ARTICLE IN PRESS

Computers & Geosciences 35 (2009) 1322–1333

Contents lists available at ScienceDirect

Computers & Geosciences

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

Tunneling Analyst: A 3D GIS extension for rock mass classification

and fault zone analysis in tunneling
Yosoon Choi a, Seo-Youn Yoon b, Hyeong-Dong Park a,
a
Department of Energy Systems Engineering, College of Engineering, Seoul National University, Seoul 151-744, Republic of Korea
b
Department of Geography, University of California, Santa Barbara, CA 93106, USA

a r t i c l e i n f o abstract

Article history: In this study, an extension called Tunneling Analyst (TA) has been developed in
Received 25 January 2008 ArcScene 3D GIS software, part of the ArcGIS software package. It dramatically extends
Received in revised form the functionalities of ArcScene because it allows: (1) estimation of the 3D distribution of
2 May 2008
rock mass rating (RMR) values using borehole and geophysical exploration data, (2) the
Accepted 8 May 2008
modeling of 3D discontinuity planes such as faults from field-based structural
measurements, and (3) analysis of 3D intersections and 3D buffer zones between
Keywords: proposed tunnel alignments and some discontinuities. Because TA can handle and
3D GIS visualize both 2D and 3D geological data in a single GIS environment, the tedious tasks
Tunneling
required for data conversion between various software packages can be reduced
Rock mass classification
significantly. The application to the Daecheong tunneling project in Korea shows that
Multiple indicator kriging
Spatial query function TA could present a rational solution to evaluating the rock mass classes along a proposed
Rock mechanics tunnel alignment and can also provide specific 3D spatial query tools to support the
tunnel design work. This paper describes the concept and details of the development
and implementation of TA.
& 2008 Elsevier Ltd. All rights reserved.

1. Introduction Traditionally, the subsurface characterization was con-

In geological and geotechnical engineering, site char-
acterization of the subsurface is necessary for the design
of underground engineering structures such as founda-
tions of buildings, underground food storage caverns and
tunnels. In routine tunnel design practice, especially
geological engineering properties of the subsurface should
be well characterized to classify rock masses near a
proposed tunnel alignment into general rock types using
rock mass classification systems such as rock mass rating
(RMR) (Bieniawski, 1989) and the Q-system (Grimstad and
Barton, 1993; Choi and Park, 2002). After the classification,
standard support patterns for the general rock types can
be developed in a tunnel design project.
 Corresponding author. Tel.: +82 2 880 8808; fax: +82 2 871 8938.
E-mail addresses: yspower7@snu.ac.kr (Y. Choi), yoon@geog.ucsb.edu
(S.-Y. Yoon), hpark@snu.ac.kr, giscity@hanmail.net (H.-D. Park).
ducted manually by geologists and required much time
and effort. However, nowadays, Geographic Information
Systems (GIS) and Computer-Aided Design (CAD) systems
are widely used in the tunneling industry and play
important roles in the process of subsurface characteriza-
tion. In particular, this new trend has been significantly
accelerated in Korea since national geosciences database
services such as National Geographic Information Clear-
inghouse (NGIC, http://www.ngic.go.kr), National Ground-
water Information Management and Service (NGIMS,
http://www.gims.go.kr), Digital Geologic Map Search
System (DGMSS, http://geoinfo.kigam.re.kr), geotechnical
information database system (http://www.geoinfo.or.kr),
Web-based geotechnical information system in Seoul
(Chang and Park, 2004), etc., started to provide large
volumes of digital-ready data to the tunneling industry.
Geologists can search and download the digital-ready data
including boreholes, cross sections, remote-sensing
images, geological maps, topographical maps, hydro-
geological maps, structural geology maps as well as

