Equipment and procedures for coring were compliant with ASTM D1587-08 (2008a) to obtain samples that were as undisturbed as possible. A rotary drilling rig equipped with Triefus triple core barrels, lined with seamless clear PET, was used in push coring mode. Local creek water was used as a drilling fluid and casing was used to stabilise the hole behind the push core barrel such that drilling fluid additives were not required. The holes were therefore fully cased to the maximum depth of push core drilling at up to 40 m BG.

The non-rotating core barrel was forced into the formation, whilst a rotating device on the outside of the tube removes the cuttings as the barrel was advanced. The cutting edge of the non-rotating sample tube projects several millimetres beyond the rotary cutters. The thin-walled core barrel complied with the standard for undisturbed sampling, with an area ratio of less than 25 % for an open-drive sampler. The area ratio of 16 % was based on a core barrel design with an external diameter of 110 mm and internal diameter of 101 mm (C size). The 1.5 m length core barrel was a composite open sampling system with a core nose screwed onto the base with a bevelled end to cut the core as the barrel pushed into the formation. After the core was extracted from the ground, an air supply was connected to the top of the core barrel to slide the core out of the barrel whilst it remained in the clear PET liner without rotation, distortion, or compression.

The cores contained within PET liners were transferred directly from the core barrels to a cool room on site, and thence to a laboratory cool room, reducing the potential for moisture loss. Semi-consolidated clay cores were selected from below the saturated zone for CP tests, at depths up to 40 m BG. Sediment core samples of lengths between 50 and 100 mm were prepared for CP testing. The moisture content and bulk density of cores was measured using methods adapted from ASTM D7263-09 (2009). These measurements were completed immediately on the drill site.

The preferred method for preservation of drill core was double plastic bagging of sections of core within their PET liners using a food grade plastic sealing system (with brief application of a vacuum to extract air from the plastic bag). Alternatively, core within PET core barrel liners were trimmed of air or fluid filled excess liner immediately after drilling, and then sealed with plastic tape. All cores were stored at 4 ∘C in a portable cool room on the drill site and then at the laboratory. Sections of cores, particularly at the nose end, that appeared to be damaged or disturbed were excluded from permeability or bulk density testing. Additional steps that were taken in the laboratory to ensure core testing was representative of in situ conditions are described in Sect. 4.1.

After coring, the holes were completed as monitoring piezometers and the casing was jacked out. The piezometers were constructed of screwed sections of 50 mm PVC casing with O-ring seals, with a 1.5 m machine slotted screen packed with pea-sized washed gravel. The annulus was then filled with a bentonite seal, backfilled to the surface and completed with a steel casing monument and cement monument pad.

Fluid for K testing (influent) should be taken from the formation at the same depth as the core. Formation water can be synthesised if it is not possible to sample directly from aquitard strata, by estimating the ionic strength, Na / Ca ratio, and pH. In this study, groundwater from piezometers at a similar depth to the core was obtained using standard groundwater quality sampling techniques (Sundaram et al., 2009). A 240 V electric submersible pump (GRUNDFOS MP1) and a surface flow cell were used to obtain representative samples after purging stagnant water to achieve constant field measurements of electrical conductivity and other parameters (Acworth et al., 2015 and unpublished data).

To ensure that cores were tested under saturated realistic conditions, drill cores were adequately preserved, stored, prepared, and set on a vacuum plate prior to centrifuge testing. Cores from PET drill core liners were trimmed and inserted into an acrylic liner for the CP using a core extruder. The custom made core extruder had five precision cutting blades driven by a motorised piston suitable for a 100 mm diameter core. Cores for CP testing in this study were 100 mm diameter C size core, with a length of 50–100 mm. A close fit between the clay core and the liner was achieved using this extruder.

