Tunnelling and Underground Space Technology 25 (2010) 614628

Contents lists available at ScienceDirect

Tunnelling and Underground Space Technology

journal homepage: www.elsevier.com/locate/tust

Geotechnical and operational applications for 3-dimensional laser scanning

in drill and blast tunnels
Stephanie Fekete a, Mark Diederichs a,*, Matthew Lato b
a
Geological Sciences and Geological Engineering, Queens University, Kingston, Ontario, Canada
b
Norwegian Geotechnical Institute, Oslo, Norway

a r t i c l e i n f o a b s t r a c t

Article history: Three-dimensional laser scanning (Lidar) techniques have been applied to a range of industries while
Received 15 October 2009 their application to the geological environment still requires development. Lidar is a range-based imaging
Received in revised form 16 April 2010 technique which collects a very accurate, high resolution 3-dimensional image of its surroundings. While
Accepted 22 April 2010
the use of Lidar in underground environments has been primarily limited to as-built design verication in
Available online 18 May 2010
the past, there is great value in the scan data collected as the excavation advances. The advantages of
employing a static Lidar system for geotechnical and operational applications have been demonstrated
Keywords:
at a drill and blast tunnel operation at the SandvikaAsker Railway Project near Oslo, Norway as well
Laser scanning
Lidar
as in two other test tunnels in Oslo. The increased scanning rate of newer systems makes it possible to
Remote sensing remotely obtain detailed rockmass and excavation information without costly delays or disruption of
Rockmass classication the construction workow with a simple tripod setup. Tunnels are non-traditional environments for laser
Structural evaluation scanners and add limitations to the scanning process as well as the in-ofce interpretation process; these
Tunnel are discussed. Operational applications of the data include: calculation of shotcrete thickness, as-built
bolt spacing, and regions of potential leakage. The authors nd that Lidar data, when correctly inter-
preted, can also provide detailed 3-dimensional characterization of the rockmass. Geometrical character-
ization of discontinuity surfaces including location, orientation, frequency and large-scale roughness can
be obtained. Discontinuity information may be synthesized for a much more representative geomechan-
ical understanding of the rockmass than was previously impossible with traditional hand mapping lim-
ited by face accessibility. The alignment of Lidar scans from successive exposed faces offers additional
interpretation and recording advantages, particularly where shotcrete is subsequently applied behind
the face. In aligning scans, larger scale features can be readily identied and rockmass trends over several
rounds may be identied. Discontinuity geometries and characteristics may be input into kinematic and
numerical models for further analysis.
2010 Elsevier Ltd. All rights reserved.

1. Introduction in tunnel operations, including photogrammetry (Gaich and Pots-

ch, 2008; Birch, 2008). The product of these endeavors has been
In order to ensure safety, long term stability and quality control a colored 3-dimensional surface model. In this technique, each
in modern tunneling operations, the acquisition of geotechnical point in the surface model must be captured by the camera in at
information about encountered rock conditions and detailed in- least two photographs. Full 360 views are then the result of post
stalled support information is required. In order to provide such processing of narrower views. Many laser scanning models, on
information and improve the precision and rate of advancement, the other hand, now scan a full 360 with one setup and capture
data collection as the tunnel progresses must make use of fast a full near-spherical eld of view. This is particularly useful in con-
and effective technology in order to surmount operational con- ned tunnel. In addition, photogrammetry requires supplementary
straints. Limited space and time in an operational tunnel environ- lighting while the laser scanner acts as its own source of illumina-
ment make the use of geophysics and remote sensing equipment tion (Kim et al., 2006; Birch, 2008).
challenging. The use of high-denition (HD) laser scanning in a Three-dimensional laser scanning (Lidar) is an imaging technol-
tunnel environment, however, shows great potential for under- ogy that has been employed in aerial, terrestrial and bathometric
ground implementation. Other imaging techniques have succeeded environments. In principle, the scanners transmit and collect laser
pulses that are reected off physical objects, recording the location
of millions of points in space. The point cloud generated is a high
* Corresponding author.
E-mail address: [email protected] (M. Diederichs).
resolution 3-dimensional dataset of the scanned scene. Normally

0886-7798/$ - see front matter 2010 Elsevier Ltd. All rights reserved.
doi:10.1016/j.tust.2010.04.008
 S. Fekete et al. / Tunnelling and Underground Space Technology 25 (2010) 614628 615

Fig. 1. Three scenes of a Lidar data set from the completed intersection of the three heading at Sandvika tunnel site. Point data is decimated in this view.

the point data (x, y, z) also includes a reection intensity value (i). typically involve scanning at approximately 300 rotations per sec-
These data sets can be manipulated with various software applica- ond, and approximately 2000 points per section prole with a deci-
tions in order to create realistic and practical surface models. meter between proles (Frhlich and Mettenleiter, 2004). While
Applications of the technology span many industries including these proling systems are appropriate for producing efcient 3-
architecture, crime scene investigation, art preservation, transpor- dimensional as-built models for construction and contracting cal-
tation infrastructure maintenance, open pit mining and geo-hazard culations (overbreak, shotcrete control and tunnel alignment), a
mitigation. In terrestrial applications, both mobile and static sys-
tems can be employed. Developing applications of mobile technol-
ogy include urban planning and railway hazard monitoring. Static
systems have traditionally been employed for parts manufacturing
inspection and chemical plant design and maintenance. The geo-
logical engineering community utilizes laser imaging for landslide
monitoring and more recently, rock outcrop characterization, rock-
fall hazard assessment and stratigraphy modeling (Bitelli et al.,
2004; Kemeny et al., 2006; Strouth and Eberhardt, 2005; Buckley
et al., 2008).
The application of laser scanning in underground excavations
has been increasing in recent years. Low density proling scanners,
often referred to as prolers, have been in use for more than a dec-
ade. Advantages of the proling scanners are their ability to scan
long tunnel sections very quickly. Such scanners have been de-
ployed in completed tunnels at up to 100 km/h providing a helical
data set. A construction control scan with this equipment would

Table 1
Applications of lidar in tunneling.