0098-3004/$ - see front matter & 2008 Elsevier Ltd. All rights reserved.
doi:10.1016/j.cageo.2008.05.002
 ARTICLE IN PRESS
Y. Choi et al. / Computers & Geosciences 35 (2009) 1322–1333
contours of the water table that are required for subsur-
face characterization. Because the digital-ready data are
stored in various formats (e.g., Excel spreadsheets, Access
personal databases, CAD drawings, ESRI shapefiles),
geologists can select suitable formats and can combine
them with newly acquired geological data using GIS
or CAD systems. Comparing the advantages of GIS with
CAD systems, GIS is preferable for modeling and analyzing
not only man-made objects but also natural geo-objects
(Turner, 2005). Consequently, GIS is becoming a useful
tool in the tunneling industry for characterizing the
geological engineering properties of the subsurface and
for supporting tunnel design work.
In response to the increased availability of GIS in the
tunneling industry, several recent applications of GIS for
tunneling projects have been reported in the literature.
Kimmance et al. (1999) tried to use GIS as a geotechnical
database system on airport line Metro Rail Transit (MRT)
construction and discussed the design and architecture of
database systems for a tunneling project. Their system
uses a well-designed graphical user interface to access the
database and to report various charts, images and tables
available from the system. However, the functionality of
the system was limited to query records stored in
relational database tables. Yoo et al. (2006) developed a
GIS-based risk assessment system for tunneling as an
extension of ESRI’s ArcGIS. The program provided
advanced functionalities for analyzing ground movement,
utility damage and groundwater drawdown that can occur
during tunnel construction. It was successfully applied to
the Seoul metro subway orange-line extension project in
Korea. However, it could only manipulate and analyze
two-dimensional (2D) data; therefore, there were great
difficulties in handling three-dimensional (3D) geological
data for the subsurface characterization. Yoo and Kim
(2007) also used GIS for predicting tunneling performance
in a high-speed railway tunnel; however, the aim of GIS
utilization was limited to visualizing the performance
prediction maps in 2D. Because most geological data used
for the subsurface characterization are in 3D, the
capability of handling 3D data is required in GIS to
support tunnel design work. Without 3D functionality in
GIS, geologists need to interpret the 2D representation of
geological data to understand 3D implications and in
doing so errors can occur during interpretation.
3D GIS can be an interpretive environment in which
3D spatial data can be queried, manipulated and repre-
sented in a meaningful manner, so as to provide insight
into geological problems (Sprague et al., 2006). 3D GIS for
characterizing the subsurface were originally developed in
mining and the oil and gas industries. Therefore, in
geological and geotechnical engineering, the first efforts
to use 3D GIS also started with external software
originally developed in the mining and oil and gas
industries. Some 10 years ago, it was expected that
3D GIS would soon be standard practice in geological
and geotechnical engineering to characterize the subsur-
face; however, the introduction has not been as fast as had
been anticipated (Hack et al., 2005). Although a few
attempts have been made to use 3D GIS for tunneling
projects (Elkadi and Huisman, 2002; Kaalberg et al., 2003;
1323
Ozmutlu and Hack, 2003), these mainly focused on
3D geological modeling with extensive use of external
software such as Lynx-GMS (Houlding, 1994), and they
used GIS for 3D visualization of block models representing
soil or rock types in the subsurface. Due to their excessive
dependence on external software, they needed much data
conversion among the various software and that is often
time consuming and tedious, especially for a large study
area. Nowadays, some commercial geological modeling
software such as EVS (www.ctech.com) and Rockworks
(http://www.rockware.com) are available for geological
and geotechnical engineering applications and provide
data conversion tools for various GIS software. Although
the time needed for data conversion can be reduced using
the data conversion tools, the fact that few stand-alone
3D GIS software exist for the tunneling industry still
makes it difficult to consider the quality control and
quality assurance when converting data among the
various software. In addition, the commercial geological
modeling software packages were usually developed as
general tools for geological and geotechnical engineering
and therefore more specific functionalities such as rock
mass classification, fault zone analysis, etc., are not
available to support tunnel design work.
This study presents a new extension, Tunneling Analyst
(TA), that has been developed in ArcScene 3D GIS software,
part of the ArcGIS software package, to improve 3D GIS
functionality for the tunneling industry. The multiple
indicator kriging method is used for estimating a
3D distribution of RMR values of the tunneling site from
borehole and geophysical exploration data. Modeling a
3D discontinuity plane is performed to generate a fault
plane in GIS, and 3D spatial queries are carried out to
identify tunnel sections with difficult tunneling conditions
due to the fault. This paper describes the concept and
details of the TA development and its application to the
Daecheong tunneling project in Korea, supporting the
tunnel design work.
2. Approaches for 3D subsurface characterization in GIS
TA includes two major approaches for characterizing
the 3D subsurface along a preliminarily proposed tunnel
alignment: (1) a volumetric modeling approach to
estimate the rock properties and (2) a geometric modeling
approach to consider any discontinuity, such as a fault.
The volumetric model describes an object by decompos-
ing the object into regular cubic cells. Therefore, it consists
of attribute records for all the blocks (e.g., voxels) making
up a space of interest (Raper, 2000). In contrast, the
geometric model describes an object by determining the
object’s boundary surfaces using a composition of math-
ematically defined surface pieces. Volumetric and geo-
metric modeling approaches can be compared with each
other in several aspects (Table 1).
2.1. Multiple indicator kriging for estimating
a 3D distribution of RMR values
To consider the rock properties in the subsurface, the
volumetric modeling approach should be used with
 ARTICLE IN PRESS
1324 Y. Choi et al. / Computers & Geosciences 35 (2009) 1322–1333

