Professional Documents
Culture Documents
Abstract
Faults can act as flow barriers or conduits to groundwater flow by introducing heterogeneity in permeability. We examine the
hydrogeology of the Sandwich Fault Zone, a 137 km long zone of high-angle faults in northern Illinois, using a large-scale historic
aquifer test. The fault zone is poorly understood at depth due to the majority of the faults being buried by glacial deposits and its
near-vertical orientation which limits geologic sampling across faults. The aquifer test—perhaps one of the largest in terms of overall
withdrawal in North American history—was conducted in 1942 at a facility adjacent to the fault zone. More than 34,000 m3 /day
was pumped for 37 days from nine multiaquifer wells open to the stratified Cambrian-Ordovician sandstone aquifer system. We
modeled the aquifer test using a transient MODFLOW-USG model and simulated pumping wells with the CLN package. We tested
numerous fault core/damage zone conceptualizations and calibrated to drawdown values recorded at production and observation
wells. Our analysis indicates that the fault zone is a low-permeability feature that inhibits lateral movement of groundwater and that
there is at least an order of magnitude decrease in horizontal hydraulic conductivity in the fault core compared to the undeformed
sandstone. Large head declines have occurred north of the fault zone (over 300 m since predevelopment conditions) and modifying
fault zone parameters significantly affects calibration to regional drawdown on a decadal scale. The flow-barrier behavior of the
fault zone has important implications for future groundwater availability in this highly stressed region.
(Ball et al. 2010), or detailed study of fault geometries a crystalline-rock aquifer and observed changes in vertical
and modeling of fracture networks (Jourde et al. 2002; flow pathways under pumping and ambient conditions.
Minor and Hudson 2006). These studies determine varia- The purpose of this study is to present an analysis
tions in permeability for various fault components to infer of a large-scale aquifer test (referred to as the Arsenal
groundwater flow conditions on a larger scale. Test) conducted adjacent to a vertical fault system
A growing number of studies have used field within the highly stressed Cambrian-Ordovician sandstone
hydrogeological approaches (e.g., slug testing, aquifer aquifer system in northeastern Illinois. To the authors’
testing, and mapping of heads), analytical modeling, or knowledge, this is perhaps one of the largest confined
numerical modeling, to understand the effect of fault aquifer tests in North American history in terms of
zones on local and regional groundwater flow (see Bense overall withdrawal. The magnitude of withdrawals and
et al. 2013 for a complete review). For example, Marler orientation of the pumping wells provides a unique
and Ge (2003) studied the permeability of the Elkhorn opportunity to study a faulted multiaquifer system at a
Fault Zone in Colorado using slug testing at boreholes considerable depth (>450 m) on a regional scale (tens
across the trace of the fault and simulated flow using of kilometers). We analyze the test both analytically and
two-dimensional numerical modeling. Ball et al. (2010) numerically to determine hydraulic conductivity for 22 km
expanded upon Marler and Ge’s study by incorporating along the length of the fault zone and use our results to
geophysical surveys to map heads across the fault zone update a regional groundwater model that simulates over
and the structure of the fault zone at depth. Both 150 years of withdrawals. We then discuss implications
studies indicate a complex conduit-barrier behavior of of our updated fault zone conceptualization for long-term
the fault zone at a local scale (tens of meters). Mayer groundwater availability.
et al. (2007) estimated conductance for a section of the
Mission Creek Fault Zone in California, which offsets
unconsolidated conglomerates and debris fan deposits, by Geology and Hydrogeology of the Study Area
calibrating a numerical groundwater model to historic The Cambrian-Ordovician (CO) sandstone aquifer
head observations. Burbey (2008) utilized a 62-day system underlies all of northern Illinois and extends into
aquifer test to infer the influence of a small (5 m wide) Wisconsin, southeastern Minnesota, and eastern Iowa. It
fault zone within a thick alluvial aquifer in Nevada. has been a principal groundwater source for municipal
Medeiros et al. (2010) also utilized pump tests to infer and industrial supplies in northeastern Illinois since
hydraulic conductivity across deformation bands within the 1860s. The CO sandstone system consists of the
a shallow sandstone aquifer, but did not detect a strong St. Peter sandstone and underlying Ironton-Galesville
flow barrier despite large differences between host rock sandstone, which are overlain by thick sequences of
and deformed rock permeabilities measured at the outcrop carbonates, shales, and glacial deposits and thus have
scale. More recently, Roques et al. (2014) conducted a limited direct recharge (Figure 1A). Since predevelopment
63-day aquifer test in a highly permeable fault zone within conditions, over 300 m of static head decline has occurred
JA, Joliet Arsenal; SP, St. Peter; IG-Ironton-Galesville; PW, Pumping Well; OW, Observation Well
Figure 2. (A) Withdrawals from production wells during the Arsenal Test. Note that several wells turned off during the test,
most notably JA-1 and JA-2. Small withdrawals during the recovery phase (after day 37) were not recorded. The withdrawal
rate for JA-8 during the recovery phase was estimated. (B) Recorded heads at the production wells and nearby observation
well (JA-5). Note the small pumping events during the recovery phase.
