University of Wollongong

Research Online
Underground Coal Operators' Conference Faculty of Engineering

2010

Global trends in coal mine horizontal stress

measurements
Christopher Mark
National Institute for Occupational Safety and Health

Murali Gadde
Peabody Energy

Publication Details
This conference paper was originally published as Mark, C and Gadde, M, Global trends in coal mine horizontal stress measurements,
in Aziz, N (ed), 10th Underground Coal Operators' Conference, University of Wollongong & the Australasian Institute of Mining
and Metallurgy, 2010, 21-39.

Research Online is the open access institutional repository for the

University of Wollongong. For further information contact Manager
Repository Services: morgan@uow.edu.au.
2010 Underground Coal Operators’ Conference The AusIMM Illawarra Branch

GLOBAL TRENDS IN COAL MINE HORIZONTAL

STRESS MEASUREMENTS

Christopher Mark1 and Murali Gadde2

ABSTRACT: Knowledge of in situ stresses is fundamental to many studies in earth sciences, and coal
mine ground control is no exception. During the past 20 years, it has become clear that horizontal
stress is a critical factor affecting roof stability in underground coal mines. The theory of plate tectonics
and the World Stress Map (WSM) project has been extremely helpful in explaining the sources and the
orientations of the horizontal stresses observed underground. Recently, WSM geophysicists studying
deep-seated stresses have developed a model of how stress magnitudes vary with depth in the crust.
They have devoted relatively little attention to near-surface stresses, however. This paper explores the
relationships between deep-seated and shallow in situ stresses in several of the world’s coalfields,
using a data base of more than 350 stress measurements from underground coal mines. The analysis
indicates that distinct regional trends exist, corresponding roughly to the regional stress fields identified
by the WSM. The paper presents equations for estimating stress magnitudes that were developed by
treating depth and elastic modulus as independent variables in regression analysis. The magnitude of
the horizontal stress increases with depth, at rates that range from 0.8 to 2.0 times the vertical stress,
just as the WSM “critically stressed crust” model predicts. Overall, it seems that the stress regimes
encountered in underground coal mines are closely linked to those that exist deep in the earth’s crust.

INTRODUCTION

As early as the 1940’s, researchers in British coal mines postulated that large horizontal stresses were
responsible for much of the roof damage experienced underground. Philips, (1946) observed that “At
depths greater than 215 m…lateral compressive forces cause fracturing along the laminations of the
roof beds…the lateral compressive forces increase at a greater rate than the vertical compressive
force, and ultimately both forces may be equal.” Once rock mechanics researchers began to measure
in situ stresses, it became clear that in many cases the horizontal stress actually exceeded the vertical,
often by factors of three or more (Dahl and Parsons, 1972; Hoek and Brown, 1978).

A number of theories were proposed to explain the presence of horizontal stress. Two of the earliest
were the “Poisson’s effect” and the “lithostatic stress state.” Both of these theories presumed a static
earth, in which the horizontal stresses were generated in response to the vertical overburden load. The
“Poisson’s effect” model predicted that the horizontal stresses should be about 1/3 of the vertical, while
the lithostatic model predicted that the three principal stresses would be approximately equal (McGarr,
1988).

These static earth theories could not explain two key characteristics of horizontal stress:

 Why the horizontal stress often exceeds the vertical (SV) in magnitude, even at depth, and;
 Why horizontal stresses are typically highly anisotropic, with the major one (SHmax) significantly
larger than the minor (Shmin).

In addition, the Poisson’s effect model suffers from severe theoretical errors, because it implicitly
assumes that soft sediment (or magma) lithifies in the absence of gravity, and then gravity is
instantaneously switched on once the rock has reached an elastic state McGarr, 1988; Zoback and

1
Senior Research Scientist, National Institute for Occupational Safety and Heath
2
Senior Manager (R&D), Peabody Energy.

11– 12 February 2010 21

2010 Underground Coal Operators’ Conference The AusIMM Illawarra Branch

Zoback, 2002). As Hoek (2007) writes, the Poisson’s effect model was “widely used in the early days
of rock mechanics, but proved to be inaccurate and is seldom used today.”1

Fortunately, during the 1970’s earth scientists were constructing a revolutionary new theory with the
breadth and depth to explain all the major tectonic processes observed on the earth. Plate tectonics
describes a dynamic earth, in which the crust of the earth consists of a number of continental plates
that are sliding across the softer rock in the mantle below. Where the plates contact each other, their
different directions of relative movement create large forces that are transmitted across the plate
interiors. The scientists associated with the world stress map (WSM) have used a variety of indicators,
including earthquake focal mechanisms, wellbore breakouts, and hydraulic fracturing stress
measurements, to identify the lithospheric stresses that result from these plate movements. They
found that the state of stress is remarkably uniform over vast regions of plate interiors, and that it is due
to present-day forces, and not due to residual stresses from past tectonic activity (Zoback and Zoback,
2007). Today there is complete consensus within the geophysics community about the general validity
of the trends identified by the WSM.

The WSM has found that at any given location, the stress direction within the crust is typically
consistent from the “upper 2-5 km (1-3 miles), where essentially all of the wellbore breakout and
hydraulic fracturing data come from, down through the lower 5-20 km (3-12 miles), where the majority
of crustal earthquakes occur” (Zoback and Zoback, 2002). However, the WSM has specifically
excluded “near surface” measurements from its database, because they have sometimes found that
there are “marked changes in stress orientations and relative magnitudes with depth in the upper few
hundred meters, possibly related to effects of nearby topography or a high degree of surface fracturing”
(Zoback, 1992). Later, Zoback and Zoback (2002) stated that “only in situ stress measurements made
at depths greater than 100 m are indicative of the tectonic stress field.”

Since most underground coal mining takes place within several hundred meters of the surface, it is
legitimate to ask how relevant the WSM is to underground coal mining. On the one hand, we do know
that topographic features can significantly affect the stresses we observe underground (Molinda et al.,
1991; Hasenfus and Su, 2006). On the other hand, since the near-surface is part of the crust, we
would certainly expect some relationship to the deep-seated stress patterns. The question is, how
closely are the two related? Do we observe the same general trends that the WSM has identified in the
deep crust, or are the topographic and other near-surface effects so powerful that they completely
mask any relationship?

