Selle 2013

Download as pdf or txt
Download as pdf or txt
You are on page 1of 10

Environ Earth Sci (2013) 69:443–452

DOI 10.1007/s12665-013-2333-z

SPECIAL ISSUE

Recharge and discharge controls on groundwater travel times


and flow paths to production wells for the Ammer catchment
in southwestern Germany
B. Selle • K. Rink • O. Kolditz

Received: 27 July 2012 / Accepted: 15 February 2013 / Published online: 6 March 2013
 Springer-Verlag Berlin Heidelberg 2013

Abstract Travel times and flow paths of groundwater Different recharge scenarios resulted in a comparable fit to
from its recharge area to drinking-water production wells observed water levels, and similar estimates of hydraulic
will govern how the quality of pumped groundwater conductivities, flow paths and travel times of groundwater
responds to contaminations. Here, we studied the 180 km2 to production wells. Travel times calculated for all scenar-
Ammer catchment in southwestern Germany, which is ios had a plausible order of magnitude which were com-
extensively used for groundwater production from a car- parable to apparent groundwater ages modelled using
bonate aquifer. Using a 3-D steady-state groundwater environmental tracers. Scenario with groundwater dis-
model, four alternative representations of discharge and charge across the entire streambed of the Ammer River and
recharge were systematically explored to understand their its tributaries resulted in a better fit to water levels than
impact on groundwater travel times and flow paths. More scenario with discharge at a few springs only. In spite of the
specifically, two recharge maps obtained from different poorer fit to water levels, flow paths of groundwater from
German hydrologic atlases and two plausible alternative the latter scenario were more plausible, and these were
discharge scenarios were tested: (1) groundwater flow supported by the observed major ion chemistry at the pro-
across the entire streambed of the Ammer River and its duction wells. We concluded that data commonly used in
main tributaries and (2) groundwater discharge via a few groundwater modelling such as water levels and apparent
major springs feeding the Ammer River. For each of these groundwater ages may be insufficient to reliably delineate
scenarios, the groundwater model was first calibrated capture zones of wells. Hydrogeochemical information
against water levels, and subsequently travel times and flow relating only indirectly to groundwater flow such as the
paths were calculated for production wells using particle major ion chemistry of water sampled at the wells can
tracking methods. These computed travel times and flow substantially improve our understanding of the source areas
paths were indirectly evaluated using additional data from of recharge for production wells.
the wells including measured concentrations of major ions
and environmental tracers indicating groundwater age. Keywords WESS  Water Earth System Science 
OpenGeoSys  OGS
B. Selle (&)  O. Kolditz
Water and Earth System Science (WESS) Competence Cluster,
c/o University of Tübingen, Hölderlinstraße 12, Introduction
72074 Tübingen, Germany
e-mail: benny.selle@uni-tuebingen.de
Travel time and flow paths of groundwater from its
K. Rink  O. Kolditz recharge area to drinking-water production wells will
Department of Environmental Informatics, Helmholtz Centre govern how the quality of pumped groundwater responds to
for Environmental Research-UFZ, Leipzig, Germany contamination, particularly from near surface sources such
as agriculture or urban wastewater (Alley et al. 2002). For
O. Kolditz
Applied Environmental Systems Analysis, many groundwater flow systems surrounding production
TU Dresden, Dresden, Germany wells, steady-state models were used to map capture zones

