Download as pdf or txt
Download as pdf or txt
You are on page 1of 8

Module 9: The Fresh water/Saltwater interface and Groundwater

Modeling
The fresh water/saltwater interface
The oceans have high total dissolved solids because they receive inflow from streams and groundwater
that have dissolved rocks. The concentration of dissolved solids, or salinity, is increased by evaporation
and reduced by precipitation of shells by organisms. Because much of the world’s population lives near
the coast, intrusion of saltwater into coastal aquifer is a major concern.

Effect of Water Density Variations on Hydraulic Head


Previously in this course, the density of the water has been assumed uniform. For some problems, it is
necessary to consider the effects of variable fluid density on groundwater flow. Variable density is
important near the coast, where fresh and saltwater interact and in some contamination studies, where
contaminating liquids are sometimes denser or
less dense than water.

Consider two wells that both are screened at an


elevation of 80 ft above sea level. The water
levels in both wells are 90 ft above sea level. If
both wells had the same water density, there
would be no horizontal flow between the two
because h1 = h2=90 ft.

But what if the water in Well 1 is fresh water,


with a density of 1000 kg/m3 and Well 2 is
seawater with a density of 1025 kg/m3?
Figure 1. Example of wells with fluids of different densities.
A useful concept that allows us to compare the
hydraulic heads at the two wells is the fresh water head. It is the equivalent height of water in a well if it
were fresh water. You can calculate the fresh water head by:

𝝆
𝒉𝒇 = 𝒛 + 𝒉𝒑 ( 𝝆𝒘 ) Equation 1
𝒇

where w is the density of the water or other fluid in the well and f is freshwater density (~1000 kg/m3),
z is the elevation and hp is the pressure head. In this case, the fresh water head of Well 1 isn’t changed
(Well 1 h =90 ft). For Well 2, the freshwater head can be calculated with Equation 1:

h2f=80 ft + 10 ft (1025/1000) = 90.25 ft

Because the fresh water hydraulic head at Well 2 is greater than at Well 1, groundwater will flow from
Well 2 to Well 1. Important note: This comparison is only accurate for calculating the horizontal flow
between two wells screened at the same elevation. Otherwise, comparisons need to consider the
“buoyancy” effect of different densities (i.e., less dense water rises and denser water sinks).
The Ghyben-Herzberg relationship

In coastal areas, groundwater discharges to the ocean. As described above, because salt water is denser
than freshwater, it flows inland – even if the water levels are exactly the same. As a result, a “wedge” of
saltwater exists beneath the freshwater in an aquifer. The Ghyben-Herzberg relationship provides a
simple means to estimate the depth to the freshwater/salt water interface beneath an aquifer. In the
equation, h is the hydraulic head relative to sea level, and zs is the depth of the fresh water/saltwater
interface below sea level.
𝜌𝑓
𝑧𝑠 = 𝜌 ℎ Equation 2 (or Eq. 3.46 in the text)
𝑠 −𝜌𝑓

Typical ocean water density, s, is 1025 kg/m3, and typical freshwater density, f, is 1000 kg/m3 . As a
result zs = 1000/(1025-1000) x h OR zs = 40h. For typical coastal conditions, there will be 40 ft of
freshwater below sea level for
every 1 ft of freshwater above
sea level. The total thickness of
the freshwater zone will be h +
z. So if the water table is 4 feet
above sea level, the interface
will be 160 feet below sea level
and the freshwater zone will be
164 feet thick. The freshwater
zone near the coast is
sometimes referred to as a Figure 2. Freshwater/saltwater interface in an unconfined aquifer. Figure from the USGS.

freshwater lens due to its shape

The Ghyben-Herzberg relationship is very


useful but is a simplification. In the figure
above, the freshwater thickness would be
zero at the shoreline. This is not realistic
because it would not allow any discharge to
the ocean. In reality, groundwater flow
pushes the interface seaward, allowing an
offshore discharge zone. The width of this
zone will be proportional to the total
discharge (Q) through the aquifer. If all else
is the same, the greater the discharge, the
wider the the zone of offshore freshwater Figure 3 The freshwater/saltwater interface has a transition zone and an
outflow face, where freshwater discharges offshore.
outflow.
Figure 32.
In addition to ignoring outflow to the ocean,
the Ghyben-Herzberg equation doesn’t account for the mixing between freshwater and saltwater.
Unlike oil and water, freshwater and saltwater can mix. As a result, there will not be a sharp boundary
between the two, such as assumed by the Ghyben-Herzberg relation. Instead, there will be transition
zone, where the salinity gradually changes from fresh to saline. This transition zone occurs due to
molecular diffusion. In addition, the fresh water/saltwater interface will shift with the tides and with
seasonal or longer-term changes in the aquifer discharge. As the interface shifts, salt and fresh water
are mixed. As a result of diffusion and mixing, some salt will be added to the freshwater and carried out
to sea. This salt is replaced by seawater that circulates inland and upwards (dashed flowlines in Figure
3).

