Content uploaded by Emil Frind
Author content
All content in this area was uploaded by Emil Frind on Jul 18, 2014
Content may be subject to copyright.
Delineating Capture Zones for Environmentally Sensitive
Features – A Model Comparison
R. Chow1, E.O. Frind1, M. Sousa1, J.P. Jones1, D.L. Rudolph1 and J.W. Molson2
1 Department of Earth and Environmental Sciences, University of Waterloo, Waterloo,
Ontario, Canada
2
Département de Géologie et de Génie Géologique, Université Laval, Québec City,
Québec, Canada
ABSTRACT
The protection of water sources requires the delineation of well capture zones. Environmentally sensitive features such
as wetlands and streams also require protection, and in principle, capture zones can be delineated using the same
methodology as that used for well capture zones. In this study, different models are applied to delineate capture zones
for two reaches of a stream. The results show that dramatically different capture zones can be obtained with different
models. The causes of these differences are examined. A key conclusion is that capture zones for environmentally
sensitive features can be highly sensitive to numerical error.
RÉSUMÉ
La protection des sources d’eau nécessite la délimitation des zones de captage. Des environnements écologiques
sensibles comme les zones humides et cours d'eau exigent également une protection qui, en principe, peuvent être
délimitées en utilisant la même méthodologie que celle utilisée pour les zones de captage. Dans cette étude, différents
modèles sont utilisés pour délimiter les zones de captage pour deux tronçons d'un ruisseau. Les résultats montrent que
les zones de captage diffèrent radicalement en fonction du modèle utilisé. Les causes de ces différences ont été
analysées. La conclusion principale indique que l’analyse des zones de captage, pour des environnements écologiques
sensibles, peuvent être très sensibles aux erreurs
numériques.
1 INTRODUCTION
Groundwater is a resource essential to life. In Ontario, the
standard for source water protection has now been set
through the Clean Water Act of 2006. The management
and protection of groundwater resources requires the
delineation of capture zones which define where
groundwater recharge originates from and where it
discharges. A number of models are available for
delineating groundwater capture zones. Although most
applications so far have been to delineate capture zones
for production wells, the same methodologies can in
principle be applied to the delineation of capture zones for
environmentally sensitive areas.
A critical issue in the delineation of a capture zone is
the accuracy of model predictions. Because capture
zones cannot be validated, except by long-term tracer
studies, it is important that model results are reasonably
accurate and reliable. In this study we apply three proven
models to consider the question of predictive accuracy in
the delineation of capture zones for different reaches of a
stream.
2 STUDY AREA – ALDER CREEK
The Alder Creek Watershed within the Regional
Municipality of Waterloo has been selected as a study
area. This watershed is an agricultural area that hosts a
number of key water supply wells for the Regional
Municipality, and it also contains numerous
environmentally sensitive areas.
The Alder Creek Watershed is a sub-watershed within
the Grand River Watershed in southern Ontario. It is
embedded within the south central area of the Waterloo
Moraine. The Alder Creek Watershed is situated in close
proximity to the cities of Kitchener and Waterloo. Both
cities are rapidly growing, and as a result, the Alder Creek
Watershed is under a great deal of development pressure.
Figure 1. Map of Alder Creek Watershed
3 METHODOLOGY
The focus of this research is to delineate capture zones of
environmentally sensitive areas. Much work has been
done on the use of models for the delineation of capture
zones for point sources such as production wells,
generally involving reverse particle tracking or reverse
transport (see for example Frind et al. 2002). However, as
far as we are aware, there are no published results as yet
on capture zones for linear features such as streams and
rivers, or sensitive areas such as wetlands.
We will here apply three models: HydroGeoSphere
(HGS) (Therrien et al. 2005), Modflow (MacDonald and
Harbaugh 1988) and Watflow (Molson et al. 2002). HGS is
a state-of-the-art model based on control volume finite
element theory which fully integrates surface water,
unsaturated groundwater flow and saturated groundwater
flow systems. Modflow is a three-dimensional finite
difference model, and is today’s industry-standard
commercially available hydrogeology software. Modflow
has no surface water or unsaturated zone capabilities, but
it has a well-proven particle tracking module (Modpath;
Pollock 1988) which is used in most well capture zone
delineation work today. Watflow is a finite element
groundwater model capable of solving three-dimensional
flow problems in heterogeneous confined/unconfined
aquifer systems. Watflow is limited to fully saturated flow
but it has a linearized unsaturated zone representation.
