Skip to main content

Dewatering analysis is common in the civil, mining, and environmental sectors. This webinar demonstrates the simulation of a groundwater dewatering problem using SEEP3D.

The ultimate goal of a dewatering analysis is to understand the transient response of the flow system and determine pumping rates, pore-water pressure distribution, and the extents of drawdown. This webinar demonstrates the creation of two three-dimensional groundwater dewatering problems and their associated drawdown conditions. The results of the SEEP3D analyses are then exported into the third party software, Python, for the creation of drawdown maps.



Marc Lebeau


29 min

See more on demand videos


Find out more about Seequent's environmental solution

Learn more

Video Transcript

<v [Instructor>Hello and welcome to this GeoStudio webinar</v>

that focuses on both de-watering,

which includes pumping and pressure relief systems,

and drawdown maps in SEEP3D.

My name is Marc LeBeau

and I’m a research and development scientist

with the Geoslope engineering team at Seequent.

Today’s webinar will be approximately 30 minutes long.

Attendees can ask questions using the chat feature

and I will respond to these questions via email

as quickly as possible.

It should be noted that a recording of the webinar

will be available so participants

can review the demonstration at a later time.

Geostudio is a software package developed

for geotechnical engineers and earth scientist

which is comprised of several products.

The range of products allows users

to solve a wide array of problems

that may be encountered in these fields.

Today’s webinar will focus on Seep3D.

Those looking to learn more about the products,

including background theory,

available features and typical modeling scenarios,

can find an extensive library of resources

on the Geoslope website.

Here you can find tutorial videos,

examples with detailed explanations,

and engineering books on each product.

To begin this webinar,

let’s start by answering three fundamental questions.

The first question is, why?

Why should we consider using a de-watering system?

The second question is, what?

What are the different types of de-watering systems?

And the third very important question is, how?

How can we implement these systems in a numerical model?

So why exactly should we consider using

a de-watering system?

Well, the construction of engineered structures,

such as buildings, dams, levees and tunnels

often requires excavation into water bearing soils

and/or consideration of underlying artesian pressures.

The reasons for installing a de-watering system

therefore include, but are not limited to,

lowering the phreatic surface and intercepting seepage

that would otherwise emerge from a slope

or bottom of an excavation,

increasing slope stability,

preventing erosion heave and uplift

that may occur at the bottom of an excavation

or at the toe of a dam,

reducing lateral loads on excavation sheeting and bracing,

and reducing required air pressures

during tunnel construction.

There are many types of de-watering systems.

Some of these systems are quite old

and were introduced prior to the advent

of modern well installation techniques

and pumping apparatus.

These systems include,

permanent and less expensive structures,

such as sumps and ditches that are well adapted

to small excavations or fine grain soils,

sheeting and open pumping,

sheeting with deep-well sumps,

which are now mostly replaced by wells with screens,

and more recent techniques,

which include well point systems in which small wells

are connected to centrifugal pumps,

deep well systems in which large diameter wells

are connected to the submersible or turbine pumps,

pressure relief systems in which well points

or deep wells are used to decrease artesian pressures,

horizontal systems in which horizontal pipes

are projected from shafts or wells,

vacuum de-watering systems for fine grain soils

in which well points are connected to vacuum pumps,

and finally, electrical systems for very fine grain soils

in which well points are combined

with the flow of electricity.

The first recorded lowering of the phreatic surface

was achieved in 1838 during the construction

of the Kilsby tunnel in the United Kingdom.

In this case, the phreatic surface was lowered

by pumping from large vertical shafts

adjacent to the tunnel.

One of the earliest examples of the use of deep wells

for pressure relief purposes dates back

to the construction of the Bremerhaven lock

or Nordshleuse in 1927.

The system consisted of 58 deep wells located outside

of the sheet piling

and resulted in a 50 foot decrease in artesian pressure.

Another early example of de-watering is the system used

for the construction of the Rutgers tunnels

in New York city.

The system consisted of five 100-foot deep wells

equipped with electrically powered deep-well turbine pumps.

The system allowed for the construction

of a large portion of the tunnels under free air

and significantly lowered the air pressure required

for tunneling purposes.

The remainder of this webinar will focus on answering

the “how” question.

And this will be done by modeling two de-watering systems.

The first is a pressure relief system

for a levee along a river bend,

and the second is a deep well system

for an excavation near a river.

Let’s start by modeling the pressure system

for the levee along a river bend.

