field evidence suggests that the bulk hydraulic conductivity of the till is significantly greater than that of the till matrix, though decreases with depth. 4. Numerical modelling 4.1. Model assumptions and mathematical formulation In order to test the conceptual model of the fieldsite developed above, and to attempt to quantify the likely magnitude and timing of the recharge flux to the Permo-Triassic sandstone through the till, numerical modelling was undertaken. Given the complex conceptual model described above, a number of assumptions and simplifications needed to be made in order to produce a tractable model of the site. These are now discussed in turn: 1. Since the water table is within the relatively permeable upper till/lower soil zone during the winter, and as the lateral variation in the water table is likely to be relatively low during this period, the hydraulic gradient between the water table and the base of the till (i.e. the piezometric surface of the confined sandstone aquifer) may be used to estimate the bulk flux through the till. 2. To explore the possibility of lateral flow to the ditch adjacent to the site from the till, a simple steady state analytical model has been used for an unconfined aquifer with fully penetrating drains under constant recharge (Todd, 1959). Using an assumed recharge to the till of 70 mm/a (1.59  109 m/s), and a conservatively high hydraulic conductivity of 1  108 m/s, for drains placed at 1 m bgl and assuming all flow occurs through the upper 1.5 m of the profile, only recharge within 1.5 m of the drain will be routed to the drain. This simple model indicates that lateral flows to the drain are not likely to be significant across the site and that an assumption of 1-dimensional vertical flow though the till within the experimental area of the site is reasonable. 3. During the simulated winter period, rainfall intensity never exceeded the measured soil permeability, and so it was decided that runoff due to infiltration excess could be omitted from the model. 4. It is assumed that the potential evapotranspiration (PE) for the site is adequately known and that actual evapotranspiration (AE) is equal to PE for the modelled winter period. This is a reasonable assumption in winter conditions given that the maximum tension recorded by T1 located within the soil zone was around 30 cm, well below the literature values of tension beyond which grass AE will begin reducing from the potential rate (Feddes et al., 1978). Monthly MORECS (UK Meteorological Office data: Hough and Jones, 1997) PE values were converted to a daily time series for input to the model by assuming constant daily values through the month. 5. Continuous daily rainfall data recorded by a ground level raingauge, situated at Bowling Green climate station approximately 2.5 km north of the site, were available for the modelled period. Data from a network of standard raingauges (set 300 mm above ground level) for the wider area indicate a 2– 3% variation in long term average (1978–2001) rainfall can be expected across the Potford Brook catchment (Cuthbert, 2006). However, the situation of the fieldsite in the corner of a field with adjacent hedgerows and overhanging oak trees is likely to substantially decrease the rainfall due to interception losses. Recent literature on this subject (including research in

the UK) suggests that leafless interception may reduce throughfall by 10–30% at distances up to several meters from a forest edge (Herbst et al., 2007) and by around 20% within 5–10 m of a typical hedgerow (Herbst et al., 2006; Ghazavi et al., 2008). Based on these data it seems reasonable to assume that the rainfall at the fieldsite may be around 80% of that recorded by the nearby ground level gauge, but that significant uncertainty is associated with this input to the model. This is consistent with time series electrical resistance tomography results for the site which indicate that parts of the site closest to the field boundaries take much longer to re-wet after the summer than areas further away (Cuthbert, 2006). Although this may also, in part, be due to enhanced evapotranspiration, it is thought this is primarily indicative of a lower incident rainfall in areas closer to the vegetated site boundaries. 6. Hydraulic disequilibrium between the matrix and fracture regions of the till is likely to be very significant during partially saturated conditions. However, assuming that the till does become completely saturated during the winter, the pressure changes in the matrix in response to those in the fractures, and vice versa, are likely to be very quick. It can be shown using an analytical model (e.g. Keller et al., 1989) that, assuming a hydraulic conductivity of 1.4  1010 m/s and a specific storage of 4  106 m1, daily periodic pressure changes will propagate across till matrix blocks of several tens of cm within hours. Thus, the pore water pressures and hydraulic gradients observed using tensiometers T1 and T3 can be used to calculate a time series for the position of the water table against which to refine the model. 7. It is assumed that, during the winter, flow predominantly occurs through the fracture network, that fracture apertures are constant but that the fracture spacing doubles with depth. Since the fractures are infilled with sediment it is also assumed that flow is Darcian, Thus