A vacuum plate system for core samples was designed to ensure fully saturated cores, remove air at the base of the core, and ensure an effective seal between the CP liner prior to testing at accelerated gravity. The vacuum plate device was designed to fit the CP liners containing the cores, drawing ponded influent from the top to the base of the cores using a standard laboratory vacuum pump at 100 kPa of negative pressure. After 12–48 h, or upon effluent flow from the base, the acrylic liners containing the prepared cores were then transferred directly to the CP module without disturbing the sample.

Furthermore, the moisture content and degree of saturation was monitored by measuring weight change of the permeameters during testing, and direct moisture tests of samples before and after CP testing. There was negligible difference observed between the moisture content of the core tests and in situ conditions, and the results were not associated with the time between sampling and testing of the core. Moisture content was not affected by the use of a vacuum to expel air from sealing bags or from the top or base of the cores fitted into the CP liners.

A self-seal was observed forming from material swelling at the interface with the liner within minutes of introducing the influent solution. Prior to the self-seal development, leakage along the liner interface was identified by a flow rate of several orders of magnitude higher than the steady-state flow Kv value. The swelling that occurred to self-seal the core was estimated at less than 0.02 % of the cross sectional area of the core by comparing flow rates through the CP drainage hole (described in Sect. S3 in the Supplement). It was calculated that this area of swelling was sufficient to seal an annulus aperture of ∼ 0.01 mm between the clay core and the acrylic liner.

Given the relatively shallow depth of these cores, and the semi-consolidated status, the maximum g level in the centrifuge was limited to prevent structural changes in the core matrix. To minimise changes in porosity of the core during testing, the g level and the weight of ponded fluid on the cores were therefore designed to ensure that total stress was less than estimated in situ stress at the depth from which the core was drilled.

Blind permeability tests were carried out by an independent laboratory, which adapted a constant/falling-head method (AS 1289 6.7.3/5.1.1, 1991) with methods from Head (1988). For these 1 g column tests, a sample diameter of 45.1 mm and length 61.83 mm was used, and a confining pressure of 150 kPa and back pressure of 50 kPa was applied, providing a vertical uniaxial stress of 100 kPa. The test time was up to 100 h. These standard 1 g column tests used deionised water as the influent.

The Broadbent CP module and some unique systems developed as part of this study are described in this section, with further details in Sects. S1 and S3. A conceptual plan of a CP is shown in Fig. 2. The CP contains a cylindrical clay sample with length L and diameter D, and is spinning in a centrifuge around a central axis at an angular velocity ω. The permeameter has an inlet face at a radius r, and a drainage plate at a radius of r0. The co-ordinate z is defined as positive from the base of the sample towards the central axis of rotation, consistent with definitions in 1 g column testing (McCartney and Zornberg, 2010). This frame of reference is in an opposite direction to that defined by Nimmo and Mello (1991), but is convenient for interpretation and comparison of column flow tests.

Influent was fed from burettes located next to the centrifuge via a pair of custom designed low voltage peristaltic pumps mounted either on the centrifuge beam, or outside the centrifuge and through the low flow rotary union. In this study, the outlet face was a free drainage boundary, and is discussed further in Sects. S2 and S3.

The K value is based on flow rate, flow area, radius, and revolutions per minute (RPM), although the method was adapted from a unsaturated/saturated flow apparatus (UFA) centrifuge to this CP system (Sect. 4.3). Importantly, both testing systems are for steady-state flow with free drainage due to zero pressure at the base of the core.

The mass of two core samples were balanced to the nearest 100 g and tested simultaneously at either end of the centrifuge beam. The CP was operated at 10 g for 30 min, and if no rapid flows due to leakage were detected, this was gradually increased to 20, 40 g, and so on, until the maximum total stress on the core approached the estimated in situ stresses of the material at the given depth in the formation. The upper permissible g level was designed to be less than the estimated in situ stress from the depth at which the core was obtained. It was also important to ensure that effective stress (Sect. 4.4) was acceptable, as variable pore fluid pressures during testing could cause consolidation of the core matrix. Influent volume was measured using both a calibrated continuous time record of pump rotations, and manual burette measurements, and effluent volumes were measured by weight. steady-state flow was defined as ±10 % change in discharge over subsequent measurements in time, provided that influent flow rate was within ±10 % of the effluent flow rate. Both of these conditions were required for the testing to be considered as a steady-state flow condition. This protocol provided additional quantitative measures to the ASTM D7664 (2010), which states that steady-state conditions have been attained “if the outflow is approximately equal to the inflow”. Section S4 discusses the uncertainty of the measured data in more detail.

