113

T
u
n
n
e
l
l
i
n
g
Assessment of ground
movement impacts on
existing tunnels
Abstract
As an increasing number of tunnelling schemes are being designed and
constructed in congested cities around the world, the complications
involved as a consequence of tunnelling on existing sub-surface
structures are becoming ever more challenging for engineers. Most
would agree that the impact of tunnelling-induced ground movement
on existing tunnels is an important issue, and the need for improved
technical knowledge for solving this complex soil-structure interaction
problem is essential. This paper initially explores the paradox of
commonly used semi-empirical methods for predicting the 3D
tunnelling-induced ground movements. The potential effects of these
ground movements on the longitudinal and transverse response on the
structural behaviour of the existing tunnels are then investigated. The
shortcomings of commonly adopted assessment methods which may
result in non-conservatism and mask critical issues of existing tunnels
at tunnel crossings are highlighted throughout the discussion. This
paper also demonstrates the use of modern computing to do powerful
complex analysis to solve the complexities of soil-structure interaction
and manage risks associated with tunnel crossings.
As mass transit and other tunnelling
schemes are constructed in increasingly
developed urban areas, the impacts
of tunnelling on existing tunnels are
becoming ever more important. The
problem of soil-structure interaction
is relatively complex, and the need for
improved knowledge of the impacts
of new tunnelling works on existing
tunnels is essential to obtain a thorough
understanding of such issues.
This paper initially examines historic
case studies that form a useful basis for
predictions of likely induced movements
of existing tunnels caused by tunnel
construction, which often provide a
database against which modelling
assumptions can be verified. The use
and limitations of simple semi-empirical
methods for predicting the three
dimensional ground displacements
caused by tunnel construction are
subsequently investigated. The discussion
then leads on to the assessment of
longitudinal and transverse (cross-
sectional) behaviour of the existing
tunnels. The complexity of ground
movement predictions and the structural
performance of the existing tunnels are
examined and various methods of analysis
are assessed. Throughout the discussion,
issues with non-conservative assumptions
and methods of analysis are highlighted,
which could potentially mask the most
significant issues of existing tunnels at
tunnel crossings.
Case histories
One of the challenges in understanding
the effect of tunnel induced settlements
on existing tunnels is the limited number
of published case histories. The problems
of predicting ground movements and
structural performance of existing tunnels
are complex and to fully understand
the behaviour at these tunnel crossings,
not only does the performance of an
existing tunnel have to be extensively
documented, details of the construction
of the new tunnel are equally important.
The lack of quality information adds to
the dilemma as, quite often, ground
movements are so small that interpreting
the exact behaviour of the tunnels at the
crossing point, given the accuracy of the
monitoring, is not reasonably possible.
Whilst such performance is without
doubt desirable from the point of view
of the tunnelling contractor and the
Introduction
Ian Turner
Senior Engineer
Water & Environment
Atkins
114
Michael Yap
Tunnel Engineer
Water & Environment
Atkins
114
asset owner, this poses a challenge for
assessors in predicting tunnel behaviour
at crossings.
Probably the most relevant example
recorded in recent history is the
construction of the Heathrow Express
tunnels as they were driven under
the Piccadilly Line at Heathrow
3
. This
example is well documented and the
movements are sufficiently large relative
to the accuracy of the monitoring that
reasonable conclusions can be drawn
from the results. In addition, the relatively
flexible behaviour of the Piccadilly Line
tunnel meant that a good understanding
of the ground movements could be
obtained. Particular modes of response
of the existing tunnels to imposed
movements that were identified include
the rotation of the tunnel as the face
of the new tunnel passes underneath
(Figure 1) and the large distortion that
occurred on the diagonals of the tunnel.
It is observed that movements of the
existing tunnel could be reasonably
assessed using a calibrated prediction of
the tunnel-induced ground movement.
Since then, a number of other tunnel
crossings have been documented. On
the Channel Tunnel Rail Link (CTRL) the
crossing under the Central Line was
notable because of the close separation
between the tunnels
12
. Back analysis of
the data obtained on the CTRL crossings
both at the Central Line and at Highbury
and Islington Stations show a wide
variation in the settlement volume losses
and trough width parameters (Table 1).
