Seismic Hazard Analysis: January 2002

See discussions, stats, and author profiles for this publication at: https://www.researchgate.
net/publication/329682244
Seismic hazard analysis
Chapter · January 2002
CITATIONS READS
34 1,055
2 authors, including:
Kenneth W. Campbell
CoreLogic, Inc.
117 PUBLICATIONS 6,285 CITATIONS
SEE PROFILE
Some of the authors of this publication are also working on these related projects:
Attenuation of Arias Intensity View project
NGA-subduction View project
All content following this page was uploaded by Kenneth W. Campbell on 18 December 2018.
The user has requested enhancement of the downloaded file.

0068_C08_fm Page 1 Tuesday, August 20, 2002 7:30 AM
8
Seismic Hazard Analysis
8.1 Introduction
8.2 Probabilistic Seismic Hazard Methodology
8.3 Constituent Models of the Probabilistic Seismic
Hazard Methodology
Seismic Sources · Earthquake Recurrence Frequency · Ground
Motion Attenuation · Ground Motion Probability
8.4 Definition of Seismic Sources
Area Sources · Fault Sources
8.5 Earthquake Frequency Assessments
Historical Frequency Assessments · Geologic Earthquake
Frequency Assessments · Conservation of Seismic Moment on
Segmented Faults · Time-Dependent Probability Modeling
8.6 Maximum Magnitude Assessments
Area Source Determinations · Individual Fault Determinations
· Mixed Source Determinations
8.7 Ground Motion Attenuation Relationships
Impact on Seismic Source Definition · Reference Site Class
8.8 Accounting for Uncertainties
8.9 Typical Engineering Products of PSHA
8.10 PSHA Disaggregation
Scaling Empirical Earthquake Spectra
8.11 PSHA Case Study
Tectonic Setting · Regional Seismicity · Great Earthquakes
· Earthquake Source Characterization
8.12 The Owen Fracture Zone–Murray Ridge Complex
Maximum Magnitude · Earthquake Recurrence Frequencies
8.13 Makran Subduction Zone
8.14 Southwestern India and Southern Pakistan
8.15 Southeastern Arabian Peninsula and Northern
Arabian Sea
8.16 Ground Motion Models
Stable Continental Interior Earthquakes · Stable Oceanic
Interior Earthquakes · Transform Plate Boundary Earthquakes
· Subduction Zone Earthquakes
8.17 Soil Amplification Factors
Paul C. Thenhaus 8.18 Results
ABS Consulting 8.19 Conclusions
Evergreen, CO 8.20 PSHA Computer Codes
Kenneth W. Campbell Defining Terms
ABS Consulting and EQECAT Inc.
References
Portland, OR Further Reading
© 2003 by CRC Press LLC

0068_C08_fm Page 2 Wednesday, July 31, 2002 7:10 AM
8-2 Earthquake Engineering Handbook
8.1 Introduction
Seismic hazard is a broad term used in a general sense to refer to the potentially damaging phenomena
associated with earthquakes, such as ground shaking, liquefaction, landslides, and tsunami. In the specific
sense, seismic hazard is the likelihood, or probability, of experiencing a specified intensity of any
damaging phenomenon at a particular site, or over a region, in some period of interest. It is the latter,
specific sense, that is the subject of this chapter.
The methodology for assessing the probability of seismic hazards grew out of an engineering need for
better designs in the context of structural reliability [Cornell, 1968, 1969], since such assessments are
frequently made for the purpose of guiding decisions related to mitigating risk. However, the probabilistic
method has also proven to be a compelling, structured framework for the explicit quantification of
scientific uncertainties involved in the hazard estimation process. Uncertainty is inherent in the estimation
of earthquake occurrence and the associated hazards of damaging ground motion, permanent ground
displacements, and, in some cases, seiche and tsunami. Scientific knowledge for the accurate quantifica-
tion of these hazards is always limited. The balance of the hazard assessment is comprised of informed
technical judgment.
Prior to the widespread use of probabilistic seismic hazard analysis (PSHA) for assessing earthquake
hazards, deterministic methods dominated such assessments. Deterministic methods consider the effect
at a site of either a single scenario earthquake, or a relatively small number of individual earthquakes.
Difficulties surrounded the selection of a representative earthquake on which the hazard assessment
would be based. These difficulties often involved the identification of an earthquake that satisfied a
codified or regulatory definition. The probabilistic methodology reduces the need for such earthquake
definitions, which typically are ambiguous at best. The probabilistic methodology quantifies the hazard
at a site from all earthquakes of all possible magnitudes, at all significant distances from the site of interest,
as a probability by taking into account their frequency of occurrence. Deterministic earthquake scenarios,
therefore, are a subset of the probabilistic methodology.
In principle, PSHA can address any natural hazard associated with earthquakes, including ground
shaking, fault rupture, landslide, liquefaction, seiche, or tsunami. However, most interest is in the
probabilistic estimation of ground-shaking hazard, since it causes the largest economic losses in most
earthquakes. The presentation here, therefore, is restricted to the estimation of the earthquake ground
motion hazard. Figure 8.1 illustrates elements of the probabilistic ground motion hazard methodology
in the context of a complete program for establishing engineering seismic design criteria for a site of
significant engineering importance. The process begins with the characterization of earthquake occur-
rence using two sources of data: observed seismicity (historical and instrumental) and geologic. The
occurrence information is combined with data on the transmission of seismic shaking (termed attenu-
ation, see Chapter 5) to form the seismotectonic model. Since uncertainty is inherent in the earthquake
process, the parameters of the seismotectonic model are systematically varied via logic trees, Monte Carlo
simulation, and other techniques, to provide the probabilistic seismic hazard model’s results. The results
may be disaggregated (also known as deaggregation) to identify specific contributory parameters to the
overall results. The results must also consider the site-specific soil properties. The final results, presented
in many different ways depending on the user’s needs, are termed seismic design criteria, if the end use
is the design of an engineering structure. Each of these aspects is discussed further below.
8.2 Probabilistic Seismic Hazard Methodology

PSHA can be summarized as the solution of the following expression of the total probability theorem:
M Max
λ[X ≥ x] ≈ ∑ ν ∫ ∫ P [ X ≥ x M , R] f
Sources i
i
Mo R|M
M (m) f R|M (r m) dr dm (8.1)

Seismic Hazard Analysis 8-3
FIGURE 8.1 Flowchart showing the elements of the probabilistic hazard methodology in the context of a seismic
design criteria methodology.
where λ[X ≥ x ] is the annual frequency that ground motion at a site exceeds the chosen level X = x ; νi
is the annual rate of occurrence of earthquakes on seismic source i, having magnitudes between Mo and
MMax; Mo is the minimum magnitude of engineering significance; MMax is the maximum magnitude
assumed to occur on the source; P[X ≥ x |M,R] denotes the conditional probability that the chosen ground
motion level is exceeded for a given magnitude and distance; fM(m) is the probability density function
of earthquake magnitude; fR|M(r |m ) is the probability density function of distance from the earthquake
source to the site of interest. In application, this expression is solved for each seismic source i of a
seismotectonic model.
Once the annual exceedance rate λ[X ≥ x ] is known, the probability that an observed ground motion
parameter X will be greater than or equal to the value x in the next t years (the exposure period) is easily
computed from the equation:
( )
P [ X ≥ x ] = 1 − exp −tλ [ X ≥ x ] (8.2)
where the “return period” of x is defined as:
1 −t
RX ( x ) = = (8.3)
(
λ [ X ≥ x ] ln 1 − P [ X ≥ x ] )

Probability values commonly used and cited in PSHA are ground motions that have a 10% probability
of being exceeded in a 50-year exposure period of engineering interest. From Equation 8.3, this gives a
return period of:
−50
RX ( x ) = = 475 years (8.4)
ln (1 − 0.1)
Thus, these specific ground motions, which have a 10% probability of being exceeded during 50 years,
are commonly termed to have an average 475-year return period. It is informative to note that setting
the exposure period equal to the return period in Equations 8.2 and 8.3 results in a 63% probability that
the ground motions will be exceeded in t years under the Poisson assumption used to develop these
relationships.
8.3 Constituent Models of the Probabilistic Seismic

Hazard Methodology
Figure 8.2 schematically illustrates the constituent models of the probabilistic approach to estimating
earthquake ground motion hazard. These are models of:
1. Seismic sources
2. Earthquake recurrence frequency
3. Ground motion attenuation
4. Ground motion occurrence probability at a site
These models are introduced here and discussed in more detail in the following sections of this chapter.
8.3.1 Seismic Sources

PSHA requires that the distribution of earthquakes be defined in space with each epicenter having a
defined distance from the site of interest. This has been traditionally accomplished through the geo-
graphic delineation of seismic source zones and seismically active faults. Definitions of these sources are
based on interpretations of available geological, geophysical, and seismological data with respect to
earthquake mechanisms and source structures that are likely to be common within specific geographic
regions (Figure 8.3). Seismic source delineation is generally premised on geoscience knowledge that
relates earthquakes to geological structure. However, where causative earthquake faults and structure
are not known with certainty, seismic source interpretations are not unique [Thenhaus, 1983, 1986] and
the geographic distribution of earthquakes largely guides the definition of sources [Electric Power
Research Institute, 1986]. Methods of seismicity smoothing have recently been introduced to avoid
arbitrary decisions regarding placement of area-source boundaries [Frankel et al., 1995, 1996, 2000] and
to better represent the fractal geometry of distributed seismicity as a self-organized critical-state process
[Woo, 1996].
8.3.2 Earthquake Recurrence Frequency

Determining the earthquake recurrence frequency of the defined seismic sources is an important explicit
task in PSHA, whereas it is either implicit or is disregarded in deterministic seismic hazard analysis.
Earthquake recurrence frequency is based largely on statistical analyses of the historical record of earth-
quakes for all but the most tectonically active areas of the world where detailed paleoseismic studies of
active faults have been performed. Paleoseismology is the geological study of prehistoric earthquakes
[McCalpin, 1996; Yeats et al., 1997] and aids the analysis of large-earthquake occurrence frequency.
Earthquake frequency estimates in PSHA typically assume independence of earthquake events, or Poisson
arrival times. However, time-dependent treatments of earthquake recurrence estimates are the basis for

FIGURE 8.2 Elements of PSHA shown in relation to constructing a uniform hazard spectrum. (From EERI, Earth-
quake Spectra, 5, 675–699, 1989. With permission.)
earthquake forecasting in areas, such as the San Francisco Bay area, where a good historical and paleo-
seismic record of past earthquakes is available.
8.3.3 Ground Motion Attenuation

Empirical ground motion attenuation relationships are widely used to establish the amplitude of
earthquake ground motion at a site of interest (see Chapter 5). In engineering applications, the ground
motion parameters of interest are typically peak ground acceleration (PGA), response spectral acceler-
ation (PSA), response spectral velocity (PSV), and spectral displacement (SD). Proper implementation
of most modern ground motion attenuation relationships requires that the seismic sources are char-
acterized by the details of a fault-rupture model including depth to the top and bottom of the earthquake
rupture zone, fault dip, and the style of fault slip (i.e., strike-slip, normal, or reverse). These fault-
rupture parameters are a product of the tectonic environment of the region in which the analysis is
being performed.

0068_C08_fm Page 6 Tuesday, August 13, 2002 3:05 PM
45°N
44°N
43°N
42°N
41°N
40°N
39°N
38°N
37°N
36°N
35°N
34°N
0 150 300
33°N Kilometers
32°N
18°E 20°E 22°E 24°E 26°E 28°E 30°E 32°E 34°E 36°E 38°E 40°E 42°E 44°E 46°E 48°E 50°E 52°E
FIGURE 8.3 Seismic source zones overlaid on major faults and tectonic features, for Turkey. (From Erdik, M. et al.,
1999, Assessment of Earthquake Hazards in Turkey and Neighboring Regions, contribution to Global Seismic Hazard
Assessment Program, available at http://seismo.ethz.ch/gshap/turkey/papergshap71.htm.)
8.3.4 Ground Motion Probability

The estimation of the probability of exceeding some amplitude of shaking at a site in some period of
interest requires that a probability distribution of the ground motion amplitudes be assumed. The Poisson
model serves as a reasonable assumption in most engineering applications except in rare cases where a
single earthquake source may dominate the hazard at a site and the earthquake occurrence model for
the source can be considered time-dependent, or non-Poissonian [Cornell and Winterstein, 1988].
Poisson models have traditionally been used throughout seismic hazard assessment. However, time-
dependent earthquake occurrence estimates have been used for earthquake forecasting in the San Fran-
cisco Bay area of California [Working Group on California Earthquake Probabilities, 1999] as well as
elsewhere in California [Working Group on California Earthquake Probabilities, 1995]. Time-dependent
probability models are discussed later in this chapter.
8.4 Definition of Seismic Sources

The fundamental assumptions of a defined earthquake source is that (1) earthquake occurrence is
uniformly distributed for given magnitude within the source, and (2) earthquake occurrence is only
considered between a minimum earthquake magnitude of engineering interest (Mmin) and a maximum
magnitude (Mmax) that is representative of the entire source [Reiter, 1990]. The seismic sources are shown
as map representations of lines (fault sources), and area source zones that are defined on the basis of a
number of different types of geological, geophysical, and seismological data (Figure 8.3). These data are
summarized in Table 8.1 along with their indicated usefulness in the definition of types of seismic sources.
The defined geographic distribution of seismic sources and the specification of all source characteristics
required for the seismic hazard analysis is termed the seismotectonic model. The seismotectonic model
provides a complete description of earthquake occurrence in time and space to an outer distance of an
engineering interest, and to a depth sufficient to encompass the seismogenic thickness of brittle crust
beneath the site. Depending on the seismotectonic regime and the application for the results, the max-
imum distance and depth considered may be several hundred kilometers and a hundred kilometers,
respectively. The entire model is summarized in an input file format appropriate to the PSHA computer
code being used.

TABLE 8.1 General Data Types and Their Applications in Identifying and Characterizing Seismic Source Zones
Zone Type
Faults Area Sources
Data Type Location Activity Length Dip Depth Style Area Depth
Geological/Remote Sensing
Detailed mapping X X X X X X
Geomorphic data X X X X X
Quaternary surface rupture X X X X
Fault trenching data X X X X
Geochronology X X
Paleoliquefaction data X X X
Borehole data X X X
Aerial photography X X X X
Low sun-angle photography X X X
Satellite imagery X X X
Regional structure X X X X X X
Balanced cross section X X X X
Geophysical
Regional potential field data X X X X
Local potential field data X X X X X
High resolution reflection data X X X X
Standard reflection data X X X
Deep crustal reflection data X X X X X
Tectonic geodetic/strain data X X X X X X X
Regional stress data X
Seismological
Reflected crustal phase data X
Historical earthquake data X X X X
Teleseismic earthquake data X
Regional network seismicity data X X X X X X X
Local network seismicity data X X X X X X
Focal mechanism data X X
8.4.1 Area Sources

Area seismic sources define regions of the Earth’s crust that are assumed to have uniform seismicity
characteristics that are distinct from neighboring zones, and are exclusive of active faults that are indi-
vidually defined. The central and eastern United States (CEUS) region is often cited as a leading example
of a region where seismic hazard is defined through the use of area seismic source zones [Thenhaus,
1983; Reiter, 1990; Coppersmith, 1991; Coppersmith et al., 1993]. This region is located interior to the
North American tectonic plate where earthquake occurrence is much less frequent than in the western
United States, the historical and instrumental seismicity generally does not correlate well with geologic
structures observable at the surface, evidence of prehistoric earthquakes that is preserved in the geologic
record is difficult to find, and strain-rates of crustal deformation are exceedingly low. With the exception
of a few areas, such as the New Madrid Seismic Zone, the result of these characteristics is a seismotectonic
environment in which the identification of geologic structures that are responsible for earthquakes is
ambiguous and not unique. Interpretations of the regional distribution of seismicity largely guide the
definition of area seismic sources in this region [Thenhaus, 1983; EPRI, 1986]. Analogous regions of the
world are areas that are located large distances from active plate boundary zones (Figure 8.4). Such
regions are referred to as “stable continental interior regions” [Johnston et al., 1994]. Stable continental
interiors, as the name implies, are some of the least active areas worldwide with regard to earthquake

Active Volcanoes, Plate Tectonics, and the “Ring of Fire”
Eurasian Plate Eurasian Plate

North American Plate
Aleutian Trench CASCADE

RANGE
San Andreas
“Ring of Fire” Fault
Mid-Atlantic
Ridge
Arabian
Hawaiian "Hot Spot" Plate
Cocos Plate
Java Trench East Pacific Nazca South

Rise
Plate American
Plate African Plate
Indo- Australian Plate
Pacific Plate
Antarctic Plate
USGS
FIGURE 8.4 Plate tectonic setting of the world (from the U.S. Geological Survey). Primary tectonic plates are labeled
by name. Heavy black lines indicate plate boundaries. “Ring of Fire” is a colloquialism referring to the earthquakes
and volcanoes that occur along the northern boundary of the Pacific tectonic plate, extending from Japan through
North America.
activity and are generally comprised of areas of very old continental crust in which the most active
geologic process operating today is that of erosion.
An alternative to defining uncertain area source boundaries within regions of low seismicity is that of
seismicity smoothing. Woo [1996] formalized a kernel-estimation method of seismicity smoothing,
noting that the practice of seismic source zonation “should not be merely routine when applied to areas
where such correlations [between geological structure and seismicity] are tenuous.” Woo [1996] argued
that use of “Euclidean zones” (standard area sources) unrealistically impose a uniform spatial distribution
of epicenters in PSHA that conflicts with the clustering habit and inter-event correlation of recorded
seismicity [Kagan and Knopoff, 1980; Korvin, 1992]. Woo goes on to say that the power-law of the
Gutenberg–Richter relationship can be explained by the theory of self-organized criticality where, over
geologic time periods, regional build-up of crustal stress is marginally balanced by the stress released
during earthquakes. Some of this stress is manifested in aseismic deformation and folding. However, the
regional fault network is self-organized to the extent that earthquakes occur as a critical chain reaction.
As Woo [1996] notes, “if the process were supercritical, it would run away, but if the process were
subcritical, it would terminate rapidly.” Noteworthy in this regard was the documented regional increase
in seismicity throughout the western United States following the magnitude 7.3 Landers, California
earthquake [Hill et al., 1993]. This was the first time that such a large regional influence has been
scientifically documented from the occurrence of a single earthquake.
Frankel et al. [1996, 1998, 2000; http://earthquakes.usgs.gov/hazards/] applied seismicity smoothing
in recent updates of the U.S. national seismic hazard maps that have been produced over the last 24 years
by the U.S. Geological Survey under the auspices of the National Earthquake Hazards Reduction Program
(NEHRP). This model separates the CEUS into two broad areas of different estimated maximum potential
earthquakes. The mid-continent region of the central United States is defined as an area in which the
maximum earthquake is judged to be magnitude MW 6.5. The region of the Appalachian Mountains
eastward, and the Gulf Coast area of the southern United States, are areas of rifted continental crust