aK a z aSa SðzÞ ¼ z

KðzÞ ¼

ð1Þ ð2Þ

where z = depth below ground surface, and Ka and Sa are the hydraulic conductivity and specific yield respectively at a representative depth, a, at which values of K and S are defined. Based on the assumptions above and the conceptual simplifications embedded in these assumptions, a model was constructed to simulate the position of the water table during the winter months (December 2004–March 2005). If the mean vertical K in fractures is estimated as a harmonic mean between the water table and the base of the till then, using Darcy’s Law:

Qt ¼

ðWT t  GWLt Þ ð2aK a Þ : ðWT t  DÞ ðWT t þ DÞ

ð3Þ

where Qt = vertical flow (L/T) through the base of the till (i.e. recharge to the underlying sandstone) at time t, WTt = water table elevation (bgl) at time t, GWLt = groundwater level in the sandstone aquifer (bgl) at time t, D = total thickness of till. Letting the unit time interval for model calculations be equal to Dt, the inflow–outflow imbalance for a given time step, DQ, is given by:

DQ ¼ Pt  AEt  Q t

ð4Þ

where Pt and AEt are the precipitation and evapotranspiration at time t. If the mean vertical S in fractures is estimated as an arithmetic mean over the interval of head changes then it can be shown that:

273

M.O. Cuthbert et al. / Journal of Hydrology 391 (2010) 263–276

input in order to reduce the evaporation from the fractured system resulting in much better model fits. Results are shown in Fig. 15 for three ‘best fit’ models alongside a table of assigned parameters for each run. The models reproduce the timing and magnitude of water table responses relatively well, enabling us to derive useful inferences regarding the magnitude and timing of winter recharge to the underlying aquifer. It is clear from the results that equally ‘fit’ models can be produced using a range of rainfall inputs yielding estimated values for the bulk K of the till in the range 8  1010–9  108 ms1, i.e. approximately 1–3 orders of magnitude greater than that of the till matrix. The depth averaged bulk K values of the till used in the model were in the range 1.6–5.8  109 ms1. For the 4 month modelled period, the recharge to the underlying sandstone is in the range 5.4 ± 3.1 mm/month and relatively constant throughout the period (standard deviation of 3–5%). It is beyond the scope of this paper to model the complexities of the matrix/fracture interactions during warmer periods when evapotranspirative demand dominates the near surface water balance. However, the water table is in the range 1.4–2.0 m bgl for the summer period. Using Eq. (3) and parameters derived from the model runs, the minimum range of summer recharge to the sandstone is likely to be 2.75 ± 1.5 mm/month, approximately 50% of the winter values. The model results suggest values for the Sy of the soil layer to be in the range 5–12%. The fracture porosity of the more open upper part of the till profile may be as high as a few percent. This is consistent with 4 mm wide open fractures at a spacing of 0.2 m assuming hexagonal matrix blocks. At 6 m bgl values may be as low as 0.1% based on the relationship defined in Eq. (2). A simple calculation for groundwater velocities combining average flux and fracture porosities assuming an average water table elevation gives values in the range 1–4 cm/d. The rainfall during the modelled period was lower than average but even during this dry winter a degree of ponding was observed on the site, and the water table was often very close to the ground

ð5Þ

where WT tþDt = water table elevation (bgl) at time t + Dt. Combining Eqs. (3)–(5) gives:

