Figure 1 depicts the main kinds of sources which generate seismic waves. These are oscillatory deformations which propagate through the earth's medium and can be recorded by seismic sensors. These sources are of different intensity and magnitude and thus of radiated seismic energy.
Fig. 1: Schematic classification of various kinds of events which generate seismic waves.
Natural earthquakes are caused by the fracturing of brittle rock when it is deformed under stress load beyond its breaking strength. Sudden rupture will occur, either along pre-existing faults or by breaking up a new fault. The crustal blocks "snap" into a new position. For very large earthquakes, the length of the ruptured zone is up to 1000 km and the slip along the fault can reach several meters.
Laboratory experiments with homogeneous consolidated rocks under surface conditions show that, depending also their porosity, a volume strain in the order of 10-2 - 10-3 (i.e. about 0.1% to 1% volume change) will break them. The strength of rocks in tension or shearing is generally much less. Shear strains in the order of about 10-4rad or even less may already cause fracturing of solid brittle rock. But the Earth's crust is already highly fractured and faulted. The strength of pre-fractured rock is much less than that of unbroken competent rock and is mainly controlled by the frictional resistance of the faults. The latter depends on the orientation of the faults with respect to the stress field and other conditions (cf. Scholz 1990) and varies in a wide range. Accordingly, deformations in the order of only 10-5 rad to 10-7rad, i.e. the bending of a lithosphere plate by about 0.1 mm to 1 cm over a distance of 1 km, may already cause shear faulting along pre-existing zones of weakness. But the shear strength depends also on the composition and fabric (anisotropy) of rock, the temperature, the containing pressure, the deformation velocity etc. and not only on the total cumulative strain. More details on the physics of earthquake faulting and related geological and seismotectonic conditions in the real earth can be found in Scholz (1990) and in the chapter on Physical and mathematical models of seismic source processes. Further recommended overview articles on lithosphere rheology stratification and their relation to crustal composition, age and heat flow were published by Meissner and Wever (1988), Ranalli and Murphy (1987) and Wever et al. (1987). They also explain the influence of these parameters on the thickness and maximum depth of the seismogenic zone in the earth crust.
The relative motion and related interaction, deformation and stress loading of the earth´s lithosphere plates is considered to be the main cause of tectonic earthquakes. The plates are driven, pushed and pulled by the slow motion of convection currents in the more plastic hot material of the earth mantle beneath the lithosphere. These relative motions are in the order of several cm per year. Fig. 2 shows the global pattern of earthquake belts and the related major lithosphere plates. Besides this there exist many smaller sub- or micro-plates, in continental regions in particular. While earthquakes within major plates and along the mid-oceanic ridges or continental rifts do generally occur within the (mostly upper) earth´s crust deep earthquakes down to a maximum of about 700 km depth may occur along ocean trenches, mainly of the Circum-Pacific earthquake and volcanic belt (e.g. along the Peru-Chile Trench, the Mexico Trench, the Ryukyu-Japan-Kurile-Aleuten Trenches, the New Hebrides Trench, the Tonga-Kermadec Trench) but also locally at some other places such as the Tyrrhenian Sea and the Aegean Sea in the Mediterranean Sea.
It is obvious that most earthquakes occur along the main plate boundaries. These boundaries constitute either zones of extension (e.g. in the up-welling zones of the mid-oceanic ridges or intra-plate rifts), transcurrent shear zones such as the west coast of North America with the famous San Andreas fault or the North Anatolian fault in Turkey, or zones of plate collision (e.g. the Himalaya thrust front) or subduction, the latter mostly along deep sea trenches. Accordingly, tectonic earthquakes may be of very different faulting type (strike-slip, normal, reverse or thrust faulting and mixed types (cf. Figs. 18 and 19 in the chapter Determination of fault-plane solutions).
The largest strain rates are observed near active plate boundaries (about 10-8 to 3¥10-10 per year). They are significantly less in active plate interiors (about 5¥10-10 to 3¥10-11 per year) or within stable continental platforms (about 5¥10-11 to 10-12 per year; cf. Giardini 1994). Accordingly, a critical cumulative strain in the pre-fractured/faulted seismogenic zone of the brittle lithosphere in the order of about 10-6 to 10-7 is reached roughly after some 100, 1000 to 10.000 or 10.000 to 100.000 years of loading, respectively. This agrees well with the order of time of so-called seismic cycles, i.e. the mean return period of the largest possible events, in these different plate environments (Muir-Wood 1993; Scholz 1990).
Fig. 2: Global distribution of earthquake epicentres according to the data catalogue of the United States National Earthquake Information Center (NEIC) 01/1977 - 07/1997 and the related major lithosphere plates.
