the nature of earthquake ground motion -...

1

Chapter 1

THE NATURE OF EARTHQUAKE GROUND MOTION

Bruce A. Bolt, D.Sc.Professor Emeritus, of Seismology, University of California, Berkeley, California

Key words: Attenuation, Causes of Earthquakes, Collapse Earthquakes, Damage Mechanisms, Directivity, Earthquakes,Earthquake Faults, Earthquake Prediction, Elastic Rebound, Fault Types, Intensity Scales, IntraplateEarthquakes, Magnitude, Magnitude Scales, Magnitude Saturation, MMI, Near-Fault Effects, PlateTectonics, Reservoir-Induced Earthquakes, Seismicity, Seismic Moment, Seismic Risk, Seismic Waves,Seismology, Source Models, Strong Ground Motion, Surface Rupture, Tectonic Earthquakes, VolcanicEarthquakes

Abstract: The aim of this chapter is to provide a basic understanding about earthquakes, their world-wide distribution,what causes them, their likely damage mechanisms, earthquake measuring scales, and current efforts on theprediction of strong seismic ground motions. This chapter, therefore, furnishes the basic informationnecessary for understanding the more detailed concepts that follow in the subsequent chapters of this book.The basic vocabulary of seismology is defined. The seismicity of the world is discussed first and itsrelationship with tectonic plates is explained. The general causes of earthquakes are discussed next wheretectonic actions, dilatancy in the crustal rocks, explosions, collapses, volcanic actions, and other likelycauses are introduced. Earthquake fault sources are discussed next. Various faulting mechanisms areexplained followed by a brief discussion of seismic waves. Earthquake damage mechanisms are introducedand different major damage mechanisms are identified by examples. Quantification of earthquakes is ofsignificant interest to seismic design engineers. Various earthquake intensity and magnitude scales aredefined followed by a description of earthquake source models. Basic information regarding the concepts ofdirectivity and near-fault effects are presented. Finally, the ideas behind seismic risk evaluation andearthquake prediction are discussed.

2 Chapter 1

1. THE NATURE OF EARTHQUAKE GROUND MOTION 3

1.1 INTRODUCTION

On the average, 10,000 people die each yearfrom earthquakes (see Figure 1-1). A UNESCOstudy gives damage losses amounting to$10,000,000,000 from 1926 to 1950 fromearthquakes. In Central Asia in this interval twotowns and 200 villages were destroyed. Sincethen several towns including Ashkhabad(1948), Agadir (1960), Skopje (1963),Managua (1972), Gemona (1976), Tangshan(1976), Mexico City (1985), Spitak (1988),Kobe (1995), cities in Turkey and Taiwan(1999) and hundreds of villages have beenseverely damaged by ground shaking.Historical writings testify to man’s longconcern about earthquake hazards.

The first modern stimulus for scientificstudy of earthquakes came from the extensivefield work of the Irish engineer, Robert Mallett,after the great Neopolitan earthquake of 1857in southern Italy. He set out to explain the“masses of dislocated stone and mortar” in

terms of mechanical principles and in doing soestablished basic vocabulary such asseismology, hypocenter and isoseismal. Suchclose links between engineering andseismology have continued ever since(1-1,1-2).

It is part of strong motion seismology toexplain and predict the large amplitude-longduration shaking observed in damagingearthquakes. In the first sixty years of thecentury, however, the great seismologicaladvances occurred in studying waves fromdistant earthquakes using very sensitiveseismographs. Because the wave amplitudes ineven a nearby magnitude 5 earthquake wouldexceed the dynamic range of the usualseismographs, not much fundamental work wasdone by seismologists on the rarer largeearthquakes of engineering importance.

Nowadays, the situation has changed. Afterthe 1971 San Fernando earthquake, hundreds ofstrong-motion records were available for thismagnitude 6.5 earthquake. The 1.2g recorded atPacoima Dam led to questions on topographic

TURKEY

PERU

TANGSHAN

GUATEMALA

MANAGUAMORACCOIRAN

IRAN

PAKISTAN

KANSUKANSU

KANTO

SAN FRANCISCO

ITALY

MESSINA

MEAN

1980197019601950194019301920191019001890

1,000,000

100,000

10,000

1,000

100

Figure 1-1. Loss of life caused by major earthquakes [After Hiroo Kanamori(1-10)].

4 Chapter 1

amplification and the construction of realisticmodels of fault-rupture and travel-path thatcould explain the strong motion patterns.Progress on these seismological questionsfollowed rapidly in studies of variation inground motions in the 1989 Loma Prietaearthquake (M 7.0), the 1994 Northridgeearthquake (M 6.8) and the 1999 Chi Chi eventin Taiwan (M 7.6). A harvest of strong motionrecordings was obtained in the latterearthquake, showing numerous horizontal peakaccelerations in the range 0.5g to 1.0g. Digitalrecorders and fast computers mean that bothseismologists and engineers can tackle morefundamental and realistic problems ofearthquake generation and ground shaking.

Knowledge of strong ground shaking is nowadvancing rapidly, largely because of thegrowth of appropriately sited strong-motionaccelerographs in seismic areas of the world.For example, in the Strong MotionInstrumentation Program in California, by theyear 2000 there were 800 instruments in thefree-field and 130 buildings and 45 otherstructures instrumented. Over 500 records hadbeen digitized and were available for use inresearch or practice (see Chapter 16). Inearthquake-prone regions, structural design oflarge or critical engineered structures such ashigh-rise buildings, large dams, and bridgesnow usually involves quantitative dynamicanalysis; engineers ask penetrating questions onthe likely seismic intensity for constructionsites and require input motions or spectra ofdefining parameters. Predicted seismograms(time-histories) for dynamic modeling instructural design or vulnerability assessmentsare often needed.

The aim of this chapter is to provide a basicunderstanding about earthquakes, theirworldwide distribution, what causes them, theirlikely damage mechanisms, earthquakemeasuring scales, and current efforts on theprediction of strong seismic ground motions.Additional helpful background on the subjectmay be found in References 1-2 through 1-34.

1.2 SEISMICITY OF THEWORLD

From the earthquake wave readings atdifferent seismographic observatories, theposition of the center of an earthquake can becalculated(1-1). In this way, a uniform picture ofearthquake distribution around the world hasbeen obtained (see Figure 1-2). Definite beltsof seismic activity separate large oceanic andcontinental regions, themselves mainly, but byno means completely, devoid of earthquakecenters. Other concentrations of earthquakesources can be seen in the oceanic areas, forexample, along the center of the Atlantic andIndian Oceans. These are the sites of giganticsubmarine mountain ranges called mid-oceanicridges. The geological strains that prevailthroughout this global ridge system areevidenced by mountain peaks and deep riftvalleys. Volcanic eruptions are frequent, andearthquakes originating along these ridges oftenoccur in swarms, so that many hundreds ofshocks are concentrated in a small area in ashort time.

Dense concentrations of earthquake centerswith some as much as 680 kilometers beneaththe surface also coincide with island arcs, suchas those of the Pacific and the easternCaribbean.

On the western side of the Pacific Ocean,the whole coast of Central and South Americais agitated by many earthquakes, great andsmall. High death tolls have ensued from themajor ones. In marked contrast, the eastern partof South America is almost entirely free fromearthquakes, and can be cited as an example oflow seismic risk country. Other seismicallyquiet continental areas can be seen in Figure 1-2.

In Europe, earthquake activity is quitewidespread. To the south, Turkey, Greece,Yugoslavia, Italy, Spain and Portugal sufferfrom it, and large numbers of people have diedin disasters throughout the years. Anearthquake off southwest Iberia on November1, 1755 produced a great tsunami, whichcaused many of the 50,000 to 70,000 deaths


occurring in Lisbon, Portugal, and surroundingareas; the shaking was felt in Germany and theNetherlands. In Alicante, Spain, on March 21,1829, a shock killed about 840 persons andinjured many hundred more. Total or partialdestruction of more than 5,000 houses wasreported in and near Torrevieja and Murcia. OnDecember 28, 1908, a devastating earthquakehit Messina, Italy, causing 120,000 deaths andwidespread damage. The most recent deadlyone to affect that country struck on May 6,1976, in the Friuli region near Gemona; about965 persons were killed and 2280 injured.

On December 27, 1939, in Erzincan,Turkey, 23,000 lives were lost from a majorearthquake. Similar killer earthquakes haveoccurred in Turkey and Iran in recent years.The Erzincan earthquake along the Anatolianfault in Turkey on March 13, 1992 causedmany building collapses and the June 21, 1990earthquake (M 7.3) devastated two Iranianprovinces, Gilan and Zanjan. August 17, 1999saw a 50 km rupture of the north Anatoliam

fault along the Marmara Sea south of Izmitproducing a magnitude 7.4 earthquake and over16,000 deaths.

North of the Mediterranean margin, Europeis much more stable. However, destructiveearthquakes do occur from time to time inRomania, Germany, Austria and Switzerland,and even in the North Sea region andScandinavia. For example, on October 8, 1927,an earthquake occurred near Schwadorf inAustria and caused damage in an area southeastof Vienna. This earthquake was felt inHungary, Germany, and Czechoslovakia atdistances of 250 kilometers from the center ofthe disturbance. The seismicity in the NorthSea is sufficiently significant to requireattention to earthquake resistant design of oilplatforms there.

In Africa, damaging earthquakes haveoccurred in historical times. A notable case wasthe magnitude 5.6 earthquake on November 14,1981 that was felt in Aswan, Egypt. Thisearthquake was probably stimulated by the

Figure 1-2. Tectonic plates and world-wide distribution of earthquakes. (From Earthquakes, by Bruce A. Bolt. Copyright1978, 1999 W. H. Freeman and Company. Used with permission.)

6 Chapter 1

impounding of water in Lake Nassar behind thehigh Aswan Dam.

An example of infrequent and dispersedseismicity is the occurrence of earthquakes inAustralia. Nevertheless, this country does havesome areas of significant present-dayseismicity. Of particular interest is a damagingearthquake of moderate size that was centerednear Newcastle and causing major damage andkilling fourteen people. It was a surprise from aseismological point of view because no faultmaps were available which showedseismogenic geological structures nearNewcastle.

During an earthquake, seismic wavesradiate from the earthquake source somewherebelow the ground surface as opposite sides of aslipping fault rebound in opposite directions inorder to decrease the strain energy in the rocks.Although in natural earthquakes this source isspread out through a volume of rock, it is oftenconvenient to imagine a simplified earthquakesource as a point from which the waves firstemanate. This point is called the earthquakefocus. The point on the ground surface directlyabove the focus is called the earthquakeepicenter.

Although many foci are situated at shallowdepths, in some regions they are hundreds ofkilometers deep. Such regions include theSouth American Andes, the Tonga Islands,Samoa, the New Hebrides chain, the Japan Sea,Indonesia, and the Caribbean Antilles. On theaverage, the frequency of occurrence ofearthquakes in these regions declines rapidlybelow a depth of 200 kilometers, but some fociare as deep as 680 kilometers. Ratherarbitrarily, earthquakes with foci from 70 to300 kilometers deep are called intermediatefocus and those below this depth are termeddeep focus. Some intermediate and deep focusearthquakes are located away from the Pacificregion, in the Hindu Kush, in Romania, in theAegean Sea and under Spain.

The shallow-focus earthquakes (focus depthless than 70 kilometers) wreak the mostdevastation, and they contribute about threequarters of the total energy released in

earthquakes throughout the world. InCalifornia, for example, all of the knownearthquakes to date have been shallow-focus. Infact, it has been shown that the great majorityof earthquakes occurring in central Californiaoriginate from foci in the upper five kilometersof the Earth, and only a few are as deep as even15 kilometers.

Most moderate to large shallow earthquakesare followed, in the ensuing hours and even inthe next several months, by numerous, usuallysmaller earthquakes in the same vicinity. Theseearthquakes are called aftershocks, and largeearthquakes are sometimes followed byincredible numbers of them. The great RatIsland earthquake in the Aleutian Island onFebruary 4, 1965 was, within the next 24 days,followed by more than 750 aftershocks largeenough to be recorded by distant seismographs.Aftershocks are sometimes energetic enough tocause additional damage to already weakenedstructures. This happened, for example, a weekafter the Northridge earthquake of January 17,1994 in the San Fernando Valley when someweakened structures sustained additionalcracking from magnitude 5.5 aftershocks. Afew earthquakes are preceded by smallerforeshocks from the source area, and it has beensuggested that these can be used to predict themain shock.