Modeling Methods and extends into southern Wisconsin, eastern Iowa, and
western Indiana. For a complete description of model
Illinois Groundwater Flow Model parameters, boundary conditions, and hydrostratigraphic
We used a transient 21-layer MODFLOW model units, see Abrams et al. (2018a).
of the CO aquifer system already developed for long- The St. Peter and Ironton-Galesville sandstones are
term water supply planning purposes as a base model to represented in the regional model with a K h of 1.5 m/day
simulate the Arsenal Test. This regional model (known and a vertical hydraulic conductivity (K v ) of 0.18 m/day.
as the Illinois Groundwater Flow Model) consists of 12 Multiaquifer wells throughout the model domain are
bedrock layers and 9 unconsolidated layers on a 762 m simulated with high K v zones in the aquitards that separate
grid and simulates annual stress periods from 1863- the sandstone units (following Mandle and Kontis 1992)
present, as well as the impacts of future demands. The and utilize the Time Variant Materials package available
regional model covers the northern two-thirds of Illinois in Groundwater Vistas 6.0 (Rumbaugh and Rumbaugh
Unrestricted flow Fault zone same as host rock 1.5 1.5 1.5
Restricted flow Damage zone K h reduced by 1.5 0.76 0.15
half, fault core K h reduced by
order of magnitude
Highly-restricted flow Damage zone and fault core K h 1.5 0.015 0.015
reduced by two orders of
magnitude
production wells on days with either 24 h of pumping and 4B), but had less of an effect at JA wells 1 to 4 and
or 24 h of no pumping. We also used drawdowns at 6 (Figure 4C, 4D, and 4E). JA-10 achieved the best cal-
surrounding municipal wells observed at the end of the test ibration in the restricted flow scenario with a root mean
as calibration targets. Observed drawdown curves were square error (RMSE) of 1.5 m and a regression line slope
compared to our simulated drawdown curves using linear of 0.93 (Figure 5E), whereas in the unrestricted or highly-
regression and by comparing the slopes of the regression restricted flow scenarios there was either not enough or too
lines. much simulated drawdown (Figure 4A) resulting in larger
Recovery observations were recorded graphically at RMSEs and deviations of the regression line (Figure 5D
the time of the test (Figure 2B), but withdrawal rates and 5F). The differences between simulated drawdowns
were not recorded when individual wells turned on and for each of the three scenarios becomes progressively less
off during the recovery phase. Given the unknown rates as the distance between the production wells and the fault
of withdrawals and the short duration of these pumping zone increases (compare hydrographs for JA wells 10 and
events (many of which were less than 12 h), we did 9 versus JA wells 4 and 2 in Figure 4). The far-field
not use these recovery observations as calibration targets. municipal wells were largely unaffected by whether the
Because withdrawal rates were not recorded for JA-8, but fault zone was represented as a flow barrier or not. In all
was noted as always pumping, we calculated the average three scenarios, the mean residual of simulated drawdown
withdrawal rate of JA-8 during the test and applied that for the far-field municipal wells was ≤0.3 m (Table 3).