Table 1 random variable Z(x) conditioned by the prior information

Comparison between geometric model and volumetric model (based on available. The posterior CDF of the unknowns can be
previous works by Kavouras, 1992; Marschallinger, 1996; Raper, 2000)
defined by N known values as follows:
Characteristics Volumetric model Geometric model F x ½zjN ¼ probfZðxÞpzjInformationðNÞg (1)
Emphasis Defining and Defining and Indicator kriging is a method that can calculate the
representing variations representing the posterior CDF at the location x as a linear combination
in internal properties of bounding surfaces
objects of 3D objects
of weighted indicator datum I(xk: z), defined at N known
Variables Continuous variables Discrete variables values Z(xk) with the same threshold value z (see Eqs. (2)
Comparison to Analogous to raster- Analogous to vector- and (3)).
2D GIS based representations based representation
in 2D GIS in 2D GIS X
N
Advantage Easy and efficient to Easy to visualize F x ½zjN ¼ lk ðx : zÞ  Iðxk : zÞ (2)
perform Boolean objects by various k¼1
operations and volume degrees of surface
computations smoothness 0 if ðZðxk Þ4zÞ
Easy to store and More attractive for Iðxk : zÞ ¼ (3)
visualize the results of visual perception 1 if ðZðxk ÞpzÞ
3D spatial interpolation
techniques Multiple indicator kriging uses NT increasing threshold
Disadvantage Large storage Difficult to represent values zt in a variability interval (a,b) of z(x) (Eq. (4)) and
requirement for high- fragmented objects determines the posterior CDF by combining NT indicator
resolution model and variations in
kriging estimates.
internal properties
Not attractive for visual Difficult to perform zt ; t ¼ 1; 2; :::; NT with aoz1 pz2 p    pzNT ob (4)
perception due to Boolean operation
‘‘jaggy’’ approximations between two objects In the rock mass classification system based on RMR
values, the variability interval of z(x) can be defined from
0 to 100, and four increasing threshold values (i.e., 20, 40,
60, 80) can be applied to represent the boundary values of
appropriate 3D geostatistical estimation techniques. If rock mass classes.
both an interpolation and an extrapolation are desired, Based on the You (2003) method, this study designed
only kriging and the inverse distance weighting (IDW) the procedures of multiple indicator kriging for estimating
method can be used easily for 3D problems (Ruhaak, a 3D distribution of RMR values from boreholes and
2006). Houlding (2000) presented an application of the geophysical exploration data (Fig. 1). After setting the
3D kriging method for characterizing the subsurface of a dimensions, both the measured RMR values imported
tunneling site. He estimated a 3D distribution of rock from boreholes and the measurements (e.g., electric
quality designation (RQD) values from several boreholes resistivity, seismic velocity) from geophysical exploration
using the 3D ordinary kriging method. Although boreholes data should be transformed into multiple indicator data
can provide quantitative and direct information for according to the four threshold values. The measured RMR
evaluating rock mass classes, in tunnel design practice values from boreholes can be directly compared with the
the number of boreholes available in a tunneling site is four threshold values; therefore, indicator transforma-
usually limited due to the investigation cost. In other tions can be easily performed using Eq. (3), and the
words, much money has to be invested to acquire indicator datum has a value of 0 or 1. For example, the
boreholes dense enough to resolve most uncertainties indicator datum at the threshold value 60 will be
for site characterization. To overcome this limitation, You determined as 1 if the measured RMR value is 50. This
(2003) suggested the 1D multiple indicator kriging means that the probability is 100% that a rock mass with a
method to estimate RMR values along a proposed tunnel RMR value of 50 will be classified into Classes III to V. In
alignment by combining boreholes and geophysical contrast, the electric resistivity or seismic velocity values
exploration data. Although the information derived from obtained from geophysical exploration cannot be directly
geophysical exploration data, such as electric resistivity compared with the four threshold values; therefore
and seismic velocity, is qualitative and an indirect means another approach is required for the indicator transforma-
of evaluating rock properties, it can cover the entire tions. Choi (2007) suggested a method for the indicator
proposed tunnel alignment including both drilled and transformations of geophysical exploration data by
undrilled sections. By using the multiple indicator kriging considering the correlation between the measured
method, boreholes and geophysical exploration data can RMR values and measurements from geophysical explora-
make up for their weak points, and then the uncertainty of tion at each tunneling site. Table 2 shows an example of
estimated rock properties, especially at the undrilled indicator transformations according to the method of Choi
sections, can be decreased (You, 2003). (2007). By considering the frequency distribution of the
The concept of multiple indicator kriging for estimat- measured RMR values at each class of electric resistivity,
ing RMR values originated from the posterior Cumulative cumulative relative frequency at each threshold value can
Distribution Function (CDF) of unknowns proposed by be calculated and then can be used as an indicator datum.
Journel (1986). Under the posterior CDF, the unknown For example, an electric resistivity of 1000 will be
value z(x) is characterized by the distribution of the model transformed to an indicator datum of 0.27 at the threshold
 ARTICLE IN PRESS
Y. Choi et al. / Computers & Geosciences 35 (2009) 1322–1333 1325

value 60, according to the transformation rules in Table 2. smooth the indicator kriging estimates by rounding to the
The indicator datum transformed from geophysical nearest admissible value (Solow, 1986). The probability P
exploration data can be a real number from 0 to 1. that a rock mass is classified into a specific class in the
After the indicator transformations, all indicator data RMR classification system can be calculated as follows.
from both boreholes and geophysical exploration should 0
Class I ¼ 1  CDFðwhere zt ¼ 80Þ
be combined and used as input values to the indicator
B Class II ¼ CDFðwhere z ¼ 80Þ  CDFðwhere z ¼ 60Þ
kriging at each threshold value. The posterior CDF at B t t
B
unknown points can be approximated by the indicator P¼B B Class III ¼ CDFðwhere z t ¼ 60Þ  CDFðwhere z t ¼ 40Þ
kriging estimates at four threshold values. The four B Class IV ¼ CDFðwhere z ¼ 40Þ  CDFðwhere z ¼ 20Þ
@ t t
indicator kriging estimates that are used for characteriz- Class V ¼ CDFðwhere zt ¼ 20Þ
ing the posterior CDF should satisfy the following two
(5)
conditions: (1) the indicator kriging estimates cannot be
less than 0 or greater than 1, and (2) the indicator kriging Finally, the RMR value at unknown points can be
estimates to a lower threshold value cannot be greater estimated as follows:
than the indicator kriging estimates to a higher threshold
RMR ¼ 90  PðClass IÞ þ 70  PðClass IIÞ þ 50  PðClass IIIÞ
value. However, in practice, inadmissible estimates may
occur; these are called order relation problems (Solow, þ 30  PðClass IVÞ þ 10  PðClass VÞ (6)
1986; You, 2003). When order relation problems occur, a
simple but inexpensive method to correct them is to