1.3 CAUSES OFEARTHQUAKES

1.3.1 Tectonic Earthquakes

In the time of the Greeks it was natural tolink the Aegean volcanoes with the earthquakesof the Mediterranean. As time went on itbecame clear that most damaging earthquakeswere in fact not caused by volcanic activity.

A coherent global geological explanation ofthe majority of earthquakes is in terms of whatis called plate tectonics(1-3). The basic idea isthat the Earth’s outermost part (called thelithosphere) consists of several large and fairlystable rock slabs called plates. The ten largest


plates are mapped in Figure 1-2. Each plateextends to a depth of about 80 kilometers.

Moving plates of the Earth’s surface (seeFigures 1-2 and 1-3) provide mechanisms for agreat deal of the seismic activity of the world.Collisions between adjacent lithospheric plates,destruction of the slab-like plate as it descendsor subducts into a dipping zone beneath islandarcs (see Figure 1-4), and spreading along mid-oceanic ridges are all mechanisms that producesignificant straining and fracturing of crustalrocks. Thus, the earthquakes in thesetectonically active boundary regions are calledplate-edge earthquakes. The very hazardousshallow earthquakes of Chile, Peru, the easternCaribbean, Central America, Southern Mexico,California, Southern Alaska, the Aleutians theKuriles, Japan, Taiwan, the Philippines,Indonesia, New Zealand, the Alpine-Caucasian-Himalayan belt are of plate-edge type.

As the mechanics of the lithospheric platesbecome better understood, long-termpredictions of place and size may be possiblefor plate-edge earthquakes. For example, manyplates spread toward the subduction zones atrates of from 2 to 5 centimeters (about one totwo inches) per year. Therefore in active arcslike the Aleutian and Japanese Islands andsubduction zones like Chile and westernMexico, knowledge of the history of largeearthquake occurrence might flag areas thatcurrently lag in earthquake activity.

Many large earthquakes are produced byslip along faults connecting the ends of offsetsin the spreading oceanic ridges and the ends ofisland arcs or arc-ridge chains (see Figure 1-2).In these regions, plates slide past each otheralong what are called transform faults.Considerable work has been done on theestimation of strong ground motion parametersfor the design of critical structures inearthquake-prone countries with eithertransform faults or ocean-plate subductiontectonics, such as Japan, Alaska, Chile andMexico. The Himalaya, the Zagros and Alpineregions are examples of mountain rangesformed by continent-to-continent collisions.These collision zones are regions of high

present day seismic activity. The estimation ofseismic hazard along continental collisionmargins at tectonic plates has not as yetreceived detailed attention.

Y E A R S A GO71 MI L L I ON



P R E S E NT


40° 70° 80° 90°

90°80°70°40°

40°

30°

20°

10°

0°

10°

20°

30°

40°

30°

20°

10°

0°

10°

20°

30°

Figure 1-3. Continued drift of the Indian plate towardsAsian plate causes major Himalayan earthquakes. (FromThe Collision Between India and Eurasia, by Molnar andTapponnier. Copyright 1977 by Scientific American, Inc.All rights reserved)

While a simple plate-tectonic theory is animportant one for a general understanding ofearthquakes and volcanoes, it does not explainall seismicity in detail, for within continentalregions, away from boundaries, largedevastating earthquakes sometimes occur.These intraplate earthquakes can be found onnearly every continent.

8 Chapter 1

One example of such earthquakes is theDashte-e-Bayaz earthquake of August 31, 1968in north-eastern Iran. In the United States, themost famous are the major earthquake series of1811-1812 that occurred in the New Madridarea of Missouri, along the Mississippi Riverand the 1886 Charleston, South Carolinaearthquake. One important group for example,which seems to bear no simple mechanicalrelation to the present plate edges, occurs innorthern China.

Such major internal seismic activityindicates that lithospheric plates are not rigid orfree of internal rupture. The occurrence ofintraplate earthquakes makes the prediction ofearthquake occurrence and size difficult in

many regions where there is a significantseismic risk.

1.3.2 Dilatancy in the Crustal Rocks

The crust of the continents is a rocky layerwith average thickness of about 30 km butwhich can be as thick as 50 km under highmountain ranges. Under the ocean, the crustalthickness is no more than about 5 km.

At a depth in the crust of 5 kilometers or so,the lithostatic pressure (due to the weight of theoverlying rocks) is already about equal to thestrength of typical uncracked rock samples atthe temperature (500° C) and pressureappropriate for that depth. If no other factorsentered, the shearing forces required to bring

A CT IV E VOL CA NO

I S L A ND A R C

S UB DUCT ION Z ONE

OCE A NI C T R E NCH

E A R TH QUA K E F OCI

F R A CTUR E Z ONE

T R A NS F OR M F A UL TMA GNE TIC S T R IP E S

MA GMA

CONTI NE NT

TR E NCH

R I S E

OCE A N

Figure 1-4. A sketch of the Earth's crust showing mid-oceanic ridges and active continental margin along a deep trench


about sudden brittle failure and frictional slipalong a crack would never be attained; rather,the rock would deform plastically. A wayaround this problem was the discovery that thepresence of water provides a mechanism forsudden rupture by reduction of the effectivefriction along crack boundaries. Nevertheless,in a normal geological situation, such as thecrust of coastal California, temperaturesincrease sufficiently fast so that at crustaldepths greater than about 16 km the elasticrocks become viscoelastic. Strain is thenrelieved by slow flow or creep rather than bybrittle fracture. The part of the crust above thistransition point is the seismogenic zone.

Studies of the time of travel of P and Swaves before the 1971 San Fernandoearthquake indicated that four years before itoccurred, the ratio of the velocity of the Pwaves to the velocity of the S waves decreasedrather suddenly by 10 percent from its averagevalue of 1.75. There was, thereafter, a steadyincrease in this ratio back to a more normalvalue. One explanation is the dilatancy model.This states that as the crustal rocks becomestrained, cracking occurs locally and thevolume of rock increases or dilates. Crackingmay occur too quickly for ground water to flowinto the dilated volume to fill the spaces so thecracks become vapor-filled. The consequentfall in pore pressure leads to a reduction mainlyin P wave velocities. Subsequent diffusion ofground water into the dry cracks increases thepore pressure, and provides water forlubrication along the walls of the cracks, whileat the same time, the P wave velocity increasesagain (see Figure 1-31 in Section 1.10 below).

The full implications and relevance of thedilatancy theory of earthquake genesis are notyet clear, but the hypothesis is attractive in thatit is consistent with precursory changes inground levels, electrical conductivity and otherphysical properties which have been noted inthe past before earthquakes. The theory has apotential for forecasting earthquakes undercertain circumstances. For example,measurement of the P wave velocity in thevicinity of large reservoirs before and after

impounding of water might provide a moredirect method of indicating an approachingseismic crisis near dams than is now available.

1.3.3 Explosions

Ground shaking may be produced by theunderground detonation of chemicals or nucleardevices. When a nuclear device is detonated ina borehole underground, enormous nuclearenergy is released. Underground nuclearexplosions fired during the past several decadesat a number of test sites around the world haveproduced substantial artificial earthquakes (upto magnitude 6.0). Resultant seismic waveshave traveled throughout the Earth’s interior tobe recorded at distant seismographic stations.

1.3.4 Volcanic Earthquakes

As Figure 1-4 shows, volcanoes andearthquakes often occur together along themargins of plates around the world. Likeearthquakes, there are also intraplate volcanicregions, such as the Hawaiian volcanoes.

Despite these tectonic connections betweenvolcanoes and earthquakes, there is no evidencethat moderate to major shallow earthquakes arenot essentially all of tectonic, elastic-reboundtype. Those earthquakes that can be reasonablyassociated with volcanoes are relatively rareand fall into three categories: (i) volcanicexplosions, (ii) shallow earthquakes arisingfrom magma movements, and (iii) sympathetictectonic earthquakes.

Among the three categories, Category (iii),tectonically associated with volcanoes, is moredifficult to tie down, as cases which may fit thiscategory, are rare. There is no report ofsignificantly increased volcanic activity in thegreat 1964 Alaska earthquake, but PuyehueVolcano in the Andes erupted 48 hours afterthe great 1960 Chilean earthquake. One mightsuppose that in a large earthquake the groundshaking would set up waves in reservoirs ofmagma; the general compression and dilatationof the gaseous liquid melt may trigger volcanicactivity.

10 Chapter 1

1.3.5 Collapse Earthquakes

Collapse earthquakes are small earthquakesoccurring in regions of underground cavernsand mines. The immediate cause of groundshaking is the sudden collapse of the roof of themine or cavern. An often observed variation isthe mine burst. This rock rupture happens whenthe induced stress around the mine workingscauses large masses of rock to fly off the mineface explosively, producing seismic waves.Mine bursts have been observed, for example,in Canada, and are especially common in SouthAfrica.

An intriguing variety of collapseearthquakes is sometimes produced by massivelandsliding. For example, a spectacularlandslide on April 25, 1974, along the MantaroRiver, Peru, produced seismic waves equivalentto a magnitude 4.5 earthquake. The slide had avolume of 1.6 x109 cubic meters and killedabout 450 people.

1.3.6 Large Reservoir-InducedEarthquakes

It is not a new idea that earthquakes mightbe triggered by impounding surface water. Inthe 1870’s, the U.S. Corps of Engineersrejected proposals for major water storage inthe Salton Sea in southern California on thegrounds that such action might causeearthquakes. The first detailed evidence of suchan effect came with the filling of Lake Meadbehind Hoover Dam (height 221 meters),Nevada-Arizona, beginning in 1935. Althoughthere may have been some local seismicitybefore 1935, after 1936 earthquakes were muchmore common. Nearby seismographssubsequently showed that after 1940, theseismicity declined. The foci of hundreds ofdetected earthquakes cluster on steeply dippingfaults on the east side of the lake and have focaldepths of less than 8 kilometers.

In Koyna, India, an earthquake (magnitude6.5) centered close to the dam (height 103

meters) caused significant damage onDecember 11, 1967. After impounding began in1962, reports of local shaking became prevalentin a previously almost aseismic area.Seismographs showed that foci wereconcentrated at shallow depths under the lake.In 1967 a number of sizable earthquakesoccurred, leading up to the principal earthquakeof magnitude 6.5 on December 11. This groundmotion caused significant damage to buildingsnearby, killed 177 persons, and injured morethan 1,500. A strong motion seismograph in thedam gallery registered a maximum accelerationof 0.63g. The series of earthquakes recorded atKoyna has a pattern that seems to follow therhythm of the rainfall (see Figure 1-5). At leasta comparison of the frequency of earthquakesand water level suggests that seismicityincreases a few months after each rainy seasonwhen the reservoir level is highest. Suchcorrelations are not so clear in some otherexamples quoted.

In the ensuing years, similar case historieshave been accumulated for several dozen largedams, but only a few are well documented.Most of these dams are more than 100 metershigh and, although the geological framework atthe sites varies, the most convincing examplesof reservoir induced earthquakes occur intectonic regions with at least some history ofearthquakes. Indeed, most of the thousands oflarge dams around the world give no sign ofearthquake induction. A poll in 1976 showedthat only four percent of large dams had anearthquake reported with magnitude greaterthan 3.0 within 16 kilometers of the dam.

Calculation shows that the stress due to theload of the water in even large reservoirs is toosmall to fracture competent rock. The bestexplanation is that the rocks in the vicinity ofthe reservoir are already strained from thetectonic forces so that existing faults are almostready to slip. The reservoir either adds a stressperturbation which triggers a slip or theincreased water pressure lowers the strength ofthe fault, or both.


Figure 1-6. Normal fault at the Corinth Canal, Greece.(Photo courtesy of L. Weiss.)

1.4 Earthquake fault sources

Field observations show that abrupt changesin the structure of rocks are common. In someplaces one type of rock can be seen butting upagainst rock of quite another type along a planeof contact. Such offsets of geological structureare called faults(1-4). Clear vertical offset oflayers of rock along an exposed fault in thewall of the Corinth canal, Greece, can be seenin Figure 1-6.