Whos Contractor/ Geological engineer/

interested tunnel engineer geologist
Applications  As-built tunnel model  Structural discontinuity
Final liner surface model geometry
Final rock surface model  Joint spacing
 High density rock/tunnel  Joint surface characteri-
proles zation
 Support installation quality  Structural overbreak
control analysis
Contoured shotcrete  Wedge identication
thickness model  Linear feature mapping
As-built bolt spacing across faces
 Potential leakage location  Geological texture
mapping geometry Fig. 2. (a) Quad-boom Jumbo at Sandvika railway tunnel, near Oslo; (b) Lidar tripod
set up at 10 m diameter face; and (c) detailed view of tripod setup.
616 S. Fekete et al. / Tunnelling and Underground Space Technology 25 (2010) 614628

lot of key geotechnical information is not available and can only be bottom scenes are viewed from inside the data, the top-right
obtained in higher-resolution scanning. scene is seen obliquely from above, i.e. from within the rockmass.
To date, the primary application of 3-dimensional laser scan- Features that are easily identiable include shotcrete lining, venti-
ning in tunneling has been the creation of as-built nal lining mod- lation, and nal excavation geometry.
els. For example, an as-built model to monitor seepage of a 5.6 km There remains a lot of unexplored potential for data collected
section of the Dallas Area Rapid Transit tunnel system was created during the active construction phase. With limited time allocated
from a static Lidar scanning system (Jackson, 2008). There have to engineering characterization during construction, tunnel design
also been examples of Lidar used to evaluate rock or liner deforma- engineers must identify specic technologies that provide the most
tion in tunnels (Van Gosliga et al., 2006; Lemy et al., 2006). Fig. 1 practical information for design while minimizing disruption to the
below shows three scenes from a scan data set taken at the com- excavation workow. Given recent advancements in the rate of
peted intersection of three headings near Oslo, Norway, one sec- data collection, phase-based laser scanners are practical additions
tion of the AskerSandvika Railway Project. The top-left and to the excavation cycle. The employment of a static Lidar scanner

Fig. 3. Processing: (a) raw point cloud, (b) triangular mesh, and (c) meshed tunnel model, Sandvika site. (d) Radial difference map showing nal shotcrete prole compared to
a cylindrical design prole (contoured on the rockshotcrete model or on the cylinder as desired).

Fig. 4. Alignment process: three rock models of successive 5 m rounds in a 10 m diameter tunnel, Sandvika site.
 S. Fekete et al. / Tunnelling and Underground Space Technology 25 (2010) 614628 617

for geotechnical assessment in a drill and blast tunnel operation For this setup, the full process was found to be easily accom-
has been demonstrated in a collaborative project involving Queens plished by one person although an additional person can provide
University (Kingston, Ontario) and Norwegian Geotechnical Insti- support for equipment transport as well as providing a safety
tute (Oslo). Four tunnel sites near Oslo, Norway have been used, guard for the scan operator, typically standing or sitting in the cen-
including two active construction headings in the AskerSandvika tre of an active heading. The scanner was set up just inside the lim-
Railway Project and two completed unlined rock tunnels. This pa- it of supported rock, for safety and best scanning practice,
per will explore the procedure of scanning in tunnels with a static approximately 7 m from the faces (10 and 15 m span), ideally
system, its challenges, limitations and adaptive solutions. Further, 0.51 diameter from the face is the optimal position. The product
the technical applications will be illustrated, both relevant to the of each Lidar scan is a 3-dimensional point cloud, in an x, y, z
contractor and design engineers as seen in Table 1. coordinate system relative to the scanner. As well as position, each
data point is assigned an intensity value and a normal vector. The
intensity value, i.e. strength of returned beam, aids in visualization
2. Hd laser scanning technology and tunneling
and characterization, while the normal vector helps in creating
surface models of the data.
2.1. Tunnel specic challenges and technological solutions
The data for this research was collected with the Leica Geo-
systems HDS6000, a phase-based scanner with a maximum range
The active tunneling environment is more demanding and chal-
of 80 m at 90% reectivity (less for rock surfaces). The scanner
lenging than those where Lidar scanning has traditionally been
has a 360 horizontal eld of view and a 310 vertical eld of
employed. Tunnels require robust systems that are effective in
view, scanning all but underneath the legs of the tripod setup.
dusty, damp, and dark conditions. One key strength of Lidar scan-
ners is that they can function without underground lighting be-
cause the laser acts as its own light source. In the past, the use
of Lidar in active tunnels has been impractical due to long scanning
times for time-of-ight systems (based on return travel time of
beam). However, the development of portable, high speed phase-
based scanners now allows for their use in active tunnels. Phase-
based (or phase offset) scanners emit a continuous wave with
modulated amplitude (AMCW) that varies according to a sinusoi-
dal function (Thiel and Wehr, 2004). The difference in phase be-
tween the transmitted and returned wave is used to calculate the
distance to the target. Because of the AMCW source, the rate of
data collection of phase-based scanners (over a million points
per second) can be 13 orders of magnitude greater than standard
time-of-ight systems. The trade-off for increased acquisition
speed is that the range of phase-based scanners is limited to
80 m. This range is nevertheless sufcient for most underground
environments including infrastructure tunnels which are the focus
of this paper.

