HYDROLOGICAL MODELLING

OF THE ALERGAARDE
CATCHMENT
Roseanna Bentley, Ciprian Ungur
1.0 INTRODUCTION
OVERALL ISSUES
The issues of water quality and quantity have been placed under increasing strain in the past
decades as human consumption is becoming more and more significant. Over-extraction, climate
change, poor management and increased contaminants are just part of the anthropic impact
upon the water reserves of our planet. Thus, we need to find a way to understand how our key
aquifer systems respond in relation to the whole hydrological cycle in order to improve our
management and exploitation strategies (Healy, et al., 2007).

Denmark is an unique situation, since 99% of their drinking water is extracted from aquifer
systems (Henrikson, et al., 2003; Scharling, et al., 2009). Moreover, 62% of Danish land is used
for agriculture (European Environment Agency, 2015). As in most modern agricultural
productions, Danish agriculture extracts groundwater for irrigation and uses chemical pesticides
and fertilizers on the land which eventually leak into the surface water and groundwater systems
and contaminate water dependent systems. In response to this, GEUS (the Geological Survey of
Denmark and Greenland) developed a national resource model for Denmark to continuously
assess the exploitation of aquifers at both catchment and national scale (Sonnenborg, et al.,
2003).

Only recently it has been realized that groundwater and surface water systems must be
approached as a whole, something that has been mandated by the EU Water Framework
Directive (European Commission, 2000; Dahl, et al., 2007). The interaction between surface
water and ground water is dynamic, with a very important major water flux pathway between
the two. This intimate relationship means that our exploitation and management of one of these
resources can have an impact on the other. An increased interest in assessing these two
hydrological components at catchment scale has led to the development and improvement of
modelling approaches, which can help in expanding the knowledge of system behaviors.
2.0 INTRODUCTION TO CATCHMENT AREA
A catchment is defined as being a specific area of land that drains to a single outlet and is
separated from other catchments by a divide. This divide can be topographical or a divide of
groundwater (U.S.G.S., 2015).

The Skjern River Catchment is located on the Jutland Peninsula, in the western part of Denmark.
It has a total area of 2500 km2 (Jensen & Illangasekare, 2011). The model for this report focuses
on the sub-catchment area, the Alergaarde Catchment, covering 1055 km2 (Karlsson et al., 2014).
The topography of the catchment has a gentle slope starting in the east from the Jutland Ridge
(125m above sea level) to sea level on the western coast. The landscape is traversed by the Skjern
River, which emerges from the eastern highlands and reaches the North Sea via the inland
Ringkøbing Fjord (Jensen & Illangasekare, 2011).

Figure 1 - Skjern Catchment. A map highlighting the model catchment area (‘Skjern River Sub-
Catchment’), including precipitation, discharge and temperature stations. (Karlsson, et al.,
2014, p.596)
 The length of the entire river system is of 96 km and the catchment including numerous lakes
(Karlsson, et al., 2014). The catchment model creates no flow boundary, being bounded by the
external topography.

Jutland is dominated by westerly winds and its passages are frequented by extratropical cyclones.
The mean annual precipitation is 990mm, with maximum levels occurring in the autumn, and
minimum in the spring. Jutland hosts mild winters and relatively cold summers with regular rain
showers. The mean annual evapotranspiration is 575 mm, while the mean annual temperature
is 8.2oC

The present-day topography was sculpted by glaciation, which fashioned the dominant
Quaternary geology. The thickness of the deposits ranges from 50m in depth in the central area
to 250 m in the south and west. Geologically speaking, these deposits consist of sand and gravel
of the Quaternary age and isolated Saalian sandy tills. Underneath these lie layers of marine,
lacustrine and fluvial deposits dating back to the Miocene, demonstrating a dynamic sea level
during that time (Jensen & Illangasekare, 2011). Thick layers of clay from Paleogene underlie the
Miocene deposits.

Figure 2 - Quaternary Geology of the Skjern Catchment (2500km2) Map displays the landscape elements of which the Skjern
river system runs through. (Jensen & Illangasekare, 2011, p.3)
The mean annual river discharge is 475 mm, with an average of 15.8 m3/s within the Alergaarde
catchment (Jensen & Illangasekare, 2011; Karlsson et al. 2014). Moraine deposits have a lower
permeability due to being poorly sorted. The outwash is generally well sorted, therefore it has
more porous spaces and higher permeability. The Quaternary and Miocene sand deposits have
formed interconnected aquifer systems within the catchment area (Jensen, 2015). The specific
properties of clay (fine, flat grains) make the clay layers to act as an aquitard, having relatively
low hydraulic conductivities. All water outside the wetlands infiltrates due to highly permeable
superficial soils, making this the dominant characteristic of the streams discharge (Jensen &
Illangasekare, 2011). The groundwater flow pattern is generally a reflection of the topography
and stream systems, having an east to west trend. The average hydraulic gradient is 0.001 (Stisen,
et al. 2011).