Faults may range in length from a fewmeters to many kilometers and are drawn on ageological map as continuous or broken lines(see Figure 1-7). The presence of such faultsindicates that, at some time in the past,movement took place along them. Suchmovement could have been either slow slip,which produces no ground shaking, or suddenrupture (an earthquake). Figure 1-8 shows oneof the most famous examples of sudden faultrupture slips of the San Andreas fault in April1906. In contrast, the observed surface faulting

Figure 1-5. The relationship between reservoir level and local seismic activity at Koyna Dam. (From Earthquakes, by BruceA. Bolt. Copyright 1978, 1999 W.H. Freeman and Company. Used with permission.)

12 Chapter 1

Figure 1-7. A simplified fault map of California. (From The San Andreas Fault, by Don L. Anderson. Copyright 1971 byScientific American, Inc. Al rights reserved.)

13

of most shallow focus earthquakes is muchshorter and shows much less offset. Indeed, inthe majority of earthquakes, fault rupture doesnot reach the surface and consequently is notdirectly visible. Geological mappings andgeophysical work show that faults seen at thesurface sometimes extend to depths of tens ofkilometers in the Earth’s crust.

It must be emphasized that most faultsplotted on geological maps are now inactive.However, sometimes previously unrecognizedactive faults are discovered from fresh groundbreakage during an earthquake. Thus, a faultwas delineated by a line of cracks in open fieldssouth of Oroville after the Oroville dam,California earthquake of August 1, 1975. Thelast displacement to occur along a typical faultmay have taken place tens of thousands or evenmillions of years ago. The local disruptiveforces in the Earth nearby may have subsidedlong ago and chemical processes involvingwater movement may have healed the ruptures,

particularly at depth. Such an inactive fault isnot now the site of earthquakes and may neverbe again.

In seismology and earthquake engineering,the primary interest is of course in active faults,along which rock displacements can beexpected to occur. Many of these faults are inwell defined plate-edge regions of the Earth,such as the mid-oceanic ridges and youngmountain ranges. However, sudden faultdisplacements can also occur away fromregions of clear present tectonic activity (seeSection 1.3.1).

Fault displacement in an earthquake may bealmost entirely horizontal, as it was in the 1906San Francisco earthquake along the SanAndreas fault, but often large vertical motionsoccur, (Fig. 1-9) such as were evident in the1992 Landers earthquake. In California in the1971 San Fernando earthquake, an elevationchange of three meters occurred across theruptured fault in some places.

Figure 1-8. Right-lateral horizontal movement of the San Andreas Fault in the 1906 earthquake the Old Sir FrancisHighway. (Photo by G.K Gilbert, courtesy of USGS.)

14 Chapter 1

Figure 1-9. Normal fault scarp associated with the 1992 Landers, California, earthquake (Photo by Dr. Marshall Lew).


The classification of faults depends only onthe geometry and direction of relative slip.Various types are sketched in Figure 1-10. Thedip of a fault is the angle that fault surfacemakes with a horizontal plane and the strike isthe direction of the fault line exposed at theground surface relative to the north.

A strike-slip fault, sometimes called atranscurrent fault, involves displacements ofrock laterally, parallel to the strike. If when westand on one side of a fault and see the motionon the other side is from left to right, the faultis right-lateral strike-slip. Similarly, we canidentify left-lateral strike-slip.

L E F T L A T E R A L F A U L T

R E V E R S E F A U L T( L E F T OB L I QU E E V E R S E F A U L T )L E F T L A T E R A L R E V E R S E F A U L T

( L E F T OB L I QU E NOR MA L F A U L T )L E F T L A T E R A L NOR MA L F A U L T

NOR MA L F A U L T

( f )( c )

( d)

( e )( b )

( a )

F A UL T L I NE

H A NGI NG WA L L

F OOTWA L L

DI P

Figure 1-10. Diagrammatic sketches of fault types

16 Chapter 1

A dip-slip fault is one in which the motionis largely parallel to the dip of the fault andthus has vertical components of displacement.A normal fault is one in which the rock abovethe inclined fault surface moves downwardrelative to the underlying crust. Faults withalmost vertical slip are also included in thiscategory.

A reverse fault is one in which the crustabove the inclined fault surface moves upwardrelative to the block below the fault. Thrustfaults are included in this category but aregenerally restricted to cases when the dip angleis small. In blind thrust faults, the slip surfacedoes not penetrate to the ground surface.

In most cases, fault slip is a mixture ofstrike-slip and dip-slip and is called obliquefaulting.

For over a decade it has been known thatdisplacement in fault zones occurs not only bysudden rupture in an earthquake but also byslow differential slippage of the sides of thefault. The fault is said to be undergoingtectonic creep. Slippage rates range from a fewmillimeters to several centimeters.

The best examples of fault creep come fromthe San Andreas zone near Hollister,California, where a winery built straddling thefault trace is being slowly deformed; in thetown, sidewalks, curbs, fences and homes arebeing offset. On the Hayward fault, on the eastside of San Francisco Bay, many structures arebeing deformed and even seriously damaged byslow slip, including a large water supply tunnel,a drainage culvert and railroad tracks thatintersect the zone.

Horizontal fault slippage has now also beendetected on other faults around the world,including the north Anatolian fault at Ismetpasain Turkey and along the Jordan Valley rift inIsrael. Usually, such episodes of fault slip areaseismic-that is, they do not produce localearthquakes.

It is sometimes argued that a large damagingearthquake will not be generated along a faultthat is undergoing slow fault slip, because theslippage allows the strain in the crustal rocks tobe relieved periodically without sudden

rupture. However, an alternative view is alsoplausible. Almost all fault zones contain aplastic clay-like material called fault gouge. Itmay be that, as the elastic crystalline rocks ofthe deeper crust stain elastically andaccumulate the energy to be released in anearthquake, the weak gouge material at the topof the fault zone is carried along by theadjacent stronger rock to the side andunderneath. This would mean that the slow slipin the gouge seen at the surface is an indicationthat strain is being stored in the basementrocks. The implication of this view is that, onportions of the fault where slippage occurs, anearthquake at depth could result from suddenrupture, but surface offset would be reduced.On the portion where slippage is small ornonexistent, offsets would be maximum. Aprediction of this kind can be checked afterearthquakes occur near places where slippage isknown to be taking place.

Sometimes aseismic slip is observed at theground surface along a ruptured fault that hasproduced an earlier substantial earthquake. Forexample, along the San Andreas fault break inthe 1966 earthquake on June 27 near Parkfield,California, offset of road pavement increasedby a few centimeters in the days following themain earthquake. Such continued adjustment ofthe crustal rock after the initial major offset isprobably caused partly by aftershocks andpartly by the yielding of the weaker surfacerocks and gouge in the fault zone as theyaccommodate to the new tectonic pressures inthe region.

It is clear that slow slippage, when it occursin built up areas, may have unfortunateeconomic consequences. This is another reasonwhy certain types of structures should not bebuilt across faults if at all possible. When suchstructures including dams and embankmentsmust be laid across active faults, they shouldhave jointed or flexible sections in the faultzone.


1.5 seismic waves

Three basic types of elastic waves make upthe shaking that is felt and causes damage in anearthquake(1-1). These waves are similar inmany important ways to the familiar waves inair, water, and gelatin. Of the three, only twopropagate within a body of solid rock. Thefaster of these body waves is appropriatelycalled the primary or P wave. Its motion is thesame as that of a sound wave, in that, as itspreads out, it alternately pushes (compresses)and pulls (dilates) the rock (see Figure 1-11).These P waves, just like sound waves, are ableto travel through both solid rock, such asgranite mountains, and liquid material, such asvolcanic magma or the water of the oceans.

The slower wave through the body of rock iscalled the secondary or S wave. As an S wavepropagates, it shears the rocks sideways at rightangles to the direction of travel (see Figure 1-12). Thus, at the ground surface S waves canproduce both vertical and horizontal motions.

The S waves cannot propagate in the liquidparts of the Earth, such as the oceans and theiramplitude is significantly reduced in liquefiedsoil.

The actual speed of P and S seismic wavesdepends on the density and elastic properties ofthe rocks and soil through which they pass. Inmost earthquakes, the P waves are felt first(1-5).The effect is similar to a sonic boom thatbumps and rattles windows. Some seconds laterthe S waves arrive with their significantcomponent of side-to-side motion, so that theground shaking is both vertical and horizontal.This S wave motion is most effective indamaging structures.

The speed of P and S waves is given interms of the density of the elastic material andthe elastic moduli. We let k be the modulus ofincompressibility (bulk modulus) and µ be themodulus of rigidity and ρ be the density. Thenwe have(1-5) for P waves;

P WAV E

COMP R E S S I ONS

DIL ATATI ONS

UNDIS TUR B E D ME DI UM

Figure 1-11. Ground Motion near the ground surface due to P waves. (From Nuclear Explosions and Earthquakes, byBruce A. Bolt. Copyright 1976 W. H. Freeman and Company. Used with Permission.)

18 Chapter 1

Vk

p =+ 4

3µ

ρ (1-1)

for S waves;

Vs = µρ

(1-2)

The third general type of earthquake wave iscalled a surface wave because its motion isrestricted to near the ground surface. Suchwaves correspond to ripples of water that travelacross a lake. Most of the wave motion islocated at the outside surface itself, and as thedepth below this surface increases, wavedisplacements become less and less.

Surface waves in earthquakes can bedivided into two types. The first is called aLove wave. Its motion is essentially the same asthat of S waves that have no verticaldisplacement; it moves the ground side to sidein a horizontal plane parallel to the Earth’ssurface, but at right angles to the direction ofpropagation, as can be seen from theillustration in Figure 1-13. The second type ofsurface wave is known as a Rayleigh wave.Like rolling ocean waves, the pieces of rockdisturbed by a Rayleigh wave move bothvertically and horizontally in a vertical planepointed in the direction in which the waves aretravelling. As shown by the arrows in Figure 1-14. Each piece of rock moves in an ellipse asthe wave passes.

Surface waves travel more slowly than bodywaves and, of the two surface waves, Lovewaves generally travel faster than Rayleighwaves. Thus, as the waves radiate outwardsfrom the earthquake source into the rocks of theEarth’s crust, the different types of wavesseparate out from one another in a predictablepattern.

DOUB L E A MP L I TUDE

WAVE L E NGTH

S WAV E

Figure 1-12. Ground motion near the ground surface dueto S waves. (From Nuclear Explosions and Earthquakes,by Bruce A. Bolt. Copyright 1976 W. H. Freeman andCompany. Used with Permission.)

An illustration of the pattern seen at a distantplace is shown in Figure 1-15. In this example,the seismograph has recorded only the verticalmotion of the ground, and so the seismogramcontains only P, S and Rayleigh waves, becauseLove waves are not recorded by verticalinstruments.

When the body waves (the P and S waves)move through the layers of rock in the crustthey are reflected or refracted at the interfacesbetween rock types, as illustrated in Figure 1-16a. Also, whenever either one is reflected orrefracted, some of the energy of one type isconverted to waves of the other type (seeFigure 1-16b).

L OV E WAV E

Figure 1-13. Ground motion near the ground surface dueto Love waves. (From Nuclear Explosions andEarthquakes, by Bruce A. Bolt. Copyright 1976 W. H.Freeman and Company. Used with Permission.)


R AY L E I GH WAV E

Figure 1-14. Ground motion near the ground surface dueto Rayleigh waves. (From Nuclear Explosions andEarthquakes, by Bruce A. Bolt. Copyright 1976 W. H.Freeman and Company. Used with Permission.)

When P and S waves reach the surface ofthe ground, most of their energy is reflectedback into the crust, so that the surface isaffected almost simultaneously by upward anddownward moving waves. For this reasonconsiderable amplification of shaking typicallyoccurs near the surface-sometimes doublingthe amplitude of the upcoming waves.

This surface amplification enhances theshaking damage produced at the surface of theEarth. Indeed, in many earthquakes mineworkers below ground report less shaking thanpeople on the surface.

b

a

RE

FL

EC

TE

D S

R E F L E CTE D P

INCID

E NT P

RE

FR

AC

TE

D S

RE

FR

AC

TE

D P

R OCK DI S CONT I NU I T Y ( OR B OU NDA R Y )

E A R T H QUA K E F OCU S

Figure 1-16. Reflection, refraction, and transformation ofbody waves. (From Nuclear Explosions and Earthquakes,by Bruce A. Bolt. Copyright 1999 W. H. Freeman andCompany. Used with permission.)