Hydraulic conductivity calculations for the CP in this study were based on ASTM D6527 (2008b) and ASTM D7664 (2010) with a form of Darcy's Law that incorporates the additional driving force within a centrifuge. The gradient in the centrifuge elevation potential (Nimmo and Mello, 1991), or the gradient in centrifuge “elevation head” (Zornberg and McCartney, 2010) due to the centrifuge inertial force driving was defined as flow away from the centre of rotation (or in the opposite direction to z in Fig. 2). The g level was defined at the mid-point of the core. A ponded influent above the top of the core prevented loss of saturation along the core (Nimmo and Mello, 1991). The centrifuge inertial (elevation) head gradient and hydraulic head gradient (stationary centrifuge at 1 g) were calculated at 0.005 m increments through the core.

Statistical analysis of the data followed a basic small-sampling theory using the Student's t distribution, following the approach of Gill et al. (2005) and extending the approach of Timms and Anderson (2015) for estimating sample numbers required for CP testing. Upper and lower confidence intervals (UCI, LCI) were calculated from the mean ± t(n-1). sn/n1/2, where sn is the sample standard deviation and t(n-1) is the value of the Student's t distribution at the selected confidence limits (CL) of 90 and 99 %. The confidence intervals were calculated for the increasing number (n) of Kv data from each core.

Fluid pressures and hydraulic gradient through the centrifuge core were determined following the approach of Nimmo and Mello (1991). The total fluid pressure P (kPa) was calculated in Eq. (1) P=ρw∫r0r′rω2dr, assuming a fluid density ρw of 1.0 g cm-3 and where r is the radius of rotation (cm), and ω is the angular velocity (s-1). The total stress S (kPa) was determined through the centrifuge core, following Eq. (2) S=ρs∫r0r′rω2dr, assuming core bulk density ρs of 1.9 g cm-3. The total stress and fluid pressure were calculated at 0.005 m increments through the core. The effective stress was then calculated as the difference between total stress and fluid pressure. An increase in effective stress associated with decreased fluid pressures near the base of the free draining core may cause consolidation of the core matrix near the boundary.

The total stress applied to the core, relative to stress, may affect the porosity of the core sample, depending on the stress history. In situ stress of the cores (Si) at the sampling depth below ground (D) was calculated using Eq. (3) Si=ρsDg.

It was assumed that the overlaying formations were fully saturated and of a similar bulk density to the supplied core samples.

Index properties for five representative cores are provided in Table 2. The cores were typically silty clay (where the clay–silt size boundary is defined at 0.002 mm), except for one sandy clay core. The large proportion of silt relative to clay is an important characteristic of this formation, with clay mineralogy dominated by smectite (Timms and Acworth, 2005; Acworth and Timms, 2009).

Moisture content varied from 24.7 to 36.4 % by weight, and was consistent with site measured data on the core (Sect. S5), although not all the cores were fully saturated as received by the external laboratory. Bulk wet density varied from 1.71 to 1.88 g cm-3 and particle density from 2.47 to 2.58 g cm-3. The Kv of cores tested in the CP module (Table 3) varied from 1.1×10-10 to 3.5×10-7 m s-1 (n=18). Accelerations up to 100 g were applied during CP testing of semi-consolidated sediment cores and were more typically limited to 30–40 g. Figure 3 shows the measured influent and effluent rates and the calculated Kv values during a typical CP test as the g level was gradually increased. Steady-state flow (±10 % change over time with influent rate equal to effluent rate) was achieved at ∼ 20 h (Fig. 3). However, a lower Kv value that was observed overnight (> 12 h interval between samples) than was observed during the day (∼ 1 h intervals between samples). The Kv values over shorter time periods, with minimal evaporative losses, were considered to be more reliable. Further experimentation and numerical modelling is required to adequately explain this anomalous data, which may be associated with evaporative losses over longer time periods of flow measurement or other transient processes within the system.