The catchment is mainly rural and has the following land-use distribution: grain & corn (55%),
grass (30%), forest (7%), heath (5%), urban areas (2%) and other (1%) (Jensen & Illangasekare,
2011). The past 50 years have brought significant anthropogenic changes to the catchment, such
as deforestation and engineering of the stream courses. The values of groundwater abstraction
for domestic and industrial purposes probably amounts to 10 mm/year, but the average
abstraction for irrigation can increase to up to 50 mm/year from a 20mm average, especially in
dry summers (Jensen & Illangasekare, 2011).

The catchment area has been subject to a substantial amount of climatic change over the past
133 years, experiencing a change in precipitation of 26% and a 1.4oC temperature change. This,
in turn, lead to an increase in river discharge and groundwater recharge of 52% and 86%
respectively (Karlsson, et al. 2014).

The Alergaarde Catchment area is part of the HOBE-hydrological observatory, being of particular
interest because of its exceptional daily precipitation records, spanning more than 75 years, and
with a great resolution. Additionally, the area has lengthy records of temperature and discharge
data (Jensen & Illangasekare, 2011; Karlsson, et al., 2014; Jensen, 2015).

All these data records and their resolution, together with a widespread ongoing monitoring of
the area, allow for detailed analysis of the changes that take place within the catchment area and
increase the ability of predicting into the future. The model of the Alergaarde Catchment can be
used as an accurate model for other catchment areas with similar physical features that lack
measurement facilities.

3. THEORETICAL CONSIDERATIONS
3.1. STREAM FLOW
Stream flow is the "volume of water that moves over a designated point over a fixed
period of time”, according to the EPA. This volume of water forms an erosional feature within a Commented [TRHCC1]: needs proper citation

drainage basin, flowing through a channel, affecting both the drainage of the stream and upper
groundwater flow. The two main components of stream flow are known as baseflow and
quickflow. Quickflow finds the quickest pathway through a stream system, while baseflow has a
slower response time to impacting variabilities such as climate. Also, these two components are
generated by separate sources. Quickflow can be flows directly on water bodies, overland flow
– where the infiltration capacity has been exceeded – or interflow, through drain pipes and
ditches. Baseflow is composed of interflow above the groundwater table and groundwater
flow. So baseflow is the relatively constant background stream flow.

3.2 GROUNDWATER FLOW

The groundwater, water which infiltrates into the ground, is captured within two main zones:
the unsaturated zone and saturated zone. They are separated by the water table. The
unsaturated zone (vadose zone) contains voids of both air and water which, in the case of the
saturated zone, are all filled with water (Winter et al., 1998). The vadose zone cannot be
pumped for water because it is held tightly by capillary forces. The upper part of this layer can
also be considered a soil-water zone because it can hold a high amount of water that can
participate in the process of evapotranspiration. On the other hand, the water in the saturated
zone can be withdrawn through pumping because of its pressure, giving it the name "ground
water".

The groundwater flow is transmitted from areas of recharge (inflow) to areas of discharge
(outflow), following paths of different lengths. Its energy is called a hydraulic head and it is
defined as the total energy per unit weight of water. Groundwater flows from areas of high
head to low, and thus is perpendicular to the line of equal head (Winter et al., 1998). The ability
for a layer of rock or other geological material to transmit water is characterized by its hydraulic
conductivity. A high hydraulic conductivity makes a rock porous, well-sorted, and fractured.
When it holds enough water that can be extracted, it is called an aquifer. Through these
aquifers, water flows horizontally, while through aquitards it flows only vertically. Aquitards are
layers of rocks impervious to water.

Groundwater flow systems are classified in three distinct flow system scales for unconfined
aquifers: local, intermediate, and regional (Winter, 1999; Dahl et al., 2007). These can also be
applied to confined groundwater systems (Winter et al., 1998). Flow systems are a major control
over the groundwater discharge distribution, and are of different sizes and depths, and they
overlie one another. The geomorphology of a catchment is essential in deciding the scale of the
flow system. Local systems are generally very shallow and short, having frequent exchange with
the surface. If there is a negligible local slope, no local flow systems are found. However, if a local
slope forms, the local flow systems develop and become deeper and more active as the slope
increases (Figure).