Figure 1-15. A seismograph record of the vertical component of a distant earthquake (third trace on bottom) on which thearrival of P, S and Rayleigh waves are marked. (Time increases from left to right.)

20 Chapter 1

Another reason for modification of theincoming seismic wave amplitudes near theground surface is the effect of layers ofweathered rock and soil. When the elasticmoduli have a mismatch from one layer toanother, the layers act as wave filtersamplifying the waves at some frequencies anddeamplifying them at others. Resonance effectsat certain frequencies occur.

Seismic waves of all types are progressivelydamped as they travel because of the non-elastic properties of the rocks and soils. Theattenuation of S waves is greater than that of Pwaves, but for both types attenuation increasesas wave frequency increases. One usefulseismological quantity to measure damping isthe parameter Q such that the amplitude A at adistance d of a wave frequency f (Hertz) andvelocity C is given by:

( )QCfdeAA

/0

π−= (1-3)

For P and S waves in sediments, Q is about 500and 200, respectively.

The above physical description isapproximate and while it has been verifiedclosely for waves recorded by seismographs ata considerable distance from the wave source(the far-field), it is not adequate to explain

important details of the heavy shaking near thecenter of a large earthquake (the near-field).Near a fault that is suddenly rupturing, thestrong ground shaking in the associatedearthquake consists of a mixture of variouskinds of seismic waves that have not separatedvery distinctly. To complicate the matter,because the source of radiating seismic energyis itself spread out across an area, the type ofground motion may be further mixed together.This complication makes identification of P, Sand surface waves on strong motion recordsobtained near to the rupturing fault particularlydifficult. However, much progress in this skill,based on intense study and theoreticalmodeling, has been made in recent years. Thisadvance has made possible the computation ofrealistic ground motions at specified sites forengineering design purposes(1-6).

A final point about seismic waves is worthemphasizing here. There is considerableevidence, observational and theoretical, thatearthquake waves are affected by both soilconditions and topography. For example, inweathered surface rocks, in alluvium andwater-filled soil, the size of P, S and surfacewaves may be either increased or decreaseddepending on wave frequency as they pass toand along the surface from the more rigidbasement rock.

Table 1-1. Magnitudes of Some Recent Damaging Earthquakes

Date Region Deaths Magnitude (MS)

December 7, 1988 Spitak, Armenia 25,000 7.0

August 1, 1989 West Iran, Kurima District 90 5.8

October 17, 1989 Santa Cruz Mountains, Loma Prieta 63 7.0

June 20, 1990 Caspian Sea, Iran Above 40,000 7.3

March 13, 1992 Erzinean, Turkey 540 6.8

July 16, 1990 Luzon, Phillipines 1,700 7.8

July 12, 1993 Hokkaido, Japan 196 7.8

September 29, 1993 Killari, India 10,000 6.4

January 17, 1994 Northridge, California 61 6.8

January 16, 1995 Kobe, Japan 5400 6.9

August 17, 1999 Izmit, Turkey 16,000 7.4

September 21, 1999 Chi Chi, Taiwan 2,200 7.6


The wave patterns from earthquake sourcesare much affected by the three-dimensionalnature of the geological structures(1-6). Clearevidence on this effect comes from recordingsof the 1989 Loma Prieta earthquake (see Table1-1). First, strong motion recordings show thatthere were reflections of high frequency Swaves from the base of the Earth’s crust at adepth of about 20 km, under southern SanFrancisco Bay. Secondly, the seismic S waves,like light waves, are polarized by horizontallayering into a horizontal component (SH type)and a vertical component (SV type). Because oflarge differences in the rock structure from oneside of the San Andreas fault to the other, therewere also lateral variations by refraction of SHwaves across this deep crustal velocity contrast.This produced significant amplitude variations,with azimuth from the seismic source, of thestrong ground shaking in a period range ofabout 1 to 2 seconds. In addition, there was

measurable scattering of shear waves by deepalluvial basins south of San Francisco Bay. Insum, the large wave amplitudes causedenhanced intensity in a region between SanFrancisco and Oakland, about 10 km wide by15 km long.

Finally, it should be noted that seismic Swaves travel through the rocks and soils of theEarth with a rotational component. Torsionalcomponents of ground motion are thought tohave important effects on the response ofcertain types of structures. Some building codesnow contain material on practices that takerotational ground motion into consideration.

1.6 EARTHQUAKE DAMAGEMECHANISMS

Earthquakes can damage structures invarious ways such as:

Figure 1-17. Tilting of buildings due to soil liquefaction during the Niigata (Japan) earthquake of 1964

22 Chapter 1

1. by inertial forces generated by severeground shaking.

2. by earthquake induced fires.3. by changes in the physical properties of the

foundation soils (e.g. consolidation, settling,and liquefaction).

4. by direct fault displacement at the site of astructure.

5. by landslides, or other surficial movements.6. by seismically induced water waves such as

seismic sea waves (tsunamis) or fluidmotions in reservoirs and lakes (seiches).

7. by large-scale tectonic changes in groundelevation.

Of the above categories, by far the mostserious and widespread earthquake damage andaccompanying loss of life are caused by severe

ground shaking. The bulk of this handbook isdevoted to design techniques and measures forreducing this type of hazard(1-7).

Fire hazards in earthquakes must also beemphasized. Vivid memories remain of thegreat conflagrations that followed the SanFrancisco 1906 earthquake and Tokyo’s 1923earthquake. In the 1906 San Franciscoearthquake perhaps 20 percent of the total losswas due directly to ground motions. However,the fire, which in three days burned 12 squarekilometers and 521 blocks of downtown SanFrancisco, was the major property hazard.

Soil related problems have caused majoreconomic loss in past earthquakes. One classicexample of this type of damage happened in the1964 earthquake of Niigata, Japan. Themaximum ground acceleration was

Figure 1-18. Faults rupturing under Managua (Nicaragua) during the earthquake of 1972


approximately 0.16g which considering theamount of damage, is not high. Expansion ofthe modern city of Niigata had involvedreclamation of land along the Shinano River. Inthe newly deposited and reclaimed land areasmany buildings tilted or subsided as a result ofsoil liquefaction (see Figure 1-17). 3,018houses were destroyed and 9,750 weremoderately or severely damaged in Niigataprefecture alone, most of the damage wascaused by cracking and unequal settlement ofthe ground. About 15,000 houses in Niigatacity were inundated by the collapse of aprotective embankment along the ShinanoRiver. The number of deaths was only 26.Precautionary and design measures againstearthquake induced soil problems are discussedin Chapter 3.

Perhaps surface fault displacements are themost frightening aspect of earthquakes to thegeneral public. However, although severe localdamage has occurred in this way, comparedwith damage caused by strong ground shaking,this type of damage is rather rare. Even in verylarge earthquakes, the area exposed to directsurface fault displacement is much smaller thanthe area affected by strong ground shaking. Oneof the clearest examples of damage caused bydirect fault displacement occurred in theManagua, Nicaragua earthquake of 1972,where four distinct faults ruptured under thecity (see Figure 1-18). The total length of faultrupture within the city was about 20 kilometers,and the maximum fault displacement on two ofthe faults reached about 30 centimeters. Evenin this case the total area damaged by directfaulting was less than one percent of the areadamaged by strong ground shaking.

Earthquake-induced landslides andavalanches, although responsible for majordevastation, are fortunately localized. The mostpronounced example of this kind of damageoccurred in Peru earthquake of May 31, 1970.This magnitude 7.75 earthquake led to thegreatest seismological disaster yet experiencedin the Western Hemisphere. An enormousdebris avalanche from the north peak ofHuascaran Mountain (see Figure 1-19)

amounting to 50,000,000 or more cubic metersof rock, snow, ice, and soil, travelled 15kilometers from the mountain to the town ofYungay with an estimated speed of 320kilometers per hour. At least 18,000 peoplewere buried under this avalanche, whichcovered the towns of Ranrahirca and most ofYungay.

Earthquake-induced changes in groundelevations (see Figure 1-20) may not causemajor injuries or loss of life. Their mostimportant threat is the damage they can causeto structures such as bridges and dams.

Seismic sea waves, or tsunamis, are longwater waves generated by suddendisplacements under water. The most commoncause of significant tsunamis is the impulsivedisplacement along a submerged fault,associated with a large earthquake. Because ofthe great earthquakes that occur around thePacific, this ocean is particularly prone toseismic sea waves.

Figure 1-19. Aerial view of Mt. Huascaran and the debrisavalanche that destroyed Yungay and Ranrahirca in May1970 Peru earthquake. (Photo courtesy of ServicioAerofotografico National de Peru and L. Cluff.)

24 Chapter 1

For earthquakes to generate tsunamis, dip-slip faulting (see Figure 1-10) seems to benecessary, and strike-slip faulting is almostnever accompanied by damaging tsunamis.History contains many accounts of great off-shore earthquakes being accompanied bydestructive tsunamis. In June 15, 1896, inHonshu region of Japan a tsunami with a visualrun-up height exceeding 20 meters (65 feet)drowned about 26,000 people. More recently,the Chilean earthquake of 1960 caused atsunami with a run-up height of 10 meters atHilo, Hawaii (see Figure 1-21). A tsunami atCrescent City, California, caused by the greatAlaskan earthquake of 1964, resulted in 119deaths and over $104,000,000 damage.

The most important scheme to prevent lossof life in the Pacific from tsunamis is theSeismic Sea Wave Warning System. Thewarning system is made up of a number of

seismological observatories including Berkeley,California; Tokyo, Japan; Victoria, Canada andabout 30 tide stations around the Pacific Ocean.The time of travel of a tsunami wave fromChile to the Hawaiian islands is, for example,about 10 hours and from Chile to Japan about20 hours. Under this system, therefore, there isample time for alerts to be followed up by localpolice action along coastlines so that peoplecan be evacuated.

Apart from the tsunami warning system, thehazard can be mitigated by using adequatedesign of wharf, breakwater and other facilitiesbased on techniques of coastal engineering.Often, however, zoning around coastlines isdesirable to prevent building in the most low-lying areas where tsunamis are known tooverwash the surface level. Sufficientinformation is nowadays usually available toallow local planners to make prudent decisions.

Figure. 1-20 Ground uplift along the fault in the 1999 Chi-Chi Earthquake (Photo by Dr. Farzad Naeim).


Before tsunami

After tsunami

Figure 1-21 Damage at Hilo, Hawaii, due to tsunami of May 23, 1960. (Photos courtesy of R. L. Wiegel.)

26 Chapter 1

1.7 QUANTIFICATION OFEARTHQUAKES

1.7.1 Earthquake Intensity

The oldest useful yardstick of the “strength”of an earthquake is earthquake intensity(1-1).Intensity is the measure of damage to works ofman, to the ground surface, and of humanreaction to the shaking. Because earthquakeintensity assessments do not depend oninstruments, but on the actual observation ofeffects in the meizoseismal zone, intensities can

be assigned even to historical earthquakes. Inthis way, the historical record becomes ofutmost importance in modern estimates ofseismological risk.

The first intensity scale was developed byde Rossi of Italy and Forel of Switzerland inthe 1880s. This scale, with values from I to X,was used for reports of the intensity of the 1906San Francisco earthquake, for example. A morerefined scale was devised in 1902 by the Italian

volcanologist and seismologist Mercalli with atwelve-degree range from I to XII. A version isgiven in Table 1-2, as modified by H.O. Wood

Table 1-2. Modified Mercalli Intensity Scale (MMI) of 1931

I. Not felt except by a very few under especially favorable circumstances.

II. Felt only by a few persons at rest, especially on upper floors of buildings. Delicately suspended objects mayswing.

III. Felt quite noticeably indoors, especially on upper floors or buildings, but many people do not recognize it asan earthquake. Standing motor cars may rock slightly. Vibration like passing of truck. Duration estimated.

IV. During the day felt indoors by many, outdoors by few. At night some awakened. Dishes, windows, doorsdisturbed; walls make cracking sound. Sensation like heavy truck striking building. Standing motor carsrocked noticeably.