The first part of the answer to this question was provided in the early 1990’s when researchers
compiled a data base of stress measurements from US coal mines. The WSM had identified the
eastern portion of North America as a stable mid-plate region with a consistent ENE horizontal stress
orientation (Zoback and Zoback, 1989). Sure enough, analysis indicated that 75% of the coal mine
stress measurement orientations fell within the NE quadrant (Mark, 1991; see Figure 1). This finding
was particularly meaningful because the data base included measurements made all over the eastern
U.S., by a variety of researchers using a number of different techniques. The observed trends were
highly significant statistically even though no attempt was made to minimize the effects of bad data by
applying a “quality ranking” to the individual measurements. Such quality rankings are normally
considered essential for discerning underlying trends in stress measurement data (Zoback, 1992;
Stacey and Wesseloo, 1998; Lee et al., 2008).

The WSM also defined the stress regime within eastern North America as either strike/slip (where the
magnitude of the vertical stress falls between the two horizontal stresses) or reverse faulting (where the
vertical stress is smaller than both the principal horizontal stresses). Here, the US stress measurement
data was in even better agreement, with the maximum horizontal stress exceeding the vertical 97% of
the time.

1
The most elegant of the static earth theories is the “spherical shell” model proposed by Sheorey (1994), based in
part on earlier work by McCutchen (1982). Sheorey’s model considers stresses arising from the geothermal
gradient in addition to the Poisson’s effect, and it predicts that the horizontal-to-vertical stress ratio may be very
large near the surface but declining with depth. Like all static earth theories, however, Sheorey’s model can
explain neither the near universal anisotropy of the horizontal stress, nor the prevalence of horizontal stresses that
are significantly greater than the vertical deep into the crust.

22 11 – 12 February 2010
2010 Underground Coal Operators’ Conference The AusIMM Illawarra Branch

W E
Western US
Coalfield W E
Illinois Basin
Stress Measurements US Coalfield
Stress Measurements
S

W E

S
Eastern US
Coalfield
Stress Measurements

Figure 1 - World Stress Map of the United States (Zoback and Zoback, 1989) compared with
stress orientations determined from coalfield stress measurements (Mark and Mucho, 1994).
The solid arrows show WSM stress direction and the dotted lines delineate stress

THE WSM AND STRESS ORIENTATIONS IN COALFIELDS AROUND THE WORLD

Stress measurement data bases have now been constructed for a number of the other of the world’s
coalfields. It makes sense to ask how well the predictions of the WSM compare in those cases.

Western US: According to the WSM, the coalfields of Utah, Colorado, Wyoming, and New Mexico fall
within regions of “extension” or normal faulting, where the vertical stress is predicted to be greater than
either horizontal stress (see Figure 1). In contrast to the eastern “midplate” region, the western regions
are active seismically. Stress directions also vary throughout the region.

Mark (1991) analysed stress measurements from 17 Western US mines, and found no significant
regional trends in orientation. The maximum horizontal stress was significantly lower than in the east,
and was approximately equal to the vertical stress in most cases.

Germany and the UK: The WSM defines the stress regime in western Europe, including the coalfields
of Germany and the UK, as a stable mid-plate area subject to uniform NNW maximum horizontal
stress. As in eastern North America, the stress is controlled by plate driving forces acting on the plate
boundaries (Muller et al., 1992). Western Europe is considered to be a strike slip environment, with the
vertical stress as the intermediate principal stress.

A series of 11 hydrofracturing stress measurements were conducted in four German coal mines
between 1989 and 1991 (Muller, 1991). These measurements confirmed that the greatest horizontal
stress was oriented NNW. The maximum horizontal stress was reported to exceed the vertical stress
down to depths of 1200 m.

Cartwright (1997) describes the results from 26 successful overcores at conducted by Rock Mechanics
Technology (RMT) at 16 mine sites in “virgin or near virgin conditions.” The depths of cover ranged
from 300-1000 m. In every overcore the maximum horizontal stress was located in the NW quadrant,
and the vast majority were oriented within a few degrees of NNW (Figure 2). The magnitude of the
stress was approximately equal to the vertical stress, but there was considerable spread.

11– 12 February 2010 23

2010 Underground Coal Operators’ Conference The AusIMM Illawarra Branch

United Kingdom Coalfield

Stress Orientations
N

W E

Figure 2 - World Stress Map of northern Europe (after Muller et al. 1992), compared with stress
orientations determined from UK coalfield stress measurements (after Cartwright, 1997)

Bowen Basin, Queensland, Australia: Australia is considered somewhat unique by the WSM
because the stress orientation varies considerably between different regions of the continent, reflecting
a variety of plate boundary forces rather than the direction of absolute plate motion (Hillis et al., 1999).
In the Bowen Basin coalfields of the central Queensland, the major horizontal stress is consistently
oriented NNE, and the vertical stress is either the minor or the intermediate principal stress. The region
is not seismically active, and evaluation of the available stress measurements found that few were
indicative of faults on the verge of movement.

Nemcik et al. (2005) presented the results of 235 measurements of pre-mining stress made by SCT in
Australian underground mines. About a third of these measurements were conducted in the Bowen
Basin, all at depths of less than 300 m. Nemcik at al. (2005) reported that “the direction of the major
lateral stress was in most cases confined to the N to NE quadrant” (Figure 3). The magnitudes of these
stresses almost always exceeded the vertical, sometimes by factors of 3 or more. Enever and Lee
(2000) drew similar conclusions from another set of stress measurement data.

Sydney Basin, NSW, Australia: The Sydney Basin appears to be the exception that proves the rule.
Early studies of horizontal stress in underground mines there found that stress directions could vary
widely, even from one section of a mine to another (Gale et al., 1984; Gale, 1986). No consistent
regional trend could be observed. As it turns out, the WSM found the same thing (Hillis et al., 1999).
The Sydney Basin is one of Australia’s most seismically active, and stress orientations vary widely
throughout it. It seems that plate margin effects superimposed on the regional stress direction have
resulted in a relatively low anisotropy between the major and minor principal horizontal stresses. As a
result, mild perturbations caused by local effects, such as density contrasts, faults, or major geologic
structures may cause local stress rotations (Hillis et al., 1999). Moreover, as many as 40% of the
stress measurements are indicative of faults at incipient failure. Nemcik et al. (2005) confirmed that
even with a very large data base, consisting of approximately 170 stress measurements from mines all
over NSW, no consistent trends in stress direction emerged.