Whilst it is not clear that this is due to
soil-structure interaction resulting in a
flatter trough. It is probable that higher
K values were noted where the ground
was stiffer due to the presence of the
existing tunnel.
Similar results have been observed in
other tunnel crossings in recent years
such as the Thameslink 2000 tunnels
under the existing Thameslink tunnel
10

and where a new tunnel is driven parallel
to existing tunnels
6
. Well documented
examples of the behaviour of tunnels in
response to ground movements have also
been published
15, 3
.
Whilst considering the movements
created by the construction of a proposed
tunnel is important, an understanding
of the existing tunnels response to
the movements is equally important.
These case histories demonstrate that
the movements of existing tunnels in
response to the construction of new
tunnels is reasonably consistent with the
movements that are predicted by using
conventional technologies for predicting
ground movements.
Ground movement
predictions
Key to understanding the ground
movement impacts on existing tunnels
is being able to predict the greenfield
ground movements. Historically semi-
empirical methods have been regularly
used in the UK to predict tunnel-induced
ground movements.
The most commonly used approach is
the assumption of a point source leading
to a settlement trough approximating
to a Gaussian distribution
14
. Whilst
Figure 1. Rotation of Piccadilly Line tunnel as new tunnel passes beneath Heathrow
3
CTRL UP CTRL DOWN
K Vs K Vs
GN & C
southbound
0.71 0.59% 0.61 0.38%
Victoria
southbound
0.55 0.44% 0.38 0.55%
GN & C
northbound
0.70 0.51% 0.67 0.52%
Victoria
northbound
0.45 0.44% 0.49 0.45%
Table 1. Results of back analysis of ground movement at the Highbury and Islington crossings
12
115
T
u
n
n
e
l
l
i
n
g
this approach has proved to be reliable
for many different scenarios, there are
situations where the approach has proved
to have limitations.
One particular situation where this is the
case is in assessing ground movements
close to the source of movement
generation where the assumptions
appear to break down and modifications
to the method are required, such as
changing the trough width parameter
9
.
The main reason for this is that when
the existing tunnel is in close proximity
to the new tunnel, the assumption of a
point source is no longer valid and the
shape and size of the tunnel will have an
effect on the result. This often poses a
challenge when considering the effects
of one tunnel on another because the
crossings are often close, with less than
two diameters between the tunnels.
At these distances, the geometry of
the tunnel causing the movement
becomes important to the shape of the
settlement trough and therefore needs to
be considered.
Solutions of predicting sub-surface
ground movements close to the source
of movement exist in a number of
different forms. A common approach
is to adopt a closed form solution for
a more sophisticated source for the
ground movement such as a ribbon sink
method
13
. This approach assumes that
the source of the ground movement is
evenly distributed over the width of the
new tunnel at invert level, rather than
axis level, and has shown to be of better
agreement to the field data obtained
from the Heathrow Express trial tunnel.
This results in a settlement trough very
similar to the Gaussian trough when the
depth of cover between tunnels is large,
but as the cover distance reduces, the
predicted ribbon sink trough becomes
flatter than the Gaussian trough, giving a
similar response to adjusting the trough
width parameter for the Gaussian trough.
An alternative approach available with
modern computing power is simply
to discretise the face into a number
of smaller point sources (Figure 2),
an approach that has been adopted
for modelling settlement troughs due
to jacked boxes. By breaking a tunnel
into a number of smaller sources,
the basic equations of the Gaussian
trough can be used to predict the
cumulative settlement due to each of
the small regions. This approach has
the advantage of taking into account
both tunnel size and geometry, allowing
unusual geometries and sequences to
be considered with ease. This approach
results in very similar results to the Ribbon
Sink method unless very close proximities
are being considered or the tunnel has a
significantly non-circular geometry.
Figure 3 shows the ground movement
predictions using the Gaussian, ribbon
sink and discretised methods. The two
ground movement graphs are shown
with depth of source taken to the axis
level and to the invert level respectively.
The discretised profiles are identical as
they have been calculated assuming
source at the centre of the each element.