2.2. Data collection in an operational environment

Laser scanning data collection becomes highly constrained by

construction workow in an active tunnel environment. The
authors used a stationary tripod setup at Sandvika and Fossvein
tunnel sites to scan the face, walls & crown after successive
rounds. During this demonstration phase, ve separate headings
(four 10 m diameter and one 15 m diameter) were scanned for
multiple face positions. The tunnel faces were scanned in rotation
according to the excavation schedule. The scans where conducted
after blasting, mechanical scaling, and manual scaling. The scan-
ning was initiated within 5 min of mechanical scaling with little
time for dust dispersal. Nevertheless, there was minimal apparent
degradation of scan quality due to dust although care must be ta-
ken at all times to protect exposed system moving parts from dust
and moisture (new hardened systems are now available). The rig-
orous mechanical scaling did cause damage the blasted rock sur-
face and it was noted in the eld that some structural features
were obscured by the scaling marks. The ability to identify rock
mass discontinuities is improved if scanning is performed before
mechanical scaling; the trade-off is that pre-scale data underesti-
mates the actual shotcrete thickness. The authors favored the abil-
ity to perform both liner thickness quality control as well as
geotechnical assessment and therefore chose to scan post-scaling.
This time frame was also favorable for the contractor as it corre- Fig. 5. (a) 60 m  8 m Bankal Test Tunnel complete Lidar model: 11 aligned scans,
sponded with the time allotted for geotechnical engineering med. resolution; (b) photograph of Lidar system with viewpoint A in aligned scans;
assessment. and (c) view from within Lidar model at A (only raw points shown).
618 S. Fekete et al. / Tunnelling and Underground Space Technology 25 (2010) 614628

The system provides high speed, high density acquisition at ca Geosystems, 2008). There is improved relative accuracy (dis-
500,000 points per second. The scanner offers the choice of dif- tance) for neighbouring points within the same scan. At this
ferent resolutions with point clouds ranging from hundreds of higher resolution setting, scanning at the tunnel face requires
thousands of points to tens of millions for a single scan. The 3 min with an additional 2 min to download to a eld computer
trade-off of high point density is extended scanning times and in- (not required for scanners with internal memory). The entire set
creased computer processing hardware requirements. When set up, scanning and take down processes were completed in under
up 7 m from the face at the Sandvika tunnel, the selected resolu- 7 min for each round, including download which is can be omit-
tion setting produced 6 mm intervals between points on the face. ted in the eld if desired. In this way, excavation work ow was
The system has a displayed positional accuracy of 6 mm and a not interrupted as the scanning could be easily t into the 30 min
distance accuracy <4 mm at distances up to 25 m (calibration Lei- time slot provided for geotechnical assessment. Fig. 2 includes

Fig. 6. (a) Rock and (b) shotcrete Lidar models of three aligned 5 m rounds, Sandvika site; (c) longitudinal and cross-section (along AA0 ) showing detailed comparison of
proles with liner thickness; (d) shotcrete thickness contoured onto model (with unsupported heading in front).
 S. Fekete et al. / Tunnelling and Underground Space Technology 25 (2010) 614628 619

photographs for one heading at the Sandvika site showing the hole orientation. Inuences of additional scanning bias has been
quad-boom drilling equipment as well as the Lidar tripod setup widely documented for rock slopes (Sturzenegger et al., 2007; Lato
at the face. et al., 2009a, in press) especially with regard to the invisibility of
discontinuities parallel to the line of sight. However, Lidar scanning
2.3. Limitations of tunnel scanning in tunnels is somewhat less susceptible to the bias observed when
scanning rock slopes. In a rectangular tunnel heading, for example,
The primary limitations of 3-dimensional laser scanning from a three roughly orthogonal planes are scanned (walls, face and
single setup position (per face) are occlusion and scan bias. Occlu-
sion, often referred to as shadowing, limits data acquisition to
what the laser can see. Because a surface view of some joints
may be limited or fully obstructed, these will not be documented
by the scan (Lato et al., in press). Sturzenegger et al. (2007) found
that a roadside rock outcrop cannot be fully sampled from one
scanning location in the comparative stereonet analysis of two
scanning locations and that the best coverage is, in principle,
achieved with multiple scans in both the vertical and horizontal
planes. However, in an operational tunnel environment, time and
space limitations may only allow for one scan per round. This
may limit the reliability for detection and location of discrete
structures with unfavourable viewing angles although the tunnel
environment provides for a wide range of viewing angles (joint ori-
entation relative to scanner) so that occlusion is reduced for overall
rockmass characterization purposes. At the Sandvika site, the pri-
mary source of occlusion (with the tripod-mounted scanner) was
a shallowly dipping joint set when it occurred in the upper face.
If this joint set was deemed critical in stability analysis, its visibil-
ity in the lower face would allow for extrapolation into the upper
portion of the face. Mounting the scanner on the scaling or drilling
boom has been considered in order to reduce vertical occlusion and
minimize set up time, although the scanning system would require
additional hardening for this application. Fig. 8. Point cloud data of shotcrete liner at boundary of exposed rock, showing:
Bias is a signicant limitation in geotechnical assessment as bolt spacing, bolt load area, and the contrasting intensity values of dry and moist
documented by Terzaghi (1965) with respect to scanline and drill shotcrete, Sandvika site.

