Proceedings of the World Tunnel Congress 2014 – Tunnels for a better Life. Foz do Iguaçu, Brazil.

How to turn geological uncertainty into manageable risk?

T. Dickmann and D. Krueger
Amberg Technologies AG, Regensdorf, Switzerland.

ABSTRACT: Risk management became an integral part of most underground construction projects
during the last decade. Still, there are parties, who may not plan for insubstantial unknown
uncertainties of ground conditions because they are so unpredictable and out of scope of normal
planning as they think. Finally, many of them end up as the most catastrophic events. Even known
geological uncertainty is a risk that certainly exists without knowing how it will affect the work.
Further, human biases form part of the way prior knowledge is being used to interpret data in a way
it’s anchored in one’s mind or in a way that is just available. If no naive assessment of the uncertain
situation is carelessly considered, risk becomes manageable if one know, detect and quantify the risk.
A reasonable and cost-effective way to tread is the application of 3D-Tunnel Seismic Prediction on a
regular base. Knowing what’s ahead (in a case study) results in manageable risk.

1 INTRODUCTION et al., 2004). Further, the insurance industry

issued the joint code of practice for risk
The very nature of tunnel projects implies that management of tunnel works in 2006 that is
any potential tunnel owner will be facing now being used worldwide and effective in risk
considerable risks when developing such a sharing and encouraging best practice of risk
project. Due to the inherent uncertainties, management procedure in tunneling (Adeyemo,
including ground and groundwater conditions, 2011). Eventually, this encouraging progress
there might be significant cost overrun and may lead to some tunnel builders’ opinion that
delay risks as well as environmental risks. Also, bearing the risk of loss and saving money by
as demonstrated by spectacular tunnel collapses being uninsured might be a charming option.
and other disasters in the recent past, there is a Very well then; the overall term “risk
potential for large scale accidents during management” is widespread and in everyone’s
tunneling work. Between 1994 and 2004, about lips in many fields of activity and expertise.
600 million US$ had been lost in 20 major Though, do we really sufficiently know about
projects where collapses had occurred. In 2006, manageable risk?
tunneling projects became uninsurable due to a In fact, manage risk stands for identify,
tremendous increase of loss ratio of 500%. It assess, analyze, eliminate, mitigate and control
could give the impression that insurance had risk, and it’s a fallacy to believe that risk
been the cheapest risk management tool. As a management could end before finishing the
result, risk management became an integral part tunnel. In particular, it’s the unknown
of most underground construction projects uncertainty that comes as a geological hazard
during the late 1990s. Since April 2003, zone. There are still owners or contractors, who
international guidelines on tunneling risk may not plan for these insubstantial events
management had been established showing how because they are so unpredictable and out of
risk management may be utilized throughout the scope of normal planning as they think. Finally,
project’s phases of design, tendering and many of them end up as the most catastrophic
contract negotiation and construction (Eskesen events.
1
 Proceedings of the World Tunnel Congress 2014 – Tunnels for a better Life. Foz do Iguaçu, Brazil.