24 11 – 12 February 2010
2010 Underground Coal Operators’ Conference The AusIMM Illawarra Branch

Bowen Basin
Coalfield
Stress Measurements
N

W E

Figure 3 - World Stress Map of Australia (after Hillis et al. 1999), compared with stress
orientations determined from Bowen Basin coalfield stress measurements
(after Nemcik et al., 2005)

Summary: Hillis (1999) concludes that “the apparent consistency between in situ stress
measurements and seismicity of the Bowen and Sydney Basins suggests that relatively shallow (300-
1000 m) data may be representative of the stress at greater, seismogenic depth.” Our quick tour
around other coalfields of the world leads to the same conclusion, at least with regards to stress
orientation and relative magnitude. In every case, the WSM model provided a reasonably accurate
prediction of (and explanation for!) the typical stress regime that is observed in underground coal
mines. The next question is whether we can also predict the magnitude of the stress using the WSM
model.

THE CRITICALLY STRESSED CRUST1

The “dynamic earth” plate tectonics model implies that lateral forces are constantly being applied to the
brittle upper crust. These forces would continue to build unless there was some mechanism for their
release. That mechanism is failure of the crust itself, through faulting. Decades of research along a
number of lines of evidence have resulted in what is, in the end, a simple but profound model of the
magnitude of the stresses that the crust can carry (Zoback and Zoback, 2002).

The model begins with Anderson’s (1951) classification scheme for relative stress magnitudes in the
earth:

 Normal faulting regions, where Sv>SHmax>Shmin (1)

 Strike slip faulting regions, where SHmax>Sv>Shmin (2)
 Reverse faulting regions, where SHmax>Shmin>Sv (3)

Research has shown that the strength of faults can be adequately described by the Coulomb criterion:

T = Co + uSn (4)

Where T= the shear strength of the fault plane, Co= the fault plane’s cohesion, u= the friction
coefficient, and Sn=the confining stress applied perpendicular to the fault plane.

1
The discussion in this section is based almost entirely on the summary provided by Zoback and Zoback (2002).

11– 12 February 2010 25

2010 Underground Coal Operators’ Conference The AusIMM Illawarra Branch

Using two-dimensional Mohr-Coulomb analysis, the shear stress at failure of an optimally-oriented fault
is a function of the difference between the minor (S3) and the major principal (S1) stresses (Jaeger and
Cook, 1979):

(S1-Po)/(S3-Po) = ((u2 + 1)1/2 +u)2 (5)

Where Po is the pore pressure, and (S-Po) is the effective stress.

Studies have shown that, at depth, the cohesion is much smaller than the frictional component of the
fault strength, the friction coefficient u is typically 0.6-1.0, and the pore pressure is hydrostatic on active
faults (Townend and Zoback, 2000). Assuming u=0.6, the following approximate relationships can be
derived:

 SHmax = 2.3 Sv in reverse faulting regions; (6a)

 SHmax = 1.6 Sv in strike-slip faulting regions (assuming SV = (SHmax+ Shmin)/2,) and; (6b)
 SHmax < Sv and Shmin = 0.6 Sv in extension faulting regions. (6c)

Stress measurements have now been conducted in several deep boreholes to depths of almost 10 km
(6 miles). Figure 4 shows that the measurements confirm the general stress gradients derived above.
In particular, in the mid-plate compressive stress regions where these measurements were made,
horizontal stresses well in excess of the vertical persist far down into the crust, and the horizontal stress
gradient (k) is fairly consistent with a value ranging from approximately 1.3 to 2.

Maximum Horizontal Stress (MPa)

0 50 100 150 200 250 300 350 400
0.0 0

1.0 1

2.0 2
Depth (km)

3.0 3
Depth (km)
4.0 4

5.0 5

6.0 6

7.0 7

8.0 8

Figure 4 - Stress measurements from boreholes deep in the earth’s crust (data
from Townend and Zoback, 2000)

Zoback and Zoback (2002) also state that worldwide research has found “no evidence” that “residual
stresses” from past tectonic events play any role in today’s stressfields. They speculate that if such
stresses exist at all, they can only be important “in the upper few meters or tens of meters of the crust
where the tectonic stresses are small.”

If plate tectonics are responsible for virtually all the stresses measured at depth, and if the critically
stressed crust model allows us to predict those stresses, then it is reasonable to expect that there is
some relationship between the deep crustal stresses and those measured in coal mines. After all, the
“near surface” is part of the same crust! In fact, if we don’t find a relationship, then we have a
significant problem. For instance, if we conclude that the horizontal stress increases less rapidly than
the vertical stress to depths of one or two thousand feet, but we know it exceeds the vertical stress at
greater depths, the implication is that there is a major discontinuity in the stressfield somewhere. Let us
then see what the actual measurements tell us.

26 11 – 12 February 2010
2010 Underground Coal Operators’ Conference The AusIMM Illawarra Branch

OVERCORING STRESS MEASUREMENTS IN COAL MINES

Overcoring has been the most common technique for measuring stress in underground coal mines. In
the US, most measurements have been made using the Bureau of Mines biaxial “borehole deformation
gage” (Bickel, 1993). Internationally, the triaxial ANZI or CSIRO HI cells have been by far the most
common (Mills, 1997; Nemcik et al., 2005; Cartwright, 1997: Lee et al., 2008). Most recently, downhole
wireline stress measurement devices have been developed (Conover et al., 2004).

One important feature of overcoring stress measurements is that interpretation of the data requires the
determination of the rock stiffness (elastic modulus, E). The modulus is not required for the other types
of stress measurement contained in the WSM data base. Studies in the layered sedimentary geology
of coal measure rock have found that the measured stresses in a single hole vary in proportion to the
rock stiffness (Aggson and Mouyard, 1988).

Cartwright (1997) pointed out that the relationships between horizontal stress and depth, like those of
Hoek and Brown (1980), have typically displayed high scatter, particularly near the surface. Within his
data base of UK stress measurements, there was a better correlation between stress and modulus than
between stress and depth. He proposed that the two factors might be combined into a single equation:

SH = B0 + B1 [(v/(1+v)) (Depth)] + B2 (Modulus) (7)

Where B0 is a constant with units of MPa, B1 is a constant with units MPa/m, v is Poisson’s ratio, and B2
is a dimensionless constant called the “tectonic strain factor” or TSF. Regression analysis provided the
following values for the constants, with an r-squared of 0.94:

 B0 = -4.0 MPa
 B1 = 0.009 MPa/m, and
 B2 = 0.78*10-3

Cartwright’s analysis indicated that for his data set, the modulus was more important than the depth for
predicting the maximum horizontal stress.