Fig. 7. (a) Two aligned tunnel scans (view from behind face), pre- and post-mechanical scaling with a cross-section AA0 showing prole where scaling has removed
material; (b) quantitative comparison of pre- and post-scaling scans, where hot colors indicate areas where material has been removed in the process. (For interpretation of
the references to color in this gure legend, the reader is referred to the web version of this article.)
620 S. Fekete et al. / Tunnelling and Underground Space Technology 25 (2010) 614628

crown). Thus, a surface near-parallel to the laser in the face may be lute positioning using a total station. Data point spacing on the
visible in the crown or walls and vice versa. Due to bias the scanner tunnel wall increases with decreasing wall-beam angle and data
may not see every discontinuity but as long as it is part of a set overlap between scans increases reliability and alignment accu-
that repeats, it will be detected on a least one of the scanned sur- racy. For scans of existing tunnels, it is recommended to use a lon-
faces. It is therefore important that in selecting a tripod position, gitudinal scanning spacing (distance between tripod setups) of
operators are aware of and note any critical discrete structural fea- between 1 and 2 tunnel radii. The nal tunnel model of the Bankal
tures that may not be visible in the scan. Test tunnel, illustrated in Fig. 5, it is assembled from 11 aligned
scans. A view from inside the tunnel from a photograph as well
3. Data processing as from within the Lidar point cloud highlighting roof and wall
structure are also included.
While the point cloud collected contains accurate geometric Various software programs can be used to view, align and mod-
information, the data set still requires some processing in order el point cloud data. For this project, Cyclone (Leica Geosystems)
to: (a) reduce the size of the data set to a manageable size, (b) cre- and Polyworks (InnovMetric Software Inc.) were used. The ability
ate continuous surface models (rather than just points) and (c) to analyze and interpret the data is a function of data quality, pro-
align the data with adjacent scans. Once this processing has been gram functionality and also computing power. The large le sizes
completed, the data can be used for analysis. Not all types of anal- have high memory requirements for visualization and processing.
yses or data sets require the all these steps but generally, this is the Further, high-end video cards are essential in order to be able to
work ow. Lidar data sets can be up to GBs in size and so in order visualize the data at a practical rate.
to manipulate the data in an efcient manner, even with a power-
ful computing station, the data set should be reduced to the region 4. Operational applications
of interest (ROI). Any unwanted objects can be deleted: scaling
equipment, muck piles, reections etc. In addition, depending on As seen with current applications of laser prolers, the point
the level of precision required and the analyses to be conducted, cloud data of lined tunnel sections have great value for contractors.
the processer may elect to decimate the data, in other words, to As already shown in Fig. 3, nal proles can be compared with de-
use only 1/4 or 1/16 points collected. If there are specic zones sign proles for contractual verication. Furthermore, by aligning
of interest identied in the decimated data set, a smaller region active face scans obtained with a 3-dimensional laser scanner, a
with higher point density may be subsequently analyzed in detail. high density model can be obtained which unites information from
The creation of the mesh or surface model is an important step in before and after support installation. This equates to a powerful
both handling the large data sets as well as in preparation for anal- quality control tool.
ysis, reducing the size of the data by assigning groups of adjacent
points appearing to lie on the same plane to triangles with a de- 4.1. Support evaluation and scaling assessment
nite centroid, area, vertices and normal vector. The mesh often im-
proves the interpreters ability to visualize the data and is required Similar to the data collected by tunnel prolers, Lidar scan data
for any analysis requiring continuous surface information. The pro- can be used to produce rock and nal lining proles. Three rounds
cess (Fig. 3) of reducing the point cloud (a) to a mesh (b) in order to from a 10 m heading at the Sandvika site are shown longitudinally
create a 3-dimensional tunnel model (c) provides for improved and in cross-section in Fig. 6. These can be easily exported for fur-
visualization, data reduction and the ability to perform geometrical ther use in CAD programs for comparison against original design.
investigations (Fig. 3d) such as the analysis of overbreak or nal Every part of the tunnel surface is scanned twice (pre- and post-
prole compared to a design prole. shotcreting). This allows direct comparison and evaluation of shot-
An alignment (registration) step is required in order to connect crete thickness in detailed prole or contoured on the whole tun-
scans and create full tunnel models. The alignment of the rock sur- nel model for visualization.
faces of three rounds at the Sandvika tunnel (i.e. the shotcrete liner In addition to lining evaluation, the Lidar data sets can be used
etc. has been edited out) is shown in Fig. 4. This can be done using to evaluate construction cycle efciency (blasting, scaling, etc.) by
relative position markers (visible bolts, pipes, etc. appearing in analyzing the resulting rock surfaces. The scan of a 10 m heading at
consecutive scans) for rough alignment or can be tied into abso- the Sandvika site was scanned after blasting, pre-scaling as well as