It is observed that the main distinction is
the reduction in maximum settlement,
and the flatter and wider trough when
the source of ground movement is
assumed to be at the invert. In hindsight,
the location of the source of ground
movement could perhaps be subjected
to further investigation, and could
potentially be influenced by various
factors including the construction method
of the new tunnel and ground conditions.
The current challenge with all of
these approaches is ensuring that the
parameters used are consistent with
the technology being adopted for the
ground movement prediction. Modern
ground movement prediction is still
fundamentally a statistical assessment
method that is a subject of various
influencing factors, and as such relies
on databases from back analyses to give
assessors reasonable input parameters for
their assessment. Although all methods
use the same parameters, there is no
reason why for a given depth and tunnel
geometry the parameters used in the
assessment, the trough width parameter,
K, and even the volume loss, V, should be
the same.
Figure 2. One method of discretising into
a number of smaller point sources - using
Delaunay triangulation (From Python code)
Figure 3. A comparison of ground movement predictions with source taken to axis level (left) and to invert level (right) on the same vertical scale
(From Python code)
114

Assessment of ground movement impacts on
existing tunnels
116
The longitudinal effect:
effects of ground
movements along the axis of
the existing tunnel
Having predicted the greenfield
ground movements around the existing
tunnel, an assessment of the impact
of the ground movement on these
tunnels is normally undertaken. One
consequence of these ground movements
is the structural response along the
length of the existing tunnel. This
section sets out to discuss the potential
longitudinal structural behaviour of the
existing tunnel.
As noted in historic case studies, there
has been evidence that the stiffness of
the existing tunnel can affect the shape
of the predicted greenfield settlement
trough, with trough being wider and
shallower as a result of the stiffness of
the existing tunnel. Whilst it is typically
conservative not to consider the soil-
structure interaction, it is possible that not
considering the interaction could result in
the need to construct extensive mitigation
measures which in themselves hold risks
of unplanned outcomes. Not considering
soil-structure interaction is also likely to
result in an underestimate of the width of
the settlement trough. For this reason, it
is important for the assessment engineer
to consider, in minimal, the implications
of soil-structure interaction, even if
the final assessment is based on the
greenfield ground movements.
The simplest method to take into account
soil-structure interaction is to adopt a
beam-spring model. This methodology has
been used for many years for assessing the
settlement impacts on existing pipelines
and other small diameter tunnels
2,7
as well
as larger diameter tunnels
12
. A beam-
spring model can be used to investigate
this soil-structure interaction, where the
existing tunnel can be modelled as a series
of beams with the appropriate sectional
properties, attached to vertical and axial
Winkler springs at the nodal ends. Springs
are usually linear elastic, but non-linear
elasto-plastic no-tension springs can be
adopted for a more sophisticated model.
Greenfield ground movements are
subsequently applied to these springs,
either as forced displacements or by
using a spring softening approach where
the spring stiffness is progressively
decreased by an amount proportional to
the Greenfield ground movement at that
Figure 4. Comparison between Greenfield and beam-spring ground movement profiles (from MS
Excel)
Figure 5. Results from beam-spring model which allows brittle cracking of the lining and a reduction in sectional capacity (from Python code)
117
T
u
n
n
e
l
l
i
n
g
location. The influence of the stiffness
of an existing structure will reduce the
deformation and flatten the resultant
curvature in comparison to Greenfield
ground movements (Figure 4).
This approach is based on assumptions
of a monolithic beam distributing loads
along the length of the tunnel. One of the
challenges the authors have identified in
using this approach has been modelling
brittle tunnels where tension cracking will
reduce the ability of the existing tunnel
to span longitudinally over the settlement
trough. To address this particular problem,
a beam spring model which allows brittle
cracking of the existing lining has been
developed. This approach allows the
tensile stress in the existing lining to reach
a limiting value at which point the section
gradually cracks, reducing the stiffness at
this location (Figure 5).