2.2. An approach to modeling a 3D discontinuity plane

A discontinuity such as a fault or joint is a critical geo-

Fig. 1. Procedures of multiple indicator kriging for estimating 3D
distribution of RMR values.
object in evaluating the stability of a rock mass concern-
ing orientation. Especially in a tunnel design project, the
stability of a rock mass that will be excavated is usually
evaluated by considering the angle between a proposed
tunneling direction and the dip directions of disconti-
nuities in the rock mass (Bieniawski, 1989; Choi and Park,
2004) and the fractured zones generally created by faults
(Waltham, 1994). There have been many studies to
develop an effective method for modeling faulted surfaces
with simple to complex geometry in 3D GIS (Egan et al.,
1999; Liu et al., 2002; Wu and Xu, 2003; Wu et al., 2005;
Sprague and De Kemp, 2005; Frank et al., 2007). Liu et al.
(2002) discussed the spatial function fitting problems of
discontinuity planes with the constraints of position and
dip/dip direction, and remarked that the simplest fitting
function in modeling a fault surface is a clear plane,
because faults can be approximately treated as clear
planes in most geological and geotechnical engineering
practice. This study adopted the simplest approach and
considered the discontinuity as a clear plane that can be
generated from constraints of position and dip/dip
direction.
Both the position of a measuring point and the dip/dip
direction (or strike/dip) of a discontinuity are usually
recorded on the 2D map after field-based structural
measurements. As shown in Fig. 2, dip direction can be

Table 2
An example of indicator transformations of geophysical exploration data at boundary values of rock mass classes in a RMR classification system

Electric resistivity (O m) Frequency Cumulative relative frequency (indicator datum)