  ðWT t  GWLt Þ:ð2K a Þ ðPt  AEt Þ WT tþDt ¼ exp lnðWT t Þ þ  ðWT t  DÞ:ðWT t þ DÞ:Sa aSa ð6Þ Eq. (6) was implemented in a spreadsheet to generate a simulated water table for given time series inputs of P and AE using a daily timestep. A starting value for WTt was assigned and then values of Ka, Sa and a were varied to refine the model. Three different scenarios were run based on rainfall values factored at 70%, 80%, and 90% of the observed values at Bowling Green. The model outputs are time series of water table elevation and groundwater recharge. 4.2. Model results and discussion While refining the models, it became apparent that with good fits to the data for the first 3 months of the modelled period the simulated water table values fall much too low during the month of March. Furthermore, fitting the March data produces poor fits to the first 3 months of observations. Warmer weather in March may have led to a change in the mode of evaporation at the near surface, most likely due to evapotranspiration gaining significance at the initiation of plant growth from this time on. Plants will thus draw moisture from both the fractures and the matrix in the soil zone whereas in the preceding months it has been assumed that evaporation is predominantly drawing water from the fracture network. Hence in March, even though the PE has increased, the proportion of the energy demand from the fracture system is still likely to be smaller than in the previous month because of the increased proportion of energy that will go towards transpiration. Therefore, for the month of March, a factor was applied to the AE 0

4

-10 3.5

3

-40 2.5 -50 -60

2

Model Ka (cm/d) 1 1.00 2 1.30 3 0.60

-70 -80

Sa a (cm bgl) %Rainfall %AE Factor Recharge (cm/d) 0.11 10.00 0.8 0.55 0.02 0.10 1.30 0.9 0.55 0.03 0.12 7.00 0.7 0.55 0.01

1.5

-90

1

-100 0.5 -110 0

Water Table

WT1

Feb-05

Jan-05

-120

WT2

WT3

Fig. 15. Model results.

Rainfall

R3

R2

R1

Rainfall (cm/d) & Recharge (mm/d)

-30

Dec-04

Water Table Depth (cm agl)

-20

Mar-05

WT tþDt

  DQ ¼ exp lnðWT t Þ  aSa

274

M.O. Cuthbert et al. / Journal of Hydrology 391 (2010) 263–276

surface following prolonged/heavy rainfall. If recharge to the sandstone is in the range implied by the modelling then during a wetter winter, runoff is likely to become more significant. Thus this process would need to be included within equivalent models for wetter years. Although an estimate of recharge to the sandstone has been made for the year 2004–2005, under certain climate and abstraction scenarios this value may be very different. For instance, if the groundwater level in the sandstone in the vicinity of the fieldsite was reduced due to increased groundwater abstraction it is likely that the recharge through the till would increase as water would drain through the till under an increased hydraulic gradient. During a series of drier than average years, the relative rate of decline of the till water table and the sandstone piezometric level will govern the hydraulic gradient driving recharge from the till to the sandstone. Without a longer period of monitoring it is difficult to predict how this would affect the rate of recharge to the sandstone. However, based on the understanding of the hydraulics outlined in this paper, it is likely that as the water table declined, two positive feedback mechanisms would begin to operate to limit the rate of decline of the water table within the till. Firstly, evapotranspiration would increasingly be limited by high soil moisture deficits in the top of the soil/till profile as the water table lowered and capillary rise to the root zone was reduced, while flow to depth through preferential flow in fractures would still occur so the decrease in evapotranspiration would allow much of the preferential flow to percolate to the water table. Secondly, due to the decrease in fracture intensity with depth the bulk hydraulic conductivity of the till may also decrease with depth. Hence during times of lower water table, recharge to the sandstone may be reduced, also slowing the rate of water table recession. The recharge to the sandstone will therefore depend on the delicate balancing of these interacting hydraulic processes. 5. Summary and conclusions 5.1. Summary The geology, material properties and hydraulics have been investigated for a fieldsite situated on till deposits in Shropshire, UK, in order to further an integrated understanding of both the unsaturated and saturated hydraulic processes controlling groundwater recharge in such contexts. The hydraulic data indicate that water table responses to rainfall during the summer occur while large tensions are present higher up the profile. Given the very low permeability of the till at the core scale, these data are highly suggestive of the occurrence of preferential flow. Furthermore, near vertical, hydraulically active fractures have been observed in a test pit extending to depths of >2 m. The fracture intensity decreases with depth and they are commonly infilled with sediment (often fine grained calcareous material or sand) derived from weathered clasts within the till. Flow within the fracture system appears to be highly variable due to variations in fracture hydraulic conductivity (even within the same fracture) and, possibly, flow focussing within the unsaturated zone. Thus, although the bulk till permeability may be related to fracture density and therefore thickness, caution must be applied in assuming general relationships in this regard, for instance in deriving estimates for groundwater models, unless variations in fracture infill are adequately known. The bulk hydraulic conductivity of the till is significantly enhanced by fracturing and, as a result, estimates of recharge based on values of hydraulic conductivity measured at the core scale in the laboratory may be very misleading. Modelling indicates that the bulk hydraulic conductivity of the till at the fieldsite is in the range 9  108–8  1010 ms1, up to three orders of magnitude