V. Felt by nearly everyone, many awakened. Some dishes, windows, etc., broken; a few instances of crackedplaster; unstable objects overturned. Disturbances of trees, poles, and other tall objects sometimes noticed.Pendulum clocks may stop.

VI. Felt by all, many frightened and run outdoors. Some heavy furniture moved; a few instances of fallen plasteror damaged chimneys. Damage slight.

VII. Everybody runs outdoors. Damage negligible in buildings of good design and construction; slight to moderatein well-built ordinary structures; considerable in poorly built or badly designed structures; some chimneysbroken. Noticed by persons driving motor cars.

VIII. Damage slight in specially designed structures; considerable in ordinary substantial buildings, with partialcollapse; great in poorly built structures. Panel walls thrown out of frame structures. Fall of chimneys, factorystacks, columns, monuments, walls. Heavy furniture overturned. San and mud ejected in small amounts.Changes in well water. Persons driving motor cars disturbed.

IX. Damage considerable in specially designed structures; well designed frame structures thrown out of plumb;great in substantial buildings, with partial collapse. Buildings shifted off foundations. Ground crackedconspicuously. Underground pipes broken.

X. Some well-built wooden structures destroyed; most masonry and frame structures destroyed with foundations;ground badly cracked. Rails bent. Landslides considerable from river banks and steep sloes. Shifted sand andmud. Water splashed (slopped) over banks.

XI. Few, if any, (masonry) structures remain standing. Bridges destroyed. Broad fissures in ground. Undergroundpipelines completely out of service. Earth slumps and land slips in soft ground. Rail bent greatly.

XII. Damage total. Practically all works of construction are damaged greatly or destroyed. Waves seen on groundsurface. Lines of sight and level are distorted. Objects are thrown into the air.


and Frank Neumann to fit conditions inCalifornia(1-7).The descriptions in Table 1-2allow the damage to places affected by anearthquake to be rated numerically. These spotintensity ratings can often be separated by lineswhich form an isoseismal map (see Figure 1-22). Such intensity maps provide crude, butvaluable information on the distribution ofstrong ground shaking, on the effect of surficialsoil and underlying geological strata, the extentof the source, and other matters pertinent toinsurance and engineering problems.

Because intensity scales are subjective anddepend upon social and construction conditionsof a country, they need revising from time totime. Regional effects must be accounted for.In this respect, it is interesting to compare theJapanese scale (0 to VII) summarized in Table1-3 with the Modified Mercalli descriptions.

1.7.2 Earthquake Magnitude

If sizes of earthquakes are to be comparedworld-wide, a measure is needed that does notdepend, as does intensity, on the density ofpopulation and type of construction. A strictlyquantitative scale that can be applied toearthquakes in both inhabited and uninhabitedregions was originated in 1931 by Wadati in

Japan and developed by Charles Richter in1935 in California.

Figure 1-22. A typical isoseismal map. Similar maps arenow plotted by the Trinet program in Southern California.

Richter(1-5) defined the magnitude of a localearthquake as the logarithm to base ten of themaximum seismic wave amplitude in microns(10-4 centimeters) recorded on a Wood-Anderson seismograph located at a distance of

Table 1-3. Japanese Seismic Intensity Scale

0. Not felt; too weak to be felt by humans; registered only by seismographs.

I. Slight: felt only feebly by persons at rest or by those who are sensitive to an earthquake.

II. Weak: felt by most persons, causing light shaking of windows and Japanese latticed sliding doors (shoji).

III. Rather strong: shaking houses and buildings, heavy rattling of windows and Japanese latticed sliding doors,swinging of hanging objects, sometimes stopping pendulum clocks, and moving of liquids in vessels. Somepersons are so frightened as to run out of doors.

IV. Strong: resulting in strong shaking of houses and buildings. Overturning of unstable objects, spilling of liquidout of vessels.

V. Very strong: causing cracks in brick and plaster walls, overturning of stone lanterns and grave stones, etc. anddamaging of chimneys and mud and plaster warehouses. Landslides in steep mountains are observed.

VI. Disastrous: causing demolition of more that 1% of Japanese wooden houses; landslides, fissures on flat groundaccompanied sometimes by spouting of mud and water in low fields.

VII. Ruinous: causing demolition of almost all houses: large fissures and faults are observed.

28 Chapter 1

100 kilometers from the earthquake epicenter(see Figure 1-23). This means that every timethe magnitude goes up by one unit, theamplitude of the earthquakes waves increase 10times. Since the fundamental period of theWood-Anderson seismograph is 0.8 second, itselectively amplifies those seismic waves witha period ranging approximately from 0.5 to 1.5seconds. Because the natural period of manybuilding structures are within this range, thelocal Richter magnitude remains of value toengineers.

It follows from the definition of themagnitude, that it has no theoretical upper orlower limits. However, the size of anearthquake is limited at the upper end by thestrength of the rocks of the Earth’s crust. Since1935, only a few earthquakes have beenrecorded on seismographs that have had amagnitude over 8.0. At the other extreme,highly sensitive seismographs can recordearthquakes with a magnitude of less thanminus two. See Table 1-4 for an averagenumber of world-wide earthquakes of variousmagnitudes.

Generally speaking, shallow earthquakeshave to attain Richter magnitudes of more than5.5 before significant widespread damageoccurs near the source of the waves.

At its inception, the idea behind the Richterlocal magnitude scale (ML) was a modest one.It was defined for Southern California, shallowearthquakes, and epicentral distances less thanabout 600 kilometers. Today, the method hasbeen extended to apply to a number of types ofseismographs throughout the world (see Figure1-24). Consequently a variety of magnitudescales based on different formulas forepicentral distance and ways of choosing anappropriate wave amplitude, have emerged:

Surface Wave Magnitude (Ms) Surfacewaves with a period around 20 seconds areoften dominant on the seismograph records ofdistant earthquakes (epicentral distances ofmore than 2000 kilometers). To quantify theseearthquakes, Gutenberg defined a magnitudescale (Ms) which is based on measuring the

amplitude of surface waves with a period of 20seconds (1-8).

max 10 L M = log A

WOOD-ANDE RS ON S E IS MOGRAPH

max Z(t ) A ( in Microns )

MAGNIF ICATION = 2800 DAMPING = 0.8 CRITICAL NATURAL PE RIOD = 0.8 S E C COL L APS E , E TC.

E X PL OS ION, E ARTHQUAK E ,

E PICE NTE R GROUND S URF ACE

k

Z(t)100 km

C m

Figure 1-23. Definition of local Richter magnitude

Body Wave Magnitude (mb) Deep focusearthquakes have only small or insignificanttrains of surface waves. Hence, it has becomeroutine in seismology to measure the amplitudeof the P wave, which is not affected by thefocal depth of the source, and therebydetermine a P wave magnitude (mb). Thismagnitude type has also been found useful incontinental regions like the eastern UnitedStates where no Wood-Anderson instrumentshave operated historically.

Moment Magnitude (MW) Because ofsignificant shortcomings of ML, mb, and to alesser degree Ms in distinguishing betweengreat earthquakes, the moment magnitude scalewas devised(1-10). This scale assigns amagnitude to the earthquake in accordance withits seismic moment (M0) which is directlyrelated to the size of the earthquake source:

MW = (logM0)/1.5 - 10.7 (1-4)

where M0 is seismic moment in dyn-cm.Magnitude Saturation As described earlier,

the Richter magnitude scale (ML) measures theseismic waves in a period range of particularimportance to structural engineers (about 0.5-1.5 seconds). This range correspondsapproximately to wave-lengths of 500 meters to2 kilometers. Hence, although theoreticallythere is no upper bound to Richter magnitude,progressively it underestimates more seriouslythe strength of earthquakes produced by the


longer fault rupture lengths. The saturationpoint for the Richter magnitude scale is aboutML = 7. The body wave magnitude (mb)saturates at about the same value.

Figure 1-24. Amplification of seismic waves by variousseismographs. (From The Motion of Ground inEarthquakes, by David M. Boore. Copyright 1977 byScientific American, Inc. All rights reserved.)

Table 1-4. World-wide Earthquakes per YearMagnitude Ms Average No. > Ms

8 1

7 20

6 200

5 3,000

4 15,000

3 > 100,000

In contrast, the surface-wave magnitude(Ms) which uses the amplitude of 20 secondsurface waves (wave-length of about 60kilometers) saturates at about Ms = 8. Itsinadequacy in measuring the size of greatearthquakes can be illustrated by comparing theSan Francisco earthquake of 1906 and theChilean earthquake of 1960 (see Figure 1-25).Both earthquakes had a magnitude (Ms) of 8.3.

However, the area that ruptured in the SanFrancisco earthquake (the dashed area) wasapproximately 15 kilometers deep and 400kilometers long whereas the area that rupturedin the Chilean earthquake (the dotted area) wasequal to about half of the state of California.Clearly the Chilean earthquake was a muchlarger event.

The moment-magnitude scale (MW) is theonly magnitude scale which does not sufferfrom the above mentioned saturation problemfor great earthquakes. The reason is that it isdirectly based on the forces that work at thefault rupture to produce the earthquake and notthe recorded amplitude of specific types ofseismic waves. Hence, as can be expected,when moment magnitudes were assigned to theSan Francisco earthquake of 1906 and theChilean earthquake of 1960, the magnitude ofthe San Francisco earthquake dropped to 7.9,whereas the magnitude of the Chileanearthquake was raised to 9.5. Ms and Mw forsome great earthquakes are compared in Table1-5. Magnitudes of some recent damagingearthquakes are shown in Table 1-1.

In light of the above discussion, applicationof different scales have been suggested formeasuring shallow earthquakes of variousmagnitudes:MD for magnitudes less than 3ML or mb for magnitudes between 3 and 7Ms for magnitudes between 5 and 7.5MW for all magnitudes

Table 1-5. Magnitudes of Some of the GreatEarthquakes

Date Region Ms Mw

January 9, 1905 Mongolia 8¼ 8.4

Jan. 31, 1906 Ecuador 8.6 8.8

April 18, 1906 San Francisco 8¼ 7.9

Jan. 3, 1911 Turkestan 8.4 7.7

Dec. 16, 1920 Kansu, China 8.5 7.8

Sept. 1, 1923 Kanto, Japan 8.2 7.9

March 2, 1933 Sanrika 8.5 8.4

May 24, 1940 Peru 8.0 8.2

April 6, 1943 Chile 7.9 8.2

Aug. 15, 1950 Assam 8.6 8.6

30 Chapter 1

Date Region Ms Mw

Nov. 4, 1952 Kamchatka 8 9.0

March 9, 1957 Aleutian Islands 8 9.1

Nov. 6, 1958 Kurile Islands 8.7 8.3

May 22, 1960 Chile 8.3 9.5

March 28, 1964 Alaska 8.4 9.2

Oct. 17, 1966 Peru 7.5 8.1

Aug. 11, 1969 Kurile Islands 7.8 8.2

Oct, 3, 1974 Peru 7.6 8.1

July 27, 1976 China 8.0 7.5

Aug. 16, 1976 Mindanao 8.2 8.1

March 3, 1985 Chile 7.8 7.5

Sep. 19, 1985 Mexico 8.1 8.0

1.8 EARTHQUAKE SOURCEMODELS

Field evidence in the 1906 Californiaearthquake showed clearly that the strainedrocks immediately west of the San Andreasfault had moved north-west relative to the rocksto the east. Displacements of adjacent points

along the fault reached a maximum of 6 metersnear Olema in the Point Reyes region.

H.F. Reid (1-11) studied the triangulationsurveys made by the U.S. Coast and GeodeticSurvey across the region traversed by the 1906fault break. These surveys made in 1851-1865,1874-1892 and just after the earthquake,showed (i) small inconsistent changes inelevation along the San Andreas fault; (ii)significant horizontal displacements parallel tothe fault trace; and (iii) movement of distantpoints on opposite sides of the fault of 3.2meters over the 50-year period, the west sidemoving north.