Anomalous flow via preferential pathways could be readily identified by a flow rate of several orders of magnitude greater than otherwise observed. Anomalous flow was often observed along the interface of the cores and the liner during the early minutes of a test before sealing occurred and steady-state conditions were established. On one occasion a preferential flow path developed during the test, which caused very fast flow at accelerated gravity that was easily detected. A test failure like this could be readily identified and excluded from further evaluation.

A small uncertainty in Kv results for the CL site was calculated at a confidence limit of 99 % using the methods described in Sect. 4.3. By increasing the number of samples, the confidence bounds for Kv were narrowed from a range of 4.8×10-10 to 2.4×10-9 m s-1 (n=5) to a range of 1.1×10-9 to 2.1×10-9 m s-1 (n=9). Increasing the number of samples from five to nine decreased the standard deviation, although a similar geometric mean occurred with the increased sample number (Table 4). However, there was less certainty at two other core sites (BF and NR) due to limited and more variable Kv data. At the BF site the 99 % confidence interval had relatively wide Kv bounds for n=6, while at NR site, a confidence interval of 90 % results in similarly wide Kv bounds for n=3. However, such a small number of samples is not considered sufficient for statistical analysis. This evaluation of the results highlights the relative Kv variability and small sample set for the BF and NR sites, and the need for further testing, particularly at the NR site. This issue will be expanded in the following discussion.

How realistic the Kv measured by CP testing is of in situ conditions depends in part on the magnitude of stress and any structural changes that occur within the core matrix. Section S2 provides background on the definition and significance of hydrostatic pore pressure, centrifuge inertial (elevation) head, and gradients driving fluid flow. Section S2 also discusses the possibility that K values reported in this study could be biased on the high side, considering total stress at the base of the core under steady-state conditions.

During centrifuge testing effective stress is maximum at the base of the free draining core, where fluid pressure is zero, and thus effective stress is equal to total stress under hydrostatic conditions (no flow). In both testing methods in this study, the total stress was less than estimated in situ stress; however, the stress history of the core sample and effective stress dynamics were uncertain. It appears that the stresses during these tests were likely within an acceptable range to minimise structural changes including swelling and consolidation. There was no evidence of significant changes in core length due to consolidation of the samples during spot checks of core length with a digital calliper. However, further attention on these processes, including instrumentation to measure fluid pressures and core matrix changes during testing is required in future studies. A separate geotechnical study of these semi-consolidated sediments, including oedometer testing is in progress to better quantify the relationship between stress and permeability in these semi-consolidated materials. In future studies of semi-consolidated materials, measurement of consolidation state (over consolidation ratio) and pre-consolidation stress is recommended prior to centrifuge testing to ensure that an appropriate centrifuge stress is applied.

A comparison of Kv from in situ and column testing methods are shown in Fig. 4 for the CL site. Results from the CP method (1.1×10-10 to 3.5×10-9 m s-1, n=9) were similar to Kv values from the independent and in situ method (1.6×10-9 to 4.0×10-9 m s-1) confirming that the sequence is of low-permeability at the CL site with a reasonable level of confidence (Table 4). However, Kv from both in situ and CP methods were higher than the 1 g column tests of cores from 11.27 to 11.47 and 28.24 to 28.33 m BG from this site (1.4×10-9, 1.1×10-10, and 1.5×10-10 m s-1, n=3).