Class V Class IV Class III Class II Class I Total z ¼ 20 z ¼ 40 z ¼ 60 z ¼ 80

0–800 2 6 8 22 6 44 0.05 0.18 0.36 0.86

800–1600 0 5 11 27 17 60 0.00 0.08 0.27 0.72
1600–2500 0 1 5 21 38 65 0.00 0.02 0.09 0.42
2500–3400 0 0 2 14 5 21 0.00 0.00 0.10 0.76
3400–16320 0 0 0 6 12 18 0.00 0.00 0.00 0.33
 ARTICLE IN PRESS
1326 Y. Choi et al. / Computers & Geosciences 35 (2009) 1322–1333
measured clockwise from north, and the dip is measured
down from the dip direction vector. The normal vector,
(Ux, Uy, Uz), which is perpendicular to the discontinuity
plane, can be calculated from the dip/dip direction as
follows:
U x ¼ sinðbÞ  sinðaÞ
U y ¼ sinðbÞ  cosðaÞ where a ¼ dip direction; b ¼ dip
U z ¼ cosðbÞ
The equation of the discontinuity plane can be obtained
from the normal vector and the coordinates of the
measuring position in 3D.
U x ðX  X 0 Þ þ U y ðY  Y 0 Þ þ U z ðZ  Z 0 Þ ¼ 0
where X0, Y0, and Z0 are the coordinates of the measuring
point in Fig. 2.
To visualize the discontinuity plane defined by Eq. (8)
in 3D GIS, a Triangular Irregular Network (TIN) can
generated from some points on the discontinuity plane
that are usually defined at the boundaries of interesting
areas. Because the X, Y coordinates of points at the
boundaries of interesting areas are already managed in
GIS, the Z coordinate of the points on the discontinuity
plane can be easily calculated by putting the X,
Y coordinates of the points into a plane equation (Eq. (8)).
Fig. 2. Orientation of a discontinuity: spatial relationship between
measured values and normal vector in XYZ space.
Fig. 3. Procedures for generating a 3D discontinuity plane from a
structural measurement.
Fig. 3 illustrates the procedures for modeling a
3D discontinuity plane in this study.
3. 3D spatial query functions for analyzing fault zones
Query is an essential functionality in GIS to select and
investigate spatial data including geo-objects. The follow-
ing examples illustrate possible queries in GIS to support
tunnel design work:
(7)
 ‘‘Select all faults and joints that are pierced by a
proposed tunnel alignment.’’ This query may be used
to search for discontinuities that need consideration of
the angle between a proposed tunneling direction and
(8) their dip direction. It can also indicate where the fault
will appear on the tunnel alignment during excavation.
 ‘‘Select the set of tunnel stations within 20 m of the
intersection between a proposed tunnel alignment and
a fault plane.’’ This type of query can be used for
identifying tunnel sections with unstable ground
conditions such as a fracture zone that needs strong
supporting systems.
These queries can be carried out in 2D GIS if the faults
are simplified to 2D straight lines on the map. However, as
mentioned above, 2D approaches for interpreting the geo-
objects in the subsurface can lead to significant errors due
to their over simplification. 3D spatial query functions in
GIS can lead to new insights about the complex geological
situations in the subsurface and can allow the more
systematic interpretation of relationships among geo-
objects (Apel, 2006).
Some authors have recently examined spatial query
functions in 3D GIS (Apel, 2006; Sprague et al., 2006; Ellul
and Haklay, 2006). Although the classification of spatial
queries required in 3D GIS was different in each case, all
authors considered common spatial functions such as
3D intersection and 3D buffer. This study also considered
both 3D intersection and 3D buffer, because these are key
spatial query functions needed to perform the above two
query examples in a tunnel design project.
3.1. 3D intersection
The 3D coordinates of the intersection, (Xi, Yi, Zi),
between a proposed tunnel alignment and a fault plane
can be determined using following equations:
U x ðX 0  X 1 Þ þ U y ðY 0  Y 1 Þ þ U z ðZ 0  Z 1 Þ
t¼ (9)
U x ðX 2  X 1 Þ þ U y ðY 2  Y 1 Þ þ U z ðZ 2  Z 1 Þ
1
X i ¼ ðX 2  X 1 Þt þ X 1
Y i ¼ ðY 2  Y 1 Þt þ Y 1 C
A (10)
Z i ¼ ðZ 2  Z 1 Þt þ Z 1
where X0, Y0 and Z0 are the coordinates of a measuring
point on the fault plane; Ux, Uy and Uz are the normal
vectors of the fault plane (see Eq. (8)); X1, Y1 and Z1 are the
coordinates at the initial point of tunnel alignment; X2, Y2
and Z2 are the coordinates at the terminal point of tunnel
alignment.
 ARTICLE IN PRESS
Y. Choi et al. / Computers & Geosciences 35 (2009) 1322–1333
There can exist no intersection between a proposed
tunnel alignment and a fault plane if the calculated
coordinates of the intersection, (Xi, Yi, Zi), are not in the
interval between (X1, Y1, Z1) and (X2, Y2, Z2). Therefore, this
rule can be used to search appropriate results about the
first query example. It should be noted that, in the case of
a curved tunnel alignment, the entire alignment should be
divided into segments short enough to be approximately
treated as straight lines, and then the intersections should
be repeatedly queried with respect to all segments.
3.2. 3D buffer
In 2D GIS, the buffer function creates a planar zone that
falls within a specific distance from an object. The concept
can be expanded into 3D GIS as shown in Fig. 4. In
particular, a plane generates a volume that has two faces
parallel to the plane using a 3D buffer function (Fig. 4(c)).
This type of 3D buffer should be considered to analyze the
fractured zones due to a fault, because it can specify the
boundaries to which a planar geo-object has influence.
If a proposed tunnel alignment and a fault plane
intersect at a point, (Xi, Yi, Zi), the tunnel stations, S, within
a specific distance D from the fault plane, can be
determined as follows:
D multiple indicator kriging method. The interface was
l¼ (11)
sinðjof  ot jÞ  sinðjyf  yt jÞ
SIB ¼ fSjX i  lpSx pX i þ lg (12)
Fig. 4. Conceptual representations of buffer functions in 3D GIS: (a) a sphere as a 3D buffer zone of a point; (b) a cylinder for a line; (c) a volume for a
plane and (d) a volume enlarged by a specific distance from an original volume.
1327
where of is the dip direction of the fault plane; ot is the
direction of a line perpendicular to the tunnel alignment
on the XY plane; yf is the dip of the fault plane; yt is the
dip of the tunnel alignment; Sx is the X coordinate of a
tunnel station; SIB is the set of tunnel stations in the 3D
buffer zone.
4. Implementation of the extension program in GIS
A 3D GIS extension, TA, was developed in the ArcScene
environment with ArcObjects, which is a developer kit for
ArcGIS based on Component Object Model (COM). The
functionalities of ArcScene can be extended using COM-
based programming languages such as VBA, Visual Basic,
C++, C#, and so on. This study selected VBA as a
programming language for the extension development,
because it can interact well with core libraries of ArcGIS
for data management and visualization.
TA is an extension in the form of a toolbar that is
tightly integrated within the ArcScene environment
(Fig. 5(a)). The toolbar of TA includes four tools to improve
the 3D GIS functionality to characterize the subsurface in
geological and geotechnical engineering and to support
the tunnel design work. Fig. 5(b) shows the graphic user
interface of the 3D RMR estimation tool based on the
designed according to the sequential procedures in
Fig. 1. Borehole, geophysical exploration data (i.e., electric
resistivity or seismic velocity), and a Digital Elevation
 ARTICLE IN PRESS
1328 Y. Choi et al. / Computers & Geosciences 35 (2009) 1322–1333

Fig. 5. Tunneling Analyst extension: (a) toolbar in ArcScene environment; (b) graphic user interface (GUI) for 3D RMR Estimation tool; (c) GUI for
3D Discontinuity Plane tool; (d) GUI for 3D Intersection tool and (e) GUI for 3D Buffer tool.