123
444 Environ Earth Sci (2013) 69:443–452

of production wells. In this context, computed velocity Materials and methods


fields and particle-tracking methods have been routinely
applied to quantify how long groundwater travels to wells Study catchment
and which flow paths it takes (Chiang 2005, p. 187).
However, these model calculations are inherently uncertain The studied catchment is the approximately 180-km2 area
due to an inaccurate representation of heterogeneity in contributing groundwater flow to the Ammer River upstream
groundwater systems, and this source of uncertainty has of Pfäffingen gauging station (Fig. 1). The Ammer River is a
received considerable attention in the literature (e.g. Gua- northern tributary to the Neckar River. It has an average
dagnini and Franzetti 1999; Harrar et al. 2003; Refsgaard discharge of *1 m3/s which mainly originates from
2011). Another important source of uncertainty—which groundwater. The geology of the catchment comprises a
has received less attention in groundwater modelling—is sequence of Triassic strata forming a landscape characterised
the insufficiently known spatial distribution of recharge to by escarpments. The topography of the catchment is hilly
and discharge from the aquifers which greatly shapes with altitudes ranging between 312 and 600 m asl. There is a
groundwater flow systems. This is of particular importance prominent escarpment of the Schönbuch forest in the eastern
for areas with fractured and karstified aquifers (Praamsma part of the catchment. Annual average precipitation is
et al. 2009; Zanini et al. 2000). For groundwater recharge, between 700 and 800 mm with a maximum during summer.
several methods have been applied to indirectly estimate Mean annual air temperature is about 8 C. The catchment
the spatial distribution of long-term average groundwater has an above-average proportion of urban (17 %) and agri-
recharge including spatially distributed hydrological mod- cultural areas (71 %). The hydrogeological setting is domi-
elling (e.g. Githui et al. 2012) and the regionalisation of nated by the Upper Muschelkalk (mo) which forms the main
catchment-scale recharge estimated from long-term aver- aquifer. It has an average thickness of 80–90 m, is partly
age baseflows (e.g. Armbruster and Leibundgut 2001). As a karstified and mainly consists of dolomitic, micritic and
consequence, different estimates of long-term average bioclastic limestone. The main aquifer slightly dips towards
recharge exist and can be applied as inputs to groundwater the southeast and is unconfined in the northwest of the study
models. For groundwater discharge, it is typically poorly catchment. It becomes confined towards the southeast where
understood how and where groundwater flows into surface it is increasingly overlain by Keuper layers (Villinger 1982).
water bodies such as rivers. Even for well-studied catch- These layers are the Lettenkeuper (ku: 20 m), the Gipskeuper
ments, it often remains unclear whether groundwater (km1: up to 100 m), the Schilfsandstein (km2: 20 m), the
mainly discharges as a localised process happening, e.g. via Bunte Mergel (km3: 25 m) and the Stubensandstein (km4:
springs or as a rather diffuse process affecting the entire 30 m) mainly consisting of terrestrial sediments including
streambed (Gleeson et al. 2009). gypsum. The latter is characteristic for km1. Groundwater
In this study, different alternative representations of flow is governed by the stratigraphic dip of the hydrogeo-
discharge and recharge were systematically explored to logical units. The units of km2, km3 and km4 are limited to
understand their impact on groundwater travel times and the Schönbuch forest where recharge tends to be low, and
flow paths. For this purpose, a 3-D steady-state groundwater hence, these units play a minor role for groundwater flow in
model was applied for the 180 km2 Ammer catchment in the catchment. The primary discharge of the aquifer system
southwestern Germany, which is extensively used for in the study area is groundwater discharge to the Ammer
groundwater production from a carbonate aquifer. Two River and its tributaries, which mainly comes from the mo
recharge maps obtained from different German hydrologic aquifer and also from the km1 aquifer. However, the relative
atlases and two alternative and plausible representations of contributions of focused groundwater discharge via major
groundwater discharge to the Ammer River and its tribu- springs and diffuse discharge processes across the entire
taries were tested in a total of four scenarios. For each streambed were unclear. Secondary discharge includes
scenario, the groundwater model was first calibrated against groundwater pumping from the mo with an average pro-
water levels and subsequently groundwater travel times and duction of 150 l/s of drinking water from four well sites.
flow paths were calculated for groundwater production Groundwater recharge from streams may be locally impor-
wells using particle tracking methods. These computed tant, e.g. for the Kochart Creek. Additional information on
groundwater travel times and flow paths were indirectly the Ammer catchment can be found in Grathwohl et al.
evaluated using data on the chemical and isotopic compo- (2013).
sition of well-water including measured concentrations of The catchment for groundwater discharge into the
major ions, and tritium and sulphur hexafluoride utilised for Ammer River has been investigated in detail by Harress
groundwater age dating. In addition to the analysis for (1973), Villinger (1982) and Plümacher (1999). However,
production wells, flow paths of groundwater to major the source area of recharge for the groundwater production
springs were evaluated for the various scenarios. wells was unknown.

123
Environ Earth Sci (2013) 69:443–452 445

Fig. 1 3-D view of the Ammer


catchment with the Ammer
River and two tributaries, the
Kochart and the Käsbach Creek,
as represented in discharge
scenario 1 (dotted black lines).
Gipskeuper springs (squares)
and Upper Muschelkalk springs
(diamonds) as represented in
discharge scenario 2. Drinking-
water production well sites
(circles W1, 2, 3, 4).
Hydrogeological units: the
Upper Muschelkalk (mo), the
Gipskeuper (km1), the
Lettenkeuper (ku) at the central
and the lower parts (anticline of
Reusten) of the figure, the
Schilfsandstein (km2), the Bunte
Mergel (km3) and the
Stubensandstein (km4) at the
right edge of figure, i.e. the
escarpment of the Schönbuch
forest. Inset in the lower right
corner displays locations of
colour coded observation wells
used for calibration. Inset in the
upper right corner shows a
cross-section; dashed line
indicates its approximate
location