In situ Kv of the clayey-silt at the CL site was based on observed amplitude and phase changes of pore pressures (at hourly or 6-hourly intervals) due to five major rainfall events over 4 years (Timms and Acworth, 2005). The phase lag at the base of the clay varied between 49 and 72 days. The phase lag pore pressure analysis resulted in a Kv value of 1.6×10-9 m s-1, while the change in amplitude over a vertical clay sequence of 18 m (from a 17 m depth piezometer to the inferred base of the aquitard at 35 m depth) resulted in a Kv value of 4.0×10-9 m s-1.

It is noted that the reliability of harmonic analysis related methods may be compromised by specific storage measurements. Jiang et al. (2013) relied on indirect specific storage values derived from downhole sonic and density log data from boreholes in the region, while Timms and Acworth (2005) calculated specific storage from barometric and loading responses that were recorded in the same groundwater level data set and boreholes used for harmonic analysis.

The reduced test times of CP testing may be attributed to the reduced time required to achieve steady-state flow with centrifugal forces driving flow. Alternatively, the relatively longer time required for 1g column testing may be attributed to deionised water interaction with clay that reduced infiltration rates into the cores (10–100 lower Kv result for 1 g column tests compared with CP tests). It is known that decreased ionic strength of influent (e.g. deionised water) causes a linear decrease in permeability, and that the relative concentrations of sodium and calcium can affect permeability due to swelling and inter-layer interactions (e.g. Shackelford et al., 2010; Ahn and Jo, 2009). Differences in Kv values from the two laboratory testing methods could be due to differences in test set-up (e.g. 45 vs. 100 mm diameter core) and stress changes that occur as discussed in Sects. 5.2 and S2.

CP testing was relatively rapid, typically with a few hours up to 24 h required for steady-state flow CP, compared with an average of 90 h (73, 96, and 100 h for the tests reported here) for 1 g column testing. In addition, an extended test of 830 h in the CP (unpublished data) verified that no significant changes occurred over extended testing periods. The CP technique can therefore reduce average testing time to ∼ 20 % of the time that would be required in 1 g laboratory testing systems, similar to the reduced time requirement of centrifuge methods for unsaturated hydraulic conductivity functions compared with 1 g column tests ASTM (2010). The relative time advantage of testing cores at accelerated gravity may be greater at lower Kv, due to the increased time required to establish steady-state flow conditions. The relative time advantage could be significant for contaminant transport testing, which requires several pore volumes of steady-state flow, compared to permeability testing where steady-state flow is established before one pore volume.

The similarity of Kv measurements with different scales at the CL site (Fig. 4) indicates that in this part of the alluvial deposit K is independent of vertical scale from centimetres to several metres. These Kv results from both in situ and laboratory methods provide an important constraint for evaluations of hydraulic connectivity, particularly as there is a general lack of Kv data for these sediments. Complimentary studies of hydraulic connectivity to quantify leakage rates include pore pressure monitoring and piezometer slug testing at various depth intervals along with hydrogeochemical and isotope tracer data. Recent geological studies of the alluvial sequence (Acworth et al., 2015) outlined in Sect. 2, and identification of dual porosity structures in the large diameter cores (Crane et al., 2015) indicate that it may be possible for vertical leakage to occur through clayey silts if a vertical hydraulic gradient were to be imposed. A diffusion dominated salt profile through the sequence suggest negligible vertical flow (Timms and Acworth, 2006); however, a proper assessment of flow connectivity requires vertical hydraulic gradients to also taken into account any salinity variations with depth and pore pressure variations that have occurred over at least the past decade.

The Kv measurements reported in this paper are important because there is a lack of aquitard data for alluvial groundwater systems globally. Even where many groundwater investigations have been completed (e.g. Murray–Darling Basin) there continues to be a lack of information on the thick clayey-silt sediments at various spatial scales.