Based on geological evidence, geodeticsurveys, and his own laboratory experiments,Reid put forth the elastic rebound theory forsource mechanism that would generate seismicwaves. This theory supposes that the crust ofthe Earth in many places is being slowlydisplaced by underlying forces. Differentialdisplacements set up elastic strains that reach

Figure 1-25. Fault rupture area for the San Francisco 1906 and Chile 1960 earthquakes. (Modified from The Motion ofGround in Earthquakes, By David M. Boore. Copyright 1977 by Scientific American, Inc. All rights reserved.)


levels greater than can be endured by the rock.Ruptures (faults) then occur, and the strainedrock rebounds along the fault under the elasticstresses until the strain is partly or whollyrelieved (see Figures 1-26 and 1-27). Thistheory of earthquake mechanism has beenverified under many circumstances and hasrequired only minor modification.

b

a

DI S P L A CE D W H I T E L I NE

D2

1D

TR

AC

EF

AU

LT

W H I T E L I NE ON R OA DD

A

B

B

A

Figure 1-26. A bird's eye view of market lines drawnalong a road AB, which crosses a fault trace at the groundsurface. (a) Elastic strain accumulation before faultrupture. (b) Final position after the fault rupture. (FromEarthquakes, by Bruce A. Bolt. Copyright 1999, W.H.Freeman and Company. Used with permission.)

The strain slowly accumulating in the crustbuilds a reservoir of elastic energy, in the sameway as a coiled spring, so that at some place,the focus, within the strained zone, rupturesuddenly commences, and spreads in alldirections along the fault surface in a series oferratic movements due to the uneven strengthof the rocks along the tear. This unevenpropagation of the dislocation leads to bursts of

high-frequency waves, which travel into theEarth to produce the seismic shaking thatcauses the damage to buildings. The faultrupture moves with a typical velocity of two tothree kilometers per second and the irregularsteps of rupture occur in fractions of a second.Ground shaking away from the fault consists ofall types of wave vibrations with differentfrequencies and amplitudes.

In 1966, N. Haskell(1-12) developed a model“in which the fault displacement is representedby a coherent wave only over segments of thefault and the radiations from adjacent sectionsare assumed to be statistically independent orincoherent.” The physical situation in thismodel is that the rupture begins suddenly andthen spreads with periods of acceleration andretardation along the weakly welded fault zone.In this model, the idea of statistical randomnessof fault slip or chattering in irregular stepsalong the fault plane is introduced.

More recently, Das and Aki(1-13) haveconsidered a fault plane having variousbarriers distributed over it. They conceive thatrupture would start near one of the barriers andthen propagate over the fault plane until it isbrought to rest or slowed at the next barrier.Sometimes the barriers are broken by thedislocation; sometimes the barriers remainunbroken but the dislocation reinitiates on thefar side and continues; sometimes the barrier isnot broken initially but, due to localrepartitioning of the stresses and possiblenonlinear effects, it eventually breaks, perhapswith the occurrence of aftershocks.

The elastic rebound model involving amoving dislocation along a fault planesegmented by barriers, over which roughnesses(or asperities) of various types are distributedstochastically, is thus the starting point for themodern interpretation of near-field records (1-14).Based on this model, there have been recently anumber of attempts to compute syntheticseismograms from points near to the source andcomparisons have been made with observations(see Section 1.10).

As mentioned earlier, there are differentkinds of fault ruptures. Some involve purely

32 Chapter 1

horizontal slip (strike-slip); some involvevertical slip (dip-slip). It might be expected thatthe wave patterns generated by fault geometriesand mechanisms of different kinds will bedifferent to a larger or lesser extent, because ofthe different radiation patterns produced. Thesedifferent geometries can be modeledmathematically by appropriate radiationfunctions.

The theory must also incorporate effects ofthe moving source. The Doppler-likeconsequences will depend on the speed of faultrupture and the directions of faulting. Thephysical problem is analogous (but moredifficult) to the problem of sound emissionfrom moving sources. The problem can beapproached both kinematically anddynamically. The acoustic problem shows thatin the far-field the pressure is the same as when

the source is at rest. However, in the near-field,the time dependence of both frequency andwave amplitude is a function of the azimuth ofthe site relative to the moving source (Figure 1-28).

In the case of a fault rupture toward a site ata more or less constant velocity (almost as largeas the sear wave velocity), most of the seismicenergy from the elastic rebound of the faultarrives in a single large pulse of motion(velocity or displacement), which occurs nearthe beginning of the record. This wave pulserepresents the cumulative effect of almost all ofthe seismic radiation from the movingdislocation. In addition, the radiation pattern ofthe shear dislocation causes this large pulse ofmotion to be oriented mostly in the directionperpendicular to the fault. Coincidence of theradiation pattern maximum for tangential

PA CIF IC P L ATE

NOR TH A ME R ICA N P L ATE

4

2

3

1

Figure 1-27. Elastic rebound model of earthquakes


motion and the wave focusing due to therupture propagation direction toward the siteproduces a large displacement pulse normal tothe fault strike.

The horizontal recordings of stations in the1966 Parkfield, California and the Pacoimastation in the 1971 San Fernando, Californiaearthquake were the first to be discussed in theliterature as showing characteristic velocitypulses. These cases, with maximum amplitudesof 78 and 113 cm/sec, respectively, consistedpredominantly of horizontally polarized SHwave motion and were relatively long period(about 2-3 sec). The observed pulses areconsistent with the elastic rebound theory ofearthquake genesis propounded by H.F. Reidafter the 1906 San Francisco earthquake. Thesevelocity and displacement pulses in thehorizontal direction near the source were firstcalled the source fling. Additional recordings inthe near field of large sources have confirmedthe presence of energetic pulses of this type,and they are now included routinely insynthetic ground motions for seismic design

purposes. Most recently, the availability ofinstrumented measured ground motion close tothe sources of the 1994 Northridge earthquake,the 1995 Kobe earthquake and the 1999 Chi-Chi earthquake provided important recordingsof the “fling” or velocity pulse.

As in acoustics, the amplitude andfrequency of the velocity and displacementpulses or “fling” have a geometrical focusingfactor, which depends on the angle between thedirection of wave propagation form the sourceand the direction of the source velocity.Instrumental measurements show that suchdirectivity focusing can modify the amplitudevelocity pulses by a factor of up to 10. Thepulse may be single or multiple, with variationsin the impetus nature of its onset and in its half-width period. A widely accepted illustration isthe recorded ground displacement of theOctober 15, 1979 Imperial Valley, California,earthquake generated by a strike-slip faultsource. The main rupture front moved towardEl Centro and away from Bonds Corner.

We now summarize the main lines of

T I MET I ME

OF P R OP A GA T I ON

R E S U L TA NT I N DI R E CT I ON

F R OM DI R E CT I ON OF P R OP A GA T I ON

R E S U L TA NT I N DI R E CT I ON A WAY

WAV E L E T SWAV E L E T S

DI R E CT I ON OF P R OP A GAT I ON

0

1

2

3

4

55

4

3

2

1

0

5432104321030 1 22100 10

Figure 1-28. Effect of direction of fault rupture on ground motion experienced at a site. [After Benioff(1-15) and Singh (1-16)]

34 Chapter 1

approach to modeling mathematically theearthquake source. The first model is thekinematic approach in which the time history ofthe slip on the generating fault is given a priori.Several defining parameters may be specified,such as the shape, duration, and amplitude ofthe source (or source time function and slip),the velocity of the slip over the fault surface,and the final area of the region over which theslip occurred. Numerous theoretical papersusing this approach have been published (seethe various discussions in Reference 1-16). Theprocess is a kind of complicated curve fittingwhereby the parameters of the source are variedin order to estimate by inspection the closenessof fit between recorded and computed near-field or far-field seismic waves. Once theseismic source is defined by this comparisonprocess, then the estimated source parameterscan be used to extrapolate from the knownground motions near to a historical source tothe future conditions required for engineeringpurposes.

A second approach is to use the differentialequations involving the forces which producethe rupture. This dynamic procedure hasreceived considerable emphasis lately. Thebasic model is a shear crack which is initiatedin the pre-existing stress field and which causesstress concentrations around the tip of thecrack. These concentrations, in turn, cause thecrack to grow. For example, analyticexpressions for particle accelerations in givendirections from a uniformly growing ellipticalcrack are derived, but the effect of crackstoppage is not always included (this unrealisticboundary condition is included in most work ofthis kind).

The key to the crack problem seems to be inmodeling the physical processes of the typicalcrack where there is interaction between thestress accumulation, rate of crack growth, andthe criterion of fracture. Most studies ondynamic shear cracks are concerned primarilywith the actual rupture process, and so thecrack is assumed to be embedded in an infinitehomogeneous medium. More realistic studiesconcerned with the seismic waves that are

recorded in the near-field need a numericalapproach, such as finite elements or finitedifferences, to handle geologic structuralconditions.

( c )( b)( a )

O O

O

OM = ( D) ( A ) ;D M = DdA ; M = DA

= S H E A R MODU L US

E QU I VA L E NT S H E A R S T R E S S = F /dA = D/ x

E QU I VA L E NT S H E A R S T R A I N = D/ x

E QU I VA L E NT COU P L E MOME NT M = ( F )( x )

F A U L T A R E AdA = E L E ME NT OF

F

F

Dx

= S H E A R S TR E S S = S H E A R S TR A I ND = F A U L T S L I P

F A U L T

L I NEL I NE

F A U L T

Figure1-29. Definition of seismic moment.

The studies mentioned under kinematic anddynamic models are built around the elasticrebound theory of slip on a fault. There are,however, more general studies that take a lessspecific view of the earthquake source(1-16).

It should be mentioned here that the scalarseismic moment (direction of force couplesalong the fault ignored) is given by

M0 = µAD (1-5)

where µ is the rigidity of the materialsurrounding the fault, A is the slipped area, andD is the amount of slip (see Figure 1-29). Theseismic moment is now the preferred parameterto specify quantitatively the overall size of anearthquake source.

Let us now summarize the physical modelfor the earthquake source generally accepted atpresent (see Figure 1-30). The source extendsover a fault plane in the strained crustal rocksby a series of dislocations, which initiate at thefocus and spread out with various rupturevelocities. The dislocation front changes speed


as it passes through patches on roughness(asperities on the fault).

Figure1-30. Simplified model of the vertical rupturesurface for the Coyote Lake earthquake of 1979 in

California

At the dislocation itself, there is a finitetime for a slip to take place and the form of theslip is an elastic rebound of each side of thefault leading to a decrease of overall strain. Theslip can have vertical components, as well ashorizontal components, and can vary along thefault. The waves are produced near thedislocation front as a result of the release of thestrain energy in the slippage(1-17).

This model resembles in many ways radiowaves being radiated from a finite antenna. Inthe far-field, the theory of radio propagationgives complete solutions for the reception ofradio signals through stratified media.However, when the receiver is very near to theextended antenna, the signal becomes mixedbecause of the finiteness of the source andinterference through end effects. The mainparameters in the model are: Rupture length L Rupture width W Fault slippage (offset) D Rupture velocity V Rise time TRoughness (barrier) distribution density φ(x)

The main work in theoretical seismology onsource properties today is to determine whichof these parameters are essential, whether theset is an optimal one, and how best to estimateeach parameter from both field observations

and from analysis of the seismograms made inthe near and the far-field.

A number of papers have now beenpublished that demonstrate that, in certainimportant cases, synthetic seismograms forseismic waves near the source can be computedrealistically(1-18). The synthetic motions can becompared with the three observed orthogonalcomponents of either acceleration, velocity, ordisplacement at a site. There remaindifficulties, however, in modeling certainobserved complexities and there is a lack ofuniqueness in the physical formulations whichlead to acceptable fits with observations (seealso Section 1.10).

1.9 SEISMIC RISKEVALUATION

Regional seismicity or risk mapsrecommended by seismic design codes (seeChapter 5) usually do not attempt to reflectgeological conditions nor to take into accountvariations due to soil properties. It is necessary,therefore, for critical construction in populatedregions to make special geological-engineeringstudies for each site, the detail and level ofconcern which is used depending on the densityof occupancy as well as the proposed structuraltype. In inhabited areas, more casualties arelikely to result from a failed dam or a damagednuclear reactor, for example, than from adamaged oil pipeline.

The factors which must be considered inassessment of seismic risk of a site have beenwell-defined in recent times(1-7). Here a briefsummary of these factors is listed.