The local flow paths are shallow and short, with short travel times, being tens to hundreds of
meters in length. Because of this frequent interchange with the ground surface, the flow paths
are more susceptible to contamination (Winter et al., 1998).
3.3 GROUNDWATER – SURFACE WATER INTERACTION
Groundwater and surface water must be approached as a single resource. This is because
almost all types of surface water features are connected to groundwater bodies through intricate
flow paths that are commonly difficult to observe and measure. The factors that influence this
interaction are the climate, geology and topography, determining flow paths time and direction,
seepage patterns and recharge/discharge areas.

Streams are key surface water systems within a catchment area, being separated into two
fundamental types based on their behavior: gaining streams and losing streams (Healy et al.,
2007). Losing streams can also be connected or disconnected to an aquifer, being highly
dependent on the thickness of the unsaturated zone (Figure 4).

Figure 4 – Illustrating the differences between groundwater and surface – water interaction, and the effect on the
subsequent streams.

Another stream-groundwater type of interaction is the bank storage, which occurs when there is
a rapid rise in stream flow due to heavy raining. This water is stored in the subsurface temporarily.
The seepage distribution is affected by the morphology of the water table and geological
conditions. The shoreline is considered a common point of intersection between the water table
and land surface. In homogenous conditions surrounding the shoreline of lakes, streams, and
wetlands, there is an exponential decrease in seepage discharge with distance from the
sediment-water interface. If the geology and the isotropic conductivity are both homogenous,
flow paths are more continuous beneath the surface. In the case of heterogeneity, highly
permeable lenses can draw water into them, making the flow paths highly irregular. Layers that
have low permeability create flow paths that are vertical through them, causing more
irregularities. These variables in flow paths affect the seepage of the surface water (Figure 4).

Any changes in the variabilities of a catchment area have an impact on the mechanism of the
groundwater-surface water interaction. Among these, we can count the pumping of groundwater
which has a local impact on the hydrological system if the wells are situated in a small area.
However, this can have regional effects if the area is increased. The over use of wells can diminish
surface water by extracting groundwater which would have discharged to it.

3.4 CONCEPTUAL MODELS

Hydrological models can take many different forms, depending on the particular
catchment it is established for. Their calibration can lead to varied results. Such a model
integrates data from the components of the hydrological cycle and simulates flow paths. The
success of a model depends on the software used for it, which in the case of this report is GMS
MODFLOW. GMS created steady-state three-dimensional groundwater flow models of an
unconfined system. Steady-state modelling, where recharge leaving the system equals the
amount coming in, makes things simpler. Physically based models such as the ones used for this
report have their flow computed using certain equations like Darcy's law:
MODFLOW is mainly based on the general equation for three-dimensional flow:

Where Ss is the storage coefficient, Qp is pumping and R is equal to recharge.

This report analyzes two models: a homogenous model and a heterogenous model.

3.5 PARTICLE TRACKING

Particle tracking is used to determine the flow paths in the stream subsystem,
establishing where groundwater discharges into the stream, the source area of flow and the
age distribution along the stream segments (Modica et al., 1997).

This tool can help predict the pathways of contaminants, which can be transferred from
terrestrial to aquatic systems, especially in the hyporheic zone. Modica (1997) analyzed how
the properties of the source area of flow changed under the influence of various factors.
Increased recharge expands the source area of flow, as well as if the aquifer anisotropy
increases. If the thickness of the aquifer decreases, the flow also expands. Modica (1997) also
analyzed the age relationship between the groundwater flow paths. He found that the further
upstream, the younger the age and shorter the distance travelled to reach the stream. The
further downstream, the more varied were the distances travelled by the flow paths. Thus, the
stream is composed of groundwater which varies in residence times. Older water is likely to
discharge into the middle of a stream, while younger water is more towards the bank side of a
stream channel. These concepts are essential to understand when dealing with contaminated
streams.

The MODPATH program used on the models is based on numerous equations used in particle
tracking, among them being the Darcy flux equation which calculates the amount of water that
flows through certain cross-sectional area of the stream at a certain point in time. This helps
calculate the pore water velocity:

Where vi is pore water velocity, qi is Darcy flux and ne is porosity.