higher than the till matrix hydraulic conductivity. This result is consistent with the range of bulk hydraulic conductivity values reported for till deposits elsewhere in the UK for example in Shropshire (Wealthall et al., 1997) and East Anglia (Klinck et al., 1996, 1997) as well as in Denmark (Cartwright, 2001) and North America (Gerber and Howard, 2000). Fracture porosities of the till may be in the range 0.1–4% with groundwater velocities of a few cm/d. The recharge to the sandstone underlying the fieldsite is likely to have been in the range 49 ± 28 mm/a during 2004–2005 and up to 50% lower during summer than winter. This is slightly lower than the estimated areally and temporally averaged value for recharge through till used in the East Shropshire Permo-Triassic Sandstone Groundwater Modelling Project (Streetly and Shepley, 2005). 5.2. Conclusions Previous studies of till hydraulics have tended to focus on either the shallow soil zone and issues concerning unsaturated flow, or the deeper saturated zone and issues concerning aquifer recharge. In contrast, this paper has presented field observations from the ground surface, the unsaturated zone and the saturated aquifer beneath, which has enabled a holistic conceptual model of the processes contributing to groundwater recharge through till to be derived and tested. This allows the link to be made between the detail of the hydraulic processes, and simplified models of recharge estimation which may be needed for larger scale studies. Furthermore, the type of numerical model presented above could be easily transferred to other sites for which a similar conceptual model has been derived and is straightforward to implement within a spreadsheet. With relevance to water resources assessments and groundwater modelling studies in till covered areas, the following may be concluded: 1. Uncertainties in climatic variables, runoff amounts and local heterogeneities can significantly limit the transferability of site based estimates of recharge for use in other geographic areas. Thus, although site based studies may give a range of likely values, recharge is likely to always remain a calibration parameter for local or regional scale groundwater modelling. However, small scale studies can significantly improve our understanding of the detailed hydraulic processes controlling the temporal distribution of recharge which should be used to inform groundwater models at larger spatial scales. 2. Even thin tills (c. 6 m) have the capacity to significantly modify the potential recharge to underlying aquifers. Where the bulk permeability of the till imposes a geological limit on the amount of recharge to underlying aquifers (rather than recharge being climate limited), simple models of recharge are often used based on the estimated bulk permeability of the till multiplied by the hydraulic gradient between the water table and the underlying aquifer piezometric surface. However, where till hydraulic conductivity and storage change significantly with depth, such an approach is unlikely to produce correct results, and the modelling approach outlined above is suggested as a more robust alternative. 3. Due to the control of the vertical hydraulic gradient on recharge, where this may be altered by future abstraction or climate scenarios, it is important that groundwater models within till covered areas explicitly model the dynamic nature of the recharge rather than relying on fixed estimates based on historic model calibrations, or calculations made outside the groundwater model for instance using soil moisture balance models. In any case, simple soil moisture balance models are likely to be very misleading for deriving reliable recharge

