Geophysical Research Letters

Thermal anomalies indicate preferential flow along faults in unconsolidated sedimentary aquifers



[1] Anomalously high (up to +8°C) and low (−2°C) groundwater temperatures, as compared to undisturbed geothermal profiles, have been observed in unconsolidated siliciclastic aquifers off-set by normal-faults in the Lower Rhine Embayment, Germany. High hydraulic head gradients, induced by pumping, over the same faults suggest that they form effective barriers to lateral groundwater flow. Numerical analysis of the geothermal data presented here shows that the observed thermal anomalies can be explained under the assumption that the faults form a sub-vertical pathway that is connecting deep and shallow aquifers that are elsewhere separated by confining units. The hydraulic head and temperature observations taken together are consistent with the hypothesis that these faults behave as a conduit-barrier systems. Such behavior would arise from clay-smearing and drag of sand along the fault plane. Most current models of fault hydrology in unconsolidated sedimentary sequences assume faults to be effective barriers to fluid flow. Therefore our findings can have important consequences for the assessment of contaminant flow or hydrocarbon migration in sedimentary aquifer systems cut by faults.

1. Introduction

[2] Anomalously high hydraulic head gradients over faults observed in shallow sedimentary aquifers [e.g., Bense and Van Balen, 2004; Mayer et al., 2007] on the one hand suggests that faults form effective barriers to groundwater flow supporting the conceptual model of fault hydrology applicable to unconsolidated sediments proposed by Rawling et al. [2001]. The barrier character of faults is usually attributed to a reduction in fault permeability as a result of the ductile folding of clay along the fault plane and cataclastic deformation reducing permeability in the damage zone around the fault core [e.g., Rawling et al., 2001; Eichhubl et al., 2005]. On the other hand, faults in unconsolidated sediments are also often described as conduits to fluid flow or combined conduit-barrier systems [e.g., Bredehoeft et al., 1992; Bense and Person, 2006]. In the latter case faults are conceptualized as zones with a strongly anisotropic permeability (vertical permeability ≫ horizontal permeability) so that fluid flow is focused along the fault plane while a considerable hydraulic head drop across the fault is maintained. A similar model has been advocated by Caine et al. [1996] for brittle faults with a cataclasite-rich fault core and fracture-dominated damage zone. Thermal anomalies arising from focused fluid flow in such a system are discussed for example by Anderson and Fairley [2008] and references therein. In this letter, however, we analyze new geothermal data from the Lower Rhine Embayment, Germany that illustrate the conduit-barrier behavior of fault zones in unconsolidated sedimentary aquifer systems.

2. Geothermal Data and Analysis

2.1. Field Area and Data Collection

[3] The Lower Rhine Embayment (LRE; Figure 1a), Germany, is part of the Roer Valley Rift System which is the southward extension of the North Sea Basin. Since the Oligocene to recent a thick sequence (up to 1200 m) of shallow marine to fluvial sediments has been deposited (Figure 1b). Interbedded with sand and clay units are laterally continuous lignite seams of Miocene age. This sequence of unconsolidated sediments is cut by numerous NW-SE striking normal-faults. Open-pit lignite mining operations have resulted in large (>300 m) hydraulic head declines since the 1960s resulting in anomalously high hydraulic head gradients across many of these faults (Figure 1a).

Figure 1.

(a) Hydraulic head map of the Lower Rhine Embayment (LRE) showing location of wells discussed in this study. Strong hydraulic head gradients across the major faults in the area suggest that these faults act as barriers to lateral groundwater movement in the area. Hydraulic head contours are representative of the present-day situation in the relatively deep (e.g., 300 m) aquifers in the area and largely based upon observations within a dense network of groundwater observation wells. (b) Schematic geological cross-section along A-A′ illustrating the structure of the LRE (based upon Schäfer et al. [1996]). Underneath the Oligocene sediments basement rock of Mesozoic and older age is found.

[4] Geothermal data in the LRE were gathered from ∼20 piezometer nests installed in boreholes. In the discussion presented here we focus on three temperature-depth profiles obtained in the Erft Block and of which two are in close proximity to the Rurrand fault (Figure 1). A silicon temperature probe (manufactured by In-Situ Inc.; listed resolution of 0.01 K with a response time of ∼30 s) was slowly lowered in each well and temperature was recorded with depth at 2 meter intervals. This approach avoided the possibility that the temperature distribution in the well bore would be artificially disturbed by convective mixing. At each depth temperature was allowed to fully stabilize, which usually took less than three minutes. Well diameters were always ≃5 cm, which is narrow enough to almost everywhere exclude the possibility of disturbing convective effects already present in the well when temperature gradients are lower than 0.1°C m−1 [e.g., Sammel, 1968]. Because piezometer nests were present it was not always possible to collect temperature readings at all depths from a single borehole.