Fig. 9. Raw point cloud from one scan illustrating 30 m section of tunnel showing contrasting intensities of exposed rock, dry shotcrete and wet shotcrete (in this case, wet
due to seepage through fractures).
 S. Fekete et al. / Tunnelling and Underground Space Technology 25 (2010) 614628 621

after mechanical scaling. These models were then compared in or- 4.2. Potential leakage mapping, post-construction
der to evaluate the degree of rock removed by the scaling process.
The two aligned pre- and post-scaling models are shown in Fig. 7a. Data collected during the construction phase can be analyzed
The cross-section highlights a section of signicant rock removal for potential prevention of leakage as well as for comparison
with overall scaling depths contoured in Fig. 7b. against later modeling. Differentiation in intensity values between
Quality control of support can also include bolt installation. The dry and wet materials is documented by Lichti and Harvey (2002).
re-evaluation of bolts positioning of can be conducted very ef- For concrete, in particular, Lichtis experimental results show that
ciently in data visualization programs in order to obtain as-built both at near and far ranges concrete intensity values decreased
bolt spacing and audit bolt density. Fig. 8 shows metal plate an- when the surface was moistened. This difference in intensity al-
chored bolts at the Sandvika tunnel, their spacing, and an estimate lows for mapping regions of potential leakage where shotcrete
of tributary load area. This permanent 3D record provides a basis has begun to absorb moisture. Work by Sturzenegger et al.
for future support analysis should stability problems arise. (2007) on seepage in rock slopes conrms this differentiability.

Fig. 10. (a) Detailed Lidar model of a wedge shaped groundfall in the Akershus Tunnel (Oslo); (b) wedge contoured by overbreak height; (c) tunnel scale overbreak map; and
(d) mechanical model of unstable block.
622 S. Fekete et al. / Tunnelling and Underground Space Technology 25 (2010) 614628

The contrasting intensity of wet and dry patches of newly applied and detailed analysis after leaving the site as well as after the rock
shotcrete is seen in Fig. 8. As anticipated, regions adjacent to bolts is obscured by support installment. Further, the scanning and
remain moist, highlighting leakage initiated by drilling. This ability alignment of subsequent rounds can be used to create more exten-
to differentiate dry from moist may be equally useful in identifying sive rockmass models, enhancing the ability to identify key discon-
key (open and water bearing) structural discontinuities now ob- tinuity features and key failure modes.
scured by the liner. As seen in Fig. 9 the differentiability of inten-
sities is signicant and does not require high point density. Train
or truck mounted mobile scanning (Lato et al., 2009b) for example, 5.1. Analysis of structurally controlled overbreak
can be used to survey rapidly for leakage after construction is com-
plete or in an older tunnel without interrupting trafc ow. Structural discontinuity geometry information can also be used
to evaluate the impact of structure on the ability to maintain the
desired excavation geometry. Both zones where too much rock
5. Geotechnical applications has been excavated (overbreak) and zones that have been insuf-
ciently blasted to meet tunnel prole requirements (underbreak)
The primary advantage for collection of Lidar data during the can be identied in comparing the nal rock model to the design
excavation process, with respect to geotechnical applications, is model. Fig. 10 shows a detailed analysis of a particular groundfall
that it allows for permanent documentation of rock conditions in the Akershus tunnel site in Oslo. The height above a semi-cylin-

Fig. 11. (a) Lidar scanning in the Akershus Tunnel; (b) two joint sets identied from Lidar model (third set facing away from viewer); (c) wedge geometry formed by three
identied planes; and (d) stereonet and wedge models from all unstable combinations of ve joint sets.
 S. Fekete et al. / Tunnelling and Underground Space Technology 25 (2010) 614628 623

drical tunnel is computed and displayed as colored contours and authors propose the use of Lidar scan data for rockmass character-
the planes bounding the failure are imported to a mechanical mod- ization, in particular for mapping structural discontinuities. Tradi-
el for back analysis. In addition, key joint sets can be identied and tional hand mapping at the face can be time consuming and is
possible combinations of intersecting structure (wedges) can be limited by safe access to the unsupported zone. The use of Lidar
analyzed with basic wedge failure modeling programs as shown data allows for detailed mapping of structural features at the ofce,
in Fig. 11 and discussed in the next section. and allows the specialist on-site to spend more time on other char-
acterization attributes, like alteration, water inow and disconti-
5.2. Structural discontinuity evaluation nuity lling. As discussed by Decker (2008), discontinuity
analysis using Lidar data has many advantages over traditional
In order to optimize the time a geologist or geotechnical engi- mapping, i.e. by hand with compass, including: digital rockmass
neer spends at the face in a drill and blast tunnel operation, the documentation that can be reinterpreted by other specialists as

Fig. 12. (a) Bare rock model, meshed, aligned Lidar data and (b) 158 identied joint surfaces (right) from three 5 m rounds, Sandvika site; (c) stereonets of traditionally hand-
mapped structural data (left) and manually extracted planes from point cloud, virtual mapping (right) where tunnel trend is NS.
624 S. Fekete et al. / Tunnelling and Underground Space Technology 25 (2010) 614628