Particle paths can be calculated using the following equation by working out the pore water
velocity continuously along the stream path:

4.0 METHODS
4.1 DATA
The Jupiter database, a Danish national borehole database, contains information
regarding hydrogeology and geology from all boreholes within Denmark (Scharling et al., 2009).
The generation of the model used in this report used resources available from this database to
decipher the geology. Only boreholes with a depth greater than 100 m were used for the
models as only large scale geological features were required. The model also contains four
stream gauging stations within the catchment, to provide discharge data. The stream beds are
assumed to be one meter thick. For the homogenous model, hydraulic conductivity was set to
be that of a block of sand, which is 20. Vertical anisotropy was computed as 10 (one tenth of
the horizontal). The horizontal anisotropy was set as 1. The initial head was assigned to be 150.
Regarding the heterogeneous model, the hydraulic conductivities were distributed according to
the lithology. The horizontal anisotropy was assigned to be 1 (isotropic), while the vertical
anisotropy was 3 (one third of the horizontal). The recharge is uniform over the entire
catchment area for both the homogeneous and heterogeneous models, being equal to the total
outflow of the stream on the western boundary. Besides inserting this data into the models, we
calibrated them afterwards to change the stream conductance and the hydraulic conductivity.

4.2 MODEL SETUP

For the models we analyzed in this report, the conceptual model was built involving GIS
tools in the Map module. This integrates parameters of the Alergaarde catchment, topography,
stream data, model boundaries, and hydraulic conductivities. The model was built using a
background map as a catchment coverage, with integrated GIS coordinates. The Default layer
range was set to be from 1 to 20, activating an upper and lower boundary. A stream coverage
was also input in the same way. Using the Arc tool, we defined the outer boundaries for the
catchment, once the catchment was activated. The boundaries of the catchment are set up by
the topography. All the shapes created are then transferred to feature objects. These are used
to build polygons which define the catchment polygon. Using the Triangulated Irregular
Network (TIN) surface, the three-dimensional elevation is simulated. TIN is a digital data
network using GIS to represent a surface. To build this, vertices were distributed with a spacing
of 250 meters. The bottom of the aquifer is defined through TIN, being generated at a uniform
distance of 100 m from the elevated surface.

Once the conceptual model was complete, a grid was then computed to fit the catchment area
of 1055 km2 , being divided into cells of 200x200x5 meters. The Z direction contains 20 defined
layers. The next step was to simulate the groundwater catchment using MODFLOW. The
recharge and streams were activated, while solids were converted to MODFLOW through the
grid overlay so that the grid followed the topography.

The uncertainty standard deviation for head observation was calculated at a 98% window. The
stream information was integrated into the 45 defined arcs. The top and bottom elevations of
the stream bed were set as stream stage -1 and -2. A value of 0.085 was set as the roughness
coefficient for all the arcs. The initial conductance was set to 0.2. The next step involved
creating arc groups for the stream, grouping them in respect to their gauging stations based on
discharge data. A 10% error is considered from the gauging stations. The model was run
through MODFLOW afterwards, not before entering the initial parameters. This process was
done for both the homogeneous and heterogeneous models.

4.3 GEOLOGICAL MODELLING

Groundwater flow is critically influenced by the subsurface heterogeneity. This
hypothesis will be investigated by the use of two heterogeneous models of different geological
complexity, dubbed “simple” and “complicated” (Fleckstein et al 2006). The heterogeneous
model was built using geological data provided by the Jupiter database. With the use of excel,
borehole data was categorized a heterogeneous model. The heterogeneous model contains
seven categories classified using hydraulic properties, age: “Soil”, “Quat. Sand”, “Quat. Clay”
“Quat. Gravel”, “Pre-quat. Sand”, “Pre-Quat. Clay”, and “Pre-Quat. Gravel”. Age is an important
variable in groundwater processes, for example, undergoing increased diagenesis can result in
the altering of hydraulic properties. Furthermore, by utilizing a score totals of 5 and 6, results
for the boreholes within the complicated model should be more precise.

Boreholes data located too close to each other or containing peculiar stratigraphic sequence
were either deleted or altered to better fit the model. Based on data input into the model,
horizons were created that matched the different layers. Afterwards, the boreholes were
snapped to the TIN surface. The next step involved reforming geological cross sections to make
geological sense. After setting the bottom and surface elevation of the domain, the horizons
were used to create solids which were incorporated into MODFLOW. The model was run after
the hydraulic conductivities of all layers were assigned appropriately regarding their lithology
type. To improve the heterogeneous model’s performance, drains were also incorporated. We
assumed that all drains are situated 1.5 meters below the surface. This data was represented
through defining a TIN. The drainage polygons were assigned appropriate conductance values.