Whilst the conventional approach to
modelling materials such as masonry
and unreinforced concrete is to allow
no tension behaviour, this approach has
some critical issues when being used in
the assessment of existing tunnels. In
particular, this implies that the tunnel has
no longitudinal axial (tensile) force and
consequently, no longitudinal bending
strength at these locations. The lack
of bending strength results in a more
flexible tunnel which closely follows
the greenfield settlement trough, with
no enhanced longitudinal spanning
capacity as well as having a large number
of small width cracks. In some cases
neither of these consequences follows
observed behaviour, and may, therefore,
not be conservative for the assessment,
so the conventional assumption of no
tension may not be appropriate for
some assessments.
One of the consequences of the
beam-spring model which accounts
for cracking of the material is that as
the cracks develop they be become
strain concentrators, as seen in Figure
5. As sections of the tunnel become
more flexible where cracks occur, an
increased settlement induced movement
is concentrated at these locations. The
implications of this can be significant to
the assessment, such as the development
of heavily damaged sections adjacent to
relatively unaffected regions, resulting
in larger crack widths than might
normally have been predicted. In some
circumstances, the model predicts that
rapid destruction of the structure through
cracking can occur for relatively small
changes in ground movement.
One approach to mitigate the build up
of longitudinal strains along an existing
tunnel is to create circumferential
movement joints prior to commencement
of tunnelling works. This approach
articulates the existing tunnel into short
segments, and localises ground movement
and rotation at engineered locations along
the length of the existing tunnel. Strains
are therefore more evenly distributed than
if uncontrolled cracking behaviour was
allowed to occur (Figure 6).
The creation of engineered movement
joints will involve cutting or the removal
of longitudinal bolts in segmental linings,
or saw cutting through brickwork in
masonry tunnels. A flexible liner and / or
elastic joint filler can be used to maintain
the water tightness of the structure if
required. This approach does, however,
have to be treated with caution, because
the introduction of movement joints
increases the flexibility of the tunnel. In
some circumstances the original tunnel
may have sufficient strength to avoid
localised failure of the lining under
longitudinal bending and so would have
suffered little, if any, damage prior to
the introduction of flexible elements.
The introduction of the movement
joints would significantly increase the
flexibility of the tunnel resulting in more
movements, with the movement being
localised at the movement joints. A
careful risk based assessment is therefore
needed to consider the best approach for
protecting the existing tunnel.
The transverse effect: effects
of ground movements on
the cross section of the
existing tunnel
Having considered the longitudinal
effects in the previous section, the other
main consequence of tunnelling-induced
ground movements on the structural
response of the tunnel is the transverse
structural behaviour of the existing
tunnel. This section sets out to discuss
the potential effects of the cross sectional
behaviour of the existing tunnel.
Case history information on the
movements of the existing tunnel in the
cross-sectional direction is less common
than for the longitudinal movement.
Figure 6. Comparison of results tensile strains with and without engineered movement joints
(from MS Excel)
114

Assessment of ground movement impacts on
existing tunnels
118
The case histories do, however, give
reasonably consistent indications that
provided the tunnel is sufficiently
flexible, the tunnel movements can
be predicted based on the greenfield
movement assumptions.
The key issue to take into account with
the transverse effect is that a large part
of the effect occurs due to the movement
of the ground around the tunnel face.
Analytically, ground movements around
the tunnel face are typically assumed
to take the form of an S-curve defined
by the cumulative normal distribution
1
.
Broadly speaking, based on this
assumption, points in the ground tend to
move towards the source of the ground
movements and points closer to the
source of the movement tend to move
more than points further away from the
source of the ground movement.
To give a complete assessment of these
effects, a number of different positions
for the source of the settlement need to
be considered in order to account for the
transitory effect as ground movements
develop. Similarly, the maximum
distortion of the existing tunnel often
does not occur on one of the principal
axes, vertical and horizontal, but on
an angle as the TBM face passes. A
complete assessment therefore requires
the distortion to be assessed in a more
comprehensive, staged manner.
A simple approach to this problem is
to undertake the ground movement
prediction at a number of evenly
spaced points on the tunnel perimeter
at opposite ends of a tunnel diameter
(Figure 7), to give a better understanding
of the distortion in shape of the tunnel.
This assessment also considers a number
of different face locations for the
new tunnel, giving a comprehensive
assessment for the temporal changes to
ground movements and consequently
the shape of the existing tunnel as the
new tunnel is driven beneath. Effects like
skew and dip in alignment can also be
considered in such an analysis.