Watflow also has an integrated automatic calibration
routine, as well as a particle tracking routine (Watrac;
Frind and Molson 2004) for capture zone delineation that
is based on the same principles as Modpath.
Figure 2. Reach A capture zone from Modflow (left) and Watflow (right), by particle tracking. Capture zones from both
models extend in the northwesterly direction. In Modflow, maroon particle tracks travel towards the bottom of the model
domain and blue particle tracks travel towards the top.
Because both groundwater and surface water are
involved, the logical methodology would be to directly
apply HGS to characterize the flow system and delineate
the capture zone. However, HGS is also difficult to use
due to greater computational and data demands. Because
most capture zone delineations involve time scales of
decades or more, variations in surface processes
generally dampen out within the relevant time frame;
therefore capture zone calculations can be based on
average steady-state flow conditions.
Accordingly, we use HGS only to generate the spatial
annual average distribution of recharge over the study
area, and we link this recharge distribution (which includes
groundwater discharge areas) to each of the Modflow and
Watflow models. With this input, each model is run to
steady state. Reverse particle tracking, with particles
placed along a plane below the streambed and across the
entire stream width, is then used to delineate the capture
zone. The number of particles is the same for both
models.
With this setup, both Modflow and Watflow will receive
identical input data. Although the distribution of hydraulic
conductivity is the same, there are some important
differences between the models. A key generic difference
between Modflow and Watflow is the type of discretization.
While Modflow uses a structured rectangular grid, Watflow
uses an unstructured grid made up of triangular prisms
that project as triangles in the plane. Modflow calculates
hydraulic heads at the cell centres, while Watflow
calculates heads at the elements nodes. An important
difference specific to this study is in the number of layers:
Watflow uses 87 layers, while Modflow was able to
accommodate only 7 layers (one layer for each
hydrostratigraphic unit) due to convergence issues and
cell wetting/rewetting problems. The layering also affects
the way the boundary conditions are specified. Therefore,
although the hydrogeologic data used are the same for
both models, the numerical representation of the data for
the two models will be somewhat different. As a result, the
capture zones obtained from the two models can be
expected to be different.
Two stream reaches (see Figures 2 and 3) were
selected for the generation of capture zones: Reach A:
Stream section on the central part of Alder Creek; and
Reach B: Stream section on the western branch of the
northern part of Alder Creek.
There are two main criteria in choosing stream
reaches to perform the capture zone delineation. Firstly,
the stream reach needs to be a gaining stream
(groundwater discharge), and secondly, it must be
sufficiently distant from the perimeter of the domain to
avoid boundary effects. Both selected reaches satisfy
these criteria.
Figure 3. Reach B capture zone from Modflow (left) and Watflow (right), by particle tracking. Capture zones differ in
direction, size and extent.
4 RESULTS AND DISCUSSION
Figures 2 and 3 show the capture zones delineated for
Reaches A and B, using the reverse particle tracking
codes available in Modflow and Watflow, respectively. All
tracking was run to steady state, where particle positions
no longer change, which occurred at about 300 years.
Particle tracks shown are horizontal plane projections of
the actual 3D tracks generated by both models.
For Reach A (Fig. 2), the capture zones created by
both models are similar in that they both extend in the
same northwesterly direction. However, they differ in
extent and size. The capture zone delineated with Watflow
is larger in size and tends to sweep farther south than the
capture zone delineated with Modflow. The southward
sweep obtained with Watflow is a result of the 3D nature
of the flow field, which directs some of the particle tracks
to deeper layers of the model. With both models, the
capture zone extends only to the west of the stream,
suggesting a throughflow situation with inflow to the
stream from the west and possibly outflow to the east.
For Reach B (Fig. 3), the capture zones delineated by
the two models are very different. The capture zone
produced by Modflow extends only to the northwest, while
that produced by Watflow extends both towards the
northwest and the southwest from the source. In this case,
Watflow produces a capture zone extending on both sides
of the stream, suggesting a normal gaining stream – a
reasonable expectation.
Why are the capture zones obtained by the two
models different? A similar result was obtained by Sousa
et al. (2010), who delineated a well capture zone using
recharge distributions from three different models,
superimposed on the groundwater model Watflow. Sousa
et al. obtained three different capture zones, all more or
less extending in the same direction, but differing
significantly in extent.