Dolinar (2003) studied stress measurements from 37 eastern US underground mines, including several
stone mines. His analysis employed a version of equation (7), with B0=0 and B1 fixed at 0.025 MPa/m.
He found that the remaining regression coefficient, the TSF (B2), varied between 400 and 900 for the
different geographic regions studied, with the highest TSF values in two small areas in central
Appalachia.

Nemcik et al. (2005) also calculated the TSF for the large SCT data base of 235 measurements from
Queensland and NSW mines. In their analyses, they also used equation (7), setting B0=0 and
B1=0.025 MPa/m. In contrast to both Dolinar and Cartwright, however, Nemcik found that:

 There was a strong correlation between depth and the SHmax in both NSW and Qld,
 The TSF also increased significantly with depth, averaging 0.4 when the depth was less than
100 m deep, but more than 1.3 for mines at depths exceeding 500 m, and;
 At any given depth, a wide range of TSF values were measured.

Nemcik et al.’s work indicates that the TSF can vary significantly within a single region, and that the
TSF cannot always explain a large proportion of the variation in SHmax.

INTERNATIONAL OVERCORING MEASUREMENTS DATA SET

If the magnitude of the near-surface stresses measured in coal mines were closely related to the deep-
seated stresses measured by the WSM, we might expect to find that:

 The depth at least as important as the modulus in predicting the horizontal stress, though
both factors together should be better still;

11– 12 February 2010 27

2010 Underground Coal Operators’ Conference The AusIMM Illawarra Branch

 The depth gradient should be somewhere between 1.0-1.6 times the vertical stress for
coalfields located in stable, a-seismic mid-plate areas, like those in the eastern US, the UK,
Germany, or central Queensland;
 The depth gradient should be higher in a seismically active compressive regime like the one
found in the Sydney Basin, and it should be lower an active extension regime like the one
found in the western US coalfields.

To test these hypotheses, a data set of 565 stress measurements was compiled. The heart of the data
set is 373 measurements from underground coal mines. The breakdown of these by region is shown in
Figure 5. Approximately two-thirds of the coal data were from Australia, and were provided by SCT.
Preliminary statistical analyses indicated that the four eastern U.S. coalfields could be combined into a
single “eastern US coal” grouping, and that the UK and German data could be combined into a
“northern European coal” grouping.

India (5)
South Africa (22) Australia -
Northern Europe Queensland (64)
(52)

Western US (20)

Eastern US (40)

Australia - New
South Wales (170)

Figure 5 - Locations of the stress measurements included in the coal data set

In addition, about 200 non-coal measurements that were readily available in the literature and at the
WSM website (Reinecker et al., 2005) were collected. The purpose of the non-coal data was to provide
an independent check on the general regional trends observed within the coal data. The non-coal data
set includes stress measurements from the same general regions as the coal data set, though it does
not include any measurements from Australia. The non-coal data also provides an opportunity to
compare stress trends within bedded coal measure strata to those in other geologic settings. However,
it is recognized that while the coal data set is easily the most comprehensive of its kind ever compiled,
the non-coal database is quite small compared to how many measurements could be available.

Preliminary analyses were conducted to see where non-coal data sets could be combined. It was
determined that the measurements from Ireland, the UK, northern Europe, and Scandinavia could be
combined into a single “All Europe” grouping. Similarly, data from the eastern US, Canada, and the
western U.S. were combined into a “North America” grouping. It was surprising that while there were
distinct differences between the western and eastern US coal data sets, the trends within the US non-
coal data did not seem to vary much by region. Figure 6 shows the number of data points within each
regional grouping.

India (16)

South Africa (14)

Northern Europe
(47)

North America (115)

Figure 6 - Locations of the stress measurements included in the non-coal data set

28 11 – 12 February 2010
2010 Underground Coal Operators’ Conference The AusIMM Illawarra Branch

Figures 7 and 8 show the range of depth and modulus within each regional group in the coal data set.
The greatest depths are encountered in the western European mines (UK and Germany), while the
highest modulus rocks are in the eastern US. Queensland has the lowest modulus rock. Note that the
small size of the data set from India (n=5) and the relatively small range of depth in the South African
data set mean that the results from these two regions would be expected to be less reliable than those
from the other regions.

1500
1200
Depth (m)
900
600
300
0

Eastern Western UK & South India NSW Qld

United States Germany Africa Australia

Figure 7 - Range of depths of the coal stress measurements, by region. The middle line
represents the median value, and the upper and lower hinges of the box represent the 75th and
25th percentiles of the data, and the dots are outliers
60
Modulus (GPa)
40
20
0

Eastern Western UK & South India NSW Qld

United States Germany Africa Australia

Figure 8 - Range of elastic modulus of the coal stress measurements, by region

The same types of summary data for the non-coal data set are shown in Figures 9 and 10. The depth
ranges are similar to coal data, except in South Africa where the non-coal data includes a number of
measurements from extremely deep gold mines. In general, the modulus values for the non-coal
measurements are about twice as great as those from the coal data.

11– 12 February 2010 29

2010 Underground Coal Operators’ Conference The AusIMM Illawarra Branch

2400
1800
Depth (m)
1200
600
0

North Northern South Africa India

America Europe

Figure 9 - Range of depths of the non-coal stress measurements, by region

120
Modulus (GPa )
90
60
30
0

North Northern South India

America Europe Africa

Figure 10 - Range of elastic modulus of the non-coal stress measurements, by region

Data Weighting: In order that the coal data not be overwhelmingly influenced by the Australian data,
all the data was weighted by the following formula:

Weight of an individual measurement= 1/ (nR)0.5 (8)

Where nR is the number of measurements from a particular region.

The result is that an individual data point from a region with few measurements will count more heavily
than one from a region with a lot of measurements, but overall the heavily populated regions will still
have more influence. For example, the data set contains 40 measurements from the eastern US and
slightly more than four (4) times as many measurements from NSW. In the weighted analysis, an
individual US measurement is given a weight about two times as great as a measurement from NSW,
but in all the NSW measurement have twice as much influence over the final equation as do all the
eastern U.S. measurements.

30 11 – 12 February 2010
2010 Underground Coal Operators’ Conference The AusIMM Illawarra Branch

As a check, all of the analyses were run using both weighted and unweighted data, and it was found
that the results did not differ significantly. The weighted results will be reported here in this paper.