CEUS maximum-magnitude zones

290¡
50¡ 240¡
280¡ 50¡
250¡ 260¡ 270¡
6.5
40¡ 40¡
7.5
7.5
6.5
30¡ 7.2 30¡
7.5
240¡ 290¡
250¡ 280¡
260¡ 270¡
FIGURE 8.5 Maximum magnitude zones for the central and eastern U.S. used by Frankel et al., 1996. Magnitudes
are in terms of the moment magnitude (MW) scale.
(areas of past episodes of extension and faulting) and defined as areas capable of sustaining earthquakes
as large as MW 7.5 (Figure 8.5). Earthquake frequency throughout these broad regions was determined
on a 0.1° grid of points using the common Gutenberg–Richter relationship of occurrence frequencies
(see Chapter 4):
log N (m) = a − bm (8.5)
where N(m) = the number of earthquake events equal to or greater than magnitude m occurring on a
seismic source per unit time, and a and b are regional constants (10a = the total number of earthquakes
with magnitude > 0, and b is the rate of seismicity; b is typically 1 ± 0.3). The a-value of this relationship
was determined for the regional grid of points that have been smoothed over a distance of 50 km using
a Gaussian smoothing function [Frankel, 1995; Frankel et al., 1996]. Thus, area sources of this model are
only used to characterize regions of uniform maximum magnitude, not uniform earthquake frequency.
While seismicity smoothing frees the PSHA analyst of subjective judgment in locating area source
boundaries, subjective judgment is not eliminated by these methodologies, and is still required in
choosing reasonable smoothing parameters and inter-event correlation distances based on the available
seismicity data [Frankel and Safak, 1998]. These choices have a large impact on the estimated ground
motion amplitudes [Perkins and Algermissen, 1987].
8.4.2 Fault Sources

Line sources are defined in PSHA ground motion analyses as map-view representations of three-dimen-
sional fault planes for the purpose of explicit representation of faults that are considered capable of
earthquake rupture. By far, these types of sources are primarily defined in active tectonic regions that
are generally located in proximity to the boundaries of the world’s tectonic plates (Figure 8.4). However,
strain from plate–boundary tectonic processes are transmitted large distances through the Earth’s crust.
Earthquake faults can, therefore, be located large distances from these plate boundaries, albeit at greatly
diminished rates of activity and geographic concentration. Faults exhibit a wide range of offset styles
that depend on the prevailing tectonic stress regime, be it compression or extension, and the three-
dimensional geometry (i.e., strike and dip) of the individual fault within that stress regime (see
Chapter 4). Many modern ground motion attenuation relationships (see Chapter 5) are specific to the

style of faulting and require specification of the three-dimensional geometry of the defined fault through
the Earth’s crust so that source-to-site distances are accurately calculated.
The definition of earthquake faults need not be restricted to faults that are observable at the surface.
Blind-faults (faults that do not break the surface), such as ruptured in the 1983 Coalinga and 1994
Northridge, California earthquakes [Wentworth and Zoback, 1990], can and should be included in a
PSHA seismotectonic model, with appropriate specification of the depth to the top and bottom of the
earthquake rupture zone.
8.5 Earthquake Frequency Assessments

There are two fundamental approaches to assess earthquake recurrence frequency of the defined seismic
sources in PSHA. These are historical and geological frequency assessments. Historical frequency assess-
ments are based on statistical analyses of the historical catalog of earthquakes that have occurred within
a region. Geological frequency assessments are generally based either on a prehistoric record of earthquake
occurrence on faults (termed paleoseismicity ), which is compiled through detailed field geologic inves-
tigations, or on physical estimates of seismic moment either on individual faults or distributed throughout
broad regions. Moment-based recurrence frequency estimates require some knowledge of the average
long-term rate at which faults are slipping, or the rate at which tectonic deformation is occurring over
a region.
8.5.1 Historical Frequency Assessments

Historical catalogs of earthquakes are heterogeneous; that is, the data typically:
• Are nonuniform with respect to time periods of complete reporting for earthquakes of various
magnitudes
• Are nonuniform with respect to magnitude measures used to quantify earthquake size
• Contain a nonuniform mix of mainshock and aftershock earthquake events
• Contain duplicate earthquake entries
• May contain man-made events (such as blasts) that are not a product of the natural tectonic
environment of a region
Significant scrutiny (“clean-up”) of these catalogs is therefore required prior to performing statistical
analyses to determine representative estimates of earthquake recurrence frequency. This is not to denigrate
the value of historic earthquake catalogs — without such catalogs, the earthquake record would only
extend back to a few decades ago. With historic catalogs, the record extends back hundreds, in some
cases thousands, of years. Recent diligent efforts to research historic earthquakes are a significant con-
tribution to PSHA [see, for example, Lee et al., 1988; Downes, 1995; Stucchi, 1993; Usami, 1981;
Ambraseys and Finkel, n.d.; and Ambraseys and Melville, 1982].
An initial project earthquake catalog typically contains subcatalogs from various international and
local seismological reporting agencies, resulting in duplicate earthquake entries. Prioritization and
ranking of the various data contributed to the catalog are required to select the most reliable entries
in terms of earthquake location, time of occurrence, and magnitude. Conversion of multiple magni-
tude measures to a single, representative magnitude for all earthquakes in a catalog is performed using
empirical correlation relationships either available in seismological research literature (see Chapter 4)
or developed directly from the various magnitude measures in the catalog itself, providing that
sufficient data are available for obtaining statistical regression conversion relationships. Dependent
events are removed from the catalog based on magnitude–time–distance parameters appropriate for
characterizing aftershock earthquake sequences [see, for example, Reasenberg and Jones, 1989].
Removal of dependent events assures that the Poisson assumption of the PSHA is not violated in
terms of earthquake recurrence.

1.0
Intensity
IV
V
VI
VII
VIII
V
MMI VI complete
this range only
λ / λ
VI
0.1
σλ =
VII
Slope - T1/2
VIII
0.01
5 10 100
Time (Years)
FIGURE 8.6 Completeness time plots in terms of earthquake epicenter intensity following the method of Stepp
[1973]. Each symbol refers to a different earthquake intensity class.
Most commonly, the statistical procedures proposed by Stepp [1972] are used to assess completeness
times of the reported magnitudes, which assume the earthquake sequence in a catalog can be modeled
as a Poisson distribution. If k1, k2, k3 … kn are the number of events per unit time interval, then:
∑
1 n
λ= ki (8.6)
n i =1
and its variance is σ2 = λ/n, where n equals the number of unit time intervals. If unit time is one year,
σλ = λ1/2/T1/2 as the standard deviation of the estimate of the mean where T is the sample length in years.
This test, then, is for stationarity of observational quality. If data for a magnitude interval are plotted as
log (σλ) vs. log (T), then the portion of the line with slope T–1/2 can be considered homogeneous
(Figure 8.6) and used with data for other magnitude ranges (but for different observational periods)
similarly tested for homogeneity to develop estimates of recurrence frequency.
Over large regions, Gutenberg and Richter [1954] found that the average recurrence frequency of
earthquakes follows an exponential distribution related to magnitude (Equation 8.5). In its cumulative
form, the Gutenberg–Richter relation of recurrence frequencies is unbounded at the upper magnitude.
In PSHA, this relationship imposes the unrealistic assumption that the maximum potential earthquake
for any region under consideration is unbounded and unrelated to the seismotectonic setting. The

0068_C08_fm Page 12 Monday, August 19, 2002 1:13 PM
truncated exponential recurrence relationship [Cornell and Vanmarcke, 1969] is therefore commonly
used in practice:
N (m) = N m ( )0 ( ( )) ( (
exp −β m − m0 − exp −β mu − m0 )) for m ≤ mu
( ( ))
(8.7)
1 − exp −β m − m u 0
where m0 is an arbitrary reference magnitude; mu is an upper-bound magnitude where n(m) = 0 for

m > mu; and β = b · ln10. In this form, earthquake frequency approaches zero for some chosen maximum
earthquake of a region. Other magnitude-frequency relations are discussed in Chapter 4.
8.5.2 Geologic Earthquake Frequency Assessments

Characterizing earthquake recurrence frequency on individual faults, as opposed to regionally distrib-
uted area sources, is a more challenging proposition. Many earthquake faults have recurrence frequencies
for significant earthquakes of hundreds to thousands of years. It is largely fortuitous, then, if even one
large-earthquake interoccurrence interval is represented in the historical record. This lack of empirical
earthquake data precludes robust assessments of earthquake frequency by statistical treatments of
historical earthquake data. Other means are required for determining earthquake frequency on seismi-
cally active faults.
Significant efforts have been made in the past 25 years to characterize the rate at which faults slip in
many seismically active regions of the world [McCalpin, 1996]. Fault slip-rate can be related to earthquake
occurrence frequency through the use of seismic moment [Molnar, 1979; Anderson, 1979]. Seismic
moment, Mo , is the most physically meaningful way to describe the size of an earthquake in terms of
static fault parameters. It is defined as:
Mo = µ Af D (8.8)
where m is the rigidity or shear modulus of the fault, usually taken to be 3 × 1011 dyne/cm2; Af is the
rupture area on the fault plane undergoing slip during the earthquake; and D is the average displacement
over the slip surface. The seismic moment is translated to earthquake magnitude according to an expres-
sion of the form:
( )
log M o M = c M (8.9)
Based on both theoretical considerations and empirical observations, c and d are rationalized as 1.5
and 16.1, respectively [Molnar, 1979; Anderson, 1979]. Actually, to be consistent with the definition of
moment magnitude, d should be set equal to 16.05 [Kanamori, 1978; Hanks and Kanamori, 1979]. The
total seismic moment rate is the rate of seismic energy release along a fault. According to Brune [1968],
the slip rate of a fault can be related to the seismic moment rate M 0T as follows:
M oT = µA f S (8.10)
where S is the average slip rate (per unit time) along the fault. The seismic moment rate, therefore,
provides an important link between geologic and seismicity data.
While the Gutenberg–Richter relationship describes the regional occurrence frequency of earthquakes,
it has been found to be nonrepresentative of large earthquake occurrence on individual faults [Schwartz
and Coppersmith, 1984; Wesnousky, 1994]. Physically, this can be attributed to the breakdown of the
power law of the Gutenberg–Richter relationship between large and small earthquakes because they are
not self-similar processes [Scholz, 1990]. Geologic investigations of faults of the San Andreas system of
western California and of the Wasatch fault in central Utah have indicated that surface-rupturing

10
Annual Number of Earthquakes, N (m)

1
–1
10
–2
10
–3
10
FIGURE 8.7 Comparison of the exponential –4

10
(solid line) and characteristic recurrence (dashed
line) frequency curves. (From Youngs, R.R. and
–5
Coppersmith, K.J., Bull. Seismol. Soc. Am., 75, 10
4 5 6 7 8
939–964, 1985.) Magnitude, m
earthquakes tend to occur within a relatively narrow range of magnitudes at an increased frequency over
that which would be estimated from the Gutenberg–Richter relationship. These have been termed char-
acteristic earthquakes. The characteristic recurrence frequency distribution reconciles the exponential
rate of small- and moderate-magnitude earthquakes with the larger characteristic earthquakes on indi-
vidual faults (Figure 8.7). The summed rate of earthquakes over many faults in a region reverts to the
truncated exponential distribution [Youngs and Coppersmith, 1985] and is therefore consistent with the
regional empirical Gutenberg–Richter relationship.
The characteristic recurrence frequency distribution can be separated into a noncharacteristic Guten-
berg–Richter relationship for small and moderate earthquakes, and a characteristic frequency part for
large earthquake occurrence. The cumulative rate of noncharacteristic, exponentially distributed earth-
quakes, Ne , is estimated from the seismic moment and seismic moment rate as follows:
1−e ( u )
− β m −0.25
N e= M oT (8.11)
Mo e
− β ( mu −0.25)  b10 − c / 2 b10 1 − 10
+
b −c/2
( ) 
 c −b 
 c 
The cumulative rate of characteristic earthquakes, Nc , is related to the cumulative rate of noncharac-
teristic earthquakes by the expression:
β Ne e ( u 0 )
− β m −m −1.5
Nc =
( )
(8.12)
2 1−e ( u 0 )
− β m −m −0.5
Similar to the truncated exponential recurrence model, frequency estimates from the characteristic
recurrence model approach zero at the defined maximum magnitude for the source. Figure 8.7 compares
the truncated exponential and characteristic frequency distributions.
If a fault can be considered truly characteristic, then only a single earthquake of specified magnitude
is expected to occur on the fault. Such a model was used by Frankel et al. [1996, 2000] and the Working
Group on California Earthquake Probabilities [1995, 1999] for large faults in California that have been
well characterized paleoseismically. In this case, the frequency of the characteristic event, which is assumed
to rupture an entire fault segment or series of segments, is given by the expression:
M 0T
N char = (8.13)
M 0 ( M char )

where Mchar is the magnitude of the characteristic event. This is referred to as the characteristic earthquake
or maximum-magnitude model. This expression becomes somewhat more complex if the characteristic
event is assumed to have a truncated Gaussian magnitude distribution (to account for inherent random-
ness in the magnitude of the event) as was assumed by the Working Group on California Earthquake
Probabilities [1999]. To conserve the mean seismic moment rate, N char must be reduced when the
truncated Gaussian distribution is used. This reduction is a function of both the standard deviation and
the truncation limit. The Working Group on California Earthquake Probabilities [1999] used a standard
deviation of 0.12 and a truncation limit of ±2 standard deviations. In this case, N char would have to be
multiplied by 0.94 to conserve the mean seismic moment rate.
The cumulative truncated Gaussian magnitude distribution is given by the equation:
 M + zσ m

N char =
νchar Fchar 
 0∫ f M (m) dm −
0 ∫
f M (m) dm
 (8.14)
  M + zσ  
2 
 0 ∫ f M (m) dm − 1
 
where N char is the annual number of events with magnitude greater than or equal to m, νchar is the mean
rate of characteristic events on the fault (equal to the inverse of the recurrence time), F char is the reduction
factor needed to conserve the mean seismic moment rate [see Field et al., 1999], M is the characteristic
magnitude on the fault (referred to as Mchar above), z is the truncation limit, σ is the standard deviation,
and fM(m) is the Gaussian (normal) probability density function, given by:
1  1  m − M  2
f M (m) = exp −    (8.15)
σM 2π  2  σ M  
The above integrals are widely available in statistics books. In the case of the application by the Working
Group on California Earthquake Probabilities [1999], F char = 0.94, σM = 0.12, and z = 2.
The truncated exponential distribution can also be characterized in terms of seismic moment rate so
that it can be used in conjunction with slip rate. This distribution is given by Equation 8.14, in which
N 0 is given by the expression [Shedlock et al., 1980; Campbell, 1983]:
10 − b(m0 −m0′ ) − 10 − b(mu −m0′ )

N 0 = N 0′ (8.16)
1 − 10 − b(mu −m0′ )
where
M 0T (c − b)
[ ]
−1
N 0′ = M 0 (mu )10 − b(mu −m0′ ) − M 0 (m0′ ) (8.17)
b[1 − 10 − b(m −m′ ) ]
u 0
and m0′ ≤ m0 is the magnitude corresponding to a physical lower limit below which earthquakes are not
expected to occur or do not contribute to the observed slip on the fault, M0( m0′ ) is the seismic moment
of this lower limit magnitude, and M0(m u) is the seismic moment of the upper bound magnitude m u.
The physical lower limit magnitude should not be confused with the lower-bound magnitude m 0 , the
lower limit of engineering significance.
If there is no physical limit to the smallest earthquake that can occur on a fault, or if mu >> m0′ , then
Equation 8.17 can be simplified considerably, resulting in the relationship [Campbell, 1983]:
M 0T (c − b)10bmu
N 0′ = for mu >> m0′ (8.18)
bM 0 (mu )

which can be used with the truncated exponential distribution above or as an estimate of the a-value
in the Gutenberg–Richter relationship given in Equation 8.5. Petersen et al. [1996] used an incremental
version of the above relationship to estimate the a-value for characteristic faults in California.
8.5.3 Conservation of Seismic Moment on Segmented Faults

The maximum magnitude and characteristic earthquake recurrence models are conceptually linked to
the idea that large fault zones rupture in a segmented manner and rarely break their entire length in a
single earthquake. The segmented nature of the mechanics of faulting was revealed through intense
paleoseismic investigations of the San Andreas and Wasatch fault zones in the western United States
[Schwartz and Coppersmith, 1984]. Relatively small displacements observed in natural exposures or
excavations of the fault zones were not laterally continuous. Larger displacements, however, were con-
tinuous over longer lengths and crossed structural discontinuities within the fault zone. The logical
confinement of ruptures between identifiable structural discontinuities, or rupture barriers, within fault
zones led to the concept of segmented fault rupture [King and Nabelek, 1985; Schwartz, 1988]. This
concept has proven very useful in explaining the extent of recent earthquake ruptures [e.g., Crone et al.,
1985], in defining paleoseismic ruptures [e.g., Machette et al., 1992], and in relating the development of
geologic structure to repeated earthquake rupture of fault zones [Cowie and Scholz, 1992].
Although fault segmentation has proven a very useful tool in earthquake hazards assessment, it is
important to recognize that segment boundaries do not repeatedly arrest all earthquake ruptures and
that their position is not necessarily constant in geologic time and space [Wheeler and Krystinik, 1992].
Nonetheless, a basic tenet of fault segmentation is that, in a relative sense, smaller earthquakes tend to
be confined to single segment ruptures, whereas larger earthquakes tend to be characterized by multi-
segment ruptures. Specific lengths of segment ruptures depend on the tectonic environment of the region
and the style of faulting that is present. Repeated faulting of all styles, over geologic time, will produce
recognizable geologic structures at segment boundaries.
Detailed documentation of fault slip rates along faults of the San Andreas system in southern California
has shown that slip rate is not constant along all segments of a single fault zone [Working Group on
California Earthquake Probabilities, 1995, 1999]. Slip rate typically varies among the various segments
and could be due to any number of physical changes that may occur along the fault. A difficulty in PSHA,
therefore, is accounting for the varying slip-rate values between different segments of individual faults.
The various slip rates could be completely accommodated in the PSHA by a series of fault sources specific
to each segment with each segment’s earthquake rate governed by the segment slip rate. This would be
an unrealistic model, however, both in terms of the resulting distribution of probabilistic ground motion
and the representation of large historical earthquake ruptures that have been documented to rupture
more than one segment on some faults.
The Working Group on California Earthquake Probabilities [1995, 1999] developed a “cascade” model
of earthquake recurrence frequency to satisfactorily account for varying slip rates and fault depths on a
single fault zone in a PSHA. The 1995 cascade model assumes that large earthquakes break multiple,
contiguous segments of a fault at a frequency that is governed by the lowest-slipping segment. Once the
moment rate (Equation 8.17) of the slowest-slipping segment is depleted in the production of these large
earthquakes, it drops from any further considerations regarding multisegment ruptures and the slip rates
of the remaining segments are reduced by the rate of the slowest-slipping segment. A new set of multi-
segment ruptures are thereby defined, and the procedure repeats until only single-segment ruptures of
the highest-slipping segments are left to rupture in single earthquakes at a rate that is determined from
the residual slip when all multisegment ruptures have been exhausted. This novel modeling approach
maintains the slip-rate and seismic-moment budget on each defined fault segment. In the 1999 model,
each possible cascade on a fault zone was assigned a relative weight by a panel of experts and the final
weights adjusted to achieve a moment-balanced model.