Groundwater modelling network consisting of the Ammer River and its two main
tributaries: the Käsbach and the Kochart Creek and (5)
OpenGeoSys locations of four groundwater production well sites. Grids of
subsurface layers were interpolated using stratigraphic data
The steady-state groundwater model for the Ammer of over 100 boreholes and a digitised map of the geologic
catchment was implemented using the software Open- layering (Reuther 1973). Using the local polynomial method
GeoSys. It is a scientific open source project for the for grid interpolation within SURFER (Golden Software
development of numerical methods to simulate thermo- 2002), we obtained a cross-validated root-mean-square error
hydro-mechanical-chemical processes in porous and frac- of\5 m for the elevation data utilised. After the meshing, the
tured media based on the finite element method (Kolditz domain discretisation of the whole study area had a total of
et al. 2012). OpenGeoSys allows users to import data from 57,980 elements (prisms and tetrahedra). All mesh elements
various sources such as geographic information systems or had a maximum edge length of 110 m and the mesh was
other modelling software and to perform validation of input incrementally refined towards the rivers and the production
data, model setup, simulation and the analysis of results all well sites. For more details on the mesh generation proce-
within one software framework (Rink et al. 2012). dure, see Rink et al. (2013). An isotropic hydraulic con-
ductivity was assigned to each element according to its
Aquifer geometry and hydraulic properties respective hydrogeological unit; and values of hydraulic
conductivity were calibrated as detailed below. Note that
For 3-D mesh generation, we used: (1) a digital elevation conductivities probably vary within each of the aquifer units,
model minus soil thicknesses, (2) the boundary of the model in particular due to likely karstification of the mo and the km1
region, i.e. the Ammer catchment, (3) raster subsurface aquifers. However, we did not have the information to
layers representing the lower boundaries of the four hydro- explicitly represent the karst network in the model, e.g. using
geological units, i.e. mo, ku, km1 and km2…4, (4) the river a double permeability model simulating coupled flow in the

123
446 Environ Earth Sci (2013) 69:443–452

fissured matrix and the karst conduits (Bauer et al. 2003; characteristics such as landuse, soils and geology. For both
Sauter et al. 2006). By applying an equivalent porous-media WaBoA and HAD, regression models were fitted using
concept, we basically assumed that the scale of heterogeneity data on baseflow indices and the dominant characteristics
was smaller than our model discretisation (see e.g. Pana- of more than 100 catchments. However, the sets of catch-
gopoulos 2012 for a recent, successful application of the ments used to fit the regression models were different. For
equivalent porous-media concept to a karstified aquifer). WaBoA, only catchments from the state of Baden-Würt-
temberg were used whereas for HAD, catchments from all
Boundary conditions over Germany were utilized. As a consequence, annual
recharge for the Ammer catchment (Fig. 2) was higher and
The bottom boundary of the mo aquifer is formed by an more variable for WABoA (average of 185 mm with
aquitard, i.e. evaporites from the Middle Muschelkalk standard deviation of 61 mm) than for HAD (average of
which function as a quasi-impermeable layer; we assumed 105 mm with standard deviation of 43 mm).
no water flux through the bottom boundary. The same was For discharge scenario 1, the entire course of the Ammer
assumed for the groundwater divide at catchment bound- River and its main tributaries, the Kochart and the Käsbach
aries (Harress 1973; Villinger 1982; Plümacher 1999). An Creek, were represented; a head boundary condition equal
average pumping rate of 35 l/s was assigned to each of the to the topographic height of the river bed was assigned. For
four production well sites. For all discharge points at the discharge scenario 2, no rivers but six large mo springs and
surface boundary (discharge scenarios 1 and 2), a constant five major km1 springs were considered. These were
head equal to the topographic height was given. Note that springs known to significantly contribute to Ammer stream
discharge points at springs for scenario 2 were not included flow (Harress 1973). A head boundary condition equal to
in mesh-building process; they were assigned to nodes in the topographic heights of the springs was assigned. Note
the mesh closest to the location of the springs. Care was that, for our steady-state model, the catchment water bal-
taken that these mesh nodes were located within the ance was reasonably simple as groundwater recharge
appropriate hydrogeological unit. Discharge points repre- equalled discharge from pumping plus discharge to surface
senting springs were mostly within \50 m of their real water bodies (discharge scenario 1) and springs (discharge
locations. For recharge to the uppermost aquifer, a flux scenario 2), respectively.
boundary was prescribed. Discharge points and recharge
data are detailed in the following section. Model calibration