4.4 FLOW BUDGET

Flow budgets have been calculated to measure the variation of flow along stream arcs, by
separating the stream into “budget zones” which are defined from the east to the west. There
are nine separated and labelled Zone Budget ID’s in total within each model, containing four
cells. Flow budget is calculated as the flow travelling in and out of these four cells. If the water
traverses from the cells to the stream, it is indicative of a gaining stream (defined in Section
3.3), whereas water going out of the cell into the stream bed represents a gaining stream
(defined in Section 3.3). This data can be viewed in the flow budget table as stream leakage,
which portrays everything from the “OUT” section as matter which leaves the groundwater and
the “IN” category adding to the groundwater.

Figure 5 – Illustrating how the geological modelling has been set up for the Heterogeneous Model, and how these are
converted into solids.
4.5 PARTICLE TRACKING
Particle tracking is an instrumental tool used to distinguish the source locations and
groundwater residence times of water discharging to a specific stream within a catchment
(Modica et al 1998). Particle tracking and subsequent analysis within MODFLOW is conducted
via another program called MODPATH. To create the simulation, one hundred particles were
generated, resulting in one hundred path lines within a cell chosen from each of the flow
budget zones along a selected stream. To provide information about where the water for each
cell originated, the configuration was set with a backward direction as opposed to a forward
trajectory, which would indicate their destination. MODPATH subsequently produces a path
line report for each of the cells, indicating the age of the particles. More than one age is
generated due to incorporating various steps in its calculation, therefore the last value
indicated the time in which particles encounter the water table. This data is then extracted and
analyzed in excel, creating one hundred values for each zone cell, for which individual
histograms are made. Furthermore, to compare the zones within a model an average age was
calculated, and plotted with a trend line (Figure 6).

Figure 6 – Illustrating the location of the 8 Budget ID zones and their subsequent particles (Extracted from the
Heterogeneous Mode)
4.6 Parameters and Calibration
By definition, a calibration adjusts parameters to minimize potential differences (U.S
Geological Survey, 2015). For this investigation, calibration is enabled to assess whether the
model accurately represents the physical system of the catchment from the combination of the
inputted independent variables. The Root Mean Squared error (RMS) for head data is employed
as a gauge of accuracy for each model and it is intended that this value must be as low as can
be achieved. Yet the RMS cannot be lower than three, as this was the level of uncertainty of the
head observations.

To obtain a stable model, in which the simulated and observed values are in agreement, trial-
and-error procedures were practiced (resulting in a minimized RMS). The variable parameters
included stream conductance and hydraulic conductivity, and after each adjustment MODFLOW
would be run to assess the correlative change in RMS. The values for the lowest RMS for each
model are shown in Table 1, and suggest that the Heterogeneous has the least accurate data –
why is this, and could it affect the results?

Model Root Square Mean

Homogeneous 3.73
Heterogeneous (Simple) 6.50
Hetergeneous (Complicated) 5.98
Table 1 – Displaying the Root Square Mean calculated for each of the Models. The simple model has been removed due to its
large error values, and this will not be comparative to other models.

Within the models, the two capital sources of error are heterogeneity within the lithologies and
scale inconsistency. Furthermore, several assumptions have been made prior to running the
model, the foremost is the observed data measured from a singular area is applicable to a much
larger horizonal plane (200x200m). However, this is most likely not a correct assumption, and
these sources of error must be taken into account when analyzing the resulting data.
5.0 Results
5.1 Homogeneous Model
5.1.1 Calibrated Parameters

Horizontal k Vertical K Horizonal Vertical

ID Name
(m/d) (m/d) Anisotropy Anisotropy

1 Material 20 0 1.0 10

Table 2 – Material properties of the Homogeneous Model

The homogeneous model is created as a control to see how the groundwater flow is
affected by changes in the geology. Therefore, all the stream segments are treated as a “sand
box” model and given uniform, standard values. To reflect this, the hydraulic conductivity has
been assigned as 20m/d, which lies within the standard value for sand in the order of 10-
100m/d and a stream conductance of 0.2((m2/d)/m).

These two characteristics lead to the calibration of the model, providing an RMS of 3.73 (Table
1).

5.1.2 Hydraulic Heads

Figure 7 – Hydraulic Heads (Homogeneous Model)

 The error bars shown demonstrate the difference between the calculated and observed
head data, with the greatest shown in red and the lowest in green. Generally, heads of the
greatest difference are to the south west of the catchment, and the lowest difference in the
north west – which is hard to explain. Furthermore, the huge difference within the south west
is troubling, as it indicates a large inaccuracy in the data. However, the catchment processes are
intricate and complicated. Many generalizations have been made in the model to account for
this, which could create skewed data and anomalous results. Despite this, the majority of the
error bars are amber or green, suggesting that the observed and calculated head values are
within ~68% normal distribution, so the model can be accepted.