2.2. Data Interpretation

[5] Well 87521 is located centrally in the Erft block over a kilometer away from any significant mapped fault zone. The temperature profile observed at this location is probably relatively undisturbed by groundwater flow effects and reflects conductive geothermal conditions (Figure 2a). As is seen here, the lignite layers are easily identified in the temperature-depth profile as a result of the their low thermal conductivity. In wells 87263 and 87581, both situated within ∼500 from the Rurrand fault zone, conspicuous negative (∼2°C) and positive (∼8°C) thermal anomalies are observed respectively as compared to the presumed purely conductive profile (Figures 2b and 2c). In well 87263 we suspect that downward flow, driven by hydraulic drawdown in deeper aquifers, of relative cool water along the Rurrand fault is causing the observed thermal anomaly. On the other hand, relatively shallow groundwater abstraction could result an up coning of deeper water along the fault zone explaining the thermal anomalies seen in well 87581. The two distinct warm zones in the latter well could indicate different groundwater flow paths associated with a complex fault zone structure. At distances larger than a ∼500 m from major fault zones we have not encountered temperature anomalies like those seen in wells 87263 and 87581.

Figure 2.

Temperature-depth profiles and lithologies for groundwater observation wells (a) 87521, (b) 87263 and (c) 87581. Well locations are indicated in Figure 1. Symbols indicate the measured temperatures in the well, dashed line indicates interpreted temperature distribution. The solid line represents the interpreted undisturbed thermal profile due to conduction only. Black dots along the lithological description indicate the depth to which each tube of the piezometer nest installed in the borehole reaches. See the text for discussion.

[6] Similar temperature reversals as those discussed above have been interpreted as transient features resulting from convective heat-transfer in fault-charged geothermal reservoirs [e.g., Ge, 1998]. Along these lines, we hypothesize that the temperature anomalies we observe in the LRE reflect transient flow conditions that probably have developed over a time-period of decades in response to mine dewatering in the Erft block.

2.3. Numerical Modeling

[7] In order to evaluate whether the hydrogeological scenarios discussed above can indeed explain the general character of the observed temperature anomalies we have used a numerical modeling approach incorporating the geological setting of the Rurrand fault in an idealized fashion (Figure 3a). A generic finite-element code [PDE Solutions, 2006] was used to solve the heat-transport equation [e.g., Bredehoeft and Papadopulos, 1965] to describe transient heat-transfer by advection through groundwater flow and conduction in a saturated porous medium:

equation image

in which equation image [ms−1] is the Darcy velocity of groundwater flow, T is temperature [°C], c (2.0·103Jkg−1 °C−1) and cw (4.2·103Jkg−1 °C−1) are the specific heat of the saturated sediment and water, respectively, ρ (2.1·103kgm−3) and ρw (1·103kgm−3) are their respective mass densities, κ is the thermal conductivity [Wm−1 °C−1] of the saturated sediment. The applied hydrogeological and heat flow boundary conditions to calculate transient fluid fluxes (equation image), using a Darcy equation, and temperatures (equation 1) are described below.

Figure 3.

Transient hydraulic head and temperature distributions calculated for scenarios of (a–c) upward and (d–f) downward fluid flow along the fault plane (F). The dashed line in Figure 3a and 3d represents a typical fluid flow path for each scenario to help understand the calculated temperature distributions. Hydraulic boundary conditions for each scenario are indicated in Figures 3a and 3d, as well as the hydraulic head distribution at t = 40 a. For both scenarios groundwater temperatures are calculated for 20 a (Figures 3b and 3e) and 40 a (Figures 3c and 3f) after fluid flow in the system started.

[8] Table 1 summarizes the permeability and thermal property values of the model units. Fault throw is set as 250 m resulting in a fault width of 10 m in accordance with published fault-width versus fault throw relationships [e.g., Shipton et al., 2006]. Vertical fault permeability is three orders of magnitude larger than across-fault permeability mimicking the effects of clay smearing and drag of sand from aquifers along the fault plane. Because the geological build-up around each of the wells 87263 and 87581 is not known in great detail, the model stratigraphy is based upon a generalized representation of the geology along the Rurrand fault. We note that as a result of this approach we will not accurately reproduce the geology found within each observation well at the places where we compare the modelled temperature-depth profiles with those observed.