As shown in the image on the left-hand side of the slide,

the shape of the river has forced

the construction of a levee with a 90 degree bend.

The levee lies on a 10-foot thick silty clay blanket

that overlays the 90-foot thick sandy aquifer.

As will later be shown,

the blanket has been eroded in parts of the river.

Their pressure relief system consists of

65 partially penetrating wells

over a one and a half mile long span of levee.

This problem is undisputably three-dimensional

so let’s go ahead and open SEEP3D,

review the current analysis

and the add remediation wells if needed.

As shown in the Project Explorer

on the left-hand side of the screen,

a three dimensional geometry named “Levee Along River Bend”

has been used to set up the model domain.

We can review the steps that were used

to create the geometry

by toggling through the design history.

The first step in the process

was to sketch the limits of the domain

onto a translated X-Z plane.

The sketch was then extruded downwards

to form the blanket and aquifer.

The center line of the levee on the west side of the bend

was then sketched onto the upper surface of the blanket.

A sketch plane was subsequently created

on the side of the domain

and the section of the levee with 1’3″ slopes

and 10-foot wide crown was sketched onto the plane.

The section was then swept along its center line.

The process was repeated for the levee

on the east side of the bend,

and the protruding portions of the levee

were removed or deleted.

The different parts of the levee

were then merged into a single solid.

The edge of the blanket

was sketched onto the upper surface of the blanket,

extruded through the blanket and aquifer,

and used to cut the blanket.

The levee was imprinted onto the ground surface

to allow for boundary condition assignments.

The elevation of the reservoir

was also sketched onto a plane on the side of the domain,

extruded and imprinted onto the surfaces of the levee.

And finally, the ditch on the land side of the levee

was created by sketching it center line

on the upper surface of the blanket,

sketching it section on a plane on the side of the domain,

sweeping the sketch along the center line,

and cutting out the ditch from the blanket material.

Once these steps completed and the geometry created,

materials were assigned to the different solids.

We can review the material assignment

by moving over to the analysis,

removing the boundary painting,

and selecting specific solids in the Geometry Explorer.

As we can see, the blanket material

was assigned to solid 89550,

which is the portion of blanket which was non-eroded,

a sandy material, similar to the aquifer,

was assumed to have replaced the eroded blanket,

and the aquifer material was applied

to both the eroded region of the blanket and the aquifer.

And the levee material was ascribed to the levee.

If we now retoggle along the boundary painting,

we see that a potential seepage

face review boundary condition in bluish gray

was applied to all surfaces on the land side of the levee,

whereas a constant head boundary condition in blue

was applied to all of the relevant surfaces

on the Riverside of the levee.

Now that the model is complete,

we can focus our attention on the de-watering system

and more specifically the pressure relief wells.

To do so, let’s open the three-dimensional editor.

Now that we’re in the editor,

we can remove the boundary condition painting,

render the aquifer invisible,

rotate the domain,

and add a schematic representation of a pressure relief well

onto the side of the domain.

Although this may look like a trash can,

I assure you it’s a pressure relief well.

The metal grill simply prevents vandalism and damage

to the top of the well.

As most of you know,

the well consists of a check valve

that prevents back flooding

and entrance of foreign material.

The casing that lies in the blanket layer is impervious,

whereas the screen in the aquifer allows water

into the well.

This system is relatively simple if not simplistic

and allows water to flow into the screen, up the well,

and out of the top of the well.

The head loss in the well

is generally considered negligible.

And the total head of the screen portion of the well

that lies in the aquifer

is set equal to the elevation at the top of the well.

A look at one of the wells reveals

that the boundary condition has been set equal

to a specified value of head

which corresponds to the elevation

at the top of that specific well.

It should be noted that we have increased

the mesh density along each well

and thus increase the mesh density

in the vicinity of the well

where hydraulic gradients are likely to be quite high.

If we now rotate the geometry,

we can see all the wells,

and more specifically, the screen portions of the wells

that extend into the aquifer.

So we’ve finished creating the geometry,

assigning materials,

and ascribing boundary conditions,

and we need to generate the mesh.

This is specific to our three-dimensional products

in which meshing is done on demand and not automatically.

So let’s switch over to the mesh view

and review the mesh that has already been created.

In this particular case,

the edge length was set equal to 30 feet,

which resulted in a mesh of approximately 150,000 nodes

and 900,000 elements.

If we relaunch the mesh process

by clicking on the OK button,

we see that the mesh is created in less than 20 seconds.