5.1.3 Flow Budgets

Figure 8 – Stream leakage over the entire system

Figure 8 displays the stream leakage over the entire system. The area in red represents the
greatest groundwater discharge, most likely they are localities where there is increased
groundwater-surface water interaction attributed to a higher water table. These points also
represent the areas where the aquifer is most sensitive to contamination from surface water.
As shown in Figure 8 stream leakage seems to be isolated to the eastern area of the catchment,
suggesting that it is in a gaining state to the east and a losing state to the west. Stream leakage
can also be described as “base flow gain” and the greatest gain usually correlated to the largest
difference in hydraulic head and stream stage. Characteristically, this occurs is in the widest,
middle section of a streams length, which is supported by the higher values of stream leakage
within this area of the model.

5.1.4 Flow paths and groundwater ages

Figure 9 – Cells and particle pathways associated with budget zones

Figure 9 displays the location of the eight particles created along the stream and the
flow paths (explained in Section 4.4 and 4.5). Due to the homogeneous nature of this model the
pathways merge, as water will always take the path of least resistance, and with no geological
heterogeneity the particles can take the same route.
 Figure 10 – Flow path in the 3D side view (Homogeneous)

The particles are tracked backwards, meaning that particles are followed upstream to their
source. The flow paths presented are typical in that the particles percolate into the ground and
flow up to the surface. The homogeneous nature of the model encourages vertical flow in the
west of the catchment, then travelling horizontally towards the east; this is comparable with
models explained by Modica (1997). It is suggested that groundwater should move parallel to
the section, flow from upstream to downstream (in this case, east to west), and discharge to
the stream – which are supported by our model. Therefore, the homogeneous model shows
that the particles following similar paths, and do not (as seen in the heterogeneous model)
contain altered routes.

Homogeneous - Mean Age of Particles

1.40E+06
1.20E+06
Average Age (Years)

1.00E+06
8.00E+05
6.00E+05
4.00E+05
2.00E+05
0.00E+00
Zone 2 Zone 3 Zone 4 Zone 5 Zone 6 Zone 7 Zone 8 Zone 9

Flow Budget Zone

Figure 11 – Graph displaying the mean age and trending like of the particles at each cell throughout the
catchment (9 closest to stream output)
Generally, the ages lie between 600,000 years to 1.2Ma and increase in age towards the output
of the stream. This is expected, as further downstream the particles will have travelled a longer
distance at a deeper depth (Figure 10). Figure 10 is mirrored by the data, as the particles seep
to the bottom depths of the aquifer before surfacing, thus, the starting ages will be higher
(although, it is still unusually high). Furthermore, the particle tracker and subsequent ages help
to identify areas which are more vulnerable to contamination. For example, groundwater has
had a longer time to undergo chemical reactions with pollutants where flow paths and ages are
highest (Modica et al 1997). The real life application of MODPATH is to compute how long a
pollutant could travel through a system, and what implications this would have.

Figure 12 - A Histogram conveying the frequency distribution of the particles recorded in Zone 4.

Histograms for each zone for the models have been computed and are presented in Appendix
1. Particle pathways with shorter ages suggest that they have stayed shallower in the aquifer
compared to their longer counterparts. Figure 12 conveys the variation of the distribution
within one cell, 52 particles are shown to be under the age of 620,000 years, indicative of deep
aquifer infiltration.

5.2 “Simple” Heterogeneous Model

Originally two heterogeneous models were created, dubbed “simple” and “complicated”.
However, only the “complicated” model will be presented due to significant errors produced in the
“simple” model. Virtually none of the heads calculated (Figure 13) for the “simple” model are within
68% of the normal distribution, and the RMS yielded a value of 6.50. This poses a considerable problem.
The “simple” model will not be included as it will not be representative of the actual processes occurring
in the Alergaarde catchment; thus, it will be incomparable to the other models.

Figure 13 – Head values for the Simple Heterogeneous Model, the large error values suggest that this model is
not representative of the catchment, and thus is not included in our results.