8.5.4 Time-Dependent Probability Modeling

Use of the previously described magnitude-frequency models in most seismic hazard applications assumes
that the occurrence of an earthquake is independent of all other earthquakes, both spatially and tempo-
rally. That is, that earthquake occurrence follows a Poisson process, which can be characterized by a mean
annual rate and a variance equal to this mean rate. These applications are therefore time-independent.
Time-dependent occurrence models, on the other hand, assume that the probability of a future earth-
quake increases with the elapsed time since the last earthquake. Faults that are early in their seismic cycle
are less likely to have an earthquake than the Poisson model would predict and faults late in their seismic
cycle are more likely to have an earthquake. Such models are called renewal models. The consequences
of time-dependent probability on seismic hazard have been investigated only to a limited extent in the
literature [Cramer et al., 2000], but it has found widespread use in engineering practice and in loss
modeling. It has also found professional acceptance in assessing short-term probabilities of large earth-
quakes on the San Andreas and related faults in California [Working Group on California Earthquake
Probabilities, 1988, 1990, 1995, 1999], which has greatly impacted hazard mitigation policy in the San
Francisco and Los Angeles areas.
Following the great 1906 San Francisco earthquake, Reid [1910] proposed the elastic rebound theory
in which earthquakes occur whenever stress builds to a certain level on a fault (Figure 8.8). The earthquake
relieves this stored stress and the earthquake cycle begins anew. With the assumption of constant stress
increase, the recurrence time in this model is perfectly periodic. Based on a long earthquake history in
the vicinity of Kyoto, Japan and measured coastal uplifts during earthquakes, Shimazaki and Nakata
[1980] proposed two other alternative forms of recurrence. The first is a time-predictable model that
predicts earthquakes will occur when stress accumulation on the fault reaches a critical level, but that
the stress drop and magnitude of the earthquakes vary among the seismic cycles (Figure 8.8). Thus, the
time to the next earthquake can be predicted from the slip in the previous earthquake assuming a constant
fault-slip rate. The second is a slip-predictable model in which earthquake failure on the fault resets the
stress on the fault to some constant level irrespective of the earthquake’s magnitude (Figure 8.8). Thus,
slip in the next earthquake can be predicted from the time since the previous earthquake.
The motivation for time-dependent probability models arose from the identification by seismologists
of “seismic gaps” along segments of major plate boundaries of the circum-Pacific region [see for example,
Sykes, 1971]. Seismic gaps were identified as unbroken sections of major plate boundaries that were
bounded on each end by ruptures from previous earthquakes. The implication from such observations
was that the unbroken sections had an increased likelihood of rupturing in a future earthquake than
those sections that had previously ruptured. Nishenko [1991] formalized the concept for 96 circum-
Pacific plate boundary segments in terms of a conditional probability that a large earthquake would occur
in future time windows of 5, 10, and 20 years, given that one had occurred at some known time in the past.
A primary issue in such earthquake forecasts is the reliable quantification of large (characteristic)
earthquake recurrence intervals. In that these large earthquakes are rare events in themselves, there is
FIGURE 8.8 Three types of earthquake occurrence

STRESS
T1 T1
models. Upper row of figures shows stress patterns of
increase and decrease through multiple earthquake T2 T2
cycles. Lower row of figures shows corresponding pat- u u u
COSEISMIC
terns of fault slip through the earthquake cycles. (a) Per-

CUM.
SLIP
fectly periodic model of constant stress increase to a

certain level (T1), and constant stress drop back to a
certain level during the earthquake (T2). (b) Time-pre- (a) t (b) t (c) t
dictable model illustrating stress buildup to a certain TIME
level and nonuniform stress drops. (c) Slip-predictable
model showing nonuniform stress buildups and stress-
drops to a certain level. (From Scholz, D.H., 1990, The
Mechanics of Earthquake Faulting, Cambridge University
Press, Cambridge. With permission.)

insufficient available data on any individual rupture segment from which to obtain a robust statistical
distribution of recurrence times. Nishenko and Buland [1987] found that normalized earthquake recur-
rence intervals for large earthquakes in the circum-Pacific region follow a lognormal distribution with
a virtually constant intrinsic standard deviation of σD = 0.205 for historic recurrence data and 0.215 for
combined historic and geologic recurrence data, when a normalizing function T/Tave was used. In this
formulation, T is the recurrence interval of an individual earthquake sequence, and Tave is the observed
average recurrence interval for the sequence. This distribution has since become a primary element of
the renewal-type time-dependent earthquake models used by the Working Group on California Earth-
quake Probabilities [1988, 1990, 1995, 1999]. A finite value for the average intrinsic standard deviation
(usually taken to be between 0.21 and 0.5) is a significant element of the model, because it introduces a
degree of aperiodicity in the occurrence of characteristic earthquakes. The physical reality of a finite value
for the intrinsic standard deviation is also significant because it may present a barrier to precise earthquake
prediction since physical reasons for this value are currently unknown [Scholz, 1990]. In part, it might
be due to stress interactions from other earthquakes in the region, as discussed later in this chapter.
The probability that a large earthquake will occur on a fault at time τ in an interval (T, T + ∆T),
assuming a probability density function for recurrence time fT(t), can be given by the integral:
T + ∆T
P (T ≤ τ ≤ T + ∆T ) =
∫T
fT (t ) dt (8.19)
If the elapsed time since the previous earthquake on the fault Te is known, the conditional probability
that the earthquake will occur in the next ∆T years is given by the ratio:
Te + ∆T
(
P Te ≤ τ ≤ Te + ∆T τ > Te = )
∫ Te
fT (t ) dt
(8.20)
∞
∫T
fT (t ) dt
It has been common to assume a lognormal probability density function for recurrence time [Nishenko
and Buland, 1987; Working Group on California Earthquake Probabilities, 1988, 1990, 1995; Cramer et
al., 2000], which is given by the equation:
 2

1 1  t Tˆ  
fT (t ) = 
exp −   (8.21)
t σ lnT 2  2  σ lnT  
where σlnT is the standard deviation of the logarithm of recurrence time and T̂ is the median recurrence
time, related to the mean by the expression:
Tˆ = T exp ( 1
2 σ ln2 T )
The standard deviation is composed of two parts, an intrinsic standard deviation σi and a parametric
standard deviation σp , where
σ lnT = σ i2 + σ 2p
The intrinsic standard deviation represents aleatory or random variability, whereas the parametric stan-
dard deviation represents epistemic or modeling uncertainty. The latter standard deviation accounts for
uncertainty in the median estimate of the recurrence interval, which comes from deriving it from events
with uncertainty dates, such as paleoseismic estimates. Epistemic uncertainty can also be included by

Monte Carlo simulation, rather than including it in σlnT [Working Group on California Earthquake
Probabilities, 1995, 1999].
While the lognormal distribution has been the most common distribution used to describe the
variability in recurrence intervals, several other distributions have been proposed and used. Most notable
are the Weibul and time-predictable distributions [Cornell and Winterstein, 1988] and the Brownian
passage time distribution [Working Group on California Earthquake Probabilities, 1999]. Other non-
Poissonian distributions are reviewed by Cornell and Winterstein [1988].
Faults are not only affected by earthquakes that occur on them, but also by earthquakes that occur on
nearby faults, through stress transfer [Stein, 1999]. The occurrence of a large earthquake modifies shear
and normal stresses on other faults in the region, although these stress changes are small, being on the
order of several bars or less. Considering that stress drops in earthquakes are on the order of 50 to 100
bars, the changes related to stress transfer are only a small fraction of the stress drops involved in fault
rupture. Nonetheless, the frequency of earthquakes has been observed to increase in areas of increased
stress transfer and decrease in areas of decreased stress transfer. For example, stress changes following
the MW 6.9 Hyogo-ken Nanbu, Kobe, Japan earthquake is believed to have produced a tenfold probability
decrease of a large earthquake in the next 30 years on the western segment of the Arima-Takatsuki Line
(the fault that terminated the Kobe rupture on the north), where stress decreased [Toda et al., 1998]. On
the other hand, it was estimated that there was a fivefold increase in probability on the eastern segment
of the Arima-Takatsuki Line, where stress was increased (Figure 8.9). Incorporating such stress changes
augments the physical bases of time-dependent earthquake probability estimates.
The permanent probability gain caused by a stress increase is amplified by a transient gain that decays
with time [Stein, 1999]. The opposite occurs for a stress decrease. The transient gain is an effect of rate-
and state-dependent friction [Dietrich, 1994], which describes behavior seen in laboratory experiments
and in natural seismic phenomena, such as earthquake sequences, clustering, and aftershocks. The
transient gain can be significant, but will decrease exponentially after the earthquake, according to
Omori’s law, until it reaches the level of the static stress change. The rate increase can be converted to a
probability gain using Equation 8.2. The seismicity rate equation is given by [Dietrich, 1994; Stein, 1999]:
r
R (t ) = (8.22)
  − ∆σ f    −t 
exp   − 1 exp  t  + 1
  Aσ n    a
where R(t) is the seismicity rate as a function of time t, following a Coulomb stress change ∆σf , A is a
constitutive parameter, σn is the total normal stress, ta is the aftershock duration (equal to ∆σ/ τ̇, where
τ̇ is the stressing rate on the fault), and r is the seismicity rate before the stress perturbation. To evaluate
this equation, the Coulomb stress change is calculated and r, ta , and τ̇ are estimated from observations,
allowing Aσn to be inferred.
Using such a model, Parsons et al. [2000] estimated a 62 ± 15% probability of an earthquake capable
of causing strong shaking in Istanbul in the next 30 years as a result of the 1999 MW 7.4 Izmit, Turkey
earthquake. This can be compared to a probability of 49 ± 15% using only the renewal model. The
Poisson model results in a 30-year probability of 20 ± 10%. The probability during the next decade is
estimated to be 32 ± 12% compared to a renewal rate of 20 ± 9%.
Large earthquakes can also have a significant quiescent affect on seismicity, called a stress shadow.
Harris and Simpson [1998] evaluated the observed suppression of MW 6 earthquakes in the San Francisco
Bay area after the 1906 MW 7.8 San Francisco earthquake, and found a set of stress and rate-and-state
parameters that were consistent with the observed rate change. Applying these parameters to the Hayward
fault, they found that the probability of a MW 6.8 earthquake during the period 2000 to 2030, such as
occurred in 1868, is 15 to 25% lower if the effect of the 1906 stress shadow is included. Stress transfer
following the great 1906 San Francisco earthquake is one of the models considered by the Working Group
on California Earthquake Probabilities [1999] in their time-dependent probability estimates for the

M ≥1.0
Aftershocks
Coulomb for Optimally Oriented Faults

Stress Change -0.5 0.0 0.5
(bars) for Major Strike-Slip Faults
FIGURE 8.9 Illustration of stress transfer following the 1995 Hyogo-ken Nanbu (Kobe), Japan earthquake and
aftershocks greater than magnitude 2.0. Star indicates the epicenter of the.earthquake located on the northeast trend
of the fault rupture. The Arima-Takatsuki Tectonic Line trends east-northeast at the northern end of the fault rupture.
63% of the aftershocks were found to occur where Culomb stress increased on optimally oriented faults by greater
than 0.1 bar. (From Toda, S. et al., J. Geophys. Res., 103, 24,543–24,565, 1985. Also http://quake.usgs.gov/research/
deformation/modeling/ papers/kobe/fig6.jpg/.)
San Francisco Bay area. Figure 8.10 illustrates the regional, as well as fault-segment-specific, conditional
probabilities of MW 6.7 earthquakes in the San Francisco Bay region made by the Working Group for a
30-year time window from 2000 to 2030. The Working Group estimated this probability to be 70 ± 10%
with individual fault segments contributing anywhere from 4 to 32% of this overall probability. Also new
in these assessments is the incorporation of a probability (i.e., 9%) that a future earthquake may occur
off the major faults in the regions that were considered in the study.
8.6 Maximum Magnitude Assessments

Assessment of the maximum magnitude earthquake for the defined seismic sources is an important,
fundamental task in PSHA. For PSHAs addressing low-probability hazard (i.e., long return-period haz-
ard), the maximum magnitude earthquakes may dominate the ground motion assessments [Bender,
1984]. In probabilistic analyses, the maximum earthquake is defined as the earthquake that is assessed
as physically capable of occurring within, or on, a defined seismic source in the contemporary tectonic

SAN FRANCISCO BAY REGION

EARTHQUAKE PROBABILITY
70%
odds (±10%) for one or more
magnitude 6.7 or greater
earthquakes from 2000 to 2030.
This result incorporates 9% odds
of quakes not on shown faults.
Expanding urban areas
21% New odds of magnitude

6.7 or greater quakes
before 2030 on the
indicated fault
Odds for faults that were

18% not previously included
in probability studies
Increasing quake odds

along fault segments
Individual fault probabilities are

uncertain by 5 to 10%
FIGURE 8.10 Time-dependent earthquake probabilities in the San Francisco Bay area, California for a time window
from 2000–2030. (From the Working Group on California Earthquake Probabilities, 1999, U.S. Geological Survey
Open-File Rep. 99-517, http://Geopubs.wr.usgs.gov/fact sheet/fs152–99.)
stress regime. Another term defining such an event is maximum credible earthquake [California Division
of Mines and Geology, 1975]. A number of diverse methods have been used to define such earthquakes,
each dependent on the purpose of the hazard assessment, the amount and kinds of seismological and
geological data that lend themselves to such assessment, and the variety of seismotectonic settings in
which PSHAs have been performed [dePolo and Slemmons, 1990]. The assessment of maximum earth-
quake magnitude is possible mainly because empirical data indicate a correlation between earthquake
magnitude and fault parameters of rupture length, rupture area, and displacement [Wells and Copper-
smith, 1994; see Chapter 4]. However, the use of several techniques can result in more reliable estimates
than any single technique by itself [Coppersmith, 1991]. Maximum magnitude assessments in PSHA are
broadly divisible into those that characterize area sources and those that are specific to individual
earthquake faults.
8.6.1 Area Source Determinations

Area sources are generally employed due to the lack of recognizable earthquake faults and seismically
active geologic structure. Empirical correlation equations between fault rupture parameters and earth-
quake size, therefore, have limited application to these source types. Maximum magnitudes for these
sources are typically assessed from an extrapolation of the historical seismicity of the region, from

compelling worldwide analogs of the regional tectonic setting, from regional paleoseismologic data and
interpretations (if available), or simply from the judgments of experts.
Lacking clear geological guides from which to estimate maximum earthquakes in regions of low
seismicity, methods of estimating maximum magnitudes from historical earthquake data have been
commonly used. Since large earthquakes in these regions have very long recurrence times, it is generally
assumed that the largest historical earthquake is the minimum value for a maximum earthquake estimate.
Nuttli [1979], for example, evaluated the largest earthquake of the 1811–1812 New Madrid earthquake
sequence in the New Madrid Seismic Zone as mb 7.4 and expressed confidence that the sum of the energy
of the three principal shocks of the sequence defines the maximum earthquake for this seismic source
(mb 7.5). Elsewhere in the central United States region, magnitude increments of between 0.5 and 1.0
unit above the historically observed maximum earthquake were judged to approximate earthquakes with
recurrence intervals of around 10,000 years in application to nuclear facility sites. Nuttli [1981] later
extended this approach for normalized areas of 100,000 km2 and a time period of 1000 years in application
to seismic zones of the central United States. Justification was based on the fact that these parameters
provided maximum magnitude estimates in general agreement with the historical maximum earthquakes
of both the New Madrid and Charleston seismic zones, which are considered to be essentially the
maximum earthquakes in these zones.
Such techniques assume that the addition of magnitude increments to historical maximum observa-
tions accounts, to some extent, for the relative shortness of the historical reporting period. For a b-value
of 1.0 in the Gutenberg–Richter recurrence relationship (see Chapter 4), the addition of 1.0 magnitude
unit to the historical maximum observation is equivalent to multiplying the length of the observation
period by a factor 10. Justification is then required for expecting the maximum magnitude event within
this period of time. A primary issue with this technique is the correct characterization of the form of
recurrence–frequency relationship. Minor changes in the Gutenberg–Richter b-value imply greatly dif-
fering time periods of catalog compensation, and other forms of recurrence–frequency relationships may
also be appropriate [Wesnousky et al., 1983; Wesnousky, 1994].
The technique of using worldwide tectonic analogs for assessing maximum magnitudes is premised
on the acquisition of more complete data by substituting space for time. In many regions, the historical
period of earthquake reporting is far too short to have probably sampled the largest possible earthquake.
However, even if the occurrence of maximum earthquakes is random in time and space, the likelihood
of having observed a maximum earthquake is greater over a collection of similar regions worldwide. This
was the basis for a far-reaching study by the Electric Power Research Institute [Johnston et al., 1994] of
stable continental interior earthquakes worldwide. From a regionalization of the world into areas of
similar geologic and tectonic histories and a detailed examination of the historical earthquakes in each,
this study was able to identify associations of maximum earthquakes with specific classes of continental
crust. Very old continental crust that had not been disturbed by regional tectonism over the last billion
years exhibited lower magnitude earthquakes than continental crust that had been disturbed over the
last 0.5 billion years. Such analogies can provide strong arguments for maximum magnitude assessments
in regions of sparse seismicity.
Paleoseismologic data, such as paleoliquefaction, can be an insightful tool in assessing the maximum
magnitudes of low-seismicity regions. Unlike tectonic analogs that substitute space for time, paleoseis-
mological data extend the record of earthquakes into prehistoric time. Munson et al. [1992, 1994]
documented widespread liquefaction features throughout southern Illinois and Indiana dating from a
single earthquake 6100 ± 200 years ago. A second strong earthquake was documented as occurring 12,000
years ago. Historical earthquakes in this region up to magnitude 5.5 have no reports of accompanying
liquefaction, and the prehistoric earthquakes therefore appear to be considerably larger than the historic
earthquakes. Obermeier et al. [1993] evaluated the earthquake that occurred 6100 years ago as a mag-
nitude 7.5 earthquake based on the physical dimensions and properties of the liquefaction features. Thus,
although the source of the prehistoric earthquakes remains unknown, assessments of the seismic hazard
in this region must assess significantly higher maximum magnitudes than have been observed historically.
A similar example can be cited for the Pacific northwest region of the United States where no large

earthquake has affected coastal Oregon and Washington historically. Yet, regional paleoseismological
evidence suggests sudden, regional submergence of coastal marshes along the Oregon and Washington
coasts [Atwater, 1987; 1992]. These phenomena have been ascribed to prehistoric great earthquakes
(moment magnitude of 8 ~ 9) and their associated tsunamis that occurred 300 years ago, and earlier,
along the Cascadia subduction zone. Recently, evidence from tsunami observations in Japan permitted
precise dating of the last earthquake at 1700 and its magnitude at ~9 [Satake et al., 1996].
8.6.2 Individual Fault Determinations