Recharge and discharge scenarios The PEST code (Doherty 2004) was applied for model
calibration, i.e. the estimation of hydraulic conductivities
The assumptions for the four scenarios used in our study including the precision of the estimates. For this study,
are summarised in Table 1. PEST was directly applied to OpenGeoSys by preparing a
Long-term average annual groundwater recharge control file and the input–output exchange files (Sun et al.
(1961–1990) was digitised from two atlases, i.e. ‘Wasser- 2012).
und Bodenatlas Baden-Württemberg’ (WaBoA 2001) and The allowable ranges for the estimation of hydraulic
‘Hydrologischer Atlas Deutschland’ (HAD 2012). WaBoA conductivities were (1e-06…1e-04 m/s] for the mo
had a spatial resolution of 500 m whereas HAD had a aquifer and (1e-10…1e-04 m/s) for Keuper units. The
resolution of 1,000 m. Both atlases used a similar method relative narrow range for mo resulted from a number of
to estimate recharge. This method is broadly based on a pumping tests conducted in the context of groundwater
regionalisation of the baseflow index which in turn uses a production from this aquifer (Plümacher 1999). Keuper
multiple linear regression on dominant catchment units were more heterogeneous and no pumping tests were
available; and we allowed a wider range of hydraulic
Table 1 Summary of recharge and discharge scenarios conductivities. The sum of the squared differences between
Discharge Recharge Discharge Recharge
observed and calculated heads at 27 locations was used as
scenario scenario the objective function for model calibration. These obser-
vation wells were relatively homogenously distributed over
1 1 Ammer River and WaBoA the catchment (Fig. 1). Note that we calibrated a steady-
tributaries
state model but the observations stem from different years
1 2 Ammer River and HAD
tributaries and different seasons, and therefore, rather large differ-
2 1 mo and km1 springs WaBoA ences were expected. Observed water levels included: 15,
2 2 mo and km1 springs HAD
5, 6 and 1 wells screened in mo, ku, km1 and km2…4,
respectively. As observation wells were not included in the

123
Environ Earth Sci (2013) 69:443–452 447

Fig. 2 Two maps of long-term


annual groundwater recharge
for the Ammer catchment from
different German hydrologic
atlases. a ‘Wasser–und
Bodenatlas Baden-Württemberg’
(WaBoA, recharge scenario 1).
b ‘Hydrologischer Atlas
Deutschland’ (HAD, recharge
scenario 2)

mesh building process, they were represented by the closest Table 2 Apparent groundwater ages for various production well
sites estimated based on concentrations of tritium and sulphur hexa-
mesh node within the appropriate hydrogeological unit.
fluoride (Hydroisotop 2004)

Particle tracking Production well site Apparent age (years)

W1 [50
For computation of flow paths and travel times for produc- W2 20–25
tion well sites and springs, the streamline routines with W3 20–25
backward particle tracking in ParaView (http://www. W4 5–20
paraview.org/) were used. An effective porosity of 0.015
was assumed for all hydrogeological units (Landesamt für
Umweltschutz Baden-Württemberg 2002). For backward insight into water–rock interactions on the flow paths of
particle tracking, sources were specified at the locations of groundwater. In addition, gypsum saturation indices were
the production well sites and four large springs near Her- calculated using the PHREEQC geochemical code (Park-
renberg (‘Ammerquellen’: 20–200 l/s and ‘Pumpwerk Her- hurst and Appelo 1999).
renberg’: 50–200 l/s) and Reusten (‘Schwärzenbrunnen’:
20–150 l/s and ‘Pumpwerk Reusten’: 5–10 l/s).
Results
Additional data for groundwater dating
and hydrochemical evaluation Model fit to observed water levels and parameter
estimation
Groundwater dating using environmental tracers
Fit to observed water levels for discharge scenario 1 was
For production well sites, apparent groundwater ages were better than for discharge scenario 2 (Fig. 3), and hence,
estimated based on one sample for each production well hydraulic conductivities had lower coefficients of variation.
with measured concentrations of tritium and sulphur For all scenarios, hydraulic conductivities were more pre-
hexafluoride. Ages were calculated using a combined cisely estimated for the mo and km2…4 units than for the
exponential-piston flow model (Małoszewski and Zuber ku and km1 units. For the latter units, estimated parameter
1982) with the admixture of an ancient and tracer-free values could vary by several orders of magnitude, which
water component (Hydroisotop 2004). The apparent ages likely reflected the imprecisely known heterogeneities for
for all four well sites are presented in Table 2. these units. However, estimated hydraulic conductivities
for different hydrogeological units were similar for both
Hydrochemical evaluation using major ion chemistry discharge scenarios, and were comparable to values
obtained by other model applications in the same ground-
Major ion chemistry for groundwater samples from pro- water region (Landesamt für Umweltschutz Baden-Würt-
duction wells, km1 and mo springs was interpreted to gain temberg 2002; Barthel et al. 2008). Some small differences