Model (DEM) are the primary input data for this tool. The
RMR measured from boreholes and the measurements
from geophysical exploration should be defined as
3D point features in shapefiles or in a geodatabase, the
standard data formats in ArcGIS, and should contain the
property values in an attribute table attached to 3D point
features. The indicator data at four threshold values can be
automatically generated according to the correlation
between the measured RMR and geophysical exploration
data. TA provides three options for selecting an appro-
priate method for estimating the posterior CDF of
indicator data (i.e., nearest neighbor, IDW, kriging);
therefore the user can choose one of the methods. Finally,
a 3D distribution of RMR values can be estimated
according to the project dimensions setting. Before storing
the final result, meaningless estimated values above the
terrain surface can be filtered out by comparing their
elevations with the DEM.
The 3D discontinuity plane tool (Fig. 5(c)) requires
the structural measurement points to be represented as
3D point features with property values (i.e., dip/dip
direction). After selecting a structural measurement point
layer from the dropdown box or directly inputting the
position and dip/dip direction of the discontinuity, the
 ARTICLE IN PRESS
Y. Choi et al. / Computers & Geosciences 35 (2009) 1322–1333
user should define the project dimensions (i.e., min and
max of X, Y and Z coordinates of the area of interest),
where the 3D discontinuity plane will be generated. The
defined min and max of X, Y and Z coordinates can make
12 combinations of coordinates with two elements as
follows: (Xmin, Ymin); (Xmin, Ymax); (Xmin, Zmin); (Xmin, Zmax);
(Xmax, Ymin); (Xmax, Ymax); (Xmax, Zmin); (Xmax, Zmax); (Ymin,
Zmin); (Ymin, Zmax); (Ymax, Zmin); (Ymax, Zmax). These
combinations are then input to Eq. (8) for generating
some 3D point features on the discontinuity plane. Only
3D point features generated inside the defined dimensions
are used to create a TIN surface that represents the 3D
discontinuity plane.
TA also provides two spatial query tools, which
are specially developed for tunneling in 3D GIS: the
3D intersection tool (Fig. 5(d)) and the 3D buffer tool
(Fig. 5(e)). These tools commonly need both a proposed
tunnel alignment divided into small straight line seg-
ments and structural measurements of discontinuities.
The straight line segments of the tunnel alignment should
be defined as 3D line features with X, Y and Z coordinates
of initial and terminal points and with property
values such as station ID, length of section, etc. The
3D intersection tool repeatedly calculates the coordinates
of intersection points on the discontinuity with respect to
all straight line segments, and then displays the station
IDs of tunnel sections that have the intersection points
within the project dimensions. To query the tunnel
sections that exist in the fault zone, users can define the
buffer distance at the 3D buffer tool based on the
experiences from site investigation. As shown in Fig. 6,
six kinds of spatial relationship are possible between a
fault zone and a tunnel section. We assumed that a
tunnel section is in the buffer zone if any part of the
tunnel section meets any boundaries of the buffer zone
qor the fault plane (i.e., (b)–(e) in Fig. 6). This assumption
was implemented in the 3D buffer tool. After performing
the 3D buffer query, only the tunnel sections within
the fault zone are visualized in ArcScene. It is useful for
quickly identifying the weak zones that need extra care
during excavation.
Fig. 6. Possible spatial relationships between a tunnel section and a
buffer zone of fault plane.
1329
5. Application
The Daecheong tunneling site, covering 0.17 km2
(1135 m  150 m) in Korea, was selected as a study area
for the application of TA extension (Fig. 7). The Daecheong
tunnel design project involved the design of twin tunnels
with diameters of 13.4 m and a length of 1.1 km under a
mountainous environment. The geology in the study area
consists of 0–4 m thick layers of colluvium followed by a
1–12 m granitic residual soil layer. Underlying the soil
layer is a 1–12 m completely weathered granite layer
followed by moderately to slightly weathered granite
layer. In site investigations, 17 boreholes were drilled and
then interpreted by geologists to determine the measured
RMR values with their 3D coordinates. 2D distributions
of seismic velocity (m/s) were also obtained by
seismic refraction survey along the two-tunnel alignments
(XZ planes) and five perpendicular lines at the initial
part of the tunnel alignment (YZ planes). It was found
that a fault plane with dip/dip direction of 40/220 exists
near the initial part of the tunnel alignment. The thickness
of the fault zone was evaluated as approximately 5–10 m.
To consider the topography, a DEM with 5 m resolution
was generated from a digital topographical map with a
contour interval of 1 m, downloaded from NGIC. Fig. 7
shows the locations of boreholes and the distribution of
log lines for the seismic refraction survey. It also includes
both 2D representations of the fault plane and proposed
tunnel alignments.
The data obtained from the field survey were recorded
in several spreadsheets and then converted to shapefiles
having 3D coordinates. Because ArcScene provides
3D visualization functionality with the Symbol Property
Editor (Fig. 8), users can change the symbols of 3D point
features to cylinder, cube and others to improve the visual
effects. Fig. 9 shows a 3D view of the measured RMR
values from boreholes and the distribution of seismic
velocity (m/s) in the study area.
The rock masses near the proposed tunnel alignments
should be classified into five general types (Classes I–V)
based on the RMR classification system to develop some
main support patterns for tunnel design. We used the 3D
RMR estimation tool in TA to estimate the 3D distribution
of RMR values in the study area. The project dimensions
were defined as 5070–6205 m with 5 m spacing in an
X-direction, 0–150 m with 5 m spacing in the Y-direction
and 100–280 m with 5 m spacing in the Z-direction, and
3D kriging was selected as a method for estimating the
posterior CDF of indicator data. Fig. 10 illustrates the
3D distribution of estimated RMR values and the proposed
twin tunnel alignments at the Daecheong tunneling site.
The block models at the lower part in the Y-direction were
cut and removed only for visualization purpose.
To design additional support patterns to cope with
sections with difficult tunneling conditions, the fault zone
was analyzed using other tools in TA. After setting
dimensions as 5070–5500 m in an X-direction, 0–100 m
in a Y-direction and 70–200 m in a Z-direction, a TIN
surface that represents the fault plane near the initial part
of the tunnel alignment could be created using the
3D discontinuity tool (Fig. 11(a)). We also queried the
 ARTICLE IN PRESS
1330 Y. Choi et al. / Computers & Geosciences 35 (2009) 1322–1333

Fig. 7. Thematic map of study area. Highest elevation in study area is 278 ML (i.e., 278 m above see level), and lowest is 29 ML.