When assessments are undertaken in this
manner, some of the key effects that are
observed include:
The maximum distortion is not on
the principal axes but is usually on a
diagonal axis;
The maximum distortion is temporal,
with the maximum final distortion less
than maximum measured distortion;
The final distorted shape of the
existing tunnel indicates that
ovalisation rather than squat of
the tunnel occurs, and the vertical
diameter increases whilst the
horizontal diameter decreases. Since
most tunnels typically are more likely
to be built with a vertical squat, the
transverse settlement effects can be
beneficial in the final condition.
Although these effects are consistent with
observed behaviour such as the Piccadilly
Line at Heathrow
4
, the assumption of a
simple S-curve longitudinal settlement
trough is, in many circumstances, a
significant simplification. There are many
documented cases where the longitudinal
settlement trough does not take the
form of a smooth S-curve but a complex
curve representing the different sources
of ground movement around the face of
a tunnel being constructed
16,5,8
. So how
does this non-linear behaviour affect the
tunnel distortion predictions?
To assess this, a simple model has been
developed to undertake assessments
consistent with the approach proposed
above, but where the tunnel settlement
is modelled by multiple faces positioned
one behind another with slightly different
trough width parameters and volumes for
each trough. This approach generates a
complex trough in a way that makes the
assessments consistent with those used
for a simple S-curve trough.
As with the longitudinal effect, one
issue that has to be addressed for the
transverse effect is the tunnel stiffness.
Whilst the approaches discussed
above are appropriate for a flexible
lining, if a stiff monolithic lining is
considered, significant variance from
the prediction could be encountered
during construction. The same issues
that apply to the longitudinal effect also
apply to the transverse effect with respect
to the stiffness of the existing tunnel.
In particular:
Stiff tunnels will provide more
resistance to movement and so the
simple application of the greenfield
ground movement predictions to a
stiff tunnel will over predict the loads
on the tunnel;
Figure 7. Simple geometric approach to cross-sectional analysis caused by tunnelling
119
T
u
n
n
e
l
l
i
n
g
Were failure to occur in a stiff, brittle
tunnel, joints in the tunnel would
act like strain concentrators and so
additional movements, and potentially
damage, would occur at these
failure locations.
When the ground movement effect is
being considered on a relatively stiff
tunnel the greenfield movements must
therefore be converted to a quantitative
value that will give a more credible
prediction of the impact on the existing
tunnel. Often the easiest way to do this
is to modify the earth pressures acting
on the tunnel. There are a number of
ways to achieve this, for example a simple
closed form solution of the stress changes
in a circular tunnel in elastic ground
11
.
Having obtained the change in the loads,
an assessment of the capacity of the
existing tunnel to resist the additional
loads due to tunnel movement can then
be undertaken.
Conclusion
The ground movement effects on existing
tunnels are clearly an important issue
for future tunnelling, as an increasing
number of urban tunnels are constructed
with ever more tunnel crossings. A good
understanding of the behaviour of these
crossings is therefore essential to manage
the risk to future construction.
The use of simple semi-empirical method
for predicting the three dimensional
ground displacements caused by tunnel
construction is an essential tool for tunnel
assessments. However, the limitations of
these predictions should be recognised,
whilst noting that they are fundamentally
a subject of various influencing factors,
and as such rely on databases from back
analyses to give assessors reasonable
input parameters for their assessment.
Ground movement prediction is closely
related to structural performance and
the interaction between the ground and
the tunnel. This complexity means that
normally conservative assumptions may
not be appropriate and in some instances
could even mask the most significant
issues with particular tunnel crossings.
In the assessment of the longitudinal
structural behaviour of an existing tunnel,
the principles of strain concentrators,
formed naturally at points of weakness or
through engineered joints are important
in the understanding of the behaviour
of an existing tunnel subjected to
tunnel-induced ground movements. The
transverse structural behaviour of an
existing tunnel should be comprehensive
to account for the most damaging effects
which are often temporal and occur
60 40 20 0 20 40 60
10
0
10
20

max
mm

min
mm
X
Figure 8. Distorted shape of existing tunnel as new tunnel progresses
beneath. Maximum distortion is temporal and on a diagonal axis. (from
MathCAD)
114

Assessment of ground movement impacts on
existing tunnels
120
on a diagonal axis as the new tunnel
is constructed.