M.O. Cuthbert et al. / Journal of Hydrology 391 (2010) 263–276

estimates for tills with shallow water tables unless they include the potential for enhanced AE due to capillary rise. 4. Data from piezometer installations within deeper till horizons should be treated with caution for deriving flux estimates based on hydraulic gradients, due to the variation in the hydraulic conductivity of the fracture network which leads to very variable pressure head distribution through the saturated till. Piezometers monitoring the water table are more suitable for this application. 5. Recharge through the till to the underlying aquifer is likely to occur throughout the year, varying in magnitude according to the interplay between a complex set of hydraulic processes. However, simple models which capture the dynamics of the till water table over the winter period, when the complications of moisture limited evaporation are likely to be minimal, may be used to derive reasonable estimates of recharge for the whole year. In addition to water resources applications, the conceptual model presented has major implications for contaminant migration. As a contaminant enters the fracture network from the ground surface, it will be dispersed to a great extent in the upper few meters of the till by the network of partially penetrating fractures. The preliminary modelling presented suggests groundwater velocities of the order of a few cm/d in the till fracture network. However, the extent to which the contaminant will manage to reach the underlying groundwater system will depend on the amount which migrates into fully penetrating fractures and the matrix diffusion that can take place. It has been noted by other authors that mineralization on fracture faces, while potentially decreasing the permeability of the fractures, may also provide additional surface reactivity for potential contaminant retardation (Corrigan et al., 2001), depending on their precise mineralogy. Further work is needed to test the applicability of the results to other sites situated on lodgement till, and for other types of till. In addition, modelling the complex interactions between the till matrix and fractures during warmer parts of the year is needed. Although matrix flow may be less significant in till systems, recent studies of recharge through the Chalk unsaturated zone may be applicable to such developments (e.g. Mathias, 2005; Ireson et al., 2009). In terms of the most relevant fieldwork to support the development of a conceptual model for a new area, the use of test pits is recommended, as a relatively low cost but highly effective way of gaining an assessment of the till structure and possible flow pathways.

Acknowledgements This work was funded under the UK Natural Environment Research Council’s Lowland Catchment Research (LOCAR) thematic programme. Climate data was kindly supplied by the LOCAR Tern catchment support team, Ian Morrisey and Jon Weller. The authors would also like to express thanks to Mr Bromley who generously gave permission for use of his land for the development of the fieldsite, to staff and students at The University of Birmingham for help with equipment design and fieldwork, Roger Livesey and Richard Greswell in particular, and to staff at the BGS and the Environment Agency for their support during the project.

References Bonell, M., 1972. An assessment of the possible factors contributing to well level fluctuations in Holderness boulder clay, East Yorkshire. Journal of Hydrology 16, 361–368.