to the recovery phase (Figure 2A). Skin factors (in addition to fault zone modifications)
were needed to achieve calibration at the JA production
wells and ranged from −2.5 to 8 (Table 1). The negative
Arsenal Test Observations and Model Results skin factor for JA-6 indicates that the well bore skin
The cone of depression formed by the Arsenal Test has a higher K h than the surrounding formation, whereas
reached a peak drawdown of over 90 m at its center positive skin factors indicate that well bore skins have
and extended outward at a distance of greater than lower K h values than the surrounding formation. A
20 km (Video S1). Review of heads, drawdown, and negative skin factor is unsurprising, given that some
recovery observations recorded during and after the test of the JA production wells (including JA-6) were
indicates that the fault zone inhibits horizontal flow in the blasted with nitroglycerine to increase fracturing in the
sandstone units. This is supported by observations that: Ironton Galesville. Skin factor adjustments resulted in
(1) drawdowns at both JA-9 and JA-10 were the same a decrease or increase in the magnitude of simulated
magnitude (or greater) in comparison to drawdown at JA drawdown but did not result in changing the shape
wells 6 to 8 (Table 2) despite lower withdrawals rates of the simulated drawdown curves (Figure S1), whereas
occurring at JA-9 and JA-10 compared to JA wells 6 to changing fault zone parameters did affect the shape
8 (Figure 2A), (2) JA-9 and JA-10 had slower recoveries of the drawdown curves at wells closes to the fault
in comparison to JA-6 and JA-7 while JA-8 continued (Figure 4A and 4B).
to pump during the recovery phase (Figure 2B), and (3) The restricted flow scenario achieved the best overall
in general, recovery occurred more quickly at the wells calibration with an RMSE of 3.37 m and a slope of
further away from the fault zone (JA wells 1-4) compared the regression line nearest to 1 in comparison to the
to the wells closer to the fault zone (JA wells 6-10) other scenarios (Figure 5B). Overall calibration results
(Figure 2B). Full recovery did not occur at the JA wells did not vary greatly between each of the three scenarios,
after 25 days and drawdown continued to occur north of however, due to the limited influence of the fault zone
the fault zone even after this period (Video S1). on drawdowns in the far-field municipal wells and the JA
Modifying the properties of the fault zone signifi- production wells located further away from the fault zone.
cantly affected calibration at production wells closest to While all nine JA wells were very productive at the
the fault zone, most notably at JA-10 and JA-9 (Figure 4A time of drilling, JA-5 (the dedicated St. Peter monitoring
well) was the exception. JA-5 was tested prior to the along the fault zone in Ogle County, lends geologic
start of the Arsenal Test and determined to have a very evidence to a reduced permeability within the fault zone.
low specific capacity, and there was also a very subdued Our modeling efforts determined an “effective”
recovery curve at JA-5 following the end of the test hydraulic conductivity (Bense et al. 2013) for the fault
(Figure 2B). Both the pump test and recovery curve zone in Will County, but the specific mechanisms that
suggest the St. Peter is not as transmissive around JA-5 cause reduced flow across the fault zone are unknown.
in comparison to the other production wells or that there Mechanisms could include: (1) the presence of gouge
was an issue with the well such as formation collapse. In within a fault core or series of fault cores (Rawling
all three scenarios, the largest systematic error occurred at et al. 2001; Shipton and Cowie 2001; Caine and Minor
JA-5 (Figure 4F), likely because this local zone of reduced 2009), (2) dissolution and cementation within the core
permeability was not incorporated into the three model or damage zone fracture networks (Eichhubl et al. 2009),
scenarios. We thus left out the results of JA-5 in our or (3) cataclasis in broad deformation bands bounded by
regression analyses. slip planes (Antonellini and Aydin 1994; Shipton et al.
2002; Fossen et al. 2007; Medeiros et al. 2010). Overall
throw along the fault zone is well defined, but there
Discussion and Implications for Groundwater may be many more upthrown and downthrown blocks
Supply within the fault zone than currently mapped, resulting in
Results of our model comparisons support previous complex structural geometries. A geophysical approach
observations (Abrams et al. 2015, 2018a) and modeling could determine some of these specific mechanisms and
efforts (Roadcap et al. 2013) that there is reduced fault structural controls (e.g., Ball et al. 2010). Clay smearing
zone permeability within the St. Peter and Ironton- (Bense and Person 2006; Egholm et al. 2008) could also
Galesville sandstones which limits groundwater flow have a significant influence on groundwater flow in the
across the fault zone. Decreasing the K h of the fault core central portion of the fault zone (Dekalb County) since
by an order of magnitude and the K h of the damage zone there is enough offset to drag the Maquoketa and Eau
by half achieved the best calibration at wells closest to Claire shales to meet the sandstone units.