The capacity of modern computing to do
complex calculations quickly and easily
can provide the tools to look into these
complex interactions. This gives us the
capability to undertake more reliable
predictions of the response of tunnels to
movements as well as the ability to better
manage the risk associated with these
types of tunnel crossings.
References
1. Attewell, P. & Woodman, J.P., 1982. Predicting the dynamics of ground settlement
and its derivatives caused by tunnelling in soil. Ground engineering, 15(8), pp.13 -
22.
2. Attewell, P., Yeates, J. & Selby, A.R., 1986. Soil Movements Induced by Tunnelling
and Their Effects on Pipelines and Structures, London: Blackie and Son Ltd.
3. Cheung, L.L.K. et al., 2010. Optical fibre strain measurement for tunnel lining
monitoring. Proceedings of the ICE - Geotechnical Engineering, 163(1), pp.1-
1. Available at: http://www.icevirtuallibrary.com/content/article/10.1680/
geng.2010.163.1.1.
4. Cooper, M.L. et al., 2002. Movements in the Piccadilly Line tunnels due to the
Heathrow Express construction. Gotechnique, 52(4), pp.243-257. Available at:
http://www.atypon-link.com/TELF/doi/abs/10.1680/geot.52.4.243.41019.
5. Dimmock, P. & Mair, R.J., 2008. Effect of building stiffness on tunnelling-induced
ground movement. Tunnelling and Underground Space Technology, 23(4), pp.438-
450. Available at: http://linkinghub.elsevier.com/retrieve/pii/S0886779807000922
[Accessed March 1, 2011].
6. Kim, S.-H., 1996. Model testing and analysis of interactions between tunnels in
clay. University of Oxford.
7. Klar, A. et al., 2008. Tunnelling effects on jointed pipelines. Canadian Geotechnical
Journal, 45(1), pp.131-139. Available at: http://article.pubs.nrc-cnrc.gc.ca/
ppv/RPViewDoc?issn=1208-6010&volume=45&issue=1&startPage=131&ab=y
[Accessed May 8, 2011].
8. Leca, E. & New, B.M., 2007. Settlements induced by tunnelling in Soft Ground.
Tunnelling and Underground Space Technology, 22(2), pp.119-149. Available at:
http://linkinghub.elsevier.com/retrieve/pii/S0886779806001180 [Accessed February
26, 2011].
9. Mair, R.J., Taylor, R.N. & Bracegirdle, A., 1993. Subsurface settlement profiles above
tunnels in clays. Geotechnique, 43(2), pp.315 - 320.
10. Mohamad, H. et al., 2010. Behaviour of an old masonry tunnel due to tunnelling-
induced ground settlement. Geotechnique, 60(12), pp.927 - 938.
11. Morgan, D.H., 1971. A contribution to the analysis of stress in a circular tunnel.
Geotechnique, 11(3), pp.37 - 46.
12. Moss, N.A. & Bowers, K.H., 2005. The effect of new tunnel construction under
existing metro tunnels. In E. A. Kwast et al., eds. Geotechnical Aspects of
Underground Construction in Soft Ground. Amsterdam: Taylor and Francis, pp.
151 - 157.
13. New, B.M. & Bowers, K., 1994. Ground movement model validation at the
Heathrow Express trial tunnel. In Tunnelling 94. London.
14. O'Reilly, M.P. & New, B.M., 1982. Settlements above tunnel in the United Kingdom
- their magnitude and prediction. In Tunnelling 82. pp. 173 - 181.
15. Rulu, W., Jiaping, L. & Jingya, Y., 2008. Analyses on the influence factors of metro
tunnel deformation and its characteristic. In ITA 2008. pp. 1-9.
16. Sugiyama, T. et al., 1999. Observations of ground movements during tunnel
construction by slurry shield method at the Docklands Light Railway Lewisham
extension - East London. Soils and Foundations, 39(3), pp.99 - 112.