Geological Input Any of the followinginvestigations may be required.1. Provision of a structural geologic map of the

region, together with an account of recenttectonic movements.

2. Compilation of active faults in the regionand the type of displacement (e.g., left-lateral, strike-slip, etc.). Fieldwork issometimes necessary here. Of particularimportance are geological criteria for faultmovements in Holocene time (the past

36 Chapter 1

10,000 years) such as displacements inrecent soils, dating by radio-carbon methodsof organic material in trenches across thefault, and other methods.

3. Mapping of the structural geology aroundthe site, with attention to scarps in bedrock,effects of differential erosion and offsets inoverlying sedimentary deposits. Such mapsmust show rock types, surface structuresand local faults, and include assessments ofthe probable length, continuity and type ofmovement on such faults.

4. In the case of through-going faults near thesite, geophysical exploration to define thelocation of recent fault ruptures and otherlineaments. Geophysical work sometimesfound useful includes measurement ofelectrical resistivity and gravity along aprofile normal to the fault. Other keygeological information is evidence forsegmentation of the total fault length, suchas step-over of fault strands, and changes instrike.

5. Reports of landslides, major settlements,ground warping or inundation from floodsor tsunamis at the site.

6. Checks of ground water levels in thevicinity to determine if ground waterbarriers are present which may beassociated with faults or affect the soilresponse to the earthquake shaking.Seismological Input Procedures for the

estimation of ground shaking parameters foroptimum engineering design are still in theearly stages and many are untested. It isimportant, therefore, to state the uncertaintiesand assumptions employed in the followingmethods:1. Detailed documentation of the earthquake

history of the region around the site.Seismicity catalogs of historical events areneeded in preparing lists of felt earthquakes.The lists should show the locations,magnitudes and maximum ModifiedMercalli intensities for each earthquake.This information should be illustrated bymeans of regional maps.

2. Construction, where the record permits, ofrecurrence curves of the frequency ofregional earthquakes, down to even smallmagnitudes (See the Gutenberg-Richterequation, Chapter 2). Estimates of thefrequency of occurrence of damagingearthquakes can then be based on thesestatistics.

3. A review of available historic records ofground shaking, damage, and other intensityinformation near the site.

4. Estimation of the maximum ModifiedMercalli intensities on firm ground near thesite from felt reports from each earthquakeof significance.

5. Definition of the design earthquakes(1-19).The geological and seismological evidenceassembled in the above sections should thenbe used to predict the earthquakes whichwould give the most severe ground shakingat the site. (Several such design earthquakesmight be necessary and prudent.) Wherepossible, specific faults on which rupturemight occur should be stated, together withthe likely mechanism (strike-slip, thrust, andso on). Likely focal depth and length ofrupture and estimated amount of faultdisplacement should be determined, withtheir uncertainties. These values are usefulin estimating the possible magnitude ofdamaging earthquakes from standard curvesthat relate fault rupture to magnitude (seeTable 1-6).Soils Engineering Input When there is

geological indication of the presence ofstructurally poor foundation material (such asin flood plains and filled tidelands), a fieldreport on the surficial strata underlying the siteis advisable. In addition, areas of subsidenceand settlement (either natural or fromgroundwater withdrawal) and the stability ofnearby slopes must be studied. We mentionhere only three factors that may require specialscrutiny.1. Study of engineering properties of

foundation soils to the extent warranted forthe type of building. Borings, trenchingsand excavations are important for such


analyses, as well as a search for thepresence of sand layers which may lead toliquefaction.

2. Measurements (density, water content, shearstrength, behavior under cyclic loading,attenuation values) of the physicalproperties of the soil in situ or by laboratorytests of borehole core samples.

3. Determination of P and S wave speeds andQ attenuation values and in the overburdenlayers by geophysical prospecting methods.

Table 1-6. Earthquake Magnitude versus FaultRupture Length

Magnitude (Richter) Rupture (km)

5.5 5-10

6.0 10-15

6.5 15-30

7.0 30-60

7.5 60-100

8.0 100-200

8.5 200-400

1.10 EARTHQUAKE ANDGROUND-MOTIONPREDICTION

Aspects of earthquake prediction that tendto receive the most publicity are: prediction ofthe place, prediction of the size, and predictionof the time of the earthquake. For most people,prediction of earthquakes means prediction ofthe time of occurrence. A more importantaspect for mitigation of hazard is the predictionof the strong ground motion likely at aparticular site (see also Chapters 2 and 3).

First, considering the status of forecastingof the time and size of an earthquake(1-20).Prediction of the region where earthquakes arelikely to occur has now been largely achievedby seismicity studies using earthquakeobservatories. Because empirical relationsbetween the magnitude of an earthquake andthe length of observed fault rupture have alsobeen constructed (see Table 1-6), rough limitscan be placed on the size of earthquakes for aregion.

Many attempts have been made to find cluesfor forewarning. Some physical clues forearthquake prediction are shown in Figure 1-31. In 1975, Chinese officials, using in part,increased seismicity (foreshocks) and animalrestlessness, evacuated a wide area before thedamaging Haicheng earthquake. However, inthe 1976 Tangshan catastrophe no forewarningswere issued. Elsewhere, emphasis has beenplaced on geodetic data, such as geodimetermeasurements of deformation of theCalifornian crust along the San Andreas fault.An ex post facto premonitory change in groundlevel was found after the Niigata earthquake,which if it had been discovered beforehand,might have served as one indication of thecoming earthquake.

Another scheme is based on detectingspatial and temporal gaps in the seismicity of atectonic region. In 1973, a prediction was madeby seismologists of the U.S. Geological Surveythat an earthquake with a magnitude of 4.5would occur along the San Andreas fault southof Hollister within the next six months. Theprediction was based on four principal shockswhich had occurred within 3 years on both endsof a 25-km-long stretch of the San Andreasfault, bracketing a 6-km-long section free fromearthquakes in that time interval. Theassumption was that the mid-section was stillstressed but locked, ready to release the elasticenergy in an earthquake. However, noearthquake occurred in the six monthspredicted. One difficulty with such methods isthe assessment of a zero epoch with which tocompare the average background occurrencerate.

A much publicized prediction experiment inCalifornia depended on the detection of a 22year periodicity in moderate magnitude (ML =5.5) earthquakes centered on the San Andreasfault near Parkfield. Similar earthquakes wererecorded in 1901, 1922, 1934, and 1966.Available seismograms in addition allowedquantitative comparison of the sourcemechanisms for the 1922, 1934, and 1966earthquakes. Many monitoring instrumentswere put in place to try to detect precursors for

38 Chapter 1

a possible 1988 repetition. These includedchanges in ground water tables, radonconcentration, seismicity, and fault slippage.The prediction of repetition of such acharacteristic earthquake in the years 1988 ± 4,proved to be unsuccessful.

Figure 1-31 Physical clues for earthquake prediction.[After predicting earthquakes, National Academy of

Sciences. (1-1).]

It has often been pointed out that even if theability to predict the time and size of anearthquake was achieved by seismologists,many problems remain on the hazard side.Suppose that an announcement were made thatthere was a chance of one in two of adestructive earthquake occurring within amonth. What would be the public response?Would the major industrial and commercialwork in the area cease for a time, thusdislocating large segments of the localeconomy? Even with shorter-term prediction,there are difficulties if work is postponed untilthe earthquake-warning period is over. Supposethat the predictive time came to an end and noearthquake occurred; who would take the

responsibility of reopening schools andresuming other activities?

Secondly, let us consider the calculation ofartificial (synthetic) seismic strong groundmotions(1-21). The engineering demand is for theestimation of certain parameters, which will beused for design and structural checking. Thereare two representations usually used. The firstis the seismogram or time history of the groundmotion at the site represented instrumentally bythe seismogram or accelerogram. The second isthe Fourier or response spectra for the wholemotion at the site. These two representationsare equivalent and are connected by appropriatetransformations between the time andfrequency domains.

In the simplest time-history representation,the major interest is in the peak amplitudes ofacceleration, velocity, and displacement as afunction of frequency of the ground motion.Another parameter of great importance is theduration of the strong ground motion, usuallygiven in terms of the interval of time above acertain acceleration threshold (say 0.05g), in aparticular frequency range. Typically, theduration of a magnitude 7 earthquake at adistance of 10 kilometers is about 25 seconds.The pattern of wave motion is also important inearthquake engineering because the nonlinearresponse of structures is dependent on thesequence of arrival of the various types ofwaves, In other words, damage would bedifferent if the ground motion were runbackwards rather than in the actual sequence ofarrival. In this respect, phasing of the groundmotion becomes very important and the phasespectra should be considered along with theamplitude spectrum.

The phasing of the various wave types onsynthetic seismograms can be determined byestimation of times of arrival of the P, S, andsurface waves. In this way, a realistic envelopeof amplitudes in the time histories can beachieved.


There are two main methods forconstructing synthetic ground motions. Thefirst is the more empirical and involvesmaximum use of wave motion parameters fromavailable strong ground motion records andapplication of general seismological theory(1-22).

The second method entails considerablecomputer analysis based on models of theearthquake source and assumptions onearthquake scaling(1-16,1-23). In the first method,the initial step is to define, from geological andseismological information, the appropriate

Figure 1-32. Ground motions recorded at Capitola, California in the 1989 Loma Prieta earthquake (a). The synthesizedground motions for site E23 (on rock) for a magnitude 7.2 earthquake, generated by shallow rupture 5 km away (b)

40 Chapter 1

earthquake sources for the site of interest. Thesource selection may be deterministic orprobabilistic and may be decided on grounds ofacceptable risk. Next, specification of thepropagation path distance is made, as well asthe P, S and surface wave velocities along thepath. These speeds allow calculation of theappropriate wave propagation delays betweenthe source and the multi-support points of thestructure and the angles of approach of theincident seismic waves.

The construction of realistic motions thenproceeds as a series of iterations, starting withthe most appropriate observed strong motion

record available, to a set of more specific timehistories, which incorporate the seismologicallydefined wave patterns. The strong motionaccelerograms are chosen to satisfy the seismicsource type (dip-slip, etc.), and pathspecifications for the seismic zone in question.The frequency content is controlled byapplying engineering constraints, such as aselected response amplitude spectrum(1-24,1-25).The target spectra is obtained, for example,from previous data analysis, often fromearthquake building codes. The fit between thefinal iteration and the target spectrum shouldfall within one standard error. Similarly, each

Figure 1-33. Response spectrum for the Capitola motion (Figure 1-32) compared with a target code spectrum (a). On theright, the ratio of frequency components is shown. Four iterations of the Fourier amplitude spectrum (phase fixed) of the

Capitola record produced the compatibility shown (b).


seismogram must maintain the specified peakground accelerations, velocities anddisplacements within statistical bounds. Theduration of each wave section (the P, S andsurface wave portions) must satisfy prescribedsource, path and site conditions. Figure 1-32shows iterations of ground motion at site E23for the east crossing of the San Francisco BayBridge for a Safety Evaluation Earthquake(SEE) of magnitude 7.2 on the nearby Haywardfault. The initial accelerogram chosen for theinput motion at the pier in question is thehorizontal ground motion recorded at Capitolaon firm ground in the 1989 Loma Prietaearthquake. The north-south component ofaccelerations are shown. These records are thenscaled for the required peak acceleration andthen the 5% damped response spectrum iscalculated. Two steps in the fitting of theresponse spectrum are shown in Figure 1-33.The ratio function between the calculated andtarget response spectrum is achieved. Theprocess produces a realistic motion for theinput site as shown in Figure 1-32. Thesynthetic motion has a wave pattern, including“fling”, duration, amplitude and spectrum thatare acceptable from both the seismological andengineering viewpoints.

At this stage, for large multi-supportstructures, account can be taken of theincoherency of ground motion. The first step isto lag, at each wave-length, the phase of theground motion to allow for the different timesof wave propagation between input points. Usemust then be made of a coherency function (seeFigure 1-34) that has been obtained fromprevious studies in similar geological regions.The process is to adjust the phase in the Fourierspectrum at each frequency so that the resultingphase spectrum for each input matches theselected coherency function.

Consider now the second method ofsynthesizing ground motions by computationalmeans using a scaling relation between a smallearthquake in the region and the ground motionrequired for engineering design or safetyevaluation.