123
448 Environ Earth Sci (2013) 69:443–452

Fig. 3 Observed versus fitted water levels for different recharge and scenario 2. Water levels are labelled using hydrogeological units.
discharge scenarios. a Discharge scenario 1 and recharge scenario 1. RMSE root mean square error. Estimated hydraulic conductivities
b Discharge scenario 1 and recharge scenario 2. c Discharge scenario for various hydrogeological units with coefficients of variation
2 and recharge scenario 1. d Discharge scenario 2 and recharge (in brackets)

between the different scenarios included lower hydraulic sure why karstification should be more advanced in the
conductivities for ku than for mo in discharge scenario 1 lower catchment as overburden tends to be thicker than that
which was the opposite for the other discharge scenario. in the upper catchment. Biased conductivity estimates
For both discharge scenarios, recharge maps had little could also be a result of an overestimation of recharge in
effect on the quality of the fit as well as on the precision of the upper catchment, which tended to be higher than that in
the parameter estimates. Calibrated hydraulic conductivi- the lower catchment for both recharge scenarios.
ties tended to be lower for recharge scenario 2 due to its
overall lower recharge. For the mo aquifer in discharge Computed flow paths and travel times
scenario 2, water levels \450 m ASL, mainly observed in
the southeastern area of the catchment, were over predicted For discharge scenarios, flow paths of groundwater to
whereas water levels [450 m ASL, observed in the production well sites were different whereas they did not
northwestern area of the catchment, tended to be under vary much for recharge scenarios (Fig. 4). However,
predicted. This could indicate higher conductivities in the recharge scenario 2 resulted in longer travel times and
lower part of the catchment, i.e. where the production wells perhaps slightly larger recharge source areas compared to
are located, compared to the upper catchment. This con- scenario 1. The former is indicated by an overall lower
ductivity difference could reflect an advanced karstification level of brightness of streamlines in recharge scenario 2
in the lower part of the catchment. However, we are not than scenario 1. It is a result of lower recharge, and thus,

123
Environ Earth Sci (2013) 69:443–452 449

overall slightly lower calibrated conductivities for recharge discharge scenario 2 than by scenario 1 (Fig. 6). This is
scenario 2 (Fig. 3). Travel times of groundwater to pro- particularly true for the two springs near Reusten where the
duction well sites calculated for all four scenarios had a catchments of springs stretch along the southwestern bor-
plausible order of magnitude (Fig. 4) which was compa- der of the study area. Travel times of groundwater to
rable to apparent groundwater ages modelled using envi- springs were overall shorter than travel times to production
ronmental tracers (Table 2). However, apparent well sites. This can be seen from an overall lower level of
groundwater ages from environmental tracers tended to be brightness for stream lines in Fig. 4 compared to Fig. 6.
higher than average travel times computed using ground- For discharge scenario 2 and recharge scenario 1, averaged
water modelling. These observed differences between mean travel times for springs were 7 years whereas travel
hydraulic and tracer ages are most obvious for well site W1 times for wells were 17 years. However, we did not
(‘Breitenholz’) which had mean hydraulic ages of a max- interpret those differences in terms of karst features as they
imum of 16 years (discharge scenario 2 and recharge sce- were not explicitly represented in our model.
nario 2) whereas tracer ages were [50 years (Table 2).
These differences may be explainable by matrix diffusion.
Part of the tracer transported through conductive units will Discussion and conclusions
be lost by diffusion processes into the low-permeability
matrix and consequently these heterogeneities can signifi- Two different recharge maps resulted in a comparable fit to
cantly bias apparent groundwater ages. As we were unable observed water levels, and similar estimates of hydraulic
to effectively parameterise these heterogeneities for our conductivity, flow path and travel time of groundwater to
karstified aquifers, we also did not attempt to calibrate our production well sites. Travel times of groundwater to
model directly by an explicit representation of transport of production well sites calculated for all four scenarios had a
environmental tracers through the fissured matrix and karst plausible order of magnitude which was comparable to
conduits. Recharge source areas of production well sites for apparent groundwater ages modelled using environmental
discharge scenario 2, tend to be more in the ku and mo tracers. Scenario with groundwater discharge over the
areas than for scenario 1 (Fig. 4). More specifically, these entire streambed of the Ammer River and its tributaries
recharge source areas were: the area north of the Schön- (discharge scenario 1) resulted in a better fit to water levels
buch forest for well sites W1 and W3, and for well sites than the scenario with discharge at a few springs only
W2 and W4 additionally the western part of the catchment. (discharge scenario 2). In contrast to the fit to water levels,
For discharge scenario 2, the km1 tended to intercept flow paths of groundwater from discharge scenario 2 were
recharge from the km2…4 area (Schönbuch forest) and more plausible than from scenario 1. This was true for both
discharged laterally through three large springs on the left flow paths to production well sites and major mo springs.
bank of the Ammer River (Fig. 1). More specifically, the observed major ion chemistry at the
Anion composition of water from production wells was production well sites indicated that the source area of
similar to the compositions from both the mo and the ku recharge tended to be outside the km1, and hence, inside
springs, whereas km1 springs displayed different chemical the ku and mo areas. Furthermore, uranine tracer tests by
characteristics (Fig. 5). The analysis using PHREEQC Harress (1973) suggested that the catchments of springs
indicated that km1 water featured a distinct geochemical near Reusten stretch along the southwestern border of the
signature with high values of the gypsum saturation index study area. All these findings supported flow paths com-
(35–99 %). The same method applied for ku and mo puted for discharge scenario 2 rather than scenario 1. How-
springs yielded saturation values of 1–7 %; a similar ever, the latter scenario could not be rejected based on
undersaturation with gypsum was calculated for all ground- observed water levels and apparent groundwater ages.
water samples from the production wells. This indicates Therefore, these data more commonly used in groundwater
that an insignificant amount of recharge to the mo aquifer modelling may be insufficient to reliably delineate capture
flows through the overlaying km1 aquifer. Therefore, the zones. Additional information relating only indirectly to
main source area of recharge for the production wells is groundwater flow such that the major ion chemistry of water
likely to be outside of the km1, and hence, inside the ku and sampled at the wells can substantially improve our under-
mo areas (Fig. 5). This interpretation also agreed with a standing of recharge source areas for production wells.
hydrogeological map by Villinger (1982) who conceptua- While testing of different plausible recharge scenarios is
lised the mo aquifer in the km1 covered area to be confined perhaps becoming more common in groundwater model-
preventing recharge through upward hydraulic gradients. ling studies (Plümacher 1999; Ye et al. 2010; Gräbe et al.
Flow paths of groundwater to major mo springs near 2013), a test of different discharge scenarios is less fre-
Herrenberg and Reusten as obtained from uranine tracer quently applied. However, in this study, we found that
tests (Harress 1973) were more closely resembled by computed flow paths and travel times of groundwater to