Guidelines for establishing maximum magnitudes for individual faults are significantly better than for
establishing maximum magnitudes for area seismic sources. Estimates of maximum magnitudes on faults
are typically computed from empirical correlation relationships between earthquake magnitude and
rupture dimensions [e.g., Wells and Coppersmith, 1994]. The most common characteristic that is cor-
related with earthquake magnitude is fault rupture length. Extensive study has been made of this corre-
lation over the past 30 years [Bonilla and Buchanan, 1970; Mark and Bonilla, 1977; Bonilla, 1980; Wells
and Coppersmith, 1994; among many others; see Chapter 4] and the database of worldwide earthquake
surface ruptures has grown rapidly [e.g., Wells and Coppersmith, 1994]. In addition, the formal statistical
treatment of the data has improved over the years. Modern correlation relationships:
1. Quantify the scatter in the data in terms of the standard deviation
2. Are sensitive to earthquake magnitude type
3. Provide unique definition of fault-type categories
4. Minimize the error in prediction by providing separate regression relationships for the various
independent variables
Wells and Coppersmith’s [1994] relationships, presented in Chapter 4, are based on data from 244
historical continental earthquakes worldwide that produced surface rupture and had focal depths shal-
lower than 40 km and magnitudes greater than 4.5.
Two primary sources of uncertainty exist in employing magnitude-rupture length correlation equa-
tions in the assessment of earthquake magnitudes. One is the variability of the regression equation itself,
which has been described in Wells and Coppersmith’s [1994] assessments in terms of the standard
deviations of the regressions and which can be considered in the application of the relationships. The
other is the uncertainty of establishing future rupture lengths. While the empirical correlation relation-
ships provide a relatively complete assessment of past earthquake ruptures related to earthquake magni-
tude, their forward application in predicting future earthquake magnitude on a fault is ambiguous owing
to the need to define the length of future fault ruptures. In applications of a few decades ago, rather
arbitrary fractional lengths of known earthquake faults were used to determine maximum magnitudes,
premised on the empirical observation that faults seldom rupture their entire lengths in single earth-
quakes, and commonly ruptured in less than half of their entire length [Albee and Smith, 1966]. However,
the concept of fault segmentation also has proven to be a useful tool in the definition of future potential
lengths of fault rupture. For example, in the cascade model of determining earthquake recurrence
frequencies on singular faults with segment-specific slip rates, the magnitude of each earthquake cascade,
whether a multi- or single-segment rupture, is defined by the individual cascade rupture dimension.
8.6.3 Mixed Source Determinations

The previous discussions of earthquake recurrence frequency and maximum magnitude assessment in
PSHA were organized around area and fault source determinations for simplicity of presentation. How-
ever, it is common practice to define a mixed-source type in which one or more earthquake faults are
defined within the boundaries of a regional area source. Such mixed-source definitions require particular
care in specifying earthquake frequencies and maximum magnitudes to avoid double counting of earth-
quake occurrences. The simplest manner of handling such sources is to define two separate sets of

minimum and maximum magnitudes, one for the area source and one for the fault sources, respectively.
Such a treatment typically defines the area source as a region of background seismicity in which small-
and moderate-magnitude earthquakes are modeled as random events up to a maximum magnitude of
about 6.5, with a recurrence frequency statistically determined from the catalog of earthquakes in the
region. Above this magnitude, the larger earthquakes can be modeled in various ways as occurring on
the defined earthquake faults with a recurrence frequency defined either from an extrapolation of the
historical data, available paleoseismic data, or some combination of both. Maximum magnitudes may
then be based on the physical parameters of the faults, as previously described.
The magnitude break in area source/fault source modeling techniques for the mixed-source type is
often rationalized around the threshold magnitude of surface faulting in the tectonic province of the site.
In the Basin and Range province of the western United States, this threshold magnitude is around 6.75
[Bucknam and Anderson, 1979]. Care should be taken to evaluate the implied seismicity and seismic
moment rates for mixed-source types to assure that the summed seismicity parameters are reasonable
within the constraints of available seismotectonic data. This is of particular importance if the modeled
magnitude range of the defined area source overlaps with those of the defined fault sources. Unrealistic
regional magnitude-frequency relations could easily result from such a treatment of seismic sources.
8.7 Ground Motion Attenuation Relationships

The ground motion attenuation relationships provide the means of estimating a strong-motion parameter
of interest from parameters of the earthquake, such as magnitude, source-to-site distance, faulting
mechanism, local site conditions, etc. [see Chapter 5]. This relationship is a particularly important
element in PSHA for three reasons:
1. It dictates the detailed requirements of the seismic source definition.
2. It dictates the ground motion parameters that may be estimated.
3. It is a major contributor to uncertainty in the PSHA results [McGuire and Shedlock, 1981;
Bender, 1984].
A wide variety of empirical ground motion attenuation relationships is available for application in
PSHA [Campbell, 1985; see also Chapter 5] and research has shown ground motion attenuation to be
regionally dependent. In large part, the choice of an appropriate relationship is governed by the regional
tectonic setting of the site of interest, whether it is located within a stable continental region or an active
tectonic region, or whether the site is in proximity to a subduction zone tectonic environment.
8.7.1 Impact on Seismic Source Definition

A fundamental aspect of applying attenuation relationships within the PSHA methodology is the distance
measure on which the chosen relationship is based. Two broad categories exist:
• Those based on a measure of an earthquake’s distance from the site, whether measured as an
epicentral or hypocentral distance
• Those based on a distance from the earthquake fault rupture
Epicentral distance is measured as an earthquake’s horizontal distance to a site irrespective of earth-
quake depth. Hypocentral distance is measured as the slant distance to the site, accounting for both the
horizontal and vertical distances from the earthquake to the site. Such types of attenuation relationships
do not require the specification of fault planes in the definition of seismic sources. Area source definitions
completely satisfy the application requirements of these attenuation relationships and point-source
models of earthquake rupture suffice in the PSHA model. If such relationships are used with explicit
linear or planar fault sources, the rupture dimensions of earthquakes on these sources must be constrained
to infinitely small rupture areas that approximate points. Otherwise, the distance measure may be applied
inappropriately and the ground motion hazard may very likely be overestimated.

Application of hypocentral distance measures requires that at least a top and bottom depth of the area
source be defined in order to constrain the depths of the modeled earthquakes. Typically, these depth
parameters are defined on the bases of either regional or local seismological network data and are a
measure of the thickness of brittle crust in a region. Generally, this is referred to as the thickness of the
seismogenic layer. In active tectonic regions, such as the western United States, the seismogenic layer
generally ranges between 10 and 20 km deep, depending on the locality. In the stable continental region
of the eastern United States, the seismogenic layer may range up to 40 km deep in some localities.
Most modern attenuation relationships for active tectonic regions are based on various distance
measures from the fault rupture zone (see Chapter 5). These relationships allow the specification of the
top and bottom of seismogenic faulting (as in the seismogenic layer), can accommodate specific defini-
tions of fault-dip (i.e., the inclination of the fault from horizontal), and require specification of the style
of faulting for each defined source. A significant aspect of empirical attenuation relationships is the large
scatter in the ground motion data on which the relationships are based, which results in large standard
errors about the mean relationship. Campbell [1997] was able to significantly reduce the standard
deviation of his relationships by segregating the strong motion database according to fault-rupture style
(among other non-fault parameters, such as soil type) prior to regression analysis. His relationships, as
well as others, therefore require explicit definition of fault-rupture style, consisting of normal or strike-
slip, and reverse faulting. As attenuation relationships have evolved into these more precise definitions
of the earthquake source, their application has challenged the PSHA analyst with more precise definitions
of seismic sources.
8.7.2 Reference Site Class

In addition to defining the earthquake source, application of ground motion attenuation relationships
is specific to a soil or rock type on which the PSHA ground motion estimate is to be made. These ground
types are referred to as the site class, and are defined in broad categories such as hard rock, soft rock,
firm soil, and soft soil. In some cases, such as Campbell [1997], the site classes are unambiguously defined
by soil shear-wave velocities. More commonly, the site classes are only qualitatively defined soil types.
The choice of site class is important in PSHA since soil tends to amplify long-period motion and deamplify
short-period motion over that of rock. Such site effects are embodied in the various site classes of modern
spectral attenuation relationships (see Chapter 5).
The site class that is chosen for application in PSHA, referred to as the reference site class, may depend
on a number of factors, not the least of which is the purpose of the PSHA. Site-specific engineering
PSHA evaluations are often performed to obtain a more precise measure of ground motion amplitude
than can be found in regional hazard maps contained in building code documents. If seismic engineering
for the site is to follow code procedures, then the site class for the PSHA must be consistent with the
reference site class on which the codified procedures are based. If dynamic site response analyses are to
be performed in order to produce synthetic time histories, either at the ground surface or at various
depths (e.g., for pile design), then the reference site class for the PSHA depends on the geotechnical data
that are available, the depth to which the geotechnical data extend, and the depth of bedrock at the site.
The reference site class in such case may be either soil or rock.
8.8 Accounting for Uncertainties

Development of a seismotectonic model and the many required input parameters to PSHA admits to a
wide range of interpretations and uncertainties. Median probabilistic seismic hazard models consist of
a single, best-estimate set of defined sources that are characterized by a single set of best-estimate
seismicity and fault-rupture parameters [Bernreuter et al., 1989]. Development of such models is generally
made with cognizance of diverse alternative choices, but individual modeling and parameter selections
are made relative to the purpose of the hazard estimate and with respect to the representative nature of
the individual selections. For example, the Algermissen et al. [1982] PSHA model for the United States

was premised on a single set of seismic sources and input parameters but only following consideration
of diverse views presented in a series of workshops held prior to developing the PSHA models [Thenhaus,
1983]. The Frankel et al. [1996] national PSHA model was similarly constructed following a series of
regional workshops across the United States to sample professional opinion on PSHA input, but also
contained a limited logic-tree approach to more explicitly represent critical uncertainties in the national
seismic hazard estimates.
Fully probabilistic seismic hazard models may employ a number of alternative seismic source inter-
pretations with probability distributions defined for the various seismicity and fault-rupture parameters
for each source of each model. Models similar to the fully probabilistic one are generally required to
establish a mean seismic hazard result [Bernreuter et al., 1989]. In application to the seismic hazard of
nuclear power plants in the eastern United States, Bernreuter et al. [1989] solicited 11 distinct PSHA
models from individual experts with each model weighted by its creator as to credibility. Feedback was
given to each of the experts on the ground motion consequences of their model so that each expert was
comfortable with the results. However, little interaction among the experts was promoted. In an alter-
native approach, the Electric Power Research Institute [EPRI, 1986] performed a very structured PSHA
that promoted much interaction and data exchange among experts of six Tectonic Evaluation Committees
(TEC), each of which was composed of at least one geologist, geophysicist, and seismologist. Defined
seismic sources by each TEC were thoroughly documented within a probabilistic framework as to their
physical characteristics that may promote the generation of earthquakes and as to their spatial association
with the cataloged earthquakes in the region. The result of this thorough probabilistic treatment was six
highly detailed regional seismotectonic models of the eastern United States from which complete prob-
abilistic descriptions of ground motion at the nuclear sites could be obtained. The EPRI [1986] PSHA
stands as probably the most comprehensive PSHA performed to date.
It should be pointed out that the Bernreuter et al. [1989] and EPRI [1986] PSHAs were designed to
address very low levels of risk associated with projects in the nuclear industry. Such comprehensive
investigations were developed to address annual risk levels at the power plant sites of 10–4 and lower.
Most PSHA applications address common buildings and industrial facilities where annual levels of risk
of 10–2 to 10–3 are generally acceptable as an industry or code standard. PSHA methodologies are therefore
scaled back from those of the nuclear industry for appropriateness to the project objectives and for cost
effectiveness.
There are two types of variability that can be included in PSHA. These are aleatory and epistemic
variabilities [SSHAC, 1997]. Aleatory variability is uncertainty in the data used in an analysis and
generally accounts for randomness associated with the prediction of a parameter from a specific model,
assuming that the model is correct. Specification of the standard deviation (σ) of a mean ground motion
attenuation relationship is a representation of aleatory variability. Epistemic variability, or modeling
uncertainty, accounts for incomplete knowledge in the predictive models and the variability in the
interpretations of the data used to develop the models. Aleatory variability is included directly in the
PSHA calculations by means of mathematical integration. Epistemic uncertainty, on the other hand, is
included in the PSHA by explicitly including alternative hypotheses and models. This uncertainty can
be accounted for through the evaluation of multiple individual seismotectonic models or through the
formulation of a logic tree that includes multiple alternative hypotheses in a single model. The logic tree
allows a formal characterization of uncertainty in the analysis by explicitly including alternative inter-
pretations, models, and parameters that are weighted in the analysis according to their probability of
being correct. Logic tree models may be exhaustively evaluated, or adequately sampled through Monte
Carlo simulation, which is computationally a more efficient procedure.
An example of a logic tree seismotectonic model is shown in Figure 8.11. Each alternative hypothesis,
indicated by a branch of the logic tree, is given a subjective weight corresponding to its assessed likelihood
of being correct. The proposed alternative hypotheses account for uncertainty in earthquake source
zonation, maximum magnitude, earthquake recurrence rate, location and segmentation of seismogenic
faults, style of faulting, distribution of seismicity between faults and area sources, and ground motion
attenuation relationships.

0068_C08_fm Page 26 Monday, August 19, 2002 1:13 PM
Attenuation Fault Seismic Segmentation Segmentation Segments Status of Total Fault Dip Maximum Recurrence Recurrence
Relationship Recurrence Sources Model Activity Length Magnitude Data Rate
Model
Unnamed
5000
Unsegmented Cominston 65° 0.04
0.2 0.5 .3333
Oquirrh Mtns Ogden 0.28
Campbell Model A Active 37 km
Wasatch SLC Fault 2500
0.4 1.0 1.0 1.0 0.48
0.333 Provo 7.0
East Coshe
Segmented 0.15 2000
45°
Nephi 7.25 0.16
Exponential West Valley
0.7
0.8 Levon 0.5
0.3 Hansel Valley 7.5 1667
0.04
Model B 0.15
Sadigh 6.0 Segment 8.5
Bear Lake 0.333
0.6 0.3 1.0
Backgnd Active
0.333 6.25 9.3
N/A N/A 1.0 N/A N/A 0.5 N/A 0.334
Characteristic 6.5 9.8
0.7 0.2 0.333
Joyer-Fumal
0.333
FIGURE 8.11 Example logic tree characterization for seismic sources. (From Youngs, R.R. et al., 1987, “Probabilistic
Analysis of Earthquake Ground Shaking along the Wasatch Front, Utah,” in P.L. Gori and W.W. Hays, Eds., Assessment
of Regional Earthquake Hazards and Risk along the Wasatch Fault, Utah, U.S. Geological Survey Open File Rep. 87-585,
pp. M1-M110.)
8.9 Typical Engineering Products of PSHA

The fundamental engineering product of PSHA is an amplitude of some ground motion parameter that
is associated with a particular return period. This probabilistic format of relating ground motion ampli-
tude to a specific return period is now commonplace in a number of seismic design codes and recom-
mended practices, including those of the National Earthquake Hazards Reduction Program (NEHRP),
the International Building Code (IBC), the National Fire Protection Association (NFPA), the American
Petroleum Institute (API), and the International Standards Organization (ISO), among others.
Probabilistic results can be presented in a number of formats. Perhaps the most widely recognized
product is that of a ground motion hazard map, such as produced by the U.S. Geological Survey under
the National Earthquake Hazards Reduction Program for the United States (Figure 8.12) and one of the
world that was recently produced under the auspices of the Global Seismic Hazard Assessment Program
(GSHAP) [Giardini, 1999; Figure 8.13]. Such maps illustrate the regional differences in ground motion
amplitude (typically peak ground acceleration, or PGA) at a constant return period (i.e., a constant
probability of exceedance). These maps allow the rapid comparison of the seismic hazard for regions
and the identification of the most hazardous regions, on a uniform basis.
A common goal of hazard models is to rapidly estimate a hazard curve for a particular engineering
site of interest. The hazard curve is a plot showing the change in ground motion amplitude relative to
return period (Figure 8.14). Ground motion amplitude always increases with increasing return period
in Poisson hazard models and, for a single fault, will asymptotically approach the mean ground motion
amplitude of the maximum magnitude earthquake at the given source-to-site distance when attenuation
variability is not included in the estimate.
Another common product in engineering PSHA is the constant-probability, or uniform hazard,
response spectrum (Figure 8.15). These curves illustrate ground motion amplitudes over a number of
oscillator periods of engineering interest at a constant return period. Comparison of such curves at a
constant return period, but for various site classes, illustrates the modification of ground motion ampli-
tudes by various types of soil and rock. Soils tend to deamplify short-period motion, but amplify long-
period motion. Rock sites have the inverse effect. The probabilistic response spectrum is generally smooth

0068_C08_fm Page 27 Wednesday, August 14, 2002 7:11 AM
Peak Acceleration (%g) with 10% Probability of Exceedance in 50 Years

50° -120
° site: NEHRP B-C boundary
-70° 50°
180
-110°
-100° -90° -80° 100
80
60
40
30
25
40°
40° 20
15
10
9
8
7
6
5
30°
30° 4
3
2
1
Nov. 1996 0
-120
° -70°
-110° -80°
-100° -90°
U.S. Geological Survey
National Seismic Hazard Mapping Project
FIGURE 8.12 Four hundred seventy-five-year return period ground motion hazard map of the conterminous United
States. (From Frankel, A., 1996, U.S. Geological Survey Open-File Rep. 96–532. With permission.)
FIGURE 8.13 Four hundred seventy-five-year return period ground motion hazard map of the world. (From
Giardini, D. Ann. Geofis., 42, 1999. http://www.gfz-potsdam.de/pb5/pb53/project/gshap/final_result.html.)