2 RELEVANT UNCERTAINTIES processing procedures with their options of

parameters. An interpretation should therefore
An owner’s or contractor’s ability to identify be not merely a section of mapped seismic
risk is limited by the certainty or uncertainty of events, but also a description of the variability
risk. A simplified grouping of risk could be tolerated within the data. Quantifying the
made like this: known certainties, known ambiguity in the prediction is the concept of
uncertainties, and unknown uncertainties. measurement uncertainty (Leahy and Skorstad,
Actually, common understanding of risk is so 2013).
closely associated to uncertainty that almost This measurement uncertainty is clearly
nobody would consider known certainties to be different from the “conceptual uncertainty” in
risks at all. If we know something is coming, we geologic interpretation, which derives from a
think of it simply as a circumstance to be range of concepts that geoscientists could apply
addressed. Known uncertainties is a risk we to a single data set.
know exists, but we do not know how it will
affect us. The unknown uncertainties are
2.2 Constraints by conceptual uncertainty
unlikely to be addressed during project
planning. However, even in the project planning Interpretations of seismic images are used to
they need to be addressed at least in order to analyze subsurface geology and form the basis
budget for measures such as geophysical or for many exploration and extraction decisions,
more specifically seismic measurements but the uncertainty that arises from human bias
necessary to identify the unknown uncertainties in seismic data interpretation has not previously
during the prospective construction phase. At been quantified.
the end of the day, there is always the one It follows that geoscientists use their prior
question: has the geology in the tunnel area knowledge to apply or generate a new concept
been adequately explored and analyzed before to data in order to construct an interpretation
and during tunneling? and geological model. The initial geological
model might be a fundamental source of
uncertainty because it is dependent on the
2.1 Constraints by measurement uncertainty
tectonic paradigm or concept used in its
In subsurface investigation, geophysicists construction. Bond et al. (2007) argue that
typically measure, process and interpret the data conceptual uncertainty can be more important
and hand over the interpretation result to the than the uncertainty inherent in the positioning
geologist or geomodeller. She or he then of boundaries or fault planes in a geological
integrates this result with other geologic data model (Figure 1). Human biases form part of
and compiles a geologic model. This the way prior knowledge is being used to
conventional workflow fails to meet the interpret data. There are three relevant bias
challenges because it is not attuned to quantify types in the context of geo-data interpretation,
uncertainty that is associated with every piece of which are known from cognitive psychology.
geologic data due to resolution, sensitivity and The Availability bias occurs taking the model or
noise. interpretation that is most dominant in one’s
In consequence, the workflow need to be mind. Anchoring bias is the failure to adjust
changed where “measurement uncertainty” from experts’ beliefs, dominant approaches, or
associated with seismic and geologic initial ideas. Interpreters expect to see a
interpretation is quantified. With uncertainty particular type of structure in a given setting
collected at the early design and planning stages such as a geographical location. Confirmation
of an underground construction project, the bias involves actively seeking out opinions and
interpretation becomes not a single geologic facts that support one’s own beliefs or
model but an ensemble of models that can be hypotheses.
used to risk and improve decision making
(Leahy and Skorstad, 2013).
The seismic image is impacted by the survey
acquisition parameters of sources, receivers and
geometry, the material response such as Figure 1. The conventional subsurface analysis workflow
produces a single deterministic structural model of the
inelasticity, anisotropy and attenuation, ground.
environmental or electronic noise and signal
2
 Proceedings of the World Tunnel Congress 2014 – Tunnels for a better Life. Foz do Iguaçu, Brazil.

Conceptual uncertainty is likely to be a major whereas a management decision to limit the

risk factor for disciplines in which decision extent of a site investigation to save money will
making is based on prior knowledge and hence pose another type of hazard. Although
concepts of interpretation of data sets containing geological assessment constraints exist, hazards
limited information. induced by saving money don’t have to be taken
into account unless a naive assessment of the
situation is carelessly considered.
2.3 Implication to tunneling
Non-destructive geophysical site
In tunneling, uncertainty could lead to fatalities investigations while tunneling have developed
and doesn’t need to be accepted. Once a and improved significantly over recent times. In
preliminary model or hypothesis has been particular, when site investigations from the
generated, the real geological conditions in a surface are limited given the topography, tunnel
tunnel project could be generally verified by seismic imaging can detect lithological
collecting further data from pre-investigations heterogeneities within distances up to hundreds
from the surface or from geological mapping in of meters ahead of the face, many times more
the tunnel during the construction phase. Here, a that of probe drilling alone. It is the most
continuous reconciliation of forecasted and effective prediction method because of its large
observed data can be obtained resulting in a prediction range, high resolution and ease of
workflow where a geologically consistent application on a tunnel construction site
structural model is updated. This process (Dickmann and Krueger, 2013).
therefore becomes a constructive process rather The Tunnel Seismic Prediction (TSP) method
than a simple mapping process that Leahy and detects changes in rock mass such as irregular
Skorstad (2013) call a concept of model-driven bodies, discontinuities, fault and fracture zones
interpretation (Figure 2). ahead of the tunnel face (Dickmann and Sander,
1996). Employed as a predictive method during
excavation process for both drill & blast and
TBM headings, no access to the face is required
to perform measurements, which are taken in
tunneling production breaks of around 60
minutes. Acoustic signals are produced by a
series of 24 shots of usually 50 to 100 grams of
detonation cord aligned along one tunnel wall
side and having an additional shot line along the
opposite tunnel wall side in cases of more
complex geology. Four sensor probes,
consisting of highly sensitive tri-axial receivers,
are contained in protection tubes whose tips are
firmly cemented into boreholes of 45-50 mm in
both side-walls. The 3-component receivers
Figure 2. Model-driven interpretation workflow where pick up the seismic signals which were being
measurement uncertainty allows for the generation of a
geologically plausible model (after Leahy and Skorstad,
2013).