In confined aquifers, the overlying confining layer helps protect the aquifer from the denser seawater.
As a result, freshwater can extend beneath the ocean. Groundwater slowly discharges upward to the
ocean through the confining layer. For example, in the figure below from Cape Cod (Figure 4), the fresh
water/saltwater interface is very close to the coast in the unconfined portion of the aquifer, but can
extend seaward where the aquifer is confined. In many locations, the continental shelves contain fresh
water that is thought to have recharged the aquifer during the last glacial period, when sea level was
>100 m lower.

Figure 4. A schematic cross section of Cape Cod showing freshwater/saltwater interfaces for an unconfined and a partially
confined aquifer. Figure from the USGS.
Saltwater Intrusion

Pumping reduces hydraulic head, and as a


result can cause upconing (Figure 5), or a
vertical rise of saltwater beneath the well.
Following the Ghyben-Herzberg equation,
1 foot of drawdown can eventually lead to
the interface rising 40 feet. The upconing
develops through time. Upward migration
of salty water will occur more quickly in
aquifers with very high hydraulic
conductivity (and low effective porosity)
Figure 5. Upconing of the interface beneath a pumping well.
and more slowly in aquifers with moderate
hydraulic conductivity (and high effective
porosity). If pumping is stopped, the
interface can take a long time to return to
its previous location.

Reduction of discharge allows the


freshwater/saltwater interface to move
laterally. The blue arrows on Figure 6
show lateral migration due to pumping in
a coastal plain aquifer system. Both
upconing and lateral migration of the
interface are termed saltwater intrusion.

A common response to saltwater intrusion


due to pumping is to develop alternative
water supplies. The pumping wells can be Figure 6. Saltwater intrusion due to lateral migration of the
moved inland or surface water resources used freshwater/saltwater interface.
instead of groundwater. Other options include
using more wells, but pumping less from each well – or using horizontal wells rather than vertical wells.
Either of these methods allows collection of the water over a larger area and results in a shallower cone
of depression, which will cause less upconing.

In some coastal locations, the aquifer is replenished with treated waste water or runoff to increase the
hydraulic head and prevent saltwater intrusion problems. In other locations, desalination is used to
remove the dissolved solids from brackish water or seawater. The cost and energy use of desalination is
high, and there is brine that must be disposed. As a result, other options are often tried first.
Numerical Groundwater Modeling

The groundwater flow equations are partial differential equations. Combined with boundary conditions
and initial conditions (for transient problems), they can be solved to find hydraulic head at any point or
to calculate flow. Analytical (or exact) solutions to groundwater flow problems are limited to relatively
simple geometries and boundary conditions. For example, Theis (1935) found an analytical solution for
transient flow to a well in a homogenous, isotropic, and infinite aquifer. This solution can be used to
solve for drawdown at any radius and time. However, many problems are too complex for analytical
solutions. In these cases, numerical approximations can be used.

Numerical approximations
Like an analytical solution, a numerical approximation for groundwater flow uses Darcy’s Law and
conservation of mass. However, the approach is somewhat different. In a numerical approximation, the
region to be modeled is divided (discretized) using a grid or mesh. The figure shows an example of a
three dimensional grid. The rows and columns divide up the model area, and the layers represent the
vertical dimension. Often the model layers correspond to layers of aquifers and confining units. There
are two common methods for numerical approximations: the finite-difference method and finite-
element method. We’ll focus on the finite-difference method as an example. It is the basis for the most
commonly used groundwater code, MODFLOW, which was developed by the USGS.

At the edges of the modeled area, boundary conditions are applied. These are either 1) specified head,
2) specified flux, or 3) head-dependent flux boundaries. Flow between each of the cells in the grid is
governed by Darcy’s law. The hydraulic gradient is calculated using the head difference between
neighboring cells divided by the distance between the centers of the cells. This hydraulic gradient is
multiplied by the hydraulic conductivity and the cross sectional area to calculate discharge. Conservation
of volume is also applied. The inflow to the cell minus the outflow must equal the change in water
storage in the cell. For steady-state conditions, inflow must equal outflow. Because the hydraulic head
in each cell is affected by the neighboring cells, hydraulic heads in every cell in the grid must be solved
simultaneously. Often this simultaneous solution is accomplished by iteration, which means that an
initial guess is made, and the matrix of equations are solved repeatedly. Iterations are ended when the
resulting hydraulic heads change very little between iterations.