Depth Gradients and Modulus Effect for the complete data sets: Multivariate regression was
conducted using the statistical package STATA. The first analyses looked at the relationship between
depth and stress, with no other variables. The form of the model is thus:

SHmax = B0 + B1 (Depth) (9)

In this analysis, B1 is the gradient of the maximum horizontal stress with depth (in MPa/m) and the
intercept B0 can be interpreted as the “excess stress” that is not associated with the depth gradient.

The results are shown in Table 1:

Table 1 - Regression results using Equation (9) as the model

Excess Stress B0 (MPa) Depth Gradient B1 MPa/m r-squared

Coal Data 7.3 0.024 0.33
Non-Coal Data 9.6 0.028 0.56
Combined Data Set 8.0 0.026 0.50

In other words, for the combined data set, we can explain half of the variation simply by the depth (and
a constant). All the other variability—regional location, modulus, proximity to the entry and other
measurement errors (Gadde et al., 2006), is only responsible for the other 50%. Although depth is not
as strong a predictor in for the coal data as it is for the non-coal data, the depth gradients are all also
approximately equal to the vertical stress gradient of 0.025 MPa/m. The similarity between the coal
and non-coal equations is striking—these are two completely independent data sets, drawn from similar
parts of the world.

When the analyses were run with just the modulus, instead of depth, the r-squared values were all
reduced. In the case of the coal data, the reduction is only from 0.33 to 0.28, but for the non-coal data
(and the complete data set) the r-squared is reduced from about 0.5 to approximately 0.20. These
results indicate that modulus effect is most pronounced in layered, coal measure geologies, but that
even there, depth explains about as much of the variation in the data as the modulus does.

In the next set of analyses, the effects of modulus and depth are explored simultaneously. The
regression equation that is used is:

SHmax = B0 + B1 (Depth) + B2 (Modulus) (10)

In this model, the excess stress consists of two components, the intercept (B0) and plus the modulus
term (B2*E). Note that with this model, the excess stress is independent of the depth. Statistical
analyses confirmed that there was minimal interaction between the modulus term and the depth within
this data set.

The results in Table 2 show that adding the “modulus factor” improves the r-squared values
considerably, from 0.33 to 0.52 for the coal data, and to above 0.60 for the non-coal data set. Note that
the values for the depth gradient drop slightly but are still close to 0.025 MPa/m. The biggest change
from Table 1, particularly for the non-coal data, is a reduction in the constant. In other words, it
appears that the two elements of the excess stress, the intercept and the modulus term, are closely
related.

It is also worth pointing out that the relative importance of the depth effect and excess stress change
with depth. For example, for a coal mine at 90 m of cover and a modulus of 21 Gap, the total predicted
stress is 11.9 MPa. The excess stress accounts for 9.9 MPa, and the depth effect contributes just 1.8
MPa to this total. If the depth increases to 450 m, the predicted stress is now 18.9 MPa, and all the
increase is due to the depth effect.

11– 12 February 2010 31

2010 Underground Coal Operators’ Conference The AusIMM Illawarra Branch

Table 2 - Regression results using equation (10) as the model.

Intercept Depth Gradient Modulus Factor r-squared

B0 (MPa) B1 (MPa/m) B2(10-3)
Coal Data 0.7 0.020 0.44 0.52
Non-Coal Data 2.0 0.026 0.19 0.62
Combined Data 2.6 0.025 0.21 0.58

Region-by-region analyses: The next step was to determine predictive equations in the form of
equation (10) for each of the individual regions. It is reasonable to expect that different regions might
have different relationships between stress, modulus, and depth, owing to different characteristics of
the crust, tectonic forces, etc. The results of the individual analyses are shown in Table 3.

Table 3 - Regression results for regional subsets of the data, using equation (10) as the model

Intercept B0 Depth Gradient Modulus Factor r-

n
(MPa) B1 (MPa/m) B2(10-3) squared
West U.S. Coal 20 -6.4 0.015 0.62 0.71
UK/Ger Coal 52 -1.7* 0.012 0.51 0.50
South Africa Coal 22 6.1 0* -0.01* 0
India Coal 5 2.6* 0.029 -0.04* 0.79
NSW Coal 170 -4.4 0.040 0.56 0.71
Qld Coal 64 -1.5 0.031 0.34 0.51
U.S./Can Non-Coal 115 -1.9 0.033 0.27 0.70
N. Europe Non-Coal 47 6.3 0.028 0.02* 0.55
South Africa Non-Coal 14 2.3* 0.046* -0.04* 0
India Non-Coal 16 -5.8 0.049 0.26 0.56

* signifies statistic is not significantly different from 0 at the 95% confidence level.

Several observations can be made on the data in Table 3:

 The r-squared for the individual equations are between about 0.50 and 0.70 for nine of the
eleven data sets, which is not significantly better than what was achieved by analysing the
entire data set.
 The most consistent coefficients are the depth gradients (B1), and these are also generally the
most statistically significant coefficients. The range is from about 0.013-0.040 MPa/m for the
larger data sets.
 The modulus factors are also fairly consistent, ranging from about 250 to 600*10-3. However,
in four of the smaller data sets, these coefficients are not significant. Moreover, the modulus
effect is more pronounced in the coal measure rocks.
 The excess stress, determined using the mean modulus value for each region, was quite
consistent, averaging about 7 MPa for the seven largest data sets. In six of those cases, the
average total excess stress (determined using the average modulus for that data set) ranged
from 6.0-9.3 MPa.

Overall, while the equations for some regions (western US, NSW) seem reasonably reliable, in many
cases the small size of the individual data sets is probably responsible for poor correlations.

Unified analyses, controlling for different regional depth gradients: A statistical technique is
available that allows us to combine the power of using the largest possible data set, while

32 11 – 12 February 2010
2010 Underground Coal Operators’ Conference The AusIMM Illawarra Branch

simultaneously allowing for some regional variation. This is accomplished by allowing the coefficients
for the depth gradient to vary region by region1.

The final equations are in the form of equation (10), but while all of them use the same coefficients for
the components of the excess stress, each region has its own depth gradient. The results are shown in
Table 4.