123
450 Environ Earth Sci (2013) 69:443–452

Fig. 4 Horizontal plane projection of flow paths and travel times to by solid black circles. b Discharge scenario 1 and recharge scenario 2.
four drinking-water production well sites. a Discharge scenario 1 and c Discharge scenario 2 and recharge scenario 1. d Discharge scenario
recharge scenario 1. Approximate locations of springs are indicated 2 and recharge scenario 2

production well sites were controlled by model represen-


tations of groundwater discharge to streams, but they were
only insignificantly affected by alternative recharge esti-
mates. Evaluation of the two alternative discharge scenar-
ios for the Ammer catchment suggested that groundwater
predominantly discharges via localised springs rather than
as a diffuse process, which also agrees with the system
perception of Schwientek et al. (2013). We believe that—
for fractured and karstified catchments—more attention
needs to be paid to the fundamental question of how and
where groundwater flows into streams and other surface
water bodies.
As suggested by flow paths for discharge scenario 2,
there appear to be two distinct source areas of recharge for
production wells. One source area of recharge is the
unconfined western part of the catchment including the
anticline of Reusten. This water source is probably most
important for well site W4 (‘Poltringen’), and also plays a
Fig. 5 Anion composition of groundwater samples from different
springs and drinking-water production wells. Some data points role for W2 (‘Altingen’) and W3 (‘Entringen’). Ground-
represent temporal averages from up to three different analyses. water age dating from environmental tracers indicated that
Chemical data are plotted as percentages of total equivalents per litre water coming from this area is likely to be younger than

123
Environ Earth Sci (2013) 69:443–452 451

Fig. 6 Horizontal plane projection of flow paths and travel times to Approximate location of springs is indicated by solid black circles.
major mo springs near Herrenberg (central part of catchment: a Discharge scenario 1 and recharge scenario 1. b Discharge scenario
‘Ammerquellen’ and ‘Pumpwerk Herrenberg’) and Reusten (southern 1 and recharge scenario 2. c Discharge scenario 2 and recharge
part of catchment: ‘Pumpwerk Reusten’ and ‘Schwärzenbrunnen’). scenario 1. d Discharge scenario 2 and recharge scenario 2

20 years (Table 2). The other water source is the uncon- are sufficient to accommodate recharge rates of up to
fined area north of the Schönbuch forest, which is likely to 10 mm per day; the large uncertainties for the estimated
be the dominant recharge source for production well site conductivities of ku obtained using PEST leave enough
W1 (‘Breitenholz’). However, it could also be important room for this kind of interpretation.
for W3 and may be W2 too. Groundwater pumped from Our groundwater model provided useful information on
this source area probably crosses the Schönbuch area flow paths and travel times to groundwater production
without significant recharge; groundwater dating for W1 wells without an explicit representation of all possible karst
suggested ages of older than 50 years (Table 2). This features which were imprecisely known for our ground-
would make production well site W1 less vulnerable to water flow system. This result was surprising given that our
contamination from near surface sources than well sites model was set up based on a porous-media concept and
W4, W2 and W3. calibrated against observed water levels; it is known that
From the available data and our modelling results, it for fractured and karstified systems, this concept may
may be inferred that the Lettenkeuper plays a key role for correctly simulate water levels yet the velocity field is
groundwater flow in the catchment. On the one hand, it incorrect (Sauter et al. 2006). Specifically, discharge sce-
serves as a confining layer overlaying the mo aquifer in the nario 2 represented important features of flow paths to
northeastern part of catchment (Villinger 1982). On the production wells that were independently inferred from
other hand, the unconfined mo area covered by ku is likely hydrogeochemical data and the analysis of environmental
to be an important source area of recharge for both the tracers. This suggested that the overall groundwater flow
production wells and the major mo springs (Figs. 4c, d, 6c, patterns in karstified aquifer systems may be reasonably
d). These two aspects of ku neither contradicted each other represented using an appropriate geometry of the different
nor disagreed with model parameter estimates. This is subsurface layers and the predominant discharge and pos-
because even low hydraulic conductivities of 1e-07 m/s sibly recharge patterns in the area.