275

Cartwright, D.J., 2001. Using hydraulic Testing to Determine Transport Properties of a Fractured Clay Aquitard. MSc project report, Department of Geological Sciences, University College, London. Christy, A.D., McFarland, L.A., Carey, D., 2000. The use of test pits to investigate fracturing and glacial stratigraphy in tills and other unconsolidated materials. Ohio Journal of Science 100 (3–4), 100–106. Corrigan, C.A., Jamieson, H.E., Remenda, V.H., 2001. Fracture wall cements and coatings from two clayey till aquitards. Ground Water 39 (5), 786–794. Cravens, S.J., Ruedisili, L.C., 1987. Water-movement in till of East-Central SouthDakota. Ground Water 25 (5), 555–561. Cuthbert, M.O., 2006. Hydraulic Processes Controlling Recharge Through Glacial Drift. PhD thesis, University of Birmingham, UK. Cuthbert, M.O., Mackay, R., Tellam, J.H., Barker, R.D., 2009. The use of electrical resistivity tomography in deriving local scale models of recharge through superficial deposits. Quarterly Journal of Engineering Geology and Hydrogeology 42, 199–209. D’Astous, A.Y., Ruland, W.W., Bruce, J.R.G., Cherry, J.A., Gillham, R.W., 1989. Fracture effects in the shallow groundwater zone in weathered Sarnia-area clay. Canadian Geotechnical Journal 26, 43–56. Domenico, P.A., Schwartz, F.W., 1998. Physical and Chemical Hydrogeology, second ed. John Wiley & Sons, New York. Feddes, R.A., Kowalik, P.J., Zaradny, H., 1978. Simulation of Field Water Use and Crop Yield. Centre for Agricultural Publishing and Documentation, Wageningen, Netherlands. Fredericia, J., 1990. Saturated hydraulic conductivity of clayey tills and the role of fractures. Nordic Hydrology 21, 119–132. Gerber, R.E., Howard, K., 2000. Recharge through a regional till aquitard: threedimensional flow model water balance approach. Ground Water 38 (3), 410– 422. Gerber, R.E., Boyce, J.I., Howard, K.W.F., 2001. Evaluation of heterogeneity and fieldscale groundwater flow regime in a leaky till aquitard. Hydrogeology Journal 9 (1), 60–78. Gerke, H.H., Kohne, J.M., 2004. Dual-permeability modelling of preferential bromide leaching from a tile-drained glacial till agricultural field. Journal of Hydrology 289, 239–257. Ghazavi, G., Thomas, Z., Hamon, Y., Marie, J.C., Corson, M., Merot, P., 2008. Hedgerow impacts on soil–water transfer due to rainfall interception and root-water uptake. Hydrological Processes 22 (24), 4723–4735. Gonthier, G.J., 2007. A graphical method for estimation of barometric efficiency from continuous data – concepts and application to a site in the Piedmont, Air Force Plant 6, Marietta, Georgia. US Geological Survey Scientific Investigations Report 2007–5111. Greswell, Richard, Ellis, Paul, Cuthbert, Mark, White, Rachel, Durand, Veronique, 2009. The design and application of an inexpensive pressure monitoring system for shallow water level measurement, tensiometry and piezometry. Journal of Hydrology 373, 416–425. Grisak, G.E., Cherry, J.A., Vonhof, J.A., Blumele, J.P., 1976. Hydrogeologic and hydrochemical properties of fractured till in the interior plains region. In: Legget, R.F. (Ed.), Glacial Till, Royal Society of Canada, Special Publication 12, pp. 304–335. Haldorsen, S., Kruger, J., 1990. Till genesis and hydrogeological properties. Nordic Hydrology 21, 81–94. Hendry, M.J., 1988. Hydrogeology of clay till in a prairie region of Canada. Ground Water 26 (5), 607–614. Herbst, M., Roberts, J.M., Rosier, T.W., Gowing, D.J., 2006. Measuring and modelling he rainfall interception loss by hedgerows in southern England. Agricultural and Forest Meteorology 141, 244–256. Herbst, M., Roberts, J.M., Rosier, T.W., Taylor, M.E., Gowing, D.J., 2007. Edge effects and forest water use: a field study in a mixed deciduous woodland. Forest Ecology and Management 250, 176–186. Hough, M.N., Jones, R.J.A., 1997. The United Kingdom meteorological office rainfall and evaporation calculation system: MORECS version 2.0–an overview. Hydrology and Earth System Sciences 1 (2), 227–239. Hvorslev, M.J., 1951. Time lag and soil permeability in groundwater observations. US Army Corps Engineers. Waterways Experiment Station, Bull 36. Ireson, A.M., Mathias, S.A., Wheater, H.S., Butler, A.P., Finch, J., 2009. A model for flow in the chalk unsaturated zone incorporating progressive weathering. Journal of Hydrology 365, 244–260. Jacob, C.E., 1940. On the flow of water in an elastic artesian aquifer. Transactions– American Geophysical Union 21, 574–586. Jakobsen, P.R., Klint, K.E.S., 1999. Fracture distribution and occurrence of DNAPL in a clayey lodgement till. Nordic Hydrology 30 (4/5), 285–300. Jansson, C., Jansson, P.E., 2003. Preferential water flow in a glacial till soil. Abstracts Volume of EGS–AGU–EUG Joint Assembly, Nice, France, EGS. EAE03-A10309. Keller, C.K., Vanderkamp, G., Cherry, J.A., 1986. Fracture permeability and groundwater-flow in clayey till near saskatoon, saskatchewan. Canadian Geotechnical Journal 23 (2), 229–240. Keller, C.K., Vanderkamp, G., Cherry, J.A., 1988. Hydrogeology of two saskatchewan tills. 1. Fractures, bulk permeability, and spatial variability of downward flow. Journal of Hydrology 101 (1–4), 97–121. Keller, C.K., Vanderkamp, G., Cherry, J.A., 1989. A multiscale study of the permeability of a thick clayey till. Water Resources Research 25 (11), 2299– 2317. Klinck, B.A., Barker, J.A., Noy, D.J., Wealthall, G.P., 1996. Mechanisms and rates of recharge through glacial till: Experimental and modelling studies from a