Table 4 - Stress prediction parameters for equation 10 determined for the individual coal
regions using the unified analysis regression technique

Intercept Depth Gradient Modulus Factor

n
B0 (MPa) B1 (MPa/m) B2(10-3)
East U.S. Coal 42 -2.1 0.037 0.41
West U.S. Coal 20 -2.1 0.022 0.41
UK/Ger Coal 52 -2.1 0.020 0.41
South Africa Coal 22 -2.1 0.027 0.41
India Coal 5 -2.1 0.029 0.41
NSW Coal 170 -2.1 0.041 0.41
Qld Coal 64 -2.1 0.031 0.41

The regression analysis used to obtain Table 4 achieves an r-squared of 0.69. Accounting for nearly
70% of all the variation in such a large and diverse data set is an impressive accomplishment. The
following further observations can be made on these results:

 The greatest depth gradient, approaching twice the vertical stress gradient, was found for the
NSW measurements, which corresponds to the prediction based on the WSM;
 The results for the eastern US, Queensland, and the Western US are also in good agreement
with WSM predictions, and;
 The depth gradient for Europe is somewhat lower than was expected based on the WSM.
This may be attributable in part to extensive past mining in the UK and German coalfields
where these measurements were made (Muller, 1991).

The equivalent results for the non-coal data are shown in Table 5. It is significant that all but one of the
regional depth gradients determined for the non-coal data are very similar to the gradients found for the
coal data. The exception is India where the coal data set is very small.

Table 5 - Stress prediction parameters for equation 10 determined for

the individual non-coal regions using the unified analysis regression
technique.

Depth Gradient Modulus Factor

n
B1 (MPa/m) B2(10-3)
U.S./Can Non-Coal 115 0.034 0.20
N. Europe Non-Coal 47 0.022 0.20
South Africa Non-Coal 14 0.021 0.20
India Non-Coal 16 0.033 0.20

*The constant (B0) was not statistically significant at the 95% level.

Figure 11 is a type of “residual plot” that helps evaluate the validity of the regression equation. It
compares the maximum horizontal coal mine stresses predicted by Table 4 with the measured ones for
all regions except Australia. The Australian data appears on Figure 12. Both figures show that the

1
The technique involves creating interaction terms involving dummy variables (Wooldridge, 2006, pp. 244-252). In
this case, a dummy variable is defined based on region, and then that dummy variable is interacted with the
variable “depth.”

11– 12 February 2010 33

2010 Underground Coal Operators’ Conference The AusIMM Illawarra Branch

residuals (discrepancies) are quite evenly distributed for all regions and stress levels, which indicates
that the regression results are valid.

50.0
US-East
US-West
40.0 UK/Germany
Measured Stress (MPa)

SAfrica
India
30.0 prediction

20.0

10.0

0.0
0.0 10.0 20.0 30.0 40.0
Predicted Stress (MPa)

Figure 11 - Residual plot for Table 4, non-Australian data

50.0

40.0
Measured Stress (MPa)

30.0

20.0

NSW
10.0 QLD
prediction

0.0
0.0 10.0 20.0 30.0 40.0
Predicted Stress (MPa)

Figure 12 - Residual plot for Table 4, Australian data only

Figure 13 plots the coal stress measurements against the depth of cover. Two ranges of predicted
stress derived from Table 4 are shown in the figure. The upper limit for the “normal” grouping is based
on the NSW depth gradient, and the lower limit follows the Queensland depth gradient. For the “low
gradient” grouping, the upper limit uses the western US gradient, while the lower limit uses the gradient
found for the UK and German coalfields. In both cases, the upper limits are based on rock with an
elastic modulus of 35 GPa, and the lower limits assume a rock modulus of 7 GPa.

34 11 – 12 February 2010
2010 Underground Coal Operators’ Conference The AusIMM Illawarra Branch

MEASURED STRESS (MPa)

0.0 10.0 20.0 30.0 40.0 50.0
0.0
DEPTH OF COVER (m)

400.0

US-East
US-West
UK/Ger
800.0
SAf
Vertical Stress
NSW
QLD
India
Norm al Gradient
1200.0 Low Gradient

Figure 13 - The coal stress measurements data base, showing predicted ranges of stress for
regions with “normal” and “low” depth gradients (see text for more details)

Analysis Using Modulus and Tectonic Strain Factor (TSF): A number of analyses were conducted
using model represented in equation (7), and employing the TSF concept as defined by Dolinar (2003)
and Nemcik et al. (2005). In these analyses, when only the modulus was used as the dependent
variable, the r-squared values were not much different than those obtained when modulus was
regressed against SHmax (less than 0.30 for the coal data, and less than 0.2 for the non-coal data). In
an alternative analysis, individual TSF values were determined for each region. This analysis found the
highest TSF in NSW, with a value of about 0.85, while the TSF determined for the Queensland, US,
and European coalfields was about 0.45. The r-squared for this analysis was only 0.50, however,
considerably lower than the 0.69 obtained with Table 4.

Analysis of the minimum principal stress, Shmin: No data was available from Australia for analysis
of the minimum principal stress. The results of the analyses for the other regions are shown in Table 6.
The model that was employed is shown in equation 11:

Shmin = B0 + B1 (Depth) + B2 (Modulus) (11)

The same “unified” regression technique used to obtain Tables 4 and 5 was used in the analysis. The
r-squared for the regression using the coal data was 0.54, and it was 0.67 for the non-coal data. The
constant was not statistically significant in either equation.

Table 6 - Stress prediction parameters for equation 11 determined for the individual non-coal
regions using the unified analysis regression technique

Depth Gradient Modulus Factor

n
B1 (MPa/m) B2(10-3)
East U.S. Coal 42 0.030 0.15
West U.S. Coal 20 0.013 0.15
UK/Ger Coal 52 0.009 0.15
South Africa Coal 22 0.005 0.15
India Coal 5 0.009 0.15
U.S./Can Non-Coal 115 0.020 0.12
N. Europe Non-Coal 47 0.011 0.12
South Africa Non-Coal 14 0.010 0.12
India Non-Coal 16 0.023 0.12

11– 12 February 2010 35

2010 Underground Coal Operators’ Conference The AusIMM Illawarra Branch

DISCUSSION

At the beginning of the last section, several predictions were made about the horizontal stress
measurements based on the WSM “critically stressed crust” model. In nearly every instance, the
prediction was confirmed by the analysis:

 The depth was as important as the modulus in predicting the horizontal stress in the coal
mine data set, and it was a much better predictor in the non-coal data set. When both factors
were combined, the accuracy of the predictions improved significantly.
 The calculated depth gradient of 1.6 times the vertical stress for the eastern US, and 1.4
times the vertical stress in the Bowen Basin, was within the range of what was predicted for
coalfields located in stable, a-seismic, mid-plate areas. The depth gradient that was
calculated for the northern European coalfields of the UK and Germany was a little lower than
expected, but even there it was still 0.9 times the vertical stress.
 The greatest depth gradient was found to be in the seismically active compressive regime of
the Sydney Basin, and one of the lowest depth gradients was in the active extension regime
of the western US coalfields.