A review of the mesh quality reveals

that the elements are very good in the aquifer,

the region of interest,

and fair in the remainder of the domain.

So we’re now ready to close the editor

and solve the analysis.

Although the analysis solves in a manner of minutes,

the process has been edited out of the presentation

due to time considerations.

Now that the analysis has been solved,

we can have an in-depth look at the results.

But before we do so,

let’s close their grid and remove the edge painting.

The software provides the ability

to generate contour surfaces of a number of variables,

including but not limited to, total head,

pressure head, hydraulic radiant, flux,

and hydraulic conductivity.

Let’s start by having a look

at the total head contour surfaces.

A very nice and useful feature is the ability

to change the opacity of the contours in specific solids.

In so doing, we quickly realize

that large values of total head appear

in neuralgic portions of the aquifer

and more specifically, on the land side of the levee

near the river bend

and near the Eastern segment of the levee.

If we toggle over to the vertical gradient contour surfaces,

we immediately note that the higher values of total head

have led to large gradients below the ditch

and at the toe of the levee near the 90 degree bend.

These gradients definitely warrant further investigation.

The vertical gradients through the blanket

at the position of the ditch

can be determined by extracting the total head

at points on the bottom surface of the blanket.

To do so, we must first open the three-dimensional editor,

move over to the mesh view,

and define a location that contains the points.

This location can then be used

in the draw graph functionality.

The data provided in the graph can then be extracted

and pasted into a spreadsheet software.

The vertical gradient computed as the head loss

across the blanket

is found to exceed the critical gradient,

which is approximately equal to one over a large portion

of the ditch.

This is quite problematic and can be addressed

by adding remediation wells.

But before doing so,

let’s extract the total head data at the top of the aquifer.

This data will serve as a reference

when we compute the drawdown associated

with the remediation wells.

This is quite easy

and can be done by selecting the total head contours,

setting the opacity of all of the solids

except for that of the aquifer equal to zero,

selecting the top surface of the aquifer,

and clicking on the expert data by Nodes button.

The data can be exported in CSV or TXT formats.

It is important to note that the functionality

is only accessible when a point, line, surface

or solid is selected.

Now that we’ve exported the data,

we can add a number of remediation wells.

To do so, we need to open the three-dimensional editor,

select the X-Z plane,

sketch the points on the plane,

and translate the plane upwards

to the elevation of the bottom of the blanket.

Once this is completed,

we can extrude the wells into the aquifer.

We can then select each edge,

set the element edge,

and assign the boundary conditions.

We are then ready to mesh and solve.

This process has been edited out of the presentation

for the sake of expediency.

So what is the effect of the remediation wells?

Well, firstly, the total head contour surfaces

no longer extend into the aquifer near the river bend.

The vertical gradient below the ditch

is also much smaller in the vicinity of the bend,

but remains unmanaged near the eastern segment of the levee.

Moving over to the Draw Graph functionality,

and extracting, pasting the data into the spreadsheet,

shows that the vertical gradient no longer exceeds

the critical gradient near the river bend.

A clear picture of the effect of the additional wells

can be assessed by extracting the total head

at the top of the aquifer and using it

to draw a drawdown map.

As before, the data is extracted

by setting the opacity of all of the solids

except for the aquifer equal to zero,

selecting the top surfaces of the aquifer,

and clicking on the export data by Nodes button.

A number of different tools can be used

to manipulate the data and create the draw down map.

The tool that I really like is Python.

And why is that?

Well, first of all, it’s free and open source,

which is great,

and it’s also easy to read, learn, and write,

and offers a vast number of libraries.

The objective here is not to show you

how to install or program in Python,

but to provide the rudiments needed

to read-in results and create a drawdown map

in a manner of minutes.

Although Python can be downloaded

from the website,

I strongly recommend installing a Python distribution,

such as Anaconda,

which will greatly simplify package management.

I also recommend installing

an integrated development environment,

such as PyCharm, Spyder, PyDev, IDLE or Wing,

which simplifies editing, running, and debugging.

So let’s open the script file

that has been saved in the folder containing the data

and review the eight steps needed

to generate the drawdown map.

The first step in the process

is to import the modules needed to accomplish

the desired tasks.

In this case, we need numpy to create tick mark

and level arrays,

CSV to read in the data,

and matplotlib to create the drawn on map.

We can then define a series of lists

that will contain the coordinates, model results

and computed values of drawdown.