123
452 Environ Earth Sci (2013) 69:443–452

Acknowledgments This work was supported by a grant from the Hydroisotop (2004) Ergebnisse der hydrochemischen und isotopen-
Ministry of Science, Research and Arts of Baden-Württemberg (AZ hydrologischen Bestandsaufnahme im Gebiet Ammertal-Rotten-
Zu 33-721.3-2) and the Helmholtz Center for Environmental burg. Hydroisotop, Schweitenkirchen
Research, Leipzig (UFZ). We would like to thank Dr. Marc Schw- Kolditz O, Bauer S, Bilke L, Böttcher N, Delfs JO, Fischer T, Görke
ientek and Dr. Karsten Osenbrück (Water & Earth System Science), UJ, Kalbacher T, Kosakowski G, McDermott CI, Park CH, Radu
Bernhard Keim (engineering company kup), Andreas Steinacker F, Rink K, Shao H, Shao HB, Sun F, Sun YY, Singh AK, Taron
(consulting company BGU) and Inge Neeb (city council of Sindelf- J, Walther M, Wang W, Watanabe N, Wu Y, Xie M, Xu W,
ingen) for technical discussions. We also thank the Ammertal- Zehner B (2012) OpenGeoSys: an open-source initiative for
Schönbuchgruppe (local water supplier) for providing data, Igor numerical simulation of thermo-hydro-mechanical/chemical
Pavlovskiy for providing Fig. 5 and analysing water quality data (THM/C) processes in porous media, Environ Earth Sci
using PHREEQC and Dr. Wenqing Wang and Dr. Jens-Olaf Delfs for 67:589–599. doi:10.1007/s12665-012-1546-x
OGS support. We are grateful for the detailed comments provided by Landesamt für Umweltschutz Baden-Württemberg (2002) Fortschrei-
3 reviewers based on which the manuscript could be significantly bung Hydrogeologische Karte und regionales Grundwassermodell
improved. ‘‘Heilbronner Mulde’’. Landesamt für Umweltschutz Baden-Würt-
temberg, Karlsruhe
Małoszewski P, Zuber A (1982) Determining the turnover time of
groundwater systems with the aid of environmental tracers. 1.
References Models and their applicability. J Hydrol 57:207–231
Panagopoulos G (2012) Application of MODFLOW for simulating
Alley WM, Healy RW, LaBaugh JW, Reilly TE (2002) Flow and groundwater flow in the Trifilia karst aquifer, Greece. Environ
storage in groundwater systems. Science 296:1985–1990 Earth Sci 67:1877–1889. doi:10.1007/s12665-012-1630-2
Armbruster V, Leibundgut C (2001) Method for spatially distributed Parkhurst DL, Appelo CAJ (1999) User’s guide to PHREEQC
modelling of evapotranspiration and fast runoff components to (Version 2)—a computer program for speciation, batch-reaction,
describe large-scale groundwater recharge. IAHS Publ 269:3–10 one-dimensional transport, and inverse geochemical calcula-
Barthel R, Jagelke K, Götzinger J, Gaiser T, Printz A (2008) Aspects tions. USGS, Reston
of choosing appropriate concepts for modeling groundwater Plümacher J (1999) Kalibrierung eines regionalen Grun-
resources in regional integrated water resources management— dwasserströmungsmodells mit Hilfe von Umwetltisotopeninfor-
examples from the Neckar (Germany) and Oueme catchment mation. Dissertation, ETH Zürich, Zurich
(Benin). Phys Chem Earth 33:92–114 Praamsma T, Novakowski K, Kyser K, Hall K (2009) Using stable
Bauer S, Liedl R, Sauter M (2003) Modeling of karst aquifer genesis: isotopes and hydraulic head data to investigate groundwater
influence of exchange flow. Water Resour Res 39(10):WR002218 recharge and discharge in a fractured rock aquifer. J Hydrol
Chiang WH (2005) 3D-groundwater modeling with PMWIN. 366(1–4):35–45
Springer, Berlin/Heidelberg/New York Refsgaard JC (2011) Review of strategies for handling geological
Doherty J (2004) PEST: model independent parameter estimation. uncertainty in groundwater flow and transport modeling. Adv
Watermark Numerical Computing, Brisbane Water Resour 36:36–50