Peak Ground Acceleration - Site Class=C - Soft Rock
10 4
Return Period (years)

10 3
10 2
Mean
5th%
16th%
50th%
84th%
95th%
10 1
0.0 0.1 0.2 0.3 0.4
Horizontal Acceleration (g)
FIGURE 8.14 Example of a peak ground acceleration (PGA) hazard curve on a soft rock site class for a logic-tree
seismotectonic model in a low seismicity region of Eurasia. Mean estimate and confidence intervals are shown by
different line types in the legend of the figure.
and broader in shape than the response spectrum obtained from an actual earthquake recording, since
the probabilistic spectrum is the result of the aggregated ground motion contributions from all magni-
tudes and distances of significance to a site, weighted by their frequency of occurrence.
Median hazard models result in a single estimate of the uniform hazard spectrum (UHS). More fully
probabilistic PSHA models can represent the spread of UHS results accounting for uncertainty (e.g.,
through a logic tree) and are often presented by percentile levels (Figure 8.16). Such a complete description
of the seismic hazard allows explicit representation of various levels of conservatism that may be applied
in seismic engineering design.
8.10 PSHA Disaggregation

The PSHA methodology aggregates ground motion contributions from earthquake magnitudes and
distances of significance to a site of engineering interest and, as such, the PSHA results are not represen-
tative of a single earthquake. However, engineering models and computer codes generally require empir-
ical or synthetic earthquake acceleration time series as input to dynamic analyses. Specific magnitudes
and distances are also often required in slope stability and liquefaction analyses. An issue, then, is the
selection of representative earthquake time acceleration series given a probabilistic uniform hazard
response spectrum. A procedure called disaggregation (or deaggregation) has been developed to examine
the spatial and magnitude dependence of PSHA hazard results. Considerable attention has recently been
focused on PSHA disaggregation in recent research literature [e.g., Stepp et al., 1993; Cramer et al., 1996;
Chapman, 1995; McGuire, 1995; Bazzurro and Cornell, 1999; Harmsen et al., 1999].

5%-Damped Horizontal Acceleration (g)
0.1
Hard Rock
Soft Rock
Firm Soil
Soft Soil
0.01
0.001 0.01 0.1 1 10
Period (sec)
FIGURE 8.15 Example median uniform hazard response spectra (UHS) for a moderate seismicity site for various
site classes.
The PSHA results are disaggregated to determine the magnitudes and distances that contribute to the
calculated exceedance frequencies (i.e., the hazard) at a given return period and at a structural period of
engineering interest (typically, the fundamental period of a structure). In this process, the hazard for a
given return period and at a specified ground motion period is partitioned into selected magnitude and
distance bins and the relative contribution to the hazard of each bin is calculated by dividing the bin
exceedance frequency by the total exceedance frequency of all bins. The bins with the largest relative
contributions — the modes — identify those earthquakes that contribute the most to the total hazard.
If there are no clear modes, the controlling or design earthquakes are typically defined by the mean
magnitude and mean distance. These results are displayed as a histogram giving the percent contribution
to the specified hazard of those earthquakes that are capable of causing ground motions equal to or
greater than that corresponding to this hazard as a function of magnitude and distance. This histogram
will be different for spectral accelerations of varying structural periods because of the difference in the
way these spectral values scale with magnitude and distance. The relative frequencies specified by these
histograms can be used to develop mean estimates of magnitude and distance, or to identify the modal
contributions to the site hazard, in order to define a set of controlling or design earthquakes corresponding
to specified structural periods and return periods. These design earthquakes can then be used as bases
to select or construct input time histories for use in a dynamic site-response analysis or for ground-
failure evaluation. Figures 8.17 and 8.18 illustrate disaggregation plots. Figure 8.17 is unimodal and the
mode of the distribution is relatively clear from the plot. Figure 8.18, however, illustrates a relatively
strong bimodal contribution to PSHA hazard at the specific return period and structural period at which
the disaggregation was performed. In such cases, two or more design earthquakes might need to be
specified in order to fully represent the spectral content of ground motion expected at the site. Note that
using a mean magnitude and distance for this bimodal distribution would result in a large magnitude
earthquake that has no physical association with a known active fault.

FIGURE 8.16 Example uniform hazard response spectrum for a single site class showing confidence limits obtained
from a fully probabilistic application of the PSHA methodology. Note that because of lognormal distributions in
many parameters used in PSHA, the mean estimate is always higher than median estimate.
The expected (median) ground motion amplitude corresponding to the disaggregated mean or modal
magnitude and distance can be calculated by substituting these values into the attenuation relationships
that were used in the PSHA. The difference between the logarithm of the ground motion value corre-
sponding to the PSHA hazard at the return period of interest and the logarithm of the ground motion
value for the disaggregated mean magnitude and distance, divided by the logarithmic standard error of
estimate of the attenuation relationship, is referred to as ε (epsilon) [McGuire, 1995]. ε is the number
of standard deviations that the probabilistically derived ground motion amplitude deviates from the
median ground motion amplitude for an event defined by the mean magnitude and distance. An ε of 1
indicates that the probabilistic value of ground motion corresponds to the one-standard-deviation value
of the deterministic ground motion. The larger the absolute value of ε, the greater the contribution of
ground-shaking variability to the calculated hazard. Very large absolute values of ε should be avoided,
since they indicate that the hazard is potentially being overly dominated by variability in the attenuation
relationship.
8.10.1 Scaling Empirical Earthquake Spectra

As previously mentioned, a common application of the design earthquake parameters resulting from a
PSHA disaggregation is the identification of suitable earthquake records to be used in dynamic engineer-
ing testing and design. The defined magnitude and distance parameters from the disaggregation serves
as a guide in the selection of three-component (two horizontal and one vertical) empirical earthquake
time series from appropriate recording stations of historical earthquakes. The design earthquake param-
eters are only a general guide, however, with other factors such as site class, earthquake mechanism (i.e.,
style of slip), and representative spectral shape also having a bearing on the record selection. All of these
parameters cannot usually be completely satisfied in the record search, and prioritization of the search

10.0
Percent Contribution to Exceedance

9.0
8.0
7.0
6.0
5.0
4.0
3.0
2.0
1.0
0.0
5
15
25
35
45
55
65
75
85
95
105
Dista
8.00
115
nce (
7.00
125
km)
135
6.00
145
e
5.00
it ud
gn
Ma
FIGURE 8.17 PSHA disaggregation plot showing a typical unimodal distribution of earthquake magnitude and
distance to ground motion exceedance frequency. The mode of this distribution is a magnitude 6.5 to 7.0 earthquake
occurring at a distance of 5 to 10 km from the site.
criteria is required. Typically, the spectral amplitudes of the identified suitable records do not match the
UHS (i.e., the target spectrum) within the period band of engineering interest, and scaling of the empirical
records is required. The scaling is commonly performed on some average of the spectra of the two
horizontal components of motion. There are a number of averaging methods that can be applied,
including a simple mean, a geometric mean, and the square root of the sum of squares (SRSS), among
others. The geometric mean is defined with respect to the two horizontal components as (H1*H2)1/2 and
may be preferred as it is the averaging method consistent with that applied in empirical ground motion
attenuation relationships that are used to establish the UHS. Once the mean spectrum of the two
horizontal components of motion is established, it can be scaled to a best-fit criterion to the target
spectrum within the period band of engineering interest (Figure 8.19). The best-fit criterion is defined
as a fit where 50% of the empirical spectral amplitudes are both above and below the target spectrum
within the engineering period band of interest. The scale factor to achieve the best-fit spectrum may
then be applied to each horizontal component time series to obtain empirical motions that are repre-
sentative of expected earthquake ground motions at the site for a return period of interest. If dynamic
response analyses are to be performed on the soil column at the site, the above procedure can be performed
for a reference site class representative of the input soil layer for the empirical time series.
8.11 PSHA Case Study

This section presents an illustrative case study of a simple, median, peak ground acceleration (PGA)
hazard assessment along the proposed offshore Oman India pipeline route. The proposed Oman India
pipeline traverses approximately 1135 km of the northern Arabian Sea floor and adjacent continental
shelves at water depths of over 3 km on its route from Ra’s al Jifan, Oman, to Rapar Gadhwali, India
(Figure 8.20). Ground-shaking hazard was quantified in terms of PGA for return periods of 200, 500,
and 1000 years using the PSHA computer program Seisrisk III [Bender and Perkins, 1987].

6.0
Percent Contribution to Exceedance
5.0
4.0
3.0
2.0
1.0
0.0
5
15
25
35
45
55
65
75
85
95
Dis
tan
105
ce
115
8.00
(km
125
)
7.00
135
6.00
145
e
ud
5.00
it
agn
M
FIGURE 8.18 PSHA disaggregation plot showing a bimodal distribution of earthquake magnitude and distance to
ground motion exceedance frequency. The primary mode of this distribution is a magnitude 6.5 to 7.0 earthquake
at 80 to 85 km from the site. The secondary mode is a magnitude 6.5 to 7.0 earthquake at 5 to 10 km from the site.
This distribution is for a site near the eastern coast of Honshu, Japan that is in close proximity to a shallow active
fault (secondary mode), although the Japan Trench subduction zone is the strongest contributor to the hazard
(primary mode).
This summary is excerpted from the original publication [Campbell et al., 1996] and the reader is
referred there for further discussions and a full citation of references on which the work is based. This
PSHA was performed in 1995. On January 26, 2001, the MW 7.9, Bhuj (Gujarat, India) earthquake struck
the Kutch region of northwestern India in the vicinity of the western terminus of the planned pipeline.
This region was identified in the PSHA as one of high hazard. Thus, the case study serves as an example
of the utility of PSHA in the engineering and planning of future development. The case study further
illustrates that generally conservative results are obtained from time-independent, stationary models of
seismicity if parameter estimates are made carefully and estimates of maximum magnitude earthquakes
are realistic and not merely based on the maximum earthquake observed in a short reporting period [see
McGuire and Barnhard, 1981].

5%-Damped Horizontal Acceleration (g)
0.1
0.01
200 Yr. UHS

H1
H2
Mean
0.001
0.01 0.1 1 10
Period (sec)
FIGURE 8.19 Example scaled empirical earthquake spectra to a target, 200-year UHS. Solid bold line is the 200-
year return period target UHS. Heavy dashed line indicates the geometric mean of the two horizontal components.
Light line styles are the H1 and H2 components. The geometric mean of the empirical spectra has been scaled to a
best-fit criterion within the period band of 0.6 ± 0.5 seconds.
55° 60° 65° 70° 75°

PAKISTAN
IRAN
25° MAGNITUDE
ge
8 to 8.2 (2)
id
Rapar Gadhwali
R
7 to 7.9 (12)
y
OMAN
ra
6 to 6.9 (40)
ur
Ra'as al Jifan
M
INDIA 5 to 5.9 (223)

e 4 to 4.9 (876)
lin
pe
Oman - India Pi 3 to 3.9 (47)
ne
20° all others (89)

Zo
ARABIAN SEA
re
ctu
Fra
0 250 500
Owen
Kilometers
FIGURE 8.20 Location map of the Oman India pipeline route shown in relation to the regional distribution of
earthquakes.

20° 30° 40° 50° 60° 70° 80° 90°
40°
40°
30°
30°
20°
20°
10°
10°
0°
40° 50° 60° 70° 80°
FIGURE 8.21 Plate tectonic setting of the region surrounding the northern Arabian Sea showing regional tectonic
features. (After Jacob, K.H. and R.L. Quittmeyer, Geodynamics of Pakistan, Geological Survey of Pakistan,
pp. 305–317, 1979. With permission.)
8.11.1 Tectonic Setting

Within the Arabian Sea basin, the Owen Fracture Zone and the Murray Ridge are two distinct structural
geologic provinces of a transform plate boundary that accommodates slow, right-lateral differential
motion between the Arabian and Indian plates (Figure 8.21). The northern tectonic boundaries of the
Arabian and Indian plates are in continental collisional contact with the Eurasian plate, giving rise to
the complex compressional tectonics of the Zagros and Himalayan Mountains. However, only in the Gulf
of Oman is subduction currently occurring (i.e., where the ocean floor of the Arabian Sea is being thrust
beneath the overriding Eurasian plate). Rupture along the contact between these two plates results in
great thrust earthquakes, as demonstrated by an MS 8.2 earthquake and accompanying tsunami that
occurred off the coast of Pakistan in 1945 (Figure 8.20).
8.11.2 Regional Seismicity

The seismicity of the Arabian Sea and surrounding areas is dominated by earthquakes along the major
tectonic plate boundaries (Figures 8.20 and 8.21). The primary sources of earthquake data used for the
study were the Catalog of Middle East Earthquakes and the Catalog of Earthquakes for Peninsular India,
1839–1900. The Middle East catalog contains more than 22,000 earthquakes that occurred from 1900 to
1983. These catalogs were supplemented with several historic earthquake accounts of India. Seismicity
between 1983 and 1992 was taken from the Preliminary Determination of Epicenters (PDE), published
monthly by the U.S. Geological Survey.
The regional earthquake catalogs were aggregated into a single catalog by removing duplicate entries
through an automated winnowing procedure. The aggregated earthquake catalog used surface-wave
magnitude (MS) to quantify earthquake size. MS was converted to moment magnitude (MW), the

magnitude measure used in the strong-motion attenuation relationships described later, using relation-
ships between magnitude measures given in Khattri et al. [1984], Hanks and Kanamori [1979], and
Ekstrom and Dziewonski [1988].
The aggregated earthquake catalog was culled of aftershocks and the remaining earthquakes were
analyzed for completeness by determining the time period over which events of various magnitudes were
found to be completely reported. The analysis indicated that earthquakes of various magnitudes are
completely reported for the following periods: MW 5.0 ± 0.2 for the past 20 years; MW 5.4 ± 0.2 for the
last 30 years; MW 5.8 ± 0.2 for the last 60 years; and MW 6.0 for the last 80 years. Complete reporting
times for larger earthquakes range from about 100 years for MW 6.2 to about 300 years, or the total period
of the historic catalog, for great earthquakes (MW ≥ 8.0).
8.11.3 Great Earthquakes

On June 16, 1819, a great (MS ≈ 8) earthquake occurred in the Rann of Kutch, India near the eastern
terminus of the pipeline route. This earthquake was not on a plate boundary but, rather, was an intraplate
earthquake similar to those that occurred in New Madrid, Missouri in 1811 and 1812. The Kutch
earthquake was accompanied by surface rupture along a zone measuring 32 km by 16 km. The Kutch
and New Madrid earthquakes are two examples of only a few historic intraplate earthquakes that have
ruptured the ground surface. The Kutch earthquake caused damage over an extensive area. In the town
of Bhuj nearly 7000 houses were “overthrown” with a reported 1140 casualties. The recent MW 7.9 Gujuarat
earthquake that occurred on January 26, 2001 also devastated the city of Bhuj as well as the larger region
of Gujuarat State in northwestern India [EERI, 2001].
Another great (MS 8.2) earthquake ruptured the plate interface zone of the Makran Subduction Zone
on November 27, 1945. Historic accounts indicate a rumbling sound accompanied the earthquake at
Karachi, Pakistan, but only a few windows were reported broken. A tsunami followed the earthquake
and was most severe on the Makran coast, northwest of Karachi, where a telegraph office was reportedly
washed to sea and a number of buildings were damaged. In the suburbs of Bombay, India, boats were
reportedly smashed at their moorings. Two new islands had appeared in the Arabian Sea about 180 miles
west of Karachi after the earthquake.
8.11.4 Earthquake Source Characterization

The earthquake source characterization of the northern Arabian Sea and surrounding areas consisted of
defining earthquake source zones and earthquake recurrence relationships that accurately modeled the
occurrence of earthquakes in space and time within the study area. This model was developed to be
consistent with the contemporary understanding of the tectonics and seismicity of the region as docu-
mented in the technical literature. The earthquake source zones are shown in Figure 8.22.
Earthquake recurrence was modeled using the Gutenberg–Richter relationship, Equation 8.5. Earth-
quakes were modeled between a minimum magnitude of 5.0 and the magnitude of the largest earthquake
judged capable of occurring within each of the earthquake source zones.
8.12 The Owen Fracture Zone–Murray Ridge Complex

The tectonic model of the Owen Fracture Zone–Murray Ridge Complex (Figure 8.22) consists of linear
representations of young faults that have been identified in a number of offshore seismic reflection and
geophysical studies.
South of latitude 20°N, the Owen Fracture Zone was modeled as a series of linear faults along the
eastern side of a series of narrow, discontinuous ridges and troughs, consistent with interpretations of
faulting in seismic reflection profiles. There are significant structural differences in the Owen Fracture
Zone north of about 29°20'N. At this latitude, the narrow, linear trend of the southern and central
segments of the Owen Fracture Zone merges with the broader structure of the Qalhat Seamount.