Although there are hundreds of thousands to millions of weak tectonic earthquakes globally every year most of them can be recorded by sensitive instruments only. But in the long-term global statistical average about 50,000 earthquakes are strong enough to be potentially perceptible by the local population (if there is any!) in the near-source area. Some 10,000 may even cause slight damage, some 100 of them (magnitude M&Mac179;6) may result in strong damage in nearby settlements and built-up areas while about 1 event every year (M > 8) may result in wide-spread devastation and disaster (Neumann et al. 1989; Lay and Wallace 1995). During the 20th century the 1995 Great Hanshin/Kobe earthquake caused the greatest economic losses (about 100 billion US$), the 1976 Tangshan earthquake the most terrible human losses (about 243,000 people killed) while the Chile earthquake of 1960 released the largest amount of seismic energy Es of about 1019 Joule. This corresponds to about 50 to 100 years of the long-term annual average of global seismic energy release of about 1 - 2¥1017 J (Lay and Wallace 1995) and to about half a year of the total kinetic energy contained in the global lithosphere plate motion (Neumann et al. 1989). The total seismic moment of the Chile earthquake was about 3¥1023 Nm. It ruptured about 800 - 1000 km of the subduction zone interface at the Peru-Chile trench in a width of about 200 km (Boore 1977; Scholz 1990).
It should be noted that most of the faulting energy Ef is required to power the growth of the earthquake fracture and the production of heat and not primarily to produce seismic waves. Accordingly, the seismic efficiency, i.e the ratio of Es/Ef is perhaps only about 10% to less than 1%. It depends both on the stress drop during the rupture as well as on the total stress in the source region (Spence 1977; Scholz 1990).
Although the total energy released by the strongest historically known volcanic eruptions, may be even larger than the total faulting energy of the Chile earthquake their seismic efficiency is generally much smaller, also due to the long duration of such eruptions. Nevertheless, in some cases, also volcanic earthquakes may locally reach the shaking strength of destructive earthquakes (e.g. magnitudes of about 6). Most of the seismic oscillations produced in conjunction with sub-surface magma flows are of the so-called tremor type, i.e. long-lasting non-coherent and more ore less monochromatic oscillations which come from a two- or three-phase (liquid- and/or gas-solid) source process which is not narrowly localised in space and time. They cannot be analysed in the traditional way of seismic recordings from tectonic earthquakes or explosions and their source parameters be determined (cf. chapter Volcano seismology). In summary: About 85% of the total world-wide seismic moment release by earthquakes occurs in subduction zones and more than 95% by shallow earthquakes along plate boundaries. The other 5% are distributed between intraplate events and deep and intermediate focus earthquakes. The single 1960 Chile event accounts for about 25% of the total seismic moment release between 1904 and 1986. Volcanic earthquakes contribute only an insignificant amount to the global seismic moment release (see Scholz 1990).
Explosions are mostly man-made (although natural explosions in conjunction with volcanic eruptions or meteorite impacts, such as the Tunguska meteorite of June 30, 1908 in Siberia, may occur) and mostly controlled, i.e. with known location and source time. While in exploration seismology aimed at (uppermost) crustal investigations explosion yields Y of a few kg to tons are sufficient to produce seismic waves which can be recorded from several km to hundreds of km distance, underground nuclear explosions of kt up to Mt TNT (Trinitrotoluol) equivalent may be seismically recorded even world-wide (1 kt TNT = 4.2¥1012 J). Nevertheless, even the strongest of all fired underground nuclear tests of an equivalent yield of about 5 Mt TNT produced body-waves of magnitude mb ª 7 only. This corresponds to roughly 0.1% of the seismic energy released by the Chile earthquake of 1960. After 1974 underground tests with only Y £ 150 kt were carried out. Only well contained underground chemical or nuclear explosions have a sufficiently good seismic coupling factor e (e ª 10-2 to 10-3, i.e. only 1% to 0.1% of the total released explosion energy is transformed into seismic energy). The coupling factor of explosions on the surface or in the atmosphere (depending on the altitude), is much less (e ª 10-3 to 10-6) (cf. Griggs and Press 1961; Pomeroy and Oliver 1960).
Figure 3 depicts schematically an idealised sub-surface explosion and tectonic earthquake (of pure strike-slip type) in a homogeneous medium. It is obvious that the explosion produces in its initial phase a homogeneous outward directed compressional motion in all directions while the tectonic earthquake produces first motions of different amplitude and polarity in different directions which are again different for longitudinal (P-) and transversal (S-) waves. These characteristics can be used to identify the type of source processes (cf. sub-chapter 4) and to discriminate between explosions and tectonic earthquakes. Implosions, e.g. of a karst caves or mining galleries produce a similar "first motion pattern" like an explosion but with opposite sign (i.e. dilatational - first motion). Contrary to this, mining induced tectonic rock bursts or tectonic events triggered/induced by high dam reservoir load and pore-pressure changes or by fluid/gas injections into or rapid withdrawal from underground reservoirs may look more similar to tectonic earthquakes. As compared to tectonic earthquakes the duration of the source process of explosions and the so-called rise time to the maximum level of displacement is much shorter (milliseconds as compared to seconds up to a few minutes) and more impulsive (Fig. 4). Accordingly, explosions of comparable body wave magnitude excite more high-frearthquakeuent oscillations (cf. seismic source spectra).