Fig. 8. Symbol Property Editor in ArcScene. This tool can allow users to change symbols of 3D features, including both shape and colors, to improve
visual effects.

Fig. 9. 3D distributions of site investigation data. (a) Measured RMR values from boreholes. (b) Seismic velocity (m/s) from seismic refraction survey.
 ARTICLE IN PRESS
Y. Choi et al. / Computers & Geosciences 35 (2009) 1322–1333 1331

Fig. 10. 3D distribution of estimated RMR values in study area.

Fig. 11. (a) 3D view of generated fault plane and (b) tunnel sections within a 10 m buffer zone of fault plane.

3D intersections between the twin tunnel alignments
and the fault plane, and could obtain the IDs and
X-coordinates of tunnel stations that have the intersec-
tions (i.e., station ID 1 at 5084 m in an X-direction along
the up line and station ID 28 at 5127 m in an X-direction
along the down line). The 3D buffer tool was used for
identifying the tunnel sections within 10 m of the fault
plane (Fig. 11(b)). It was found that tunnel sections located
from 5070 m (initial station) to 5108 m along the up line
and from 5103 to 5151 m along the down line are within
the possible fracture zones that could be created due to
the fault. By summarizing the results from four tools in TA,
we could determine the rock mass classes along the
proposed tunnel alignments, both up and down lines, and
could identify some tunnel sections with difficult tunnel-
ing condition as shown in Fig. 12.
6. Conclusions
This paper presented the development of a 3D GIS-
based tunnel design support system, named Tunneling
 ARTICLE IN PRESS
1332 Y. Choi et al. / Computers & Geosciences 35 (2009) 1322–1333
Fig. 12. Evaluated rock mass classes along top of proposed tunnel alignments. (a) Up line. (b) Down line. Gray box indicates fault zone, which needs extra
care during tunnel construction.
Analyst (TA). TA provides useful modeling and analysis
tools to determine the rock mass classes in the subsurface.
Because TA is fully embedded within the ArcScene
environment, the strong capabilities of ArcScene in
visualization can be completely utilized with TA’s ex-
tended 3D functionalities for supporting routine tunnel
design practice. The results from an application to the
Daecheong tunnel design project in Korea has suggested
that TA is efficient in bringing the following remarkable
advantages: (1) boreholes and geophysical exploration
data can be used complementarily to estimate a
3D distribution of RMR values by multiple indicator
kriging. This can decrease the uncertainty of estimated
RMR values, especially at the undrilled tunnel sections
that occupy the major portion of proposed tunnel
alignments in a tunnel design project. Although a
comparative study about the estimated RMR values from
3D ordinary kriging that uses only boreholes and from 3D
multiple indicator kriging using both boreholes and
geophysical exploration data is beyond the scope of this
study, the merits of the multiple indicator kriging
approach for rock property estimation has already been
discussed and validated elsewhere (You, 2003; Choi,
2007). (2) True 3D intersection and buffer functions can
be used to analyze the fault zone, which is a critical part in
tunneling. They can improve the reliability of interpreta-
tion about complex spatial relationships between a tunnel
alignment and discontinuity in the subsurface. Because
they have been implemented in TA based on simple
mathematical equations, a complex 3D topological data
structure is not required in GIS for their utilization. (3) TA
allows users to complete entire procedures (i.e., data pre-
processing, volumetric modeling, geometric modeling, 3D
spatial query, visualization, etc.) in a single framework;
therefore tedious tasks for data conversion among various
software can be significantly reduced or are not required.
Although TA can provide important information to
the tunneling engineers, the current version does not
include the analysis of the traces of intersection of a fault
on the tunnel walls due to the limitation of the ArcScene
data structure and data model. In future work, it would be
interesting to extend TA’s functionalities for considering
data acquired from the tunnel sections and walls.
Acknowledgments
This study was supported by the rapid excavation and
safe tunnel research center in the KICT funded by the
Ministry of Land, Transport and Maritime Affairs (Project
No. C105A1020001–07A050200420), the Brain Korea
21 Project, and the Research Institute of Engineering Science,
Seoul National University, Korea. The current version of
TA can be ordered from the corresponding author.
 ARTICLE IN PRESS
Y. Choi et al. / Computers & Geosciences 35 (2009) 1322–1333
References
Apel, M., 2006. From 3d geomodelling systems towards 3d geoscience
information systems: data model, query functionality and data
management. Computers and Geosciences 32 (2), 222–229.
Bieniawski, Z.T., 1989. Engineering Rock Mass Classifications. Wiley, New
York, USA, 272pp.
Chang, Y.S., Park, H.D., 2004. Development of a web-based geographic
information system for the management of borehole and geological
data. Computers and Geosciences 30 (8), 887–897.
Choi, J.Y., 2007. Statistical modelling of discontinuity size distribution