It is reassuring that the depth gradients that were determined empirically seem to match up well with
those that have been identified in the deep crustal measurements made by the WSM. The findings of
this study indicate that we neither have to look for nor explain a discontinuity in stress at the transition
between the “near surface” and the “deep” crust.1

The study also confirmed that the “excess stress” is one significant component of the near-surface
stress that does not fit directly into the WSM model. The excess stress typically adds 3.5-10.5 MPa to
the horizontal stress that would be predicted by the depth gradient alone. At depths of less than 300 m,
the excess stress is usually responsible for at least 50% of the measured maximum horizontal stress.

Some important characteristics of the excess stress that were observed in the analysis are:

 Its magnitude, averaging about 7 MPa, is quite consistent over a broad range of independent
data sets;
 Its magnitude does not noticeably increase with depth.
 The value of the modulus factor was about half as large in the non-coal data set as it was in
the coal data set. However, since the average modulus in the coal data set is about twice the
average non-coal modulus, the overall excess stress is about the same in both geologic
regimes.
 The modulus factors determined from the analysis of the minimum principal stresses are
consistently about half as large as those calculated for the maximum principal stress. In other
words, the modulus effect is just as anisotropic as the vertical stress gradients.

Taken together, these observations indicate that the source of the excess stress is likely the same
current tectonic regime that is responsible for the other aspects of today’s stressfields, not “residual
stress” or some other cause that would likely vary from locality to locality. One possible explanation is
that while the “critically stressed crust” model normally assumes that the cohesion on fault planes is
very small relative to the stresses at depth, near the surface that assumption may not be valid (Zoback
et al. 2003). The cohesional strength of critically stressed faults may be enough to allow the crust to
carry 7 MPa of stress more than would otherwise be the case.

1
It is also reassuring that other researchers working with mining stress measurement data appear to reaching very
similar conclusions. Lee et al. (2008) present results from an analysis of 258 high-quality, three-dimensional stress
measurements from Australian metal mines. Stress gradients well in excess of the vertical are apparent to depths
of 1.6 km. Lee et al. (2008) also show that the ratio of each principal stress to the first invariant (S1+ S2+S3) is
relatively constant, an observation Windsor (2007) confirmed independently working with another hard rock mining
data set. Harrison et al. (2007) concluded that this last observation could be explained by a Mohr-Coulomb
relationship between S1 and S3, and is “entirely consistent with the hypothesis that the Earth’s crust is in a state of
limiting equilibrium.”

36 11 – 12 February 2010
2010 Underground Coal Operators’ Conference The AusIMM Illawarra Branch

One further observation is that the modulus factor was more important to predicting the stress in the
coal mine data set than it was in the non-coal data. In fact, for the coal data, the modulus was almost
as powerful a predictor of the stress as the depth. It makes sense that in layered, sedimentary coal
measure strata there would be a strong tendency for the stiffer rock units to carry more load. On the
other hand, the horizontal strain within the rock is certainly not, in general, constant regardless of depth,
and it may not even be fully equalized at any given stress measurement location.

While the stress prediction equations that were developed during this study are quite robust statistically,
it should be recognized that there are significant unexplained variability in the data. This is most clearly
evident near the surface in mountainous terrain, where topographic effects are likely to be substantial.
In situ stresses are also much less predictable in seismically active coalfields like NSW and the
Western U.S. Site specific stress measurements are still the only technique that can provide assurance
of the local stress conditions.

A final comment is that there now seems to be no justification for employing the “Poisson’s effect” or
other static earth theories in any aspect of the analysis of in situ stress. The horizontal stresses
measured underground are caused by the large plate tectonic forces that are currently being carried by
the earth’s crust. Poisson’s effect can play a role when loads are actively applied to the ground, as in a
longwall tailgate (Frith and Colwell, 2006), but there is no theoretical or empirical basis for using it to
explain any aspect of the in situ stresses that develop over geologic time.

CONCLUSIONS

Using the largest data base of in situ stress measurements from coal mines ever assembled, this study
found that the data fits the “critically stressed crust” model based on plate tectonics theory surprisingly
well. Past studies had already shown that plate tectonics explained the orientation of the stresses
observed underground, now it seems it explains their magnitudes as well. In particular, the maximum
horizontal stress was found to increase with depth at rates of approximately 0.8-2.0 times the vertical
stress. The study also identified a second component of the measured stress, called the “excess
stress,” which adds approximately 3.5-10.5 MPa to the total. The excess stress is apparently
independent of the depth.

The study makes a major contribution by providing equations for estimating the maximum horizontal
stress in several major coalfields around the world. More importantly, the study provides a framework
for understanding the source of the horizontal stresses encountered underground every day. By
showing that these “near surface” stresses are tied to those found in the deep crust, it links the worlds
of rock engineering and geophysics. Hopefully, future research can build on this foundation to develop
better tools for understanding and predicting in situ stress, and for using those predictions to design
safer underground structures.

ACKNOWLEDGEMENTS

The authors would like to express their sincere appreciation to Dr. Winton Gale for making the Strata
Control Technologies (SCT) stress measurement data base available for this analysis. The SCT data
base is undoubtedly the most extensive of its kind in the world, and its inclusion has added immensely
to the value of this paper. We would also like to thank Rock Mechanics Technology (RMT) for making
a large portion of their very impressive world-wide data base available to other researchers via their
website. Thanks are also due to Russel Frith, Hamid Maleki, Dennis Dolinar, and Leo Gilbride for their
contributions to the data base.

REFERENCES

Aggson, J R and Mouyard, D P, 1988. Geomechanical evaluation of a coal mine arched entry, in
International Journal of Mining Geological Engineering, Vol. 6, pp 185-193.
Anderson, E M, 1951. The dynamics of faulting. Edinburgh: Oliver and Boyd.
Bickel, D L, 1993. Rock stress determinations from overcoring—an overview. USBM Bulletin 694.
Cartwright, P B, 1997. A review of recent in-situ stress measurements in United Kingdom coal
measures strata, in Proceedings of the International Symposium on Rock Stress, Kumamoto,
Japan, pp 469-474.