5.3 “Complicated” Heterogeneous Model

5.3.1 Calibrated Parameters
The different stream segments have been assigned a stream conductance of 0.2
((m2/d)/m) and a drain conductance of 0.0001 ((m2/d)/m). Furthermore, the hydraulic
conductivity was altered to minimize the RMS head, calibrated at 5.98. This high RMS head
could be due to using the score total of 5-6 within the boreholes. However, as this was the
lowest possible value achieved the model will be used, despite erring on the side of caution
when drawing conclusions. The hydraulic conductivity varies between lithological units to best
suit their realistic hydraulic characteristics.
 Horizontal k Vetical K Horizonal Vertical
ID Name
(m/d) (m/d) Anisotropy Anisotropy
1 Quat. Sand 12.5 0.0 1.0 10.0
2 Quat. Clay 0.1 0.0 1.0 10.0
3 Quat. Gravel 17.5 0.0 1.0 10.0
4 Pre-Quat. Sand 10 0.0 1.0 10.0
5 Pre-Quat. Clay 0.1 0.0 1.0 10.0
6 Pre-Quat. Gravel 15 0.0 1.0 10.0
7 Soil 10 0.0 1.0 10.0
Table 3 – Material properties of the Heterogeneous Complicated Model

5.2.2 Hydraulic Heads

Figure 14 – Hydraulic Heads, Heterogeneous Model

Much like the homogeneous model, the error bars displaying the least difference (shown in
green) are concentrated in the north west of the catchment. However, there seems to be an increased
difference between the observed and calculated difference towards the centre and north of the
catchment. This could be due to the change in geology, the west is predominantly moraine deposits as
opposed to outwash.

5.2.3 Flow Budgets

Figure 15 - Stream leakage over the stream system

Much like the homogeneous model, stream leakage can be referred to as “base flow gain”. The
areas of greatest leakage are shown in red, and are indicative of localities with the largest groundwater-
surface water interaction. There appears to be substantial differences to the homogeneous model. This
can be attributed to the change in conductance values, alongside the distinct geological lithologies
which affect the thickness of the aquifer. As mentioned above, the greatest base flow gain occurs here
the water table is high, therefore, clay layers (aquitards and impermeable beds) must be thin or vacant.

The negative values suggest that the stream arc is gaining, and groundwater must be infiltrating to the
stream. Moreover, much like the homogeneous model the highest values of leakage are present in the
middle of the stream segment, and decrease toward the catchment boundary. Similarities must be due
to the same assigned stream conductance, which would keep the stage of the stream constant; thus,
differences must be attributed to the changing of hydraulic heads.

5.2.4 Flow Paths and Groundwater Ages

Figure 16 – Flow paths in cross sectional (side) view and birds eye (top) of the catchment. Pathways are dark
blue, and cells in black.

The particles are tracked backwards towards the source. Overall, the younger lithologies with
higher hydraulic conductivity (Sand and Gravel) have the greatest flow and predominantly horizontal
flow paths; whereas, units with lower hydraulic conductivity (Clay) have the lowest flow and are
dominated by vertical flow. It is clear from Figure 16 that geological units greatly affect the pathway of
particles, with distinct lithologies of variable characteristics paths become more convoluted and
discontinuous, as expected from the discussion by Winter (1999).
By comparing the ages of the budget zones throughout the model (Figure 17), it is obvious that the
general trend increases in age further downstream (towards Zone 9), like that of the homogeneous
model. However, there is much greater variability in the age of the particles, for example, in Zone 8 age
suddenly decreases. This could be due to a sudden change in lithology, or an alteration in the source
location. Moreover, the maximum age is within Zone 6 and 7, suggesting that they have the deepest
aquifer flow.

Heterogeneous (Complicated) - Mean Age of

Particles
3.00E+07

2.50E+07
Mean Age (Years)

2.00E+07

1.50E+07

1.00E+07

5.00E+06

0.00E+00
Zone 2 Zone 3 Zone 4 Zone 5 Zone 6 Zone 7 Zone 8 Zone 9
Zone

Figure 17 – A line graph showing the evolution of the age of particles over the zones throughout the catchment.

Figure 18 – A line graph illustrating the frequency distribution of ages of particles from Zone 4
This diagram shows that the majority of particles are ~390,000 to 680,000 years old, generally this is
seeming old and diverse. Although, there is a much larger range of ages for the heterogeneous model in
comparison to the homogeneous.