and analysis of groundwater flow in discontinuous rock. Unpub-
lished PhD Dissertation, Seoul National University, Seoul, Korea,
260pp (in Korean with English abstract).
Choi, S.Y., Park, H.D., 2002. Comparison among different criteria of RMR
and Q-system for rock mass classification for tunnelling in Korea.
Tunnelling and Underground Space Technology 17 (4), 391–401.
Choi, S.Y., Park, H.D., 2004. Variation of rock quality designation (RQD)
with scanline orientation and length: a case study in Korea.
International Journal of Rock Mechanics and Mining Sciences 41
(2), 207–221.
Egan, S.S., Kane, S., Buddin, T.S., Williams, G.D., Hodgetts, D., 1999.
Computer modelling and visualization of the structural deformation
caused by movement along geological faults. Computers and
Geosciences 25 (3), 283–297.
Elkadi, A.S., Huisman, M., 2002. 3D-GSIS geotechnical modelling of
tunnel intersection in soft ground: the second Heinenoord tunnel,
Netherlands. Tunnelling and Underground Space Technology 17 (4),
363–369.
Ellul, C., Haklay, M., 2006. Requirements for topology in 3D GIS.
Transactions in GIS 10 (2), 157–175.
Frank, T., Tertois, A.L., Mallet, J.L., 2007. 3D-reconstruction of complex
geological interfaces from irregularly distributed and noisy point
data. Computers and Geosciences 33 (7), 932–943.
Grimstad, E., Barton, N., 1993. Updating of the Q-system for NMT. In:
Proceedings of the International Symposium on Sprayed Concrete,
Fagernes, Norway, pp. 44–66.
Hack, R., Orlic, B., Ozmutlu, S., Zhu, S., Rengers, N., 2005. Three and more
dimensional modeling in geo-engineering. Bulletin of Engineering
Geology and the Environment 65, 143–153.
Houlding, S.W., 1994. 3D Geoscience Modeling, Computer Techniques for
Geological Characterization. Springer, Berlin, Germany, 309pp.
Houlding, S.W., 2000. Practical Geostatistics: Modeling and Spatial
Analysis. Springer, Berlin, Germany, 161pp.
Journel, A.G., 1986. Constrained interpolation and qualitative informa-
tion—the soft kriging approach. Mathematical Geology 18 (3),
269–286.
Kaalberg, F., Haasnoot, J., Netzel, H., 2003. What are the end-user issues?
Settlement risk management in underground construction. In:
Rosenbaum, M.S., Turner, A.K. (Eds.), Characterisation of the Shallow
1333
Subsurface: Implications for Urban Infrastructure and Environmental
Assessment. Springer, Dusseldorf, Germany, pp. 69–83.
Kavouras, A., 1992. A spatial information system with advanced
modeling capabilities. In: Turner, A.K. (Ed.), Three-Dimensional
Modeling with Geoscientific Information Systems. Kluwer Academic
Publishers, Dordrecht, Netherlands, pp. 59–67.
Kimmance, J.P., Bradshaw, M.P., Seetoh, H.H., 1999. Geographical
Information System (GIS) application to construction and geotech-
nical data management on MRT construction projects in Singapore.
Tunnelling and Underground Space Technology 14 (4), 469–479.
Liu, D., Zhang, J., Wang, S., 2002. Constrained fitting of faulted bedding
planes for three-dimensional geological modeling. Advances in
Engineering Software 33 (11–12), 817–824.
Marschallinger, R., 1996. A voxel visualization and analysis system based
on AutoCAD. Computers and Geosciences 22 (4), 379–386.
Ozmutlu, S., Hack, R., 2003. 3D modeling system for ground engineering.
In: Rosenbaum, M.S., Turner, A.K. (Eds.), Characterisation of the
Shallow Subsurface: Implications for Urban Infrastructure and Envir-
onmental Assessment. Springer, Dusseldorf, Germany, pp. 253–260.
Raper, J., 2000. Multidimensional Geographic Information Science. Taylor
& Francis, London, UK, 300pp.
Ruhaak, W., 2006. A Java application for quality weighted 3-d interpola-
tion. Computers and Geosciences 32 (1), 43–51.
Solow, A.R., 1986. Mapping by simple indicator kriging. Mathematical
Geology 18 (3), 335–352.
Sprague, K., De Kemp, E.A., 2005. Interpretive tools for 3-d structural
geological modeling. Part II: Surface design from sparse spatial data.
GeoInformatica 9 (1), 5–32.
Sprague, K., De Kemp, E.A., Wong, W., McGaughey, J., Perron, G., Barrie, T.,
2006. Spatial targeting using queries in a 3-D GIS environment with
application to mineral exploration. Computers and Geosciences 32
(3), 396–418.
Turner, A.K., 2005. Challenges and trends for geological modelling and
visualisation. Bulletin of Engineering Geology and the Environment
65 (2), 109–127.
Waltham, A.C., 1994. Foundations of Engineering Geology. Blackie
Academic & Professional, London, UK, 104pp.
Wu, Q., Xu, H., 2003. An approach to computer modeling and
visualization of geological faults in 3D. Computers and Geosciences
29 (4), 503–509.
Wu, Q., Xu, H., Zou, X., 2005. An effective method for 3D geological
modeling with multi-source data integration. Computers and
Geosciences 31 (1), 35–43.
Yoo, C., Kim, J.M., 2007. Tunneling performance prediction using integrated
GIS and neural network. Computers and Geotechnics 34 (1), 19–30.
Yoo, C., Jeon, Y.W., Choi, B.S., 2006. IT-based tunneling risk management
system (IT-TURISK)—development and implementation. Tunnelling
and Underground Space Technology 21 (2), 190–202.
You, K.H., 2003. An estimation technique of rock mass classes for a
tunnel design. Geotechnical Engineering 19, 319–326 (in Korean with
English abstract).