As an example of the differences between analytical solution and a numerical approximation, consider a
10-cm long Darcy cylinder full of homogenous sand. Hydraulic head is 5 cm at one end and 4 cm at the
other end. Flow is 1D and steady state through the cylinder. The appropriate groundwater flow equation
2h
is: 0
x 2

And the boundary conditions are:

 h=5 cm at x=0 cm
 h=4 cm at x=10 cm
The analytical solution for this case would be:
h=5 cm +0.1x. The potentiometric surface is linear. This is an analytical solution, and you could easily
calculate h at any x location in the cylinder.
Now consider a numerical approximation for the same problem. The cylinder is discretized into cells. We
will use cell dimensions of 1 cm. We will enter the boundary conditions at x= 0 cm and x=10 cm in the
cells on the left and the right, respectively.

For a homogenous 1D steady-state problem, the finite-difference approximation for the equation above
is that each cell’s hydraulic head is the average of the cell to the left and right. You could calculate this
by hand or plug this into Excel as a formula. However, because each cell depends on the hydraulic head
of the neighboring cells, the problem is “circular”, and you will have to have Excel iterate to find the
solution.

5 ? ? ? ? ? ? ? ? ? 4

Although this is a simple example, it demonstrates that sometime an analytical solution is much simpler
and quicker than a numerical solution. You can solve for one point without having to iterate repeatedly
at at every grid cell.
The analytical solution is often more accurate because the numerical approximation breaks up the
potentiometric surface into linear segments, which introduces error. For example consider if a cone of
depression’s potentiometric surface (blue) were coarsely discretized. The red segments show the
gradient in a finite-difference approach. The resulting flow
would differ from the actual flow along a curving
potentiometric surface, introducing error in the calculations.

On the other hand, numerical approximations can allow simulation of very complicated geometries and
aquifer heterogeneity. This strength often makes up for the limitations of numerical approximations.

Modeling Procedures
Purpose
The first step in solving any groundwater problem, whether it be an analytical or numerical solution is to
determine your objective. Often, the purpose of modeling is predictive, to quantify how an aquifer will
respond to a possible future change such as new pumping. Modeling can also be interpretive, to better
understand the behavior of a system. For example, you could simulate how past recharge changes might
have affected flow in the Great Artesian Basin. Or modeling could help you determine where it is most
useful to collect new field data. Finally, modeling can be generic to understand a basic process. For
example how does groundwater cut across the meanders in meandering streams.

Data Gathering and Conceptual Model Formation


Much of this past semester has been
about how data is gathered and
interpreted to understand a system.
The geologic and hydrologic
information are gathered to put
together a picture of the geometry
of a system, the system’s water
budget, and where water is entering
and leaving the system.
Characterizing hydraulic conductivity
and storage properties is part of this
step. In some cases, this will involve
a literature review, but often new
field data are needed. This step is
often the most time-consuming and
can take many months for a large Figure 7. An example of a numerical groundwater model where the aquifers
and confining units are discretized into layers, rows, and columns. Figure
study area (such as the example of a from the USGS.
model geometry for the Lake
Superior regions which is shown in the Figure 7).
Selection of Governing Equation and Code
The code is the set of computer instructions used for the numerical approximation. MODFLOW is an
example of a code. A code is different from a model, which is the application of a code to a particular
problem.

It is important to understand the system and the model purpose before deciding on a model code and
governing equation. For example, the occurrence of variable-density flow or unsaturated zone flow will
change which governing equation is appropriate. As another example, you might need a code that can
couple groundwater flow to deformation if you want to simulate subsidence.

Model Design
After a code is selected, the model geometry, boundary conditions, and hydraulic properties need to be
assigned. For steady-state models, hydraulic conductivity needs to be designated at every cell. For
transient models, storage properties are also needed, as are initial conditions. Boundary conditions
(specified head, specified flux, or head-dependent flux) are entered to represent surface water,
pumping, or physical boundaries, such as the limits of aquifers. A graphical user interface (GUI) helps
enter the information and assign it to the correct model cell. The GUI then makes input files that are
processed by the code. After the code has run, the GUI can also help with visualizing the results.

Model Calibration and Sensitivity


We can never know hydraulic conductivity, recharge, and storage properties perfectly. Therefore,
models need to be calibrated. Model results are compared to observations of hydraulic heads and flow
rates, and the input parameters are adjusted so that the model results best reproduce observations.
This step is necessary for predictive models, and many interpretive models. Generic models don’t
require calibration because they are not representing a particular location. The calibration process
includes testing how sensitive results are to uncertainty. This can help with future data collection by
determining which data affect hydraulic head the most.

Verification
If possible, the model should be tested by seeing if it can reproduce a second set of observations that
was not used for the calibration.

Prediction and Prediction Sensitivity


After the model is calibrated it can be used for predictions. For example, a new pumping well is added
and the future drawdown can be examined. A prediction could also concern past behavior. For example,
you could use a model to predict whether contaminants could have been sourced at a particular site.
Prediction sensitivity refers to varying uncertain parameters (recharge, hydraulic conductivity, future
pumping or recharge scenarios) and seeing how the results of the prediction change.

You might also like