276

M.O. Cuthbert et al. / Journal of Hydrology 391 (2010) 263–276

Norfolk site. British Geological Survey. Fluid processes series technical report WE/96/1. Klinck, B.A., Hopson, P.N., Morigi, A.N., Bloodworth, A.J., Inglethorpe, S.D.J., Entwistle, D.C., Wealthall, G.P., 1997. The hydrogeological classification of superficial clay: the hydrogeological characterisation of glacial Till in East Anglia. Environment Agency. R&D technical report W28. Klint, K.E.S., Gravesen, P., 1999. Fractures and biopores in Weichselian clayey till aquitards at Flakkebjerg, Denmark. Nordic Hydrology 30 (4/5), 267–284. Lerner, D.N., Issar, A.S., Simmers, I. (Eds.), 1990. Groundwater Recharge: A Guide to Understanding and Estimating Natural Recharge. International Contributions to Hydrogeology, Verlag Heinz Heise. Lloyd, J.W., 1980. The importance of drift deposit influences on the hydrogeology of major British aquifers. Journal of the Institution of Water Engineering Science 34, 346–356. Mase, C.W., Cherry, J.A., McKay, L.D., Unger, A.J.A., 1990. Hydrogeology of Desiccated Fractured Clays. American Geophysical Union Fall Meeting, San Francisco, California, US. Mathias, S.A., 2005. Modelling Flow and Transport in the Chalk Unsaturated Zone. PhD thesis, Imperial College London, UK. McKay, L.D., Fredericia, J., 1995. Distribution, origin and hydraulic influence of fractures in a clay-rich glacial deposit. Canadian Geotechnical Journal 32, 957– 975. McKay, L.D., Cherry, J.A., Gillham, R.W., 1993. Field experiments in a fractured clay till. 1. Hydraulic conductivity and fracture aperture. Water Resources Research 29 (4), 1149–1162. McKay, L.D., Fredericia, J., Lenczewski, M., Morthorst, J., Klint, K.E.S., 1999. Spatial variability of contaminant transport in a fractured till, Avedore Denmark. Nordic Hydrology 30 (4/5), 333–360. Nilsson, B., Sidle, R.C., Klint, K.E., Boggild, C.E., Broholm, K., 2001. Mass transport and scale-dependent hydraulic tests in a heterogeneous glacial till-sandy aquifer system. Journal of Hydrology 243, 162–179.

Ostendorf, D.W., DeGroot, D.J., Shelburne, W.M., Mitchell, T.J., 2004. Hydraulic head in a clayey sand till over multiple timescales. Canadian Geotechnical Journal 41 (1), 89–105. Ruland, W.W., Cherry, J.A., Fenestra, S., 1991. The depth of fractures and active groundwater-flow in a clayey till plain in southwestern ontario. Ground Water 29 (3), 405–417. Rushton, K.R., 2005. Estimating recharge for British aquifers. Water and Environment Journal 19, 115–124. Streetly, M., Shepley, M.G., 2005. East Shropshire Permo-Triassic Sandstone Groundwater Modelling Project. Task 8: final report. Report of the Environment Agency of England and Wales, UK. Thatcher, K.E., 2009. Spatial Characterisation and Stochastic Analysis of Unsaturated Flows and Groundwater Recharge Through Glacial Drift. PhD thesis, University of Birmingham, UK. Todd, D.K., 1959. Ground Water Hydrology. John Wiley and Sons Inc., New York. Villholth, K.G., Jensen, K.H., 1998b. Flow and transport processes in a macroporous subsurface-drained glacial till soil. 2. Model analysis. Journal of Hydrology 207, 121–135. Villholth, K.G., Jensen, K.H., Fredericia, J., 1998a. Flow and transport processes in a macroporous subsurface-drained glacial till soil. 1. Field investigations. Journal of Hydrology 207, 98–120. Wealthall, G.P., Brandon, A., Inglethorpe, S.D.J., Entwistle, D.C., 1997. The hydrogeological characterisation of glacial till and glacio-lacustrine sediments in Shropshire. British Geological Survey. R&D Technical Report W29. White, Rachel, Rivett, Michael O., Tellam, John H., 2008. Paleo-roothole facilitated transport of aromatic hydrocarbons through a Holocene clay bed. Environmental Science and Technology 42 (19), 7118–7124. doi:10.1021/ es800797. Williams, R., 1967. The influence of joints on the movement of groundwater through glacial till. Journal of Hydrology 5, 163–170.