the fault zone and overall. Sections of brecciated quartzite We analyzed how incorporating the restricted flow
within the St. Peter, observed by the authors at outcrops fault zone conceptualization into the Illinois Groundwater
1980 2014
OFZ RFZ RFZ + Ss OFZ RFZ RFZ + Ss
Ironton Galesville1
Mean error (m) −3.85 0.87 −0.14 −8.36 3.09 −0.92
Absolute mean error (m) 13.12 11.69 12.51 16.29 12.30 13.23
RMSE (m) 17.69 15.36 15.73 19.74 14.96 15.81
St. Peter2
Mean error (m) −8.87 −8.31 −8.35 −23.64 −22.19 −19.93
Absolute mean error (m) 20.54 17.21 18.62 27.50 24.80 24.55
RMSE (m) 24.26 21.19 22.67 31.98 28.76 28.56
OFZ, original fault zone conceptualization; RFZ, restricted flow fault zone conceptualization; RFZ + Ss, restricted flow fault zone conceptualization plus updated Ss
and Maquoketa Shale K v values.
1 n = 62 for 1980, n = 42 for 2014.
2 n = 47 for 1980, n = 16 for 2014.
Conclusions Production wells further from the fault zone, and the
We used a numerical modeling approach to simulate far-field observation wells, were less affected by modi-
a large-scale historic aquifer test that occurred adjacent to fying fault zone properties.
a poorly understood near-vertical fault zone system. The We then analyzed how the restricted flow fault
fault zone system is structurally complex with variable zone conceptualization affects calibration to heads in
the CO sandstone system post-1963, when there is
offset and numerous fault traces bounding upthrown and
considerably more uncertainty in the system because of
downthrown blocks. Observations of head offset and low
multiaquifer wells and desaturated conditions. Updating
specific capacities near the fault zone indicate that the fault
the Illinois Groundwater Flow Model with the restricted
zone is an important structural control on groundwater
flow scenario significantly improved calibration to
movement in the CO sandstone aquifer system.
CO sandstone water levels for years in which mass
The magnitude of groundwater withdrawals and the
measurement of sandstone water levels took place (1980
duration of the Arsenal Test allowed insight into the and 2014). The flow-barrier behavior of the fault zone
hydraulic properties for a large area of the fault zone has important implications for groundwater availability
(>20 km) at a considerable depth (>450 m) for a confined in the highly stressed CO sandstone aquifer system.
multiaquifer sandstone system. We conceptualized the Communities near the fault zone continue to rely on the
fault zone in the sandstone units as a low-permeable fault deep sandstone system despite increasing drawdown, and
core surrounded by a damage zone based on previous are expected to increase withdrawals before alternative
research of deformation bands in faulted sandstones. The supplies can be utilized. Projected desaturation of the
K h of the fault core and damage zone were modified until Ironton-Galesville sandstone may occur sooner than the
the magnitude and shape of the simulated drawdown previously estimated year of 2040 given our updated
curves matched observed records at the production fault zone conceptualization, and many wells could face
wells and at surrounding observation wells. We presented serious issues under pumping conditions by 2030. Results
results from three scenarios: an unrestricted flow scenario, from this study may inform the hydraulic properties of
a restricted flow scenario, and a highly-restricted flow other faulted sandstone aquifers in confined settings and
scenario. help in long-term groundwater management.
Our study determined an effective hydraulic conduc-
tivity for the fault zone and indicates that there is at least
Acknowledgments
an order of magnitude reduction in K h of the fault core
compared to the undeformed sandstones. We achieved This project was funded in part by the Illinois
Department of Natural Resources under the purview of
the best calibration to drawdown levels (overall and at
water supply planning. We thank Chris Langevin, John
wells closest to the fault) under the restricted flow sce-
Quinn, and an anonymous reviewer for their positive and
nario, in which the K h of the undeformed sandstones,
helpful comments that improved this manuscript.
damage zone, and fault core were set to 1.5, 0.76, and
0.15 m/day, respectively. The unrestricted flow scenario
and the highly-restricted flow scenario either allowed too Authors’ Note
much or too little flow across the fault zone to achieve The authors do not have any conflicts of interest or
calibration at the production wells closest to the fault. financial disclosures to report.