6.0 Discussion and Conclusion

The two models suggest that geology is a paramount variable when recording the hydraulic
properties of a catchment. GMS, MODFLOW and MODPATH accomplished the simulation of the
Alergaarde catchment, however some of the results yielded require skepticism. For example, generally
the ages of the particles in very old and there is an absence or local or short lived- groundwater systems,
which is peculiar. Most likely, there are local groundwater systems with shorter lives present, but an
error within GMS (either with the calibration or setup) has prevented this from being recorded.
Therefore, applications for the model on the local scale are sparse.

Respectively, the root mean square error values were 3.73 and 5.98 for the homogeneous and
heterogeneous model, therefore heterogeneous model has a greater difference between the calculated
and observed head. This could result in the observations being less accurate or less representative of the
Alergaarde population. Error bars calculated for the hydraulic head suggest that both the models have
their uncertainties (as any simulation does), but the greatest differences are to the west. Stream leakage
for the heterogeneous and homogeneous seem to support one another, with the greatest flow in the
middle of the stream, and suggested that the streams are gaining. Management of contaminants in the
intermediate and regional groundwater system are now possible due to the knowledge of groundwater
leakage and groundwater-surface water interaction. As mentioned in Section 2.0, the Alergaarde
catchment is mostly used for agricultural purposes, which could indicate a higher use of pesticides and
fertilizers. Areas with the greatest groundwater-surface water interaction would be most vulnerable to
leaching of these chemicals into the aquifer.

Finally, flow paths and particle tracking indicated that the homogeneous model has a moderately simple
path and relatively low variation in age. By including and defining the geology, flow paths were altered –
particles moved horizontally through units of higher hydraulic conductivity and vertically through those
of lower values, creating a wider distribution of residence times. Generally, both models conveyed that
closer to the outlet particle age increases.
In conclusion, despite the many imitations and errors with this program the models provide information
about the relationship of water fluxes within the catchment, which is vital for the management of this
resource.

Reference List
DAHL, M., NILSSON, B., LANGHOFF, J. and REFSGAARDE, J., 2007. Review of classification systems
and new multi-scale typology of groundwater-surface water interaction. Journal of Hydrology, 344,
pp. 1-16.

EUROPEAN COMMISSION, 2000. EU Frameworkd Directive. European Parliament and Commission,

Official Journal (OJ L 327), Directive 2000/60/EC.

EUROPEAN ENVIRONMENT AGENCY, Oct 9, 2015-last update, Land use - State and impacts
(Denmark). Available: www.eea.europa.eu/soer/countries/dk/land-use-state-and-impacts-
denmark2017].

HEALY, R., WINTER, T. and LABAUGH, JAMES, FRANKE, LEHN, 2007. Water Budgets: Foundations for
Effective Water-Resources and Environmental Management . USGS, C ircular 1308, pp. 90.

JENSEN, K.H., 2015, 13, 10-last update, Hydrogeological Observatory Description: Retrieved from the
PArtnership for European Environmental Research.
Available: www.peer.ue/fileadmin/user_upload/projects/flagship_projects/PEER_Euraqua/Skjern%20D
enmark.pdf2017].

JENSEN, K.H. and ILLANGASEKARE, T.H., 2011. HOBE: A hydrological Observatory. Vadose Zone
Journal 1-7. Vadose Zone Journal, , pp. 1-7.

JENSEN, K.H., SONNENBORG, T., JENSEN, K.H. and REFSGAARDE, J., 2014. Historical trends in
precipitation and stream discharge at the Skjern River Catchment, Denmar. Hydrology and Earth
System sciences Discussions, European Geosciences Union, , pp. 595-610.

SCHARLING, P., RASMUSSEN, E., SONNENBORG, T., ENGESGAARD, P. and HINSBY, K., Hydrogeology
Journal. Three - Dimentional Regional - Scale hydrostratigraphic modeling based on sequence
stratigraphic mothods: a case study of the Miocene succession in Denmark . 2009, 17, pp. 1913-
1933.

SONNENBORG, T., CHRISTENSEN, B., NYEGAARD, P., HENRIKSEN, H. and REFSGAARDE, J., 2001.
Transient modeling of regional groundwater flow using parameter estimated from steady-state
automatic calibration. Journal of Hydrology, 273, pp. 188-204.

U.S. GEOLOGICAL SURVEY, 2015,10-last update, Model Calibration. Available:

pubs.ugs.gov/sir/2007/5044/section5.html2017].

WINTER, T., ROSENBERRY, D. and LABAUGH, J., 2003. Where does the Ground Water in Small
Watersheds Come From? . National Ground Water Asoociation, 41, pp. 989-1000.
APPENDIX 1 – Histograms of Model Data
APPENDIX 3 – Heterogeneous (Categorisation) Model Data