well as increased quantity and accuracy of measurements with less The authors have selected manual feature extraction which is
time spent at face/in hazardous zones. The extraction of disconti- conducted by selecting the data points falling on a surface that
nuity surfaces from point cloud data can be done either interac- the interpreter deems to be a joint. While this method place more
tively and automatically. Fig. 12 shows a 15 m length of exposed onus on the interpreter, authors feel this is preferred over auto-
rock, three rounds of excavation with 158 joint measurements mated discontinuity extraction which automatically identies
identied manually. The results of traditional mapping (compass patches in the data, i.e. groups of mesh triangles with similar nor-
or clinorule) and interactive mapping of the Lidar data, are found mal vectors. Automated discontinuity feature extraction is attrac-
to be very comparable. The two main joints sets obtained had tive in its ability to objectively detect discontinuities. However,
dip and dip direction: 60/078, 90/180 (traditional mapping) and the authors nd that for underground applications where the rock
62/072, 88/177 (Lidar data). It is important to note that a greater face has been damaged during blasting, scaling, etc., current algo-
quantity of measurements is obtained by virtual mapping than rithms are unable to provide acceptable structural information. Re-
hand mapping and more signicantly, the variation of discontinu- sults from automated joint detection algorithms are often noisy
ity surface is better represented. The accuracy of the Lidar mea- (due to rock damage) with may cause any signicant structural
surements is greater since it is not affected by compass deviation information to be masked while excess automated ltering may re-
or by approximations made for hard-to-reach structures. The var- move key features. Mesh type is also critical for structural charac-
iation within a joint set as well as random fractures are therefore terization (Lato et al., 2009a,b). Simple draping mesh algorithms
far better represented by virtual mapping conducted in the ofce incorporated in some software limit the user to either interpreting
with the evaluation of tunnel face Lidar data. the face or the crown, but not both simultaneously. This leads to

Fig. 13. Extrapolation of major joint surfaces and colored according to orientation, in order to indentify key structural concerns (where grey are random orientation), three
5 m rounds in 10 m diameter tunnel, Sandvika site.

Fig. 14. (Left) 3-dimensional block model constructed actual joints indentied from Lidar data. (Right) Exaggerated block movement in mechanical model, highlighting
potential instability modes. See text for note about bias (lines radiating from left springline are model construction lines not joints).
 S. Fekete et al. / Tunnelling and Underground Space Technology 25 (2010) 614628 625

severe bias challenges while more advanced and truly 3D tessella- tract the exact location and orientation of discontinuities in 3-
tion algorithms provide a better solution (used in the examples dimensional space can also be advantageous for the creation of a
presented). very complete database for block modeling such as that shown in
Fig. 14. It is very important to note that for joints within a partic-
5.3. Discontinuity spacing and 3-dimensional models ular set, a bias against detection will exist in certain portions of the
scanners eld of view. This is not a problem for statistical collec-
The advantage of virtual structural mapping extends beyond tion since different joint orientations will dominate different por-
statistical structural characterization but positions each disconti- tions of the view. For discrete models (placing joints in their
nuity feature in 3D space, thus creating a rockmass model. This actual location) this is a critical issue. Fig. 14 illustrates this point.
model can serve to evaluate discontinuity interaction and joint The steeply dipping joint set in the image was observed to have
spacing. Extrapolations of joint features identied in Fig. 12 are similar frequency and spacing across the tunnel prole. In the Lidar
shown in Fig. 13. The features are colored by common orientation, model, however, the detection of this joint set is hindered across
where seemingly random orientations remain in grey. This sort of the upper left part of the prole. Work is ongoing to understand
analysis can be useful in efciently evaluating joint spacing, poten- and compensate for this bias.
tial locations of wedge failure as well as the prevalence of random Setting aside concerns about this bias for the moment, a joint
joint orientations. Furthermore, this analysis can become part of spacing analysis is shown for the same bedding-dened set found
the permanent digital rockmass documentation. The ability to ex- in the siltstone-sandstone unit at the Sandvika site. Fig. 15a shows

Fig. 15. (a) Identied and extrapolated joint planes and (b) joint spacing distribution of joint planes identied.
626 S. Fekete et al. / Tunnelling and Underground Space Technology 25 (2010) 614628

the planes identied along bedding and their extrapolation In this plot, the larger spacing records may be the result of the bias
through the right side of the tunnel. A histogram is shown in mentioned previously and would require correction.
Fig. 15b, created from the measured joint spacings in this model.

Fig. 16. Large joint surface and 10 cm surface prole, Bankal tunnel, compared to JRC roughness proles (from Barton and Choubey (1977).

Fig. 17. (a) Deviation of rock surface models from a tted geometrical plane (eld of view is 2 m). Note half barrels rendered in the data; (b) contouring of joint surface with a
best t plane through a large data window and (c) a more local t (higher data density as well). Field of view is 2 m.
 S. Fekete et al. / Tunnelling and Underground Space Technology 25 (2010) 614628 627

5.4. Surface characterization the Bankal tunnel in northern Oslo obtained for comparison to
the standard roughness proles for JRC (Barton and Choubey,
The high density of Lidar data also allows for detailed analysis 1977). This is equivalent to performing a manual measurement
of joint surfaces themselves. Research in past years has attempted with a prole gauge.
to both correlate 3-dimensional Lidar models to easily applied 2- Joint surface models from Lidar data can also be analyzed by
dimensional prole characterization like JRC (Barton and Choubey, comparison against geometrical best-t planes through the local
1977). Researcher have also proposed 3-dimensional fracture char- point cloud. This method allows for an estimation of roughness
acterization methods in response to the increasing ability to collect amplitude to be made, i.e. the deviation from the average plane.
and manipulate high density point cloud data sets (Haneberg, This planar analysis allows for a more 3-dimensional characteriza-
2007; Rahman et al., 2006; Fardin et al., 2001). Large-scale rough- tion where anisotropy in the surface variation may be noted. If
ness is distinctly identiable but the ability to differentiate small- anisotropy on the fracture surface is present, particular attention
scale roughness from noise has yet to be demonstrated. Fig. 16 should be paid to the selection of roughness proles if they are
shows a 10 cm vertical roughness prole along a joint surface at to be used as a classication tool. The rock surface model of a large
fracture planes are compared to a best t (average) plane and the
deviation is contoured in Fig. 17a. The approach is sensitive to
the sizing of the measurement window (directly related to the
scale of roughness) and the data density. The comparisons (devia-
tion) between a second joint surface (Bankal tunnel) to a larger and
a smaller best t plane are shown in Fig. 17b and c.