These are usually, of course, much smallermagnitude sources than required. Such smallerrecorded ground motions contain essentialproperties of the particular earthquakemechanism involved, however, as well as theeffects of the particular geological structurebetween the source and the station. In terms ofthe theory of the response of mechanicalstructures, they are called empirical Green’sfunctions(1-16). They can be considered as theresponse of the local geological system to anapproximate impulse response of short durationapplied at the rupturing fault. If such empiricalGreen’s functions appropriate to the study ofthe site in question are not available, they mustbe constructed making certain mathematicalassumptions and introducing appropriate tensoranalysis (see Section 1.8).

Figure 1-34. The coherency function commuted fromadjacent (1750m separation) recordings of strong groundmotion in the Gilroy strong ground motion array duringthe 1989 Loma Prieta main shock. The heavy line is asmooth representation of the coherency effect.

The size of the earthquake is then selectedin terms of its seismic moment which, as wasseen in Section 1.8, is given quantitatively interms of the area of the fault slip and theamount of slip. An appropriate fault area isthen mapped in terms of an elementary mesh offinite elements. The empirical Green’s function

42 Chapter 1

mentioned above is then applied to eachelement of the mesh in the sequence required toachieve the appropriate rupture velocity acrossthe whole fault surface, as well as maintain thespecified overall moment for the largerearthquake. This superposition can be done interms of amplitude and phase spectra byavailable computer programs and the syntheticseismogram calculated at a point on the surfacein the vicinity of the original recordedempirical Green’s function. Variousmodifications of the process described abovehave been explored and a number of test caseshave been published.

As can be seen, the prediction of groundmotions in this way involves a number ofassumptions and extrapolations. A particularvalue of the method is to permit the exploration

of the effect of changing some of the basicparameters on the expected ground motions. Amajor difficulty is often the lack of knowledgeof the appropriate wave attenuation for theregion in question (see Section 1.5). Because ofthe importance of the application of attenuationfactors in calculation of predicted groundmotion at arbitrary distances, a great deal ofwork has been recently done on empiricalattenuation forms(1-26,1-27). The usual form forthe peak value at distance x is given by:

2xceae

ydM

bM

+=

Where a, b, c, and d are constants and M isthe magnitude. Recent empirical fits are givenin Reference 1-31.

As an example of the type of attenuationcurve that is obtained from actual ground

Figure 1-35. Acceleration ground motion recorded at Sylmar, California in the 1994 Northridge earthquake (a).Acceleration ground motion recorded at Pacoima Dam in the 1971 San Fernando earthquake (b).


motion recordings, a relation for the peakground displacements, D, in the 1989 LomaPrieta and 1992 Landers earthquake is given inthe following equation.

log D = 1.27 + 0.16 M - log r + 0.0004 r

Where D (cm) is the displacement, M is themoment magnitude, and r (km) is the distanceto the nearest point of energy release on thefault.

It is usual that attenuation changessignificantly from one geological province toanother and local regional studies need to bemade to obtain the parameters involved. Adiscussion is given in the book by Bullen andBolt(1-12) and in Reference 1-28.

The Northridge earthquake of California,January 17, 1994 allows an importantcomparison between the theoreticalseismological expectations and actual seismicwave recordings and behavior of earthquakeresistant structures. This magnitude 6.8earthquake struck southern California at 4:31AM local time on January 17, 1994. Theearthquake rebound occurred on a southerlydipping blind-thrust fault (see Section 1.4). Therupture began at a focus about 18 km deepunder the Northridge area of the San FernandoValley. The rupture then propagated along a45° dipping fault to within about 4 km of thesurface under the Santa Susannah Mountains.No major surface fault rupture was observedalthough the mountainous area sustainedextensional surface fracturing at various placesand were uplifted by tens of centimeters. Thecausative fault dipped in the opposite sense tothat which caused the neighboring 1971 SanFernando earthquake.

Like the 1971 earthquake, the 1994 shakingtested many types of design such as baseisolation and the value of the latest UniformBuilding Codes. Notable was, again, the failureof freeway bridges designed before 1971 andthe satisfactory seismic resistance of Post-1989(Loma Prieta earthquake) retrofitted freewayoverpasses. The peak accelerations, recordedby many strong motion accelerometers in Los

Angeles and the San Fernando Valley area,were systematically larger than average foraverage curves obtained from previousCalifornia earthquakes. It is notable that theground motions at the Olive View Hospital (seeFigure 1-35) are similar to those obtained at thePacoima Dam abutment site in the 1971 thrustearthquake (1-29).

1.11 CONCLUSIONS

The state of the art in strong motionseismology is now such that prediction of keyparameters, such as peak ground accelerationand duration of the significant portion ofshaking at a given site, is relatively reliable.Recent earthquakes, such as Chi Chi, Taiwan1999, have provided many recordings of thestrong motion and various site conditions ofrock and soil. In addition the great earthquakesof 1985 in Chile and Mexico (Ms ≈ 8) yieldedaccelerograms for large- subduction-zoneearthquakes. There are still, however, no clearrecordings of ground motion in the near-fieldfrom earthquakes with Ms >7.5 so thatextrapolations to synthetic ground motions inextreme cases of wide engineering interest arenot available.

To meet this need and others of engineeringimportance, more strong motion instrumentsare being placed in highly seismic areas of theworld. Of special interest is the recentoperation of clusters of digital instruments inurban areas. These allow the intensity ofseismic waves involved in strong motionshaking in the near-field to be rapidlycomputed and distributed on the world wideweb (see http://www.trinet.org).

Of special importance is the instrumentationof large structures (such as large dams and longbridges). But structural analysis requiresrealistic predictions of free field surfacemotions at all interface points on the supportingfoundation under design earthquake conditions.In the past, engineers have normally carried outseismic analyses under the incorrectassumption that the motions of all support

44 Chapter 1

points are fully correlated, i.e. “rigidfoundation” inputs are assumed(1-30).

As recent strong motion data have come tohand, these observations allow the study of theeffects of magnitude, epicentral distance, focaldepth, etc. on such characteristics(1-31).

The seismological problems dealt with inthis chapter will no doubt be much extended insubsequent years. First, greater sampling ofstrong-ground motions at all distances fromfault sources of various mechanisms andmagnitudes will inevitably become available.An excellent example is the wide recording ofthe 1999 Chi-Chi, Taiwan, earthquake(1-32).Secondly, more realistic three dimensionalnumerical models will solve the problem of thesequential development of the wave mixtures asthe waves pass through different geologicalstructures. Two difficulties may persist: thelack of knowledge of the roughness distributionalong the dislocated fault and, in many places,quantitative knowledge of the soil, alluvium,and crustal rock variations in the region.

REFERENCES

1-1 Bolt, B.A., Earthquakes, W.H. Freeman andCompany, New York, 4th edition 1999.

1-2 Udias, A., 1999, Principles of Seismology,Cambridge University Press, Cambridge.

1-3 Fowler, C.M.R., The Solid Earth, C.V.P.,Cambridge, England, 1990.

1-4 Yeats, R.S., Sieh, K. and Allan, C.R., 1997, TheGeology of Earthquakes, Oxford University Press,Oxford.

1-5 Richter, C.F., 1958, Elementary Seismology, W.H.Freeman and Co., San Francisco, California.

1-6 Lomax, A. and Bolt, B.A., 1992, broadbandwaveform modelling of anomalous strong groundmotion in the 1989 Loma Prieta earthquake usingthree-dimensional geological structures, Geophys.Res. Letters, 19: 1963-1966.

1-7 Reiter, L., 1990, Earthquake Hazard Analysis—Issues and Insights, Columbia University Press,New York. 254pp.

1-8 Gutenberg, B., and Richter, C.F., “Seismicity of theEarth and Associated Phenomena,” PrincetonUniversity Press, Princeton, 1954.

1-9 Real, C.R., and Teng, T., “Local Richter Magnitudeand Total Signal Duration in Southern California,”

Bull. Seism. Soc. Am., Vol. 63, No. 5, October,1973.

1-10 Kanamori, H., “The Energy Release in GreatEarthquakes,” Jour. Geo. Res., 82, 20, 1977.

1-11 Reid, H.F., “The Elastic Respond Theory ofEarthquakes,” Bull. Dept. Geol. Univ. Calif., 6, pp413-444, 1911.

1-12 Bullen, K.E. and Bolt, B.A., An introduction to theTheory of Seismology, Cambridge University Press,1985..

1-13 Das, S., and Aki, K., “Fault Plane with Barriers: AVersatile Earthquake Model,” J. Geophys. Res. 82,5658-5670, 1977.

1-14 Joyner, W.B. and D.M. Boore, 1988, Measurement,characterization, and prediction of strong groundmotion, in Von Thun, J.L. Ed., Proceedings, Conf.On Earthq. Engrg. And soil Dynamics II, RecentAdvances in Ground-Motion Evaluation, ASCEGeotechnical Special Publication No. 20. pp. 43-102.

1-15 Benioff, H., “Mechanism and Strain Characteristicsof the White Wolf Fault as Indicated by theAftershock Sequence, California Division of Minesand Geology, Bulletin 171, 199-202, 1955.

1-16 Bolt, B.A. (ed), Seismic Strong Motion Synthetics,Academic Press, 1987.

1-17 Scholtz, C.H., 1990, The Mechanics of Earthquakesand Faulting, Cambridge University Press,Cambridge.

1-18 O’Connell, D.R., 1999, Replication of apparentnonlinear seismic response with linear wavepropagation model, Science, 293: 2045-2050.

1-19 Bolt, B.A., 1996, From earthquake acceleration toseismic displacement, Firth Mallet-Milne Lecture,SECED, London, 50 pp.

1-20 Rikitake, T., 1990, Earthquake Prediction,Amsterdam Elsevier.

1-21 Frankel, A.D., 1999, How does the ground shake?,Science, 283: 2032-2033.

1-22 Zeng, Y., Anderson J.G., and Yu, G., 1994, Acomposite source model for computing realisticsynthetic strong ground motions, Geophys. Res., 21,pg 725.

1-23 Wald, D.J. Helmberger, D.V., and Heaton, T.H.,1991, Rupture model of the 1989 Loma Prietaearthquake from the inversion of strong-motion andbroadband teleseismic data, Bull.Seism. Soc. Am.,81: 1540-1572.

1-24 Ambraseys, N.N., Simpson, K.A., and Bommer, J.J.,1996, Prediction of horizontal response spectra inEurope, Earthq, Engrg. Struc. Dyn., 25: 371-400

1-25 Abrahamson, N.A., and Silver, W., 1997, Wmpiricalreponse spectral attenuation relations for shallowcrustal earthquakes, Seism. Res. Letters, 68: 94-127.

1-26 Vidale, J. and Helmberger, D.V., 1987, Path effectsin strong ground motion seismology, In: Seismic


Strong Motion Synthetics, B.A. Bolt, ed., AcademicPress, New York.

1-27 Boore, D.M., Joyner, W., and Fumal, T., 1997,Equations for estimating horizontal response spectraand peak accelerations form western NorthAmerican earthquakes: A summary of recent work,Seism. Res. Letters, 68: 128-153.

1-28 Boore, D.J. and Joyner, W., 1997, Siteamplifications for generic rock sites, Bull. Seism.Soc. Am., 87: 327-341.

1-29 Somerville, P.G., Smith, N.F. Graves, R.W., andAbrahamson, N.A., 1997, Modification of empiricalstrong ground motion attenuation relations toinclude the amplitude and duration effects of rupturedirectivity, Seism, Res. Letters, 68(1).

1-30 Bolt, B.A., Tsan, Y.B., Yeh, Y.T., and Hsu, M.K.,1982, Earthquake strong motions recorded by alarge near-source array of digital seismographs,Earthq. Engrg Struc. Dyn., 10: 561-573

1-31 Abrahamson, N.A. and Shedlock, J.M., 1997,Overview, Seism. Res. Letters, 68: 9-23.

1-32 Shin, T.C., Kuo, K.W., Lee, W.H.K., Teng., T.L.,and Tsai, Y.B., 2000, “A Preliminary Report on the1999 Chi-Chi (Taiwan) Earthquake,” Seism. Res.Letters, 71:23-2

46 Chapter 1

the nature of earthquake ground motion -...

Documents