Githui F, Selle B, Thayalakumaran T (2012) Recharge estimation Reuther CD (1973) Zur Schichtlagerung und Tektonik im Oberen
using remotely sensed evapotranspiration in an irrigated catch- Gäu. Master thesis, University of Tübingen, Tübingen
ment in southeast Australia. Hydrol Process 26:1379–1389 Rink K, Kalbacher T, Kolditz O (2012) Visual data management for
Gleeson T, Novakowski K, Cook PG, Kyser TK (2009) Constraining hydrological analysis. Environ Earth Sci 65(5):1395–1403
groundwater discharge in a large watershed: integrated isotopic, Rink K, Fischer T, Selle B, Kolditz O (2013) A data exploration
hydraulic, and thermal data from the Canadian shield. Water framework for validation and setup of hydrological models.
Resour Res 45(8):W08402 Environ Earth Sci. doi:10.1007/s12665-012-2030-3
Gräbe A, Rink K, Fischer T, Sun F, Wang W, Rödiger T, Siebert C, Sauter M, Kovács A, Geyer T, Teutsch G (2006) Modellierung der
Kolditz O (2013) Numerical analysis of the groundwater regime Hydrodynamik von Karstgrundwasserleitern—Eine Übersicht.
in the western dead sea escarpment. Environ Earth Sci. doi: Grundwasser 11(3):143–156
10.1007/s12665-012-1795-8 Schwientek M, Osenbrück K, Fleischer M (2013) Investigating
Grathwohl P, Rügner H, Wöhling T, Osenbrück K, Schwientek M, hydrological drivers of nitrate export dynamics in two agricul-
Gayler S, Wollschläger U, Selle B, Pause M, Delfs J-O, tural catchments in Germany using high-frequency data series.
Grzeschik M, Weller U, Ivanov M, Cirpka OA, Maier U, Kuch Environ Earth Sci. doi:10.1007/s12665-013-2322-2
B, Nowak W, Wulfmeyer V, Warrach-Sagi K, Streck T, Attinger Sun F, Shao H, Kalbacher T, Wang W, Yang Z, Huang Z, Kolditz O
S, Bilke L, Dietrich P, Fleckenstein JH, Kalbacher T, Kolditz O, (2012) Groundwater drawdown at Nankou site of Beijing plain:
Rink K, Samaniego L, Vogel H-J, Werban U, Teutsch G (2013) model development and calibration. Environ Earth Sci
Catchments as reactors—a comprehensive approach for water 64(5):1323–1333
fluxes and solute turn-over. Environ Earth Sci 69(2). doi: Villinger E (1982) Grundwasserbilanzen im Karstaquifer des Oberen
10.1007/s12665-013-2281-7 Muschelkalks im Oberen Gäu (Baden-Württemberg). Geologis-
Guadagnini A, Franzetti S (1999) Time-related capture zones for ches Jahrb: Reihe C 32:43–61
contaminants in randomly heterogeneous formations. Ground WaBoA (2001) Wasser- und Bodenatlas Baden-Württemberg.
Water 37(2):253–260 http://www.hydrology.uni-freiburg.de/forsch/waboa/. Accessed
HAD (2012) Hydrologischer Atlas von Deutschland. http://www. 26 July 2012
hydrology.uni-freiburg.de/forsch/had/. Accessed 26 July 2012 Ye M, Pohlmann KF, Chapman JB, Pohll GM, Reeves DM (2010)
Harrar WG, Sonnenborg TO, Henriksen HJ (2003) Capture zone, Model-Averaging Method for Assessing Groundwater Concep-
travel time, and solute-transport predictions using inverse tual Model Uncertainty. Ground Water 48(5):716–728
modeling and different geological models. Hydrogeol J 11(5): Zanini L, Novakowski KS, Lapcevie P, Bickerton GS, Voralek J,
536–548 Talbot C (2000) Ground water flow in a fractured carbonate
Harress HM (1973) Hydrogeologische Untersuchungen im Oberen aquifer inferred from combined hydrogeological and geochem-
Gäu. Dissertation, University of Tübingen, Tübingen ical measurements. Ground Water 38(3):350–360

123

You might also like

pFad - Phonifier reborn

Pfad - The Proxy pFad of © 2024 Garber Painting. All rights reserved.

Note: This service is not intended for secure transactions such as banking, social media, email, or purchasing. Use at your own risk. We assume no liability whatsoever for broken pages.


Alternative Proxies:

Alternative Proxy

pFad Proxy

pFad v3 Proxy

pFad v4 Proxy