5.5. Identication of discrete and textural geological features

Lidar data sets are highly advantageous in their ability to docu-

ment discrete features in the rockmass, their orientation and posi-
tion. This information is valuable for focusing monitoring efforts in
these zones post-construction, as well as, for predicting conditions
in adjacent excavations (i.e. twin tunnels or benched excavation).
As seen in Fig. 18 which is a raw point cloud with intensity (laser
Fig. 18. A large scale discrete discontinuity crossing the face and right wall seen in reectivity) displayed, large scale discrete features maybe easily
raw point cloud (using laser reectivity (intensity) data), Sandvika site. identied and documented without any processing.

Fig. 19. Left: photograph of calcied shear zone in round 0 + 10 m. Right: aligned face scans with intersecting planar shear zone, Sandvika site.

Fig. 20. Left: raw Lidar point cloud of tunnel face with large planar surface on right. Right: zoom-in of smooth planar surface where geological texture is visible due to
alternating intensities of units, Sandvika site.
628 S. Fekete et al. / Tunnelling and Underground Space Technology 25 (2010) 614628

However, not all discrete features may be as obvious as this. A References

key advantage of collecting Lidar data during the excavation cycle
is the ability to simultaneously analyze rock that is now exposed Barton, N.R., Choubey, V., 1977. The shear strength of rock joints in theory and
practice. Rock Mech. 10 (12), 154.
and rock that has since been obscured by support installment. Birch, J.S., 2008. Using 3DM analyst mine mapping suite for underground mapping.
The alignment of scans from successive blast rounds is advanta- In: Tonon, F. (Ed.). Proceedings of Laser and Photogrammetric Methods for Rock
geous for identifying features which may only be visible as linea- Tunnel Characterization Workshop. 42nd US Rock Mechanics Symposium
ARMA, San Francisco, June 2829, 2008. Cdrom.
tions in the face. Features that may have been overlooked or Bitelli, G., Dubbini, M., Zanutta, A., 2004. Terrestrial laser scanning and digital
dismissed after one or two excavation rounds may reveal them- photogrammetry techniques to monitor landslide bodies. In: International
selves as signicant over multiple rounds. A calcied shear zone Society for Photogrammetry and Remote Sensing: XXth Congress Commission,
vol. 5. Istanbul, Turkey.
is dened geometrically as a planar feature through three face
Buckley, S., Howell, J., Enge, H., Kurz, T., 2008. Terrestrial laser scanning in geology:
scans and indicated in a photograph of the third round in Fig. 19. data acquisition, processing and accuracy considerations. Journal of the
In addition to the ability of Lidar interpretation to identify fea- Geological Society, London 165 (3), 625638.
tures of differing relief and geometry, the intensity values of the Decker, J. 2008. Laser scanning techniques in devils slide tunnels. In: Proceedings of
Laser and Photogrammetric Methods for Rock Tunnel Characterization
point cloud may also offer textural information the surfaces Workshop. Tonon, F. (Ed.). 42nd US Rock Mechanics Symposium. American
scanned. Such textural information, similar to an oriented photo- Rock Mechanics Association, San Francisco, June 2829, Cdrom.
graph, may offer additional geological characterization potential. Fardin, N., Stephansson, O., Jing, Lanru, 2001. The scale dependence of rock joint
surface roughness. International Journal of Rock Mechanics and Mining
Fig. 20 below shows the raw Lidar point cloud of the face at the Sciences 38 (2001), 659669.
Sandvika site. A large smooth failure surface reveals the ability to Frhlich, C., Mettenleiter, M., 2004. Terrestrial laser scanning new perspectives in
differentiate between interbedded sandstone and siltstone layers 3-dimensional surveying. Laser-scanners for forest and landscape assessment.
In: Proceedings of ISPRS Working Group VIII/2. International Archives of
based on differing intensities. Photogrammetry, Remote Sensing and Spatial Information Sciences, vol. XXXVI,
Part 8/W2.
Gaich, A., Potsch, M. 2008. Computer vision for rock mass characterization in
6. Conclusions
underground excavations. In: Proc. Laser and Photogrammetic Methods for
Rock Tunnel Characterization. Tonon, F. (Ed.). 42nd US Rock Mechanics
The application of Lidar technology has been expanding and is Symposium ARMA, San Francisco, June 2829, Cdrom.
rapidly becoming applied to new industries and projects. However, Haneberg, W.C., 2007. Directional roughness proles from three-dimensional
photogrammetric or laser scanner point clouds. In: Eberhardt, E. et al. (Eds.).
it is only starting to realize its potential in the underground envi- Rock Mechanics: Meeting Societys Challenges and Demands: Proceedings.
ronment, especially in the tunneling industry. Numerous applica- 1st CanadaUS Rock Mechanics Symposium, Vancouver, May 2731, pp. 101
tions exist for the high density, high accuracy 3-dimensional 106.