3 FROM GEOLOGICAL UNCERTAINTY

TO MANAGEABLE RISK

The geotechnical risks that can affect projects

result from a range of hazards associated with
geological conditions, but also from hazards
associated with the geo-engineering process. Figure 3.Measurement layout of the 3D Tunnel
Seismic Prediction method consisting of usually 4
For example, active faults identified during pre- receivers (RCV) and 24 shot points.
feasibility studies will pose one type of hazard,
3
 Proceedings of the World Tunnel Congress 2014 – Tunnels for a better Life. Foz do Iguaçu, Brazil.

reflected from any kind of discontinuity in the cause 3 days downtime at its best, not
rock mass ahead. A highly sophisticated infrequently one month. Figure 4 illustrates the
processing & evaluation software has been TSP expenses in time-related site costs. For
devised for ease of operation. The capability of TBM operation, a minimum heading length of 3
the system to record the full wave field of km is assumed. Here, a reduction of only three
compressional and shear waves in conjunction to five days of downtime pays off TSP
with the intelligent analysis software enables a operations on a regular base. In conclusion also
determination of rock mechanical properties from an economic point, does it really make
such as Poisson ratio and Young’s Modulus sense to even think about limit site
within the prediction area. The final 2D- and investigations to save money? The answer is
3D-summary results produced by the system clear with regard to ITA’s guidelines for
software present as well detected events and tunneling risk management: any implementation
boundary planes crossing the tunnel axis of measures to eliminate or mitigate risks where
coordinates ahead of the face. economically feasible or required according to
The owner and contractor can make know the specific risk objectives or health and safety
their risk, because they can detect and quantify legislation is to be ensured (Eskesen et al.,
the geological hazard. Unknown uncertainties 2004).
should belong to the past and they become By way of example, the next chapter
known uncertainties and in some cases even illustrates how these operations contribute to
certainties. Once a geological risk zone is measures economically mitigate risk.
identified, the contactor in agreement with the
Engineer is able to decide, what measures are to
be taken. 4 CASE STUDY
With a regular tunnel seismic operation, you
identify your geological risk detecting hazards The objective of this case study is a geological
and quantifying their impact to your tunneling prediction of minimum 100 m ahead of the
job. By this means, the tunnel builder can tunnel face and in addition, the verification of
understand the risk as chance or thread and even the results of an existing probe drill. Since
very economical. Depending on the heading geology is known from an extrapolation
length and type of the project, the investment in approach from a parallel tunnel, TSP was
knowing the risk by a regular TSP operation is requested verify the appearance and
just between 0.7 % and 1.8 % of the time- characteristics of fault and fracture zones and
related site costs such as labor costs, provision their crossing to the planned tunnel axis.
of installations and energy expenses. In other
words, all investments in this technology are
4.1 Site Location
already paid after saving 3 to 7 days of
downtime. How easily may happen one The geology encountered within the seismic
unforeseen incident of water ingress that would layout between receiver reference location and
face location at meter 57 is dominated by
weathered volcanic breccia. The geomechanical
classification is class III-IV according to RMR
rock mass rating. The rock behavior at the tun-
nel face is seen as instable. With on-going exca-
vation following rock behavior is assumed:
meter 60 to meter 75 - Fault zone or heavily
fractured, Class IV, instable face,
meter 75 to meter 161 - Volcanic breccia, Class
III, stable face,
Figure 4. TSP expenses when applied on a regular
meter 161 to meter 168 - Fracture zone, Class
base dependent on heading length and excavation method. IV, short-term stable face,
Expenses are based on time-related site costs. Numbers in from meter 168 onwards - Volcanic breccia,
bars give TSP expenses in percentage of total heading Class II-III, stable face.
days.