55° 60° 65° 70° 75°

PAKISTAN
n
Chama
IRAN
ZN 5 ZN 4
Omach-Nai
25°
Makran Subduction Zone ZN 3
ZN 2
North
ge
South
Rapar Gadhwali
id
(Dalrymple
R
Trough)
OMAN
y
ra
ur
Ra'as al Jifan
INDIA
M
North (Qalhat
e
Seamount) l in
pe
Oman - India Pi
20°
ne
lt
Zo
u
Fa
Central
ah
ARABIAN SEA
re
alr
ctu
qu
ZN 6 ZN 1
Si
Fra
Owen
South
FIGURE 8.22 Seismic sources used in assessing probabilistic peak ground acceleration (PGA) hazard along the
Oman India pipeline. Faults are shown as bold lines. Area sources are ruled. Vertical ruled area of the Makron
Subduction Zone interface thrust fault dips northward beneath Zone 5.
At the northern end of the Owen Fracture Zone, a low basement ridge defines a southwestern extension
of the Qalhat Seamount, some 60 km northwest of the main plate boundary. Geophysical data suggest
that this western margin of the Qalhat Seamount could be fault-bounded. Because this seamount is an
integral part of the transform plate boundary, we modeled bounding faults along both the eastern and
western flanks of this feature. Short, east-trending cross-faults were modeled between the east and west
flanking faults, consistent with interpretations of geophysical data.
Bathymetric and geophysical data indicate that most of the ridges and intervening troughs associated
with the Murray Ridge are fault-bounded. Although the overall structure of the Murray Ridge is a
bathymetric high, it is split by deep, fault-controlled troughs that are floored by very recent sediments.
The Dalrymple Trough is a prime example of one of these troughs. The present floor of the Dalrymple
Trough has down-dropped 800 m below the adjacent Arabian Sea floor. Based on these interpretations,
we modeled all linear bathymetric ridges and intervening trough with bounding faults.
8.12.1 Maximum Magnitude

Burr and Soloman [1978] provide a comprehensive account of the source parameters of all large earth-
quakes that had occurred on oceanic transforms up to the mid-1970s. Of the 36 earthquakes studied, 35
had M ≤ 7. The largest earthquakes that have occurred in the vicinity of the Owen Fracture Zone–Murray
Ridge Complex are three MS 5.9 events that occurred from 1933 to 1935. Since 1964, the largest reported
earthquake on this zone was MS 5.7. We adopted a maximum magnitude of 6.8 for all segments of the
oceanic plate boundary  a value only slightly smaller than that observed in association with oceanic
transform earthquakes. Because a slip-rate model used to determine earthquake frequencies uses maxi-
mum magnitude as one of its parameters, an upper bound of 6.8 rather than 7.0 was chosen to give
frequencies for moderate earthquakes that are consistent with those observed since 1964. A larger max-
imum magnitude allows too much of the seismic moment release along the transform boundary to be
taken up by infrequent, large earthquakes, resulting in unreasonably long repeat times for magnitude 5
to 6 earthquakes.

8.12.2 Earthquake Recurrence Frequencies

A moment-based frequency model was used to estimate earthquake recurrence rates from tectonic slip
rate, Mmax , b-value, crustal shear rigidity, and the seismogenic area of the fault. We used a slip rate of
1.75 mm/year for this analysis, which is the average rate determined from worldwide models of rigid
plate interactions. Earthquake recurrence rates were calculated for each structural segment along the
length of the oceanic transform plate boundary using the length of each structural segment, a crustal
thickness of 8 km, and a shear rigidity of 3 ×1011 dyne/cm typical of crustal rock. A b-value of 0.9 was
determined from a maximum likelihood fit of 39 earthquakes of M ≥ 4.8 that have occurred on, or in
the vicinity of, the oceanic transform boundary.
Except for the northern (Qalhat Seamount) segment of the Owen Fracture Zone, the calculated
recurrence rates were distributed equally among all modeled faults that define each structural segment
of the transform boundary. The recurrence rate for the eastern and western flanking faults of the northern
Qalhat Seamount segment was distributed in such a way as to give twice as much seismicity to the eastern-
bounding fault. This was done to account for both the greater degree of fault offset and the higher historic
rate of earthquakes along the eastern margin of this portion of the plate boundary (Figure 8.20).
A comparison of the earthquake recurrence rates determined from the assumed slip rate and the
maximum likelihood fit of the 39 earthquakes that have been observed along the entire length of the
Owen Frature Zone–Murray Ridge Complex indicates that the estimated annual frequency of earthquakes
of 4.8 ≤ M ≤ 6.2 given by both procedures agrees to within ±10%.
8.13 Makran Subduction Zone

The Makran Subduction Zone megathrust boundary between the Arabian and Eurasian plates dips gently
northward beneath the coasts of Iran and Pakistan from its origin at a sediment-filled trench off the
Makran continental slope [Jacob and Quittmeyer, 1979] (Figure 8.22). Earthquakes located approximately
400 km north of the trench at a depth of about 60 km have focal mechanisms indicative of extensional
stresses in the subducting slab of Arabian Sea floor. Normal-faulting earthquakes in this “slab-bend”
region of the subducting plate is a common feature of subduction zones throughout the world. They
represent a transition region within the subducting slab from compression trenchward of this zone to
extension in the deeper part of the slab, where it is being consumed within the Earth’s mantle. This region
marks the maximum possible extent of the seismogenic portion of the plate interface beneath Pakistan
and Iran.
A cross-sectional plot of seismicity across the Makran region was used to identify the top of the down-
going slab as it penetrates the Earth’s mantle. This cross section indicates that the seismogenic portion
of the plate interface extends some 200 to 300 km north from the oceanic trench offshore southern
Pakistan and southeastern Iran to a depth of approximately 30 km. Based on this observation, we modeled
the megathrust interface of the Makran Subduction Zone as a 250-km-wide, shallow (~7°) planar fault.
Based on tectonic considerations, the Makran Subduction Zone was estimated to be approximately 1000
km long, representing that portion of the Arabian–Eurasian convergence zone between the Ornach-Nal
fault on the east and the “Oman line” on the west.
The convergence rate between the Arabian and Eurasian plates across the Makran Subduction Zone
has been estimated to be about 5 cm/year from regional and worldwide geodynamic models of plate
interactions. This convergence rate reflects the total deformation rate between the two plates, which is
manifested in the subduction of the Arabian plate beneath the Eurasia plate. Some of this convergence
is very likely accommodated aseismically along the interface, and in folding of the overriding southern
edge of the Eurasian plate in the Makran Mountains of southern Pakistan and southeastern Iran. Geologic
studies that might indicate how this overall convergence rate is partitioned between seismic and aseismic
compressional deformation processes have not yet been performed.


The largest known earthquake on the Makran Subduction Zone was an MS 8.2 earthquake in 1945.
However, there is no reason to believe that the entire plate interface cannot rupture in a single “mega-
thrust” earthquake. Based on this hypothesis, we assigned a maximum magnitude of 9.2 to this zone,
consistent with both empirical rupture area–magnitude relationships and the magnitude of the great
1964 Prince William Sound, Alaska earthquake that ruptured a similar length.

Using seismic moment-based relationships between convergence rate, crustal shear rigidity, and the
cross-sectional area of the converging plates, we estimated the recurrence rate of earthquakes for
various magnitudes that would be expected to occur on the Makran Subduction Zone assuming that
all 5 cm/year of the estimated convergence rate was released in earthquakes. The resulting recurrence
rates were found to be significantly larger than those observed historically. Probable reasons for this
inconsistency are that some of the convergence is released by earthquakes occurring within the over-
riding Eurasian plate and some is accommodated through aseismic deformation.
Because of this inconsistency, we adopted area-normalized earthquake recurrence rates that were
developed from the historical occurrence of earthquakes in this zone [Khattri et al., 1984]. These rates,
which were defined in MS , were adjusted to correspond to MW using common empirical relationships
between these two magnitudes’ measures. Khattri et al.’s [1984] recurrence frequencies were normalized
to a 40-year time period, not the 1-year time period that is usually used to develop earthquake recurrence
relationships. For consistency with other recurrence relationships developed in this study, we converted
these rates to an annual rate.
We partitioned these recurrence frequencies into two parts. Earthquakes of 7.6 ≤ M ≤ 9.2 were assumed
to occur on the plate interface. Earthquakes of M < 7.6 were assumed to occur within the shallow crust
of the overriding Eurasian plate. This partitioning results in an average recurrence interval of 200 years
for earthquakes of similar size to the 1945 earthquake, an interval that is consistent with the observation
that only one event of this size has occurred on this zone in 300 years for which the earthquake catalog
is considered to be complete in this region. Using a range of magnitudes to model the occurrence of
earthquakes on the plate interface was consistent with other seismological investigations that suggested
the plate interface is a segmented thrust fault.
8.14 Southwestern India and Southern Pakistan

The Ornach-Nal and Chaman faults accommodate left-lateral movement between the Eurasian and
Indian plates in southern Pakistan (Figure 8.22). Earthquakes of M ≥ 6.4 on these faults were modeled
as linear ruptures. Throughout a broader zone, mostly east of the Ornach-Nal and Chaman faults (ZN 4,
Figure 8.22), earthquakes smaller than 6.4 were modeled as point sources uniformly distributed through-
out the zone. The area source encompasses a region of diffuse seismicity related to a broader zone of
deformation within southern Pakistan.
Shallow earthquakes of M ≤ 7.6 occurring within the Eurasian plate were placed in a broad source
zone in southern Pakistan and southeastern Iran that encompasses the Makran Mountains (ZN 5, Figure
8.22). The Makran Mountains form the leading edge of crustal deformation along the southern boundary
of the Eurasian plate above the Makran Subduction Zone. Crustal faults of this zone characteristically
trend east–west and have both thrust and strike-slip components of slip. Compressional deformation
and seismicity are widespread and are not limited to a few major faults. We have modeled earthquakes
of M ≥ 6.4 within this zone as linear ruptures on a series of uniformly spaced faults that follow the
predominant east–west strike of the structural trend of the region. Smaller earthquakes are modeled as
point sources distributed uniformly throughout the zone.

We adopted the earthquake source zones proposed by Khattri et al. [1984] for our source zones in
western India (ZN 1 to 3, Figure 8.22). These zones were developed based on the association of clusters
of historic earthquake occurrences with ancient tectonic trends of the Indian subcontinent  a common
technique used throughout the world for developing seismic source zones in intraplate environments.
Earthquakes of M < 6.4 were modeled as point sources uniformly distributed within the source zones.
Larger-magnitude earthquakes in the Kutch and West Coast of India source zones (ZN 1 and 2) were
modeled as linear ruptures on faults representing the major tectonic trends in these regions. In the Kutch
zone, these modeled faults strike east-west following the Kutch Rift-Delhi Trend. Notably, the January
26, 2001 Bhuj (Gujarat, India) earthquake exhibited east-west-trending surface rupture along this same
trend. In Zone 1 (Figure 8.22), we modeled a single north-trending fault to coincide with the Panvel
Flexure-West Coast fault zone.

The maximum magnitude associated with shallow seismicity in the Makran Mountains (ZN 5,
Figure 8.22) was estimated to be about 7.6 (earthquakes larger than this were constrained to occur on
the plate interface of the Makran Subduction Zone). This is about one-half magnitude higher than the
largest observed crustal earthquake in this zone. It is, however, generally consistent with calculated
maximum magnitudes for faults with lengths of up to 120 km that have been mapped in the Pakistani
portion of this zone.
Khattri and others [1984] assigned a maximum magnitude of 8 to both the Kutch and West Coast of
India source zones (ZN 1 and 2, Figure 8.22) based on the historic occurrence of the 1819 Kutch
earthquake (MS ~ 8) and the presence of Quaternary deformation along the older Panvel Flexure tectonic
trends in the West Coast of India zone. They assigned a relatively small maximum magnitude of 6 to the
Arravali source zone (ZN 3, Figure 8.22), consistent with its relatively small source dimensions and
historical seismicity. We adopted these maximum magnitudes for this study.

We adopted the area-normalized earthquake recurrence frequencies given in Khattri et al. [1984] for all
of the source zones in this region (ZN 1 to 5, Figure 8.22). These rates were annualized and adjusted to
MW as discussed previously for the Makran Subduction Zone. Earthquake recurrence rates for M ≥ 6.4
earthquakes in the Kutch and West Coast of India source zones (ZN 1 and 2) were uniformly distributed
among the modeled faults in these zones in proportion to their total fault lengths.
8.15 Southeastern Arabian Peninsula and Northern Arabian Sea

The Arabian Sea basin and eastern Arabian Peninsula were placed in a broad background zone of diffuse
seismicity (ZN 6, Figure 8.22). This zone is characterized by the infrequent occurrence of small to
moderate earthquakes. The total area of comparable background seismicity is quite large, extending
beyond the study area westward across the entire Arabian Peninsula to the Red Sea Rift and the Lavant
Transform plate boundaries. Earthquakes of M ≥ 6.4 within the study area were constrained to two faults
on the Omani continental shelf that have been identified as being potentially active from offshore
boreholes and geophysical investigations. We included the Siquirah fault (Figure 8.22) based on an
interpretation of reflection profiles that suggests that it is a profound structural feature of the southeast
Omani outer-continental shelf. The fault displaces sea-bottom reflectors, indicating recent movement,
although no earthquake epicenters plot near the fault (Figure 8.20). The northward extend of the Siquirah
fault was based on the interpretation that the entire coastline and linear continental slope of eastern
Oman is possibly the result of strike-slip faulting. The lack of epicenters in the vicinity of the Siquirah
fault might only indicate a recurrence interval of significant earthquakes that is longer than the historic
period of observation.

A second, shorter fault located east of the Siquirah fault (Figure 8.22) and near the base of the Oman
continental slope bounds a narrow linear basement ridge. The ridge forms the eastern structural boundary
of a sedimentary trough between the continental shelf and floor of the Oman basin. Consistent with our
interpretation of similar features within the Murray Ridge, we modeled this basement ridge as being
fault-bounded. We refer to this fault as the Oman Basin fault.

We assigned a maximum magnitude of 6.4 to the background zone (ZN 6, Figure 8.22). This value is
three quarters of a magnitude higher than that observed historically. We believe that the higher maximum
magnitude unit was warranted because of the large degree of uncertainty associated with the extremely
low level of seismicity in this region.
We adopted a maximum magnitude of 7.6 for the Siquirah fault based on empirical rupture
length–magnitude relationships and the assumption that one half of the total length of the fault could
rupture in a single event. Similarly, we assigned a maximum magnitude of 7.2 to the much shorter Owen
Basin fault based on the assumption that the entire length of that fault could rupture in a single event.

Since only five historic earthquakes have occurred in this region, it was not possible to develop earthquake
recurrence frequencies using a formal statistical procedure. Instead, we adopted the b-value that was
determined from the maximum likelihood fit of 39 earthquakes that occurred within the Owen Fracture
Zone–Murray Ridge Complex. We estimated the earthquake recurrence rate in this zone from the
observation that three MS 4.1 to 5.0 earthquakes had occurred in this region within the last 22 years.
Lacking available geologic or geodynamic data for the Siquirah and Owen Basin faults, we simply used
the recurrence relationship developed for the background zone to model the cumulative occurrence of
M ≥ 6.4 earthquakes on these faults. Based on this model, the calculated cumulative recurrence interval
for earthquakes of M ≥ 6.4 is 400 years. This estimate is consistent with the lack of observed earthquakes
along this portion of the Omani continental shelf within the last few hundred years.
8.16 Ground Motion Models

PGA was estimated from attenuation relationships of the form:
log(PGA) = b 1+ b 2M – b 3logR – b 4R + ε (8.23)
where PGA is the mean horizontal component of peak ground acceleration (g), M is earthquake mag-
nitude (MW), R is distance from the earthquake source to the site (km), ε is a random error term with
a mean of zero and a standard deviation equal to the standard error of estimate of log (PGA), and b1
through b4 are parameters dependent on the tectonic environment.
The attenuation relationships adopted for this study were chosen to represent as closely as possible
the earthquake and propagation characteristics of the major tectonic environments encountered in the
Arabian Sea and surrounding regions. Separate relationships were used to model the stable oceanic and
continental interior regions of the Arabian and Indian plates; the oceanic and continental transform
boundaries between the Eurasian, Arabian, and Indian plates; and the megathrust interface of the Makran
Subduction Zone.
8.16.1 Stable Continental Interior Earthquakes

Stable continental interior regions, such as the Arabian Peninsula (Oman) and the Indian subcontinent
(southwestern India), are composed of very old, Precambrian crust, often referred to as stable continental
interiors. These regions are known to occasionally produce earthquakes that have relatively high stress

drops and relatively low anelastic attenuation. The best studied of these regions is the stable continental
interior of eastern North America. Since these regions are characterized by relatively infrequent earth-
quakes, there are an insufficient number of strong-motion recordings with which to develop reliable
empirical attenuation relationships. As a result, attenuation relationships developed for these regions
have been based on intensity data (usually characterized in terms of the Modified Mercalli Intensity
[MMI] scale) or, more recently, on simple seismological models of the earthquake source and propagation
medium.
One of the most recent and best-documented theoretical attenuation relationships for stable conti-
nental regions is that developed by the Electric Power Research Institute [EPRI, 1993] for eastern North
America. Because of its thorough review and sound seismologic basis, this model was selected to represent
the attenuation of PGA in the stable continental-interior regions of Oman and India. The model was
developed to represent the attenuation characteristics of sites on hard rock using a median stress drop
of 120 bars and an anelastic attenuation parameter (Q) consistent with observed earthquakes in eastern
North America. The standard error of estimate (σln PGA) associated with the EPRI attenuation relationship,
averaged over magnitude and distance, was estimated to be 0.7 for the near-source distances of most
concern in this study.
8.16.2 Stable Oceanic Interior Earthquakes

There are no attenuation relationships available for stable oceanic interior regions. However, the oceanic
lithosphere is known to be an extremely efficient waveguide for high-frequency seismic energy. For
example, the anelastic attenuation of high-frequency waves in the Ngendei region of the Southwest Pacific
has been found to be consistent with Q of 450 for that part of the lithosphere above the Moho (6.65 km
below mudline) and Q of 1000 for that part of the lithosphere below this depth. These values are consistent
with Q observed in the stable continental-interior regions of eastern North America [EPRI, 1993]. Based
on this similarity, the EPRI attenuation relationship was used to model the attenuation of PGA from
earthquakes occurring in the intraplate regions of the Arabian Sea.
8.16.3 Transform Plate Boundary Earthquakes

Transform plate boundary earthquakes, such as those typical of the strike-slip San Andreas fault system
in western California, have lower average stress drops than intraplate earthquakes [EPRI, 1993]. However,
unlike California, plate boundary earthquakes associated with the Owen Fracture Zone–Murray Ridge
Complex and the Ornach-Nal and Chaman faults occur on a transform boundary between intraplate
regions of low anelastic attenuation, similar to the craton of eastern North America. Therefore, the source
characteristics of these plate-boundary earthquakes are expected to be similar to those in other plate
boundary environments (e.g., the San Andreas fault system); whereas, the attenuation characteristics of
these earthquakes are expected to be typical of the attenuation characteristics of an intraplate environment.
The majority of strong-motion recordings from plate-boundary earthquakes are from California,
where anelastic attenuation is much greater than in stable continental interiors. Therefore, attenuation
relationships from these regions are not applicable for this region. Instead, we modified the EPRI
attenuation relationship to predict the attenuation of PGA from plate-boundary earthquakes in the
northern Arabian Sea and southern Pakistan regions.
EPRI [1993] gives stress drops for earthquakes in interplate and intraplate tectonic environments that
have been calculated from strong-motion and standard seismograph recordings. These calculations
indicate a median stress drop of 120 bars for intraplate regions. On the other hand, the data for interplate
strike-slip and normal faulting earthquakes, typical of the plate-boundary earthquakes in the northern
Arabian Sea and southern Pakistan regions, were found to be consistent with a median stress drop of
about 85 bars. Research has indicated PGA ∝ ∆σ0.8. Therefore, PGA predicted by the EPRI attenuation
relationship was reduced by 25% for plate boundary earthquakes in the northern Arabian Sea and
southern Pakistan regions.