The core samples for testing were randomly selected from the same lithostratigraphic formation, the upper 30 m of the alluvial sequence as described in Sect. 2. Although the alluvial sequence extends to over 100 m depth, we focussed this study on sediments defined by a low net-to-gross ratio (Larue and Hovadik, 2006) of < 0.4 that reflects that clay-rich part of the sequence (Timms et al., 2011). We assumed a log-normal distribution of Kv within this formation, which as noted by Fogg et al. (1998) might be justified within individual facies, but not over the full stratigraphic section. It was also assumed that the standard deviation of the samples tested is similar to the standard deviation of the total population of Kv results from the formation, which may only be known if a significantly large number of samples are tested.

Kv values for cores from the NR and BF sites were significantly larger than for the CL site, although additional data from the NR site are required to increased confidence intervals (Tables 3 and 4). Based on the data set currently available for each site there did not appear to be any significant Kv trend with depth, except at the CL site, with a possible decrease of Kv by a factor of 3 with depth increasing from 11 to 28 m BG. Further testing is in progress to better identify any spatially significant trends in Kv.

Kv results obtained from the CP for these clayey-silt aquitards were significantly larger than Kv for consolidated rock cores tested in this system (Bouzalakos et al., 2013). The relatively low g levels in this study (up to 80 g), compared to rock core testing (up to 520 g; Bouzalakos et al., 2013) were necessary for the shallow and semi-consolidated nature of the clayey-silt cores. In fact, steady-state flow was achieved at low g levels for Kv values that were at least 100 times higher than the current detection limit and uncertainty of the CP system (Sect. S4).

The vertical permeability of the clayey-silt aquitards in this region, and the relative importance of matrix flow and preferential flow through fractures and heterogeneities are critical to the sustainability of the groundwater resource. The Kv data reported in this study for these silty and semi-consolidated sediments are higher than reported for regional flow modelling in this area (McNeilage, 2006), indicating that the aquitards allow for significant recharge to underlying aquifers.

A regional groundwater flow model developed by McNeilage (2006), with a two-layer MODFLOW code, determined the dominant source of recharge to be diffuse leakage through the soil (and aquitards) in the Breeza groundwater management area. As in typical groundwater modelling practice (Barnett et al., 2012), the aquitard was not explicitly modelled, with water instead transferred from a shallow to a deeper aquifer using a vertical leakance value (units in day-1).

The calibrated groundwater model indicated that approximately 70 % of the long-term average groundwater recharge (11 GL year-1) was attributed to diffuse leakage in this area that included the CL and NR sites. This volume is equivalent to 20 mm year-1, or a Kv of ∼6×10-10 m s-1 assuming a unit vertical hydraulic gradient over an area of approximately 500 km2. The actual Kv or leakance values were not reported. The calibrated leakance values were found to vary over 3 orders of magnitude across the Breeza area, with relatively high values in isolated areas in the south, centre, and north. In comparison, the Kv results on clayey-silt cores appear to be higher than the apparent Kv of the regional groundwater model, but with a similar degree of heterogeneity. The reasons for this discrepancy are not yet clear, but may be attributed to non-unique calibration of the regional flow model (e.g. underestimation of inter-aquifer leakance) or the lack of representative Kv values for this aquitard at a scale that accounts for heterogeneities and preferential flow paths.

The Kv results in this study are within the range of values reported elsewhere for semi-consolidated clay-silt sediments. For example, Neuzil (1994) reviewed aquitard Kv values for intact muds and lacustrine clays (10-8–10-11 m s-1) compared to consolidated materials such as shale with values as low as 10-16 m s-1 for argillite. A detailed study of a clayey-marl and limestone aquitard in France (Larroque et al., 2013) found a quasi-systematic bias of 1 order of magnitude between petrophysical Kv estimates (10-8–10-10 m s-1), compared with values (10-9–10-11 m s-1) from hydraulic diffusivity monitoring between 30 and 70 m BG. However, the empirical petrophysical relationships between porosity, pore size and intrinsic permeability do not adequately account for structural effects of clay materials. Field piezometer rising head tests (n=225) indicated that Kv of a lacustrine clay aquitard around Mexico City was 10-8–10-9 m s-1 in two areas, 100 times greater than matrix-scale permeability (Vargas and Ortega-Guerrero, 2004). However, in a third area of the Mexico City aquitard, the field tests were 10-10 m s-1 indicating the regional variability that can occur within clayey deposits.