4
 Proceedings of the World Tunnel Congress 2014 – Tunnels for a better Life. Foz do Iguaçu, Brazil.

4.2 Probe drill result & geological forecast Figure 5 illustrates the SH-wave velocity
distribution in a cropped display of a computed
Figure 7a summarizes the geological forecast
data cuboid of 200 x 50 x 50 metres in
based on the 30 m probe drill result between
tunnelling direction and in each vertical and
tunnel face at meter 57 and end of drill at meter
horizontal direction, respectively. The tunnel
87. The on-going geological forecast represents
alignment is centred in the cuboid. The copping
the extrapolation of the encountered geology of
reduces the display to velocities smaller than
the parallel tunnel until meter 182. The already
2,050 m/s. Around the tunnel, SH-wave
excavated tunnel and the first 3 m of the probe
velocities of more than 2050 m/s exist and
drill is placed in a weathered Volcanic breccia
represent rock mass of weathered volcanic
where the geomechanical rock conditions shows
breccia. Just few meters in front of the tunnel
fair to poor rock mass quality (RMR
face, a low velocity zone up-dipping and left
classification III- IV). Between meter 60 and
striking and an extension of approx. 20m is
meter 75 a Fault zone with decreased and poor
indicated where highly fractured rock mass
rock mass quality (RMR classification III- IV)
occurs. Behind this zone, the velocity increases
is embedded. At meter 75 the rock mass
and returns to values of good rock mass
improves significantly and a change to good
conditions. About 100 meters ahead of the
quality (RMR classification III- II) was found,
tunnel face the velocity drops down again and a
although with a fractured contact zone of 4 m
second low velocity zone becomes visible,
length until meter 79. The remaining section of
almost cross-cut striking the prospective tunnel
the probe drill to meter 87 revealed Volcanic
axis from approx. meter 160 to 170. Further
Breccia. The further geological forecast, based
ahead, intact rock mass returns to good
on information of the parallel tunnel, doesn’t
conditions and retains till the end of the forecast
indicate any other tunneling relevant or
range at meter 200.
significant rock mass change, except a smaller 7
With the combined velocity information of
m long fracture zone included between meter
both P- and S-waves, further rock mechanical
161 and meter 168.
parameters of interest such as Young’s modulus,
The geo-hydrological conditions can’t be
Poisson’s ratio, shear modulus etc. can be
considered critical. However, dripping water
calculated using empirical relationships
with low permeability of 1.25 l/s was present in
depending on the rock group or user-defined
the weathered volcanic breccia and fault zone
formulae for the density (Figure 7).
between meter 45 and meter 75 and extends into
Figure 7b) shows five graphs, which describe
the fractured contact zone up to meter 79.
the predicted curve progression of the P-wave
The subsequent volcanic breccia doesn’t
show any presence of water and is considered to velocity, S-wave velocity, Vp/Vs ratio, Poisson
ratio, Density and Dyn. Young’s Modulus along
be dry to the area of the forecasted embedded
the tunnel axis. The graphs are also colour
fracture zone, where a low permeability might
shaded below their respective chart line. Figure
become possible again.
7c) represents the longitudinal model view of
the 3D-TSP result with reflectors and boundary
4.3 3D investigation shading according to Young’s modulus values.
A colour change takes place at a reflector
The measurement was carried out with the latest
element extension from its location and
TSP 303 technology. This novel system
integrates 3D data acquisition and processing
software containing routines for optimal seismic
imaging with respect to tunnelling requirements.
It exploits the information in the seismic wave
field by separate compression (P) and shear (S)
wave analysis and the 3D-Velocity based
Migration & Reflector Extraction technology
(3D-VMR). The 3D-VMR technology provides
an adequate and detailed 3D image of the Figure 5. Full space perspective view of the cropped 3D
ground leading to a more reliable interpretation velocity distribution (SH-wave) revealing fault and fracture
compared to conventional 2D approaches zone.
(Dickmann and Krueger, 2013).
5
 Proceedings of the World Tunnel Congress 2014 – Tunnels for a better Life. Foz do Iguaçu, Brazil.