8.16.4 Subduction Zone Earthquakes

Subduction-zone earthquakes, similar to those associated with the Makran Subduction Zone, have been
found to have significantly different attenuation characteristics from either shallow crustal interplate or
intraplate earthquakes. Therefore, it is important to model these earthquakes with an attenuation rela-
tionship appropriate for this tectonic environment. The attenuation relationship selected for this purpose
was one developed from rock recordings of large subduction-zone earthquakes throughout the world.
The standard error of estimate (σln PGA) associated with this relationship for the larger earthquakes of
interest in this study is 0.55.
8.17 Soil Amplification Factors

Estimates of PGA from the attenuation relationships described above are for hard rock. Therefore, it is
necessary to multiply these estimates by an appropriate site amplification factor in order to account for
the existing predominant soil conditions at mudline along the pipeline route. This adjustment was done
using the amplitude-dependent amplification factors given in Borcherdt [1993]. The classification of the
existing soils along the pipeline route was based on preliminary geotechnical and seismic data collected
along the pipeline route.
The soil amplification factors for each of the soil classifications given by Borcherdt [1993] were
normalized to hard rock and the amplitude of PGA for the existing soil conditions are obtained by
multiplying the estimate of PGA on hard rock by these normalized factors. The maximum value of PGA
on soft soils was limited to 0.45 g based on site-response studies of Holocene Bay Mud in the San
Francisco Bay area. If shear strains large enough to cause significant cyclic degradation (e.g., liquefaction)
are induced in these deposits, then actual values of PGA for these soft soils may be further limited to
values on the order of 0.2 to 0.25 g. Of course, for such large strains, ground failure will become an
important issue.
8.18 Results
Values of PGA on hard rock were calculated at 130 locations along the pipeline route for return periods
of 200, 500, and 1000 years (Figure 8.23). However, much of the pipeline route is characterized by soft
pelagic and detrital deposits that are subject to submarine turbidite flows if disturbed by strong ground
shaking from earthquakes. A map showing 500-year values of PGA on hard rock and on the existing soil
conditions at mudline (in parentheses) at selected locations along the pipeline route and Indus Canyon
is given in Figure 8.24. The existing soil conditions at all of the selected sites were classified as soft soil.
For all return periods, the highest values of PGA were obtained at the intersection of the pipeline route
with the Owen Fracture Zone, where calculated values on hard rock (soft soil) were 0.35 g (0.40 g), 0.56 g
(0.45 g), and 0.77 g (0.45 g) for return periods of 200, 500, and 1000 years, respectively. Intermediate
values of PGA were calculated for the eastern pipeline terminus at the India coast and at the crossing of
the southwest extension of the Qalhat Seamount, located approximately 60 km west of the Owen Fracture
Zone, where values on hard rock (soft soil) ranging between 0.17 g (0.32 g) and 0.31 g (0.40 g) were
calculated for a 200-year return period and values ranging between 0.45 g (0.45 g) and 0.57 g (0.45 g)
were obtained for a 1000-year return period. The lowest values of PGA were calculated for the Arabian
Sea Abyssal Plain, east of the Owen Fracture Zone, where estimates for hard rock (soft soil) ranged from
0.03 g (0.08) for a 200-year return period to 0.09 g (0.22 g) for a 1000-year return period.
8.19 Conclusions
The computed ground-shaking hazard along the Oman India pipeline was found to be relatively high in
the vicinity of the Owen Fracture Zone–Murray Ridge Complex and at the India coast in the Kutch
region. These values are high enough to potentially trigger geologic hazards such as liquefaction, slope

0.80
0.70 OMAN Owen Fracture INDIA

Zone
Kutch Zone
0.60 200-yr
SWExtension,
0.50 Qalhat 500-yr
PGA (g) Seamount
1,000-yr
0.40
Siquirah
0.30
Fault
0.20 Indus Fan
Crossing
0.10
0.00
59.72
60.22
60.73
61.26
61.63
62.16
62.78
63.40
64.02
64.64
65.23
65.88
66.43
66.93
67.37
67.79
68.18
68.56
Longitude (deg. E)
FIGURE 8.23 Longitudinal profile of PGA on hard rock along the pipeline route for three return periods.
60° 65° 70°
IRAN
PAKISTAN
25°
Rapar Gadhwali
OMAN 0.11
(0.25)
Ra'as al Jifan INDIA
0.07
(0.18) 0.44
e
lin
(0.45)
pe
0.10 0.08
Pi
Oman - India
20° (0.23) (0.20)
0.30 0.56 0.05
(0.39) (0.45) (0.13)
ARABIAN SEA
FIGURE 8.24 Calculated PGA on hard rock and soft soil (in parentheses) with a return period of 500 years at
selected locations along the pipeline route and at the Indus Canyon.
instability, and turbidity flows in areas that are susceptible to these hazards. Notably, liquefaction and
ground failures were widespread throughout Western Gujuarat State in the 2001 Bhuj earthquake.
Although the computed ground-shaking hazard elsewhere along the pipeline route was found to be
relatively low, estimates of PGA are high enough offshore to also potentially trigger geologic hazards in
areas that are highly susceptible to these hazards (e.g., unstable channel slopes).
8.20 PSHA Computer Codes

There are a number of PSHA computer codes that are available to the analyst. A few are distributed free,
or at low cost [McGuire, 1976, 1978; Bender and Perkins, 1987; U.S. Geological Survey (http://geohaz-
ards.cr.usgs.gov/eq/html/hazsoft.html)]. An excellent source for software is the National Information

Service for Earthquake Engineering at the University of California, Berkeley (http://nisee.berkelye.edu/

software_and_data/eng_soft/index.html).
Some codes are not necessarily user-friendly as they are primarily products of scientific/engineering
research efforts and are not created for easy use with an end-user in mind. Several commercially available
codes are available and are user-friendly in terms of the data input interface and overall ease of use. They
also offer a wide variety of output enhancements in addition to the basic PSHA results.
Prior to selection of a specific computer code for use in engineering applications, an analyst should
have a clear understanding of his or her specific needs. Some codes are specific to area source modeling,
while others are specific to the requirements of fault modeling. Some codes are appropriate for creating
grids of PSHA values suitable for contouring in ground motion maps, while others are better adapted
for site-specific analyses. Commercial codes often come with a variety of databases and built-in models,
including ground motion attenuation relationships and seismotectonic models specific to certain regions,
and generally have the flexibility of modeling both area and fault sources, as well as mixed sources. There
is likely to be a suitable code available for any practical application required by the PSHA analyst.
Defining Terms
Aleatory — Uncertainty in the data used in an analysis; generally accounts for randomness associated
with the prediction of a parameter from a specific model, assuming that the model is correct.
Specification of the standard deviation (σ) of a mean ground motion attenuation relationship
is a representation of aleatory variability.
Characteristic earthquake — Surface-rupturing earthquakes occurring on a known tectonic structure,
within a relatively narrow range of magnitudes at an increased frequency over that which would
be estimated from the Gutenberg–Richter relationship.
Epistemic — Modeling uncertainty; accounts for incomplete knowledge in the predictive models and
the variability in the interpretations of the data used to develop the models.
Intensity — A metric of the effect, or the strength, of an earthquake hazard at a specific location,
commonly measured on qualitative scales such as MMI, MSK, and JMA (see Chapter 4).
Paleoseismicity — Prehistoric earthquakes — since there is no human record, these earthquakes are
identified (i.e., location, magnitude, etc.) via geologic trenching and other evidence.
Return period — The reciprocal of the annual probability of occurrence — earthquake probabilities of
occurrence are commonly stated in terms of a return period, which misleads some people since
they infer the earthquake occurs on a regular cycle equal to the return period.
Seismic hazard — The likelihood or probability of experiencing a specified intensity of any damaging
phenomenon, at a specific site, or over a region.
Seismotectonic model — An analytical model combining models of earthquake sources, occurrence,
and attenuation.
References
Albee, A.L. and J.L. Smith, 1966, “Earthquake Characteristics and Fault Activity in Southern California,”
in Engineering Geology in Southern California, R. Lung and T. Proctor, Eds., Association of
Engineering Geologists, Sudbury, MA, pp. 9–34.
Algermissen, S.T., D.M. Perkins, P.C. Thenhaus, S.L. Hanson, and B.L. Bender, 1982, Probabilistic Esti-
mates of Maximum Acceleration and Velocity in Rock in the Contiguous United States, U.S. Geological
Survey Open-File Rep. 82-1033.
Ambraseys, N.N. and C.F. Finkel, The Seismicity of Turkey and Adjacent Areas, EREN, Istanbul.
Ambraseys, N.N. and C.P. Melville, 1982, A History of Persian Earthquakes, Cambridge Earth Science
Series, Cambridge, U.K.
Anderson, J.G., 1979, “Estimating the Seismicity from Geological Structure for Seismic Risk Studies,”
Bull. Seismol. Soc. Am., 69, 135–158.

Atwater, B.F., 1987, “Evidence for Great Holocene Earthquakes along the Outer Coast of Washington
State,” Science, 236, 942–944.
Atwater, B.F., 1992, “Geologic Evidence for Earthquakes during the Past 2000 Years along the Copalis
River, Southern Coastal Washington,” J. Geophys. Res., 97, 1901–1919.
Bazzurro, P. and C.A. Cornell, 1999, “Disaggregation of Seismic Hazard,” Bull. Seismol. Soc. Am., 89,
501–520.
Bender, B., 1984, “Incorporating Acceleration Variability into Seismic Hazard Analysis,” Bull. Seismol.
Soc. Am., 74, 1451–1462.
Bender, B. and D.M. Perkins, 1987, “Seisrisk III, A Computer Program for Seismic Hazard Estimation,”
U.S. Geol. Survey Bull. 1772.
Bernreuter, D.L., J.B. Savy, R.W. Mensing, and J.C. Chen, 1989, “Seismic Hazard Characterization of 69
Nuclear Plant Sites East of the Rocky Mountains,” Report prepared for the U.S. Nuclear Regulatory
Commission, Lawrence Livermore National Laboratory, Report No. NUREG/CR-5250, Washing-
ton, D.C.
Bonilla, M.G., 1980, Comment and Reply on “Estimating Maximum Expectable Magnitudes of Earth-
quakes from Fault Dimensions,” Geology, 8, 162–163.
Bonilla, M.G. and J.M. Buchanan, 1970, “Interim Report on Worldwide Historic Surface Faulting,” U.S.
Geol. Survey Bull. Open File Rep.
Borcherdt, R.D., 1993, “On the Estimation of Site-Dependent Response Spectra,” Proceedings of the
International Workshop on Strong Motion Data, Vol. 2, Menlo Park, CA, The Port Harbour Research
Institute, Kanagawa, Japan.
Brune, J.N., 1968, “Seismic Moment, Seismicity and Rate of Slip along Major Fault Zones,” J. Geophys.
Res., 73, 777–784.
Bucknam, R.C. and R.E. Anderson, 1979, “Estimation of Fault-Scarp Ages from a Scarp-Height-Slope-
Angle Relationship,” Bull. Seismol. Soc. Am., 7, 11–14.
Burr, N.C. and S.C. Soloman, 1978, “The Relationship of Source Parameters of Oceanic Transform
Earthquakes to Plate Velocity and Transform Length,” J. Geophys. Res., 83, 1193–1204.
California Division of Mines and Geology, 1975, “Recommended Guidelines for Determining the Max-
imum Credible Earthquake and the Maximum Probable Earthquakes,” California Division of Mines
and Geology, Sacramento, CA, Note 43, p. 1.
Campbell, K.W., 1983, “Bayesian Analysis of Extreme Earthquake Occurrences. II. Application to the San
Jacinto Fault Zone of Southern California,” Bull. Seismol. Soc. Am., 73, 1099–1115.
Campbell, K.W., 1985, “Strong Motion Attenuation Relations: A Ten-Year Perspective,” Earthquake Spec-
tra, 1, 759–804.
Campbell, K.W., 1997, “Empirical Near-Source Attenuation Relationships for Horizontal and Vertical
Components of Peak Ground Acceleration, Peak Ground Velocity, and Pseudo-Absolute Acceler-
ation Response Spectra,” Seismol. Res. Lett., 68, 154–179.
Campbell, K.W., P.C. Thenhaus, J.E. Mullee, and R. Preston, 1996, “Seismic Hazard Evaluation of the
Oman India Pipeline,” Proceedings of the Offshore Technology Conference, May 6–9, 1996, Houston,
TX, OTC 8135, pp. 185–195.
Chapman, M.C., 1995, “A Probabilistic Approach to Ground-Motion Selection for Engineering Design,”
Coppersmith, K.J., 1991, “Seismic Source Characterization for Engineering Seismic Hazard Analysis,” in
Proceedings of the Fourth International Conference on Siesmic Zonation, Vol. 1, Aug. 25–29, 1991,
Stanford, CA, Earthquake Engineering Research Institute, Oakland, CA, pp. 3–60.
Coppersmith, K.J., P.C. Thenhaus, J.E. Ebel, W.K. Wedge, J.H. Williams, and N.C. Hester, 1993, “Regional
Seismotectonic Setting,” in Seismic Hazard Assessment in the Central and Eastern United States,
Monograph 1: Hazard Assessment, S.T. Algermissen and G.A. Bollinger, Eds., Central United States
Earthquake Consortium, Memphis, TN, pp. 35–80.
Cornell, C.A., 1968, “Engineering Seismic Risk Analysis,” Bull. Seismol. Soc. Am., 58, 1583–1606.

Cornell, C.A., 1969, “Bayesian Statistical Decision Theory and Reliability-Based Design,” Proceedings of
the International Conference on Structural Safety and Reliability, A.M. Freudenthal, Ed., Smithsonian
Institution, Washington, D.C., April 9–11, pp. 47–66.
Cornell, C.A. and E.H. Vanmarcke, 1969, “The Major Influences on Seismic Risk,” in Proceedings of the
Fourth World Conference of Earthquake Engineering, Vol. 1, Santiago, Chile, pp. 69–83.
Cornell, C.A. and S.R. Winterstein, 1988, “Temporal and Magnitude Dependence in Earthquake Recur-
rence Models,” Bull. Seismol. Soc. Am., 78, 1522–1537.
Cowie, P.A. and C.H. Scholz, 1992, “Growth of Faults by Accumulation of Seismic Slip,” J. Geophys. Res.,
97, 11085–11095.
Cramer, C.H. and M.D. Petersen, 1996, “Predominant Seismic Source Distance and Magnitude Maps for
Los Angeles, Orange and Ventura Counties, California,” Bull. Seismol. Soc. Am., 86, 1645–1649.
Cramer, C.H., M.D. Petersen, T. Cao, T.R. Toppozada, and M. Reichle, 2000, “A Time-Dependent Seismic-
Hazard Model for California,” Bull. Seismol. Soc. Am., 90, 1–21.
Crone, A.J., M.N. Machette, M.G. Bonilla, J.J. Lienkaemper, K.L. Pierce, W.E. Scott, and R.C. Bucknam,
1985, “Characteristics of Surface Faulting Accompanying the Borah Peak Earthquake, Central
Idaho,” in R.S. Stein and R.C. Bucknam, Eds., Proceedings of Workshop XXVIII: The Borah Peak,
Idaho Earthquake, Vol. A, October 3–6, 1984, U.S. Geol. Surv. Open-File Rep., pp. 43–58.
dePolo, C.M. and D.B. Slemmons, 1990, “Estimation of Earthquake Size for Seismic Hazards,” in E.L.
Krinitzsky and D.B. Slemmons, Eds., Nectonics in Earthquake Evaluation, Geological Society of
America, Reviews in Engineering, Vol. 8, pp. 1–28.
Dietrich, J.A., 1994, “A Constitutive Law for Rate of Earthquake Production and its Application to
Earthquake Clustering,” J. Geophys. Res., 99, 2601–2618.
Downes, G.L., 1995. Atlas of Isosiesmal Maps of New Zealand Earthquakes, Institute of Geological and
Nuclear Sciences, Lower Hutt.
Earthquake Engineering Research Institute (EERI), 1989, “The Basics of Seismic Risk Analysis,” Earth-
quake Spectra, 5, 675–699.
Earthquake Engineering Research Institute (EERI), 2001, “Preliminary Observations on the Origin and
Effects of the January 26, 2001 Bhuj (Gujarat, India) Earthquake,” EERI Newsletter Special Earth-
quake Rep., Vol. 35, no. 4, p. 16.
Ekstrom, G. and A.M. Dziewonski, 1988, “Evidence of Bias in Estimations of Earthquake Size,” Nature,
332, 319.
Electric Power Research Institute (EPRI), 1986, Seismic Hazard Methodology for the Central and Eastern
United States, EPRI NP-4726, Electric Power Research Institute, Palo Alto, CA.
Electric Power Research Institute (EPRI), 1993, “Methods and Guidelines for Estimating Earthquake
Ground Motion in Eastern North America,” Guidelines for Determining Design Basis Ground
Motions, Report No. EPRI TR-102293, Vol. 1, Electric Power Research Institute, Palo Alto, CA.
Erdik, M. et al., 1999, Assessment of Earthquake Hazard in Turkey and Neighboring Regions (contribution
to GSHAP, available at http://seismo.ethz.ch/gshap/turkey/papergshap71.htm).
Field, E.H., D.D. Jackson, and J.F. Dolan, 1999, “A Mutually Consistent Seismic-Hazard Source Model
for Southern California,” Bull. Seismol. Soc. Am., 89, 559–578.
Frankel, A., 1995, “Mapping Seismic Hazard in the Central and Eastern United States,” Bull. Seismol. Soc.
Am., 66, 8–21.
Frankel, A., 1996, National Seismic Hazard Maps: Documentation June 1996, U.S. Geological Survey Open-
File Rep. 96-532.
Frankel, A. and E. Safak, 1998, “Recent Trends and Future Prospects in Seismic Hazard Analysis, in
P. Dakoulas, M. Yegian, and R. Holtz, Eds., Geotechnical Earthquake Engineering and Soil Dynamics
III, Vol. 1, Geotechnical Special Publication No. 75, American Society of Civil Engineers, Reston,
VA, pp. 91–115.
Frankel, A., C. Mueller, T. Barnhard, D. Perkins, E.V. Leyendecker, N. Dickman, S. Hanson, and
M. Hopper, 2000, “USGS National Seismic Hazard Maps,” Earthquake Spectra, 16, 1–19.