Once this is completed,

we can open this CSV file of the original analysis

without remediation wells and read-in each line of the file

except for the header.

We can then repeat the process for the second file

without reading the coordinates.

We can then pair each item in the lists

that contain the values of total head using the zip function

and compute the draw down for each pair.

We are now ready to create the figure

that will contain the contour plot.

To do so, we can define the limits of the X and Y axes,

define an aspect ratio, generate the figure,

and add a title using matplotlib.

We can then set the limits of the X and Y axes,

add evenly spaced tick marks and labels

using the arange function provided in numpy,

and add titles to the axes.

We can then add field contours

on an unstructured triangular grid

using the tricontourf function

and add a color bar with ticks and title.

The final step in the process consists

in adding contour lines using the tricontour function.

We can then run the script by clicking on the green arrow.

And voila!

We have a drawdown map with contours, contour lines,

color bar, and title.

It should be noted that there are thousands of color maps

to choose from.

The drawdown map can be used to create

a smooth creative and insightful representation

of the effect of the remediation wells.

As shown in the slide,

the effect of the additional wells

is limited to the extent of the wells,

but extends well into the land side

and riverside portions of the aquifer.

Let’s now move on to the second example

that focuses on a deep well system for a 770 foot long,

370 foot wide and 40 foot deep excavation

in a 90 foot deep unconfined sandy aquifer.

The de-watering system consists of 12 pumping wells

located five feet from the crown of the excavation slopes.

The screen portions of the wells are 10 inches in diameter

and 34 feet long.

Each centrifugal pump has a pumping rate

of 1,150 gallons per minute

or 2.56 cubic feet per second.

As with most of the watering problems,

this problem is undeniably three-dimensional.

So let’s go ahead and open Seep3D,

review the current analysis,

and create a drawdown map.

As shown in the Project Explorer

on the left-hand side of the screen,

a three-dimensional geometry named “excavation”

has been used to set up the model domain.

The geometry is a very simple,

but very large 20,000 foot long,

11,000 foot wide and 90 foot high cuboid

that allows us to capture the full effect of the river.

Toggling off the material shading reveals the wells

in the cylindrical bodies or liners

that were used to increase mesh density

in the vicinity of the wells.

Opening the boundary condition definitions reveals

that the pumping flow rate was converted

into a prescribed flux with associated perimeter.

If we open the three-dimensional editor

and select one of the wells,

we see that the edge length has been set equal to five feet.

A look at the mesh shows an increase in mesh density

in the vicinity of the wells

which is mostly due to the presence

of the meshed cylindrical bodies.

We are now ready to close the editor and solve the analysis.

Although the analysis solves in a manner of minutes,

the process has been edited out of the presentation

due to time considerations.

Now that the analysis has been solved,

we can have an in-depth look at the results.

But before we do so,

let’s remove the grid and the edge painting.

Toggling on the total head contour surfaces

and changing the opacity of the solids

reveals a significant drop in total head

in the vicinity of the pumping wells.

The effect of the pumping wells on their phreatic surface

can also be assessed by defining an ISO surface.

The shape of their phreatic surface or ISO surface

is more clearly defined when adding contour lines

and labels.

A very nice and useful feature

is the ability to add a clipping plane

at the center line of the excavation.

A clear picture of the effect of the wells can be assessed

by extracting the elevation of the phreatic surface

and using it to draw a drawdown map.

To do so,

the clipping plane is suppressed.

The grid cell size is set equal to 10 feet.

And the data is extracted using the export results

by Grid Sampling button.

The drawdown map is created with a slightly modified script

in which drawdown is computed

as the difference between the initial elevation

of the phreatic surface

and the final elevation of the phreatic surface.

For the sake of clarity,

we’ve also added a contour line at 35 feet

which corresponds to the bottom of the excavation.

Running the script reveals that the drawdown

is sufficient for excavation to proceed.

This was a desired outcome

and was very easily identified using the thicker line

that appears in the drawdown map.

In this webinar,

a short review of the watering systems was presented,

two three-dimensional groundwater de-watering problems

were presented, reviewed, and interpreted,

and a simple and efficient mean

of creating insightful drawdown maps was presented.

We have now reached the end of this webinar.

A recording of the webinar will be available to view online.

Please take the time to complete the short survey

that appears on your screen

so we know what types of webinars

you’re interested in attending in the future.

Thank you very much for joining us

and have a great rest of your day.