start and end point of interrelated reflectors

within same rock mass characteristics derived
from the rock properties.
Conclusively, it is shown that the seismic
prognosis is in very good agreement with the
geological findings of the probe drill and the
further geological forecast. The result points out
the fault zone and fracture zone as rock mass
change critical for the tunnelling. In contrast to
the probe drill, TSP found a widened fractured
rock mass in the contact area between the fault
zone and stable volcanic breccia. In addition,
the result confirms the stable rock conditions
Figure 6. 3D-TSP geological model highlighting fault and after the excavation will have passed the fault
fracture zones in their environment of volcanic breccia. zone and before entering the fracture zone.

orientation in space and its intersection point

with the tunnel axis. 5 CONCLUSION
The measured reference velocity of the direct
P-wave in the area of the measurement layout It is well proved that a sound knowledge on
was 3,540 m/s (S-wave 2,050 m/s, Vp/Vs 1.73), measurement uncertainties and the consistent
corresponding to the fair to poor rock mass of way of a model-driven interpretation won’t pose
the weathered volcanic rock. hazards caused by a wrong understanding of
Beyond the tunnel face location at meter 60, cost savings. Generating and updating
the values of the mentioned parameters begin to geological plausible models with means of
decline. P-wave velocity (2,840 m/s) declines continuous cost effective 3D-Tunnel Seismic
more than S-wave velocity (1,800 m/s) and both Prediction applications during tunneling is the
considerably show the fault zone extension until right way to turn geological uncertainty into
meter 75. Its relative low Vp/Vs ratio (1.60- manageable risk.
1.58) represents a zone of mostly
unconsolidated breccia in a higher stress regime
where water presence does not influence the REFERENCES
shear wave, significantly.
A contact zone between meter 75 and 90 Adeymo, A. 2011. Risk Management of Tunnelling
projects. IRM Construction SIG Meeting, 26th July
indicates a strong fractured not weathered 2011.
volcanic breccia. Within this transition zone, the Bond, C.E. ; Gibb, A.D.; Shipton, Z.K. and Jones, S.
situation (Dyn. Young’s Modulus) slowly 2007. What do you think this is? "Conceptual
changes to better and good condition until meter uncertainty” in geoscience interpretation. GSA Today,
90, according to the decreasing fracturing that v.17, no.11, p.4-11.
was being found. This better rock mass Dickmann, T. and Krueger, D. 2013. Is geological
uncertainty ahead of the face controllable? World
condition persist for about 70 m. Tunnel Congress 2013 Geneva, Underground – the
About 100m ahead of the tunnel face, at way to the future! G. Anagnostou & H. Ehrbar (eds),
meter 161, the P-wave velocity drops again CRC Press
slightly (from 3,760 m/s to 3,630 m/s), while the Dickmann, T. and Sander, B.K. 1996. Drivage concurrent
S-wave velocity drops stronger (from 2,150 m/s Tunnel Seismic Prediction (TSP). Felsbau. v. 14, no.
to 1,940 m/s) indicating a fracture zone 6, p.406-411
forecasted for this area. This zone extends over Eskesen, S.D.; Tengborg, P.; Kampmann, J. and Holst
Veicherts, T. 2004. Guidelines for tunnelling risk
about 5m where the higher increasing Vp/Vs management: International Tunnelling Association,
ratio (from 1.75 to 1.87) also indicates possible Working Group No. 2. Tunnelling and Underground
water bearing. Space Technology, 19, p.217–237
As a last step of interpretation, the seismic Leahy, G.M. and Skorstad, A. 2013. Uncertainty in
geological model is presented in Figure 7d) as subsurface interpretation: a new workflow. First
the plan view of the rock mass changes along Break, v.31, no.9, p.87-93.
the tunnel axis. The geological borders are the
6
 Proceedings of the World Tunnel Congress 2014 – Tunnels for a better Life. Foz do Iguaçu, Brazil.

Figure 7. a) Geological forecast with 30 m long probe drilling from tunnel face b) Rock property charts derived from
TSP measurements c) TSP result with reflectors and boundary shading according to Young’s modulus values in
longitudinal view d) Seismic geological model in plan view.