Giardini, D., Ed., 1999, “The Global Seismic Hazard Assessment Program (GSHAP) 1992–1999,” Ann.
Geofis., 42(6), 957–1230.
Gutenberg, B. and C.F. Richter, 1954, Seismicity of the Earth, 2nd ed., Princeton University Press, Prin-
ceton, NJ.
Hanks, T.C. and H. Kanamori, 1979, “A Moment-Magnitude Scale,” J. Geophys. Res., 84, 2348–2350.
Harmsen, S., D. Perkins, and A. Frankel, 1999, “Disaggregation of Probabilistic Ground Motions in the
Central and Eastern United States,” Bull. Seismol. Soc. Am., 89, 1–13.
Harris, R.A. and R.W. Simpson, 1998, “Suppression of Large Earthquakes by Stress Shadows: A Com-
parison of Coulomb and Rate-and-State,” J. Geophys. Res., 103, 24,439–24,451.
Hill, D.P., P.A. Reasenberg, A. Michael, et al., 1993, “Seismicity Remotely Triggered by the Magnitude 7.3
Landers, California, Earthquake,” Science, 260, 1617–1623.
Jacob, K.H. and R.L. Quittmeryer, 1979, “The Makran Region of Pakistan and Iran: Trench-Arc System
and Active Plate Subduction,” in A. Farah and K.A. DeJong, Eds., Geodynamics of Pakistan, Geo-
logical Survey of Pakistan, pp. 305–317.
Johnston, A.C., K.J. Coppersmith, L.R. Kanter, and C.A. Cornell, 1994, The Earthquakes of Stable Conti-
nent Interiors, Vol. 1, Assessment of Large Earthquake Potential, Electric Power Research Institute,
Palo Alto, CA.
Kagan, Y.Y. and L. Knopoff, 1980, “Spatial Distribution of Earthquakes: The Two Point Correlation
Function,” Geophys. J. R. Astron. Soc., 62, 303–320.
Kanamori, H., 1978, “Quantification of Earthquakes,” Nature, 271, 411–414.
Khattri, K.N., A.M. Rogers, S.T. Algermissen, and D.M. Perkins, 1984, “A Seismic Hazard Map of India
and Adjacent Areas,” Tectonophysics, 108, 93–134.
King, G. and J. Nabelek, 1985, “Role of Fault Bends in the Initiation and Termination of Earthquake
Ruptures,” Science, 228, 984–987.
Korvin, G., 1992, Fractal Models of the Earth, Elsevier, Amsterdam.
Lee, W.H.K. et al., 1988, Historical Seismograms and Earthquakes of the World, Academic Press, New York.
Machette, M.N., S.F. Personius, and A.R. Nelson, 1992, “Paleoseismology of the Wasatch Fault Zone: A
Summary of Recent Investigations, Interpretations, and Conclusions,” in P.L. Gori and W.W. Hays,
Eds., Assessment of Regional Earthquake Hazards and Risk along the Wasatch Front, Utah, U.S.
Geological Survey Prof. Paper. 1500-A-J, pp. A1-A71.
Mark, R.K. and M.G. Bonilla, 1977, Regression Analysis of Earthquake Magnitude and Surface Fault Length
Using the 1970 Data of Bonilla and Buchanan, U.S. Geological Survey Open-File Rep. 78-1007.
McCalpin, J.P., Ed., 1996, Paleoseismology, Vol. 62 in the International Geophysics Series, Academic Press,
New York.
McGuire, R.K., 1976, Fortran Computer Program for Seismic Risk Analysis, U.S. Geological Survey Open-
File Rep. 76-67.
McGuire, R.K., 1978, FRISK: Computer Program for Seismic Risk Analysis Using Faults as Earthquake
Sources, U.S. Geological Survey Open-File Rep. 78-1007.
McGuire, R.K., 1993, The Practice of Earthquake Hazard Assessment, IDNDR Monograph, International
Association of Seismology and Physics of the Earth’s Interior/European Seismological Commission,
University of Colorado Department of Physics, Boulder, CO.
McGuire, R.K., 1995, “PSHA and Design Earthquakes: Closing the Loop,” Bull. Seismol. Soc. Am., 85,
1275–1284.
McGuire, R.K. and T.P. Barnhard, 1981, “Effects of Temporal Variations in Seismicity on Seismic Hazard,”
McGuire, R.K. and K.M. Shedlock, 1981, “Statistical Uncertainties in Seismic Hazard Evaluations in the
United States,” Bull. Seismol. Soc. Am., 71, 1287–1308.
Molnar, P., 1979, “Earthquake Recurrence Intervals and Plate Tectonics,” Bull. Seismol. Soc. Am., 69,
115–134.

Munson, P.J., C.A. Munson, N.K. Bleuer, and M.D. Labitzke, 1992, “Distribution and Dating of Prehistoric
Earthquake Liquefaction in the Wabash Valley of the Central U.S.,” Bull. Seismol. Soc. Am., 63,
337–342.
Munson, P.J., C.A. Munson, and N.K. Bleuer, 1994, Late Pleistocene and Holocene Earthquake-Induced
Liquefaction in the Wabash Valley of Southern Indiana, U.S. Geological Survey Open-File Rep.
94-176, pp. 553–557.
Nishenko, S.P., 1991, “Circum-Pacific Seismic Potential: 1989–1999,” Pure Appl. Geophys., 135, 169–259.
Nishenko, S.P. and R. Buland, 1987, “A Generic Recurrence Interval Distribution for Earthquake Fore-
casting,” Bull. Seismol. Soc. Am., 77, 1382–1399.
Nuttli, O.W., 1979, “Seismicity of the Central United States,” in Geology in the Siting of Nuclear Power
Plants, A.W. Hathaway and C.R. McClure, Jr., Eds., Reviews in Engineering Geology, Vol. 4,
pp. 67–94.
Nuttli, O.W., 1981, “On the Problem of Maximum Magnitude of Earthquakes,” in W.W. Hays, Ed.,
Evaluation of Regional Seismic Hazards and Risk, U.S. Geological Survey Open-File Rep. 81-437.
Obermeier, S.F., J.R. Martin, A.D. Frankel, T.L. Youd, P.J. Munson, C.A. Munson, and E.C. Pond, 1993,
Liquefaction Evidence for One or More Strong Holocene Earthquakes in the Wabash Valley of Southern
Indiana and Illinois, with a Preliminary Estimate of Magnitude, U.S. Geological Survey Prof.
Paper 1536.
Parsons, T., S. Toda, R.S. Stein, A. Barka, and J.H. Dietrich, 2000, “Heightened Odds of Large Earthquakes
Near Istanbul: An Interaction-Based Probability Calculation,” Science, 288, 661–665.
Perkins, D.M. and S.T. Algermissen, 1987, “Seismic Hazards Maps for the U.S.: Present Use and Prospects,”
in K.H. Jacob, Ed., Proceedings from the Symposium on Seismic Hazards, Ground Motions, Soil
Liquefaction and Engineering Practice in Eastern North America, Technical Report NCEER-87–0025,
pp. 16–25.
Petersen, M.D. et al., 1996, Probabilistic Seismic Hazard Assessment for the State of California, U.S.
Geological Survey Open-File Rep. 96-706.
Reasenberg, P.A. and L.M. Jones, 1989, “Earthquake Hazard after a Mainshock in California,” Science,
243, 1173–1176.
Reid, H.F., 1910, “The Mechanics of the Earthquake,” in The California Earthquake of April 18, 1906,
Report to the State Earthquake Investigation Commission, Publication No. 87, Vol. II, Carnegie
Institution of Washington, D.C.
Reiter, L., 1990, Earthquake Hazard Analysis: Issues and Insights, Columbia University Press, New York.
Satake, K., K. Shimazaki, Y. Tsuji, and K. Ueda, 1996, “Time and Size of Giant Earthquake in Cascadia
Inferred from Japanese Tsunami Records of January 1700,” Nature, 379, 246–249.
Scholz, D.H., 1990, The Mechanics of Earthquake Faulting, Cambridge University Press, Cambridge.
Schwartz, D.P., 1988, “Geologic Characterization of Seismic Sources: Moving into the 1990s,” in Engi-
neering and Soil Dynamics II: Recent Advances in Ground Motion Evaluation, J.L. v.Thun, Ed.,
Geotechnical Special Publication No. 20, American Society of Civil Engineering, New York.
Schwartz, D.P. and K.J. Coppersmith, 1984, “Fault Behavior and Characteristic Earthquakes: Examples
from the Wasatch and San Andreas Fault Zones,” J. Geophys. Res., 89, 5681–5698.
Shedlock, K.M., R.K. McGuire, and D.G. Herd, 1980, Earthquake Recurrence in the San Francisco Bay Region,
California, from Fault Slip and Seismic Moment, U.S. Geological Survey Open-File Rep. 80-999.
Shimazaki, K. and T. Nakata, 1980, “Time-Predictable Recurrence Model for Large Earthquakes,” Geophys.
Res. Lett., 7, 279–282.
SSHAC (Senior Seismic Hazard Assessment Committee), 1997, “Recommendations for PSHA: Guidance
on Uncertainty and Use of Experts,” Report NUREG/CR-6372, U.S. Nuclear Regulatory Commis-
sion, Washington, D.C.
Stein, R.S., 1999, “Role of Stress Transfer in Earthquake Occurrence,” Nature, 402, 605–609.
Stepp, J.C., 1972, “Analysis of Completeness of the Earthquake Sample in the Puget Sound Area and Its
Effects on Statistical Estimates of Earthquake Hazard,” Proceedings of the First Microzonation Con-
ference, Seattle, WA, pp. 897–909.

Stepp, J.C., 1973, “Analysis of Completeness of the Earthquake Sample in the Puget Sound Area,” in S.T.
Harding, Ed., “Contributions to Seismic Zoning,” Technical Report ERL 267-ESL 30, pp. 16–28,
National Oceanic and Atmospheric Administration, Washington, D.C.
Stepp, J.C., W.J. Silva, R.K. McGuire, and R.W. Sewell, 1993, “Determination of Earthquake Design Loads
for a High Level Nuclear Waste Repository Facility,” in Proceedings of the Natural Phenomena
Hazards Mitigation Conference, Vol. 2, pp. 651–657, October 19–22, Atlanta, GA.
Stucchi, M., Ed., 1993, Historical Investigations of European Earthquakes, CNR – Istituto di Ricerca sul
Rischio Sismico, Milano.
Sykes, L.R., 1971, “Aftershock Zones of Great Earthquakes, Seismicity Gaps, and Earthquake Prediction
for Alaska and the Aleutians,” J. Geophys. Res., 76, 8021–8041.
Thenhaus, P.C., 1983, “Summary of Workshops Concerning Regional Seismic Source Zones of Parts of
the Conterminous United States, Convened 1979–1980, Golden, Colorado,” U.S. Geol. Surv.
Circular 898.
Thenhaus, P.C., 1986, “Seismic Source Zones in Probabilistic Estimation of the Earthquake Ground
Motion Hazard: A Classification with Key Issues,” Proceedings of Conference 34: Workshop on
Probablistic Earthquake Hazards Assessments, pp. 53–71.
Toda, S., R.S. Stein, P.A. Reasonberg, and J.H. Dietrich, 1998, “Stress Transferred by the Mw = 6.5 Kobe,
Japan, Shock: Effects on Aftershocks and Earthquake Probabilities,” J. Geophys. Res., 103,
24,543–24,565.
Usami, T., 1981, Nihon Higai Jishin Soran (List of Damaging Japanese Earthquakes), University of Tokyo
Press (in Japanese).
Wells, D.L. and K.J. Coppersmith, 1994, “New Empirical Relationships among Magnitude, Rupture
Length, Rupture Width, Rupture Area, and Surface Displacement,” Bull. Seismol. Soc. Am., 84,
974–1002.
Wentworth, C.M. and M.D. Zoback, 1990, “Structure of the Coalinga Area and Thrust Origin of the
Earthquake,” in M.J. Rymer and W.L. Ellsworth, Eds., The Coalinga, California, Earthquake of May
2, 1983, U.S. Geological Survey Prof. Paper 1487, pp. 41–68.
Wesnousky, S.G., 1994, “The Gutenberg–Richter or Characteristic Earthquake Distribution, Which Is it?,”
Wesnousky, S.G., C. Scholz, K. Shimazaki, and T. Matsuda, 1983, “Earthquake Frequency Distribution
and the Mechanics of Faulting,” J. Geophys. Res., 87, 6829–6852.
Wheeler, R.L. and K.B. Krystinik, 1992, “Persistent and Nonpersistent Segmentation of the Wasatch Fault
Zone, Utah: Statistical Analysis for Evaluation of Seismic Hazard,” in P.L. Gori and W.W. Hays,
Eds., Assessment of Regional Earthquake Hazards and Risk along the Wasach Front, Utah, U.S.
Geological Survey Prof. Paper. 1500-A-J, pp. B1–B47.
Woo, G., 1996, “Kernel Estimation Methods for Seismic Hazard Area Source Modeling,” Bull. Seismol.
Soc. Am., 86, 353–362.
Working Group on California Earthquake Probabilities, 1988, Probabilities of Large Earthquakes Occurring
in California on the San Andres Fault, U.S. Geological Survey Open-File Rep. 88-398.
Working Group on California Earthquake Probabilities, 1990, “Probabilities of Large Earthquakes in the
San Francisco Bay Region, California,” U.S. Geol. Surv. Circular 1053.
Working Group on California Earthquake Probabilities, 1995, “Seismic Hazards in Southern California:
Probable Earthquakes, 1994 to 2024,” Bull. Seismol. Soc. Am., 85, 379–439.
Working Group on California Earthquake Probabilities, 1999, Earthquake Probabilities in the San Francisco
Bay Region: 2000 to 2030 — A Summary of Findings, U.S. Geological Survey Open-File Rep. 99-517.
Yeats, R.S., K. Sieh, and C.R. Allen, 1997, The Geology of Earthquakes, Oxford University Press, New York.
Youngs, R.R. and K.J. Coppersmith, 1985, “Implications of Fault Slip Rates and Earthquake Recurrence
Models to Probabilistic Seismic Hazard Estimates,” Bull. Seismol. Soc. Am., 75, 939–964.

Youngs, R.R., Swan, F.H., Powers, M.S., Schwartz, D.P., and Green, R.K., 1987, “Probabilistic Analysis of
Earthquake Ground Shaking along the Wasatch Front, Utah,” in P.L. Gori and W.W. Hays, Eds.,
Assessment of Regional Earthquake Hazards and Risk along the Wasatch Fault, Utah, U.S. Geological
Survey Open File Report 87-585, pp. M1-M110.
Further Reading
There is an extensive literature on seismic hazard analysis. Reiter’s book [1990] is a good introduction,
as is Cornell’s classic paper [Cornell, 1968], which started the field, and Cornell and Vanmarcke [1969].
McGuire has made a number of contributions over the years [1995; many others] which are very
instructive. McGuire [1993] is an excellent compendium of international seismic hazard assessment
practice, as is the more recent GSHAP project and papers, available on-line at [http://seismo.ethz.ch/
gshap/]. McCalpin’s [1996] book is an excellent summary of the state of the art in the relatively new field
of paleoseismology. Yeats et al.’s [1997] book provides an excellent overview of geology related to earth-
quakes and synopses of worldwide seismotectonic settings. Sholz’s [1990] book is highly instructive in
the theory of fault rupture mechanics and earthquake generation, and provides an insightful synopsis of
hazard analysis and earthquake prediction.
View publication stats

Seismic Hazard Analysis: January 2002

Caricato da

Informazioni sul documento

Titolo originale

Copyright

Formati disponibili

Condividi questo documento

Condividi o incorpora il documento

Opzioni di condivisione

Hai trovato utile questo documento?

Questo contenuto è inappropriato?

Copyright:

Formati disponibili

Seismic Hazard Analysis: January 2002

Caricato da

Copyright:

Formati disponibili

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

Seismic hazard analysis

Chapter · January 2002

Attenuation of Arias Intensity View project

NGA-subduction View project

The user has requested enhancement of the downloaded file.

© 2003 by CRC Press LLC

8-2 Earthquake Engineering Handbook

8.2 Probabilistic Seismic Hazard Methodology

© 2003 by CRC Press LLC

Seismic Hazard Analysis 8-3

where the “return period” of x is defined as:

© 2003 by CRC Press LLC

8-4 Earthquake Engineering Handbook

8.3 Constituent Models of the Probabilistic Seismic

8.3.1 Seismic Sources

8.3.2 Earthquake Recurrence Frequency

© 2003 by CRC Press LLC

Seismic Hazard Analysis 8-5

8.3.3 Ground Motion Attenuation

© 2003 by CRC Press LLC

8-6 Earthquake Engineering Handbook

8.3.4 Ground Motion Probability

8.4 Definition of Seismic Sources

© 2003 by CRC Press LLC

Seismic Hazard Analysis 8-7

8.4.1 Area Sources

© 2003 by CRC Press LLC

8-8 Earthquake Engineering Handbook

Active Volcanoes, Plate Tectonics, and the “Ring of Fire”

Eurasian Plate Eurasian Plate

Aleutian Trench CASCADE

Java Trench East Pacific Nazca South

© 2003 by CRC Press LLC

Seismic Hazard Analysis 8-9

CEUS maximum-magnitude zones

30¡ 7.2 30¡

log N (m) = a − bm (8.5)

8.4.2 Fault Sources

© 2003 by CRC Press LLC

8-10 Earthquake Engineering Handbook

8.5 Earthquake Frequency Assessments

8.5.1 Historical Frequency Assessments

© 2003 by CRC Press LLC

Seismic Hazard Analysis 8-11

© 2003 by CRC Press LLC

8-12 Earthquake Engineering Handbook

where m0 is an arbitrary reference magnitude; mu is an upper-bound magnitude where n(m) = 0 for

8.5.2 Geologic Earthquake Frequency Assessments

© 2003 by CRC Press LLC

Seismic Hazard Analysis 8-13

Annual Number of Earthquakes, N (m)

FIGURE 8.7 Comparison of the exponential –4

© 2003 by CRC Press LLC

8-14 Earthquake Engineering Handbook

10 − b(m0 −m0′ ) − 10 − b(mu −m0′ )

© 2003 by CRC Press LLC

Seismic Hazard Analysis 8-15

8.5.3 Conservation of Seismic Moment on Segmented Faults

© 2003 by CRC Press LLC

8-16 Earthquake Engineering Handbook

8.5.4 Time-Dependent Probability Modeling

FIGURE 8.8 Three types of earthquake occurrence