There are several reasons for the differences. One
reason, specific to our case, is the contrast in the vertical
discretization. As with any numerical scheme, the finer the
grid, the better the accuracy, so the 87 layers in Watflow
should provide better accuracy than the 7 layers in
Modflow. The horizontal discretization also differs,
although an effort was made at equivalent grid spacing.
Another more generic factor is numerical precision:
while groundwater models are calibrated on point values
of head, it is the gradients, not the heads, that are used in
the capture zone calculation. Small calibration errors in
head can thus be magnified to become large errors in the
gradients. This means that capture zone delineations can
be extremely sensitive to numerical errors.
The impact of numerical errors in capture zone
calculation will also depend on the magnitude of the
hydraulic gradient. In the case of a well, the induced
gradients near the well are normally larger than the natural
gradients, meaning that numerical errors may be
swamped. On the other hand, gradients occurring near
environmentally sensitive areas are natural gradients,
much smaller than those near pumping wells, so
calibration errors will appear in the calculations.
Therefore, capture zone delineations for environmentally
sensitive areas are expected to be more sensitive to
numerical errors than those for well capture zones.
5 CONCLUSIONS
The following conclusions can be drawn from this study:
• Capture zones for environmentally sensitive
areas or features can, in principle, be delineated
using the same methodology as that used for the
delineation of well capture zones.
• Capture zones can be highly sensitive to
numerical error, and different models can
produce dramatically different capture zones.
• Numerical errors can be case-specific (i.e.
discretization dependent), or generic as a result
of the magnification of calibration errors in the
calculation of gradients.
• Capture zone calculations for environmentally
sensitive features can be more sensitive to
numerical error than calculations for well capture
zones.
The implication of this work is that the predictive
accuracy of available models in the delineation of capture
zones for environmentally sensitive features may not
always be guaranteed without careful analysis. This is a
concern, because most capture zone delineations today
are done by running only one model or one scenario. By
relying on only one model/scenario, a practitioner may not
realize that different solutions may exist.
Sousa et al. (2010) managed uncertainty in the
capture zone prediction by drawing an envelope around
the different capture zones. The same approach may be
used here also. More work is required to fully understand
and manage the sources of error that may be involved.
The next step in this study will be to apply HGS to the
entire capture zone delineation.
ACKNOWLEDGEMENTS
We would like to thank Steve Holysh for suggesting this
study direction. Nicolas Benoit of the GSC is thanked for
translating the abstract. The work was supported by the
Natural Sciences and Engineering Research Council of
Canada through a grant to the second author.
REFERENCES
Frind, E.O. and Molson, J.W., 2004. A new particle
tracking algorithm for finite element grids. Keynote
lecture, FEM_MODFLOW Conference (spons.
IAHS/USGS), Carlsbad, Czechia.
Frind, E.O., Muhammad, D.S. and Molson, J.W. 2002.
Delineation of Three-Dimensional Well Capture Zones
for Complex Multi-Aquifer Systems. Ground Water,
40(6): 586-598.
Molson, J.W., Beckers, J., Frind, E.O. and Martin P.J.
2002. WATFLOW/3D: A Three-Dimensional Numerical
Model for Coupled Groundwater Surface Water Flow,
Version 4.0. Department of Earth Sciences, University
of Waterloo.
MacDonald, M.G. and Harbaugh, A.W. 1988. A modular
three-dimensional finite-difference groundwater flow
model. In: US Geological Survey Techniques of Water
Resources Investigations, 06-A1, USGS.
Pollock, D.W., 1988. Semianalytical computation of
pathlines for finite-difference models, Ground Water,
26(6): 743-750.
Sousa, M., Frind, E.O. and Rudolph, D.L. 2010. Capture
probability maps for addressing uncertainty: Protection
vs mitigation. In Proc.GQ10: Groundwater Quality
Management in a Rapidly Changing World. 7th
International Groundwater Quality Conference, Zurich,
Switzerland.
Therrien, R., McLaren, R.G., Sudicky, E.A. and Panday,
S.M. 2005. HydroGeoSphere: A three-dimensional
numerical model describing fully-integrated subsurface
and surface flow and solute transport. Groundwater
Simulations Group. University of Waterloo, Ontario,
Canada.