Table 1. Horizontal (kh) and Vertical Permeability (kv), and Thermal Conductivity (κ) Values of the Units Considered in the Models Presented in Figure 3a
Unitlog(kh)[m2]log(kv)[m2]κ [Wm−1°C−1]
  • a

    Permeability values are based upon work presented by Wallbraun [1992]. Thermal property values in this table are constrained by in situ measurements reported by Koch et al. [2008].


[9] In the first model scenario, hydraulic drawdown (incorporated by application of a flux boundary) occurs at relatively shallow depth in the foot-wall of the fault zone as a result of which flow is predominantly upward along the fault (Figure 3a). In the second scenario, groundwater abstraction takes place in deeper aquifers generating downward flow along the fault (Figure 3d). Starting conditions for both model scenarios are identical. Initially (t = 0), no fluid flow is occurring and the temperature field is in a steady-state governed by heat-conduction only in combination with the thermal boundary conditions. Surface temperature is fixed to 10°C, and groundwater recharge is limited to 10−3 m/day. A 1-D thermal model of pure heat-conduction for well 87521 was used to infer a basal heat flow of 75·10−3 W/m2 which is used as the lower thermal boundary condition. The total simulation time is 40 a covering the time-frame over which regional scale drawdown of hydraulic heads in the LRE developed in combination with. In the first 30 a of the simulation pumping rates in the aquifers, i.e. the fluid flux boundary values along the right-hand side and left-side of the model domain in Figures 3a and 3d respectively, are linearly increasing with time until the values shown in Figure 3, and are kept constant after t = 30 a. Figures 3a and 3d show the hydraulic head distributions at t = 40 a. Strong hydraulic head drops develop across the fault zone similar in magnitude to those observed in the LRE. The calculated transient temperature distributions are considered for selected time slices (t = 20, 40 a) and at specific locations (1 and 2; indicated in Figure 3).

[10] For scenario 1, model results show that a temperature-depth profile in the footwall block close to the fault (profile 1; Figure 4a) shows temperature reversals comparable in magnitude to those observed in well 87581 (Figure 2c). Large negative anomalies are observed in case groundwater flow along the fault is assumed to be predominantly downward. Profile 2 (Figure 4c), taken 300 m downstream of the fault zone, shows 4 to 8°C of cooling respectively 20 and 40 a after the start of the simulation. The negative reversals we find along a temperature-depth profile at this location are very similar in magnitude and pattern to those observed in well 87263 that is located downstream of the Rurrand fault (Figure 1).

Figure 4.

Calculated temperature-depth profiles associated with upwelling (a) and downwelling (b) of groundwater flow along faults after 0 (solid black lines), 20 (dashed line), and 40 (gray line) years of water withdraws associated with lignite mining at different depths. The temperatures at the start of the simulation reflect conductive heat transfer conditions. The location of profiles 1 and 2 is indicated in Figures 3b and 3e, respectively. The lithological logs accompanying each profile are based upon the conceptualized stratigraphy used in the model simulations.

3. Discussion and Conclusion

[11] The analysis presented above suggests that the strong positive thermal anomalies (up to +8°C; well 87581) and negative anomalies (e.g. −2°C; well 87263), are transient features that result respectively from upward- and downward-groundwater flow, resulting from by recent mine de-watering, along the Rurrand fault and subsequent lateral flow into the aquifers flanking the fault zone. In such system steady-state temperature distributions would only be reached after thousands of years [Ziagos and Blackwell, 1986]. Major faults in the LRE thus likely form the loci for vertical pathways between aquifers at different depths that are elsewhere not hydraulically connected. However, at the same time strong horizontal hydraulic head gradients indicate that faults are hampering lateral fluid flow. We conclude that these faults do behave as conduit-barrier systems in line with, the fault hydrology model proposed by Bense and Person [2006]. The pure barrier model advocated by Rawling et al. [2001] will fail to explain the observed thermal anomalies in the LRE. Groundwater fluxes in the system we studied are artificially enhanced. In systems where fluid flow rates along faults are lower thermal anomalies to the extend we observed in the LRE might not develop. In such setting the double behaviour of faults as conduit-barrier systems could go unnoticed. Nevertheless, the long term leakage of contaminants or saline water between aquifers along fault zones that are presumed to be efficient barriers can have an unexpected negative impact on the quality of water resources.


[12] This study is being supported by the National Science Foundation (NSF-EAR 0609809) awarded to VB and MP. Review comments by Jerry Fairley and two anonymous reviewers improved the quality of the original manuscript.