Jackson, T., 2008. Tunnel Vision Point of Beginning Magazine. February 25, 2008.
point cloud models created. Data obtained during the active con- <http://www.pobonline.com/Articles/Features/BNP_GUID_9-5-
struction phase can be useful for both construction contractors 2006_A_10000000000000269278> (accessed 02.06.08).
and geotechnical personnel. In particular, the alignment of succes- Kemeny, J., Turner, K., Norton, B., 2006. LIDAR for rock mass characterization:
hardware, software, accuracy and best-practices. In: Proceedings of the Laser
sive scans provides valuable information on both the rockmass and and Photogrammetric Methods for Rock Face Characterization Workshop June
the supported excavation. The authors have been successful in the 2006. Tonon, F., Kottenstette, J.T. (Eds.). American Rock Mechanics Association,
implementation of static Lidar scanning at four sites near Oslo, pp. 4962.
Kim, C., Ghanma, M., Habib, A. 2006. Integration of photogrammetric and LIDAR
Norway, including two active tunneling operations. Lidar scanning
data for realistic 3-dimensional model generation. In: 1st International
at these tunnel sites and later in-lab analysis of the Lidar data Workshop on Mobile Geospatial Augmented Reality, Banff, Alberta, Canada,
demonstrate: May 2930. Cdrom.
Lato, M., Diederichs, M., Hutchinson, D.J., Harrap, R., 2009a. Optimization of LiDAR
scanning and processing for automated structural evaluation of discontinuities
 the ability to collect good quality data with a tripod setup in an in rockmasses. International Journal of Rock Mechanics and Mining Sciences 46
active tunnel environment without disruption of construction (1), 194199.
workow, Lato, M., Hutchinson, J., Diederichs, M., Ball, D., Harrap, R., 2009b. Engineering
monitoring of rockfall hazards along transportation corridors: using mobile
 the ability to obtain detailed quality control information on the LiDAR. Natural Hazards and Earth System Sciences 9, 935946.
precision of excavation and installed support, for use during Lato, M., Diederichs, M., Hutchinson, D.J., Harrap, R., in press. Bias correction for
construction as well as for permanent documentation, and view-limited Lidar scanning of rock outcrops for structural characterization.
Rock Mechanics and Rock Engineering. doi:10.1007/s00603-010-0086-5.
 the ability to retrieve representative rockmass information Lemy, F., Yong, S., Schulz, T., 2006. A case study of monitoring tunnel wall
remotely and perform geotechnical analysis, both at the meso- displacement using laser scanning technology. International Association for
and macro-scale, Engineering Geology (Paper 482).
Lichti, D.D., Harvey, B.R., 2002. In: The Effects of Reecting Surface Material
 research is ongoing in this eld including: the semi-automation Properties on Time-of-ight Laser Scanner Measurements Symposium on
of some of the processes discussed, the evolving methods of 3D Geospacial Theory, Processing and Applications, Ottawa, Canada.
roughness classication with point cloud data, the evaluation of Rahman, Z., Siefko, S., Hack, R., 2006. Deriving roughness characteristics of rock
mass discontinuities from terrestrial laser scan data. The Geological Society of
additional rockmass classication parameters such as surface
London. International Association of Engineering Geology.
alteration or lling, and improving the ability to complete vol- Strouth, A., Eberhardt, E., 2005. The use of LiDAR to overcome rock slope hazard data
ume calculations. collection challenges at Afternoon Creek, Washington. In: 41st US Symposium
on Rock Mechanics, Golden, Colorado. American Rock Mechanics Association,
Cdrom
Sturzenegger, M., Yan, M., Stead, D., Elmo, D., 2007. Application and limitations of
Acknowledgments ground-based laser scanning in rock slope characterization. In: Eberhardt, E.,
Stead, D., Morrison, T. (Eds.). Proceedings of the rst Canadian US Rock
Mechanics Symposium, vol. 1. Taylor & Francis, London, 2007, pp. 2936.
The authors also wish to thank the Norwegian Geotechnical Terzaghi, R.D., 1965. Sources of error in joint surveys. Geotechnique 15, 287304.
Institute for their support and collaboration, in particular Eystein Thiel, K.H., Wehr, A., 2004. Performance capabilities of laser scanners an overview
Grimstad, Maureen White, Elin Morgan, Arnstein Aarset, and Su- and measurement Principle analysis. In: Laser-Scanners for Forest and
Landscape Assessment, Proceedings of ISPRS Working Group VIII/2, vol. 36,
zanne Lacasse. Thanks are due to Jean Hutchinson of Queens Uni- Part 8/W2.
versity and to Linda Nesje and Odd-Andr Rustad at the Sandvika Van Gosliga, F., Lindenbergh, R., Pfeifer, N., 2006. Deformation analysis of a bored
tunnel site. The support of the Sandvika-Asker Tunnel Project (Nor- tunnel by means of terrestrial laser scanning. In: Image Engineering and Vision
Metrology. ISPRS Commission, vol. 36, pp. 167172.
wegian Railroad Authorities, JBV) for the eld trials is appreciated.
Funding was provided by NSERC, NGI, PREA and GEOIDE.