11– 12 February 2010 37

2010 Underground Coal Operators’ Conference The AusIMM Illawarra Branch

Conover, D, Gray, I and Vandergrift, T, 2004. Downhole overcoring stress measurements at a western
underground coal mine, in Proceedings of the 23rd International Conference on Ground Control in
Mining, pp 86-91.
Dahl, H D and Parsons, R C, 1972. Ground control studies in the Humphrey No. 7 Mine, Christopher
Coal Div., Consolidation Coal Company, in Transactions, Vol. 252, SME-AIME, pp 211-222.
Dolinar, D R, 2003. Variation of horizontal stresses and strains in mines in bedded deposits in the
eastern and midwestern United States, in Proceedings of the 22nd International Conference on
Ground Control in Mining, pp 178-185.
Enever, J R and Lee, M F, 2000. On the prediction of rock stress, in Proceedings of GEOENG,
Melbourne.
Frith, R and Colwell, M, 2006. Why uniaxial compressive strength and young’s modulus are potentially
poor indicators of roadway roof stability—except in the tailgate, in Proceedings of the 25th
International Conference on Ground Control in Mining, pp 198-207.
Gadde, M M, Peng, S S and Rusnak, JA, 2006. A critical analysis of overcoring stress measurements
from U.S. Underground Coal Mines, in SME Transactions 320:25-30.
Gale, W J, 1986. The application of stress measurements to the optimization coal mine roadway
driveage in the Illawarra coal measures, in Proceedings of the International Symposium on Rock
Stress and Rock Stress Measurements, pp 551-560.
Gale, W J and seven others, (1984). An investigation of the stress field and its effect upon mining
conditions at Tahmoor Colliery, NSW, Australia. CSIRO Report No. 49.
Harrison, J P, Hudson, J A and Carter, J N, 2007. Is there a relationship between the in situ principal
stress magnitudes in rock masses?, in Proceedings of the 1st Canada-U.S. Rock Mechanics
Symposium.
Hasenfus, G J and Su, D W H, 2006. Horizontal stress and coal mines: twenty-five years of experience
and perspective, in Proceedings of the 25th International Conference on Ground Control in Mining,
pp 256-267.
Hillis, R R, Enever, J R and Reynolds, S D, 1999. In situ stress field of eastern Australia. Australian
Journal of Erath Sciences, Vol. 46, pp 813-825.
Hoek, E and Brown, E T, 1980. Underground excavations in rock. International Mining and Minerals,
London.
Hoek, E, 2007. Practical rock engineering, Chapter 11, In Situ and Induced Stresses. Available on line
at http://www.rocscience.com/hoek/Hoek.asp.
Jaeger, J C and Cook, N G W, 1979. Fundamentals of rock mechanics [ Chapman and Hall, London].
Mark, C, 1991. Horizontal stress and its effects on longwall ground control, in Mining Engineering,
November, pp. 1356-1360.
Lee, M, Mikula, P, Mollison, L and Litterbach, J, 2008. Stresses in the Australian continental tectonic
plate-variability and likely controls. Presentation to the Australian Earthquake Engineering Society
Annual Meeting, Ballarat, VIC.
McCutchen, WR, 1982. Some elements of a theory for in-situ stress, in Journal of the South African
Institute of Mining and Metallurgy, Abstr., 19, pp 201-203.
McGarr, A, 1988. On the state of lithospheric stress in the absence of applied tectonic forces, in
Journal of Geophysical Research, Vol. 93(B11), pp 13609-13617.
Mills, K, 1997. In situ stress measurements using the ANZI stress cell, in Proceedings International
Symposium on Rock Stress, pp 149–154.
Molinda, G M, Heasley, K A, Oyler, D and Jones, J R, 1992. Effects of horizontal stress related to
stream valleys on the Stability of Coal Mine Openings, USBM RI 9413.
Mucho, T P and Mark, C, 1994. Determining horizontal stress direction using the stress mapping
technique, in Proceedings 13th International Conference on Ground Control in Mining, pp 277-289.
Muller, W, 1991. The stress state in the Ruhr Basin, in Proceedings of the ISRM 7th International
Conference on Rock Mechanics.
Muller, B, Zoback, M L, Fuchs, K, Mastin, L, Gregersen, S, Pavoni, N, Stephansson, O and Ljunggren,
C, 1992.. Regional patterns of tectonic stress in Europe, in Journal of Geophysical Research
95(B8), pp11783-11804.
Nemcik, J, Gale, W and Mills, K, 2005. Statistical analysis of underground stress measurements in
Australian Coal Mines, in Proceedings of the Bowen Basin Symposium, pp 117-122.
Phillips, D W, 1946. Research on strata control in Great Britain, in Transactions of the Coal Division,
AIME 168:27-50.
Reinecker, J, Heidbach, O, Tingay, M, Sperner, B and Müller, B, 2005. The release 2005 of the World
Stress Map (available online at www.world-stress-map.org).
Sheorey, P R, 1994. A theory for in situ stresses in isotropic and transversely isotropic rock, in
International Journal of Rock Mechanics, Mining Science and Geomechanics, Abstr., 31, pp 23-34.

38 11 – 12 February 2010
2010 Underground Coal Operators’ Conference The AusIMM Illawarra Branch

Stacey, T R. and Wesseloo, J, 1998. In situ stress field of eastern Australia, in Journal of the South
African Institute of Mining and Metallurgy, Nov-Dec, pp 365-368.
Townend, J and Zoback, M D, 2000. How faulting keeps the crust strong, in Geology, Vol. 28, pp 399-
402.
Windsor, C, 2008. Structural and strength limitations on the in situ stress tensor in Earth’s crust, in
Proceedings 3rd World Stress Map Conference.
Wooldridge, J, 2006. Introductory econometrics: a modern approach. Thomson-Southwestern,
Mason, OH.
Zoback, M L, 1992. First- and second-order patterns of stress in the lithosphere: the world stress map
project, on Journal of Geophysical Researc, Vol. 5(B8), pp 11703-11728.
Zoback, M L and Zoback, MD, 1989. Tectonic stress field of the continental United States in Geological
Society of American Memoir 172, pp 523-539.
Zoback, M L and Zoback, M D, 2002. State of stress in the earth's lithosphere. International Handbook
of Earthquake and Engineering Seismology, Vol. 81A, pp 559-568.
Zoback, M D and eight others, 2003. Determination of stress orientation and magnitude in deep wells,
in International Journal of Rock Mechanics and Mining Science, Vol. 40, pp 1049-1076.
Zoback, M L and Zoback, M D, 2007. Lithosphere stress and deformation. Chapter 6.06 in Treatise on
Geophysics, vol. 6, Crust and Lithosphere Dynamics, pp 253-273.

11– 12 February 2010 39