Studies of glacial till aquitards in Canada, the US, and Denmark find that regional permeability is typically at least 2 orders magnitude greater than laboratory tests (van der Kamp, 2001; Fredericia 1990; Bradbury and Muldoon, 1990; Gerber and Howard, 2000), although one study (Husain et al., 1998) showed that for a thick glacial till aquitard in southern Ontario, Canada, the regional permeability is similar to the laboratory-obtained measurements, indicating the absence of significant permeable structures.

There is evidence of fracturing near the surface of the clayey aquitards that are the focus of this study. Fracture flow to a shallow pit and the freshening of porewater in the aquifers at 16 and 34 m depth during the irrigation season indicated rapid leakage had occurred at the BF site (Acworth and Timms, 2009). The dynamics of fracturing within ∼ 2 m of the ground surface in these sediments was described by Greve et al. (2012). However, beyond the zone of fracturing near the ground surface, there appears to be insignificant groundwater flow. Solute profiles through the 30 m thick clayey deposit at the CL site indicate that downwards migration of saline water is limited to diffusion and that flow is insignificant (Timms and Acworth, 2006). On the basis of available evidence, the clayey sediments in this region may lack preferential flow paths at some sites, and in other areas preferential flow may occur through features such as fractures and heterogeneity at a range of scales (Crane et al., 2015). Further work is required to determine permeability at a range of scales, and to better understand preferential flow paths. The current conceptual model on which the numerical models in this region are based (simple layered aquitard overlying an aquifer) do not allow for spatial variability in connectivity mechanisms that could be important across a large valley alluvial fill sequence. Multiple mechanisms for vertical connectivity, including matrix flow, fracture flow, and sedimentary heterogeneity, could be important in aquitards. The relative importance of each of these pathways for vertical flow would depend on the spatial scale and local setting in each aquitard.

To determine if accelerated flow conditions are realistic for hydrogeological environments, the linear flow velocity for various CP set-ups was compared with other flow scenarios. The rationale behind this comparison was that if the flow rate was consistent with scaling laws for physical modelling, with a unit gradient as a point of reference, then the results could be consider realistic for atypical in situ vertical hydraulic gradient. In Table 5, an in situ hydraulic gradient of 0.5 is compared with CP set-ups for 100 and 65 mm diameter cores of various lengths for an aquitard material with a Kv of 10-8 m s-1. The vertical flow rate varies from 0.3 mL h-1 under in situ conditions to 8.5 mL h-1 in the CP, such that linear flow velocities remain very low at 10-8–10-6 m s-1. The flow rate during centrifugation was N times greater than if a hydraulic gradient of 1 was applied to the core samples at 1 g. This increase in flow rate is consistent with scaling laws for physical modelling (Tan and Scott, 1987) providing further evidence that the Kv results are realistic.

The accelerated flow conditions, whilst realistic for hydrogeological environments, can also be an advantage for experimental studies of solute transport. Kv results of the order of 10-9 m s-1 were obtained in ∼ 20 % of the time required for 1 g column permeameter tests. Solute breakthrough experiments require longer testing periods of steady-state flow than for permeability testing. For example, Timms and Hendry (2008) and Timms et al. (2009) described continuous CP experiments over 90 days to quantify reactive solute transport during several pore volumes (PV) of flow. The comparisons of time required for 1 PV provided in Table 5 illustrate the possible advantages of CP for contaminant flow that may affect the structural integrity of the material.