Fig. 3: Schematic sketches of an idealised underground explosion and of a pure strike-slip earthquake along a vertically dipping fault. The fault motion is "left-lateral", i.e. counter-clockwise. The arrows show the directions of compressional (outward, +, grey shaded) and dilatational (inward, -, white areas) motions. The patterns shown on the surface indicate the azimuthal variation of observed amplitudes at seismic stations and their polarity in seismic records. While point-like explosions in an isotropic medium should show no azimuth-dependent amplitudes and compressional first motions only, amplitudes and polarities vary for a tectonic earthquake. The dotted amplitude lobes in Fig. 3, right side, indicate qualitatively the different azimuth dependence of shear (S-) waves as compared to longitudinal (P-) waves (rotated by 45°) but their absolute values are much larger (about 5 times) than that of P-waves.
Fig. 4: Schematic diagrams of the different source functions of explosions (left) and earthquakes (right). P - pressure in the explosion cavity, D - fault displacement, t - time, to - origin time of the event, tr - rise time of P or D to its maximum (or in case of multiple events intermediate max.) values, trf- rise time of fast rupture, trs - rise time of slow rupture; the step function in the right diagram would correspond to an earthquake with infinite velocity of crack propagation vcr. Current rupture models assume vcr to be about 0.6 to 0.9 of the velocity vs of shear wave propagation.
Very different seismic signals are produced by storms over oceans or large water basins (seas, lakes, reservoirs) as well as wind action on the topography and vegetation or built-up surface cover, so-called storm microseisms, and due to human activities such as rotating or hammering machinery, traffic etc., so-called cultural seismic noise. Rushing waters or gas/steam (in rivers, water falls, dams, pipelines, geysers) may be additional sources of natural or man-made seismic noise. They are mostly not well localised in space and fixed to a defined origin time. Accordingly, they produce more or less permanent on-going non-coherent interfering signals of more or less random amplitude fluctuations in a very wide frequency range of about 16 octaves (about 50 Hz to 1 mHz) which are often controlled in their intensity by the season (natural noise) or day time (man-made noise). Despite of the large variation of noise amplitudes world-wide by about 6 to 10 orders of magnitude they are generally much smaller than those of earthquakes and not felt by men. Because of all these significant differences to coherent seismic sources microseisms and seismic noise are not dealt with in this chapter(see chapter 4: Seismic signals and noise).
Rock falls may last for several minutes and cause seismic waves but generally with less distinct onsets and separation of wave groups.
The collapse of karst caves, mining induced rock bursts or collapses of mining galleries are generally of an implosion type. Accordingly, their first motion patterns should show dilatation in all azimuth directions if not a secondary tectonic event has been triggered by the collapse. The strongest events may reach magnitudes up to about M = 5.5 and be recorded world-wide (e.g. Bormann et al. 1992).
Reservoir induced (not caused!) earthquakes have been frequently observed in conjunction with the impoundment or rapid water level changes of high dams. Since they are triggered events along pre-existing and pre-stressed tectonic faults they show the typical polarity patterns of tectonic earthquakes. The strongest events reported so far reached magnitudes up to 6.5 (e.g. Koyna earthquake in 1967). But strong events of this type are very rare.
The size of a seismic source may be characterised via its macroseismic intensity I. The latter describes the strength of the resulting shaking in terms of human perceptions, damages to buildings and other structures as well as changes in the surrounding environment. I depends on the distance from the source and the underground conditions and is mostly classified according to scales of 12 degrees (e.g. Grünthal 1998). From an analysis of the areal distribution of perceptions and damages one can estimate the intensity Io in the (epicentral) source area as well as the source depth h. There exist correlation relationships between Io and other instrumentally determined measures of the earthquake size such as the magnitude as well as between I and ground acceleration. For more details see the chapter Macroseismic and strong-motion parameters.
The magnitude is a logarithmic measure of the size of an earthquake or explosion based on instrumental measurements. The magnitude concept was first proposed by Richter (1935). Magnitudes are derived from instrumental recordings of ground motion amplitudes and periods or from signal duration. There is no a-priory scale limitation or classification of magnitudes as for macroseismic intensities. Therefore, magnitudes are often incorrectly referred to in the press as "... according to the open-ended RICHTER scale...". In fact, nature limits the maximum size of tectonic earthquakes which is controlled by the maximum size of a brittle fracture in the lithosphere. The largest moment magnitude Mw observed so far was that of the Chile earthquake in 1960 (Mw ª 9.5; Kanamori 1977). On the other hand, the magnitude scale is open to the lower end. Since nowadays high sensitive instrumentation close to the sources may record much smaller events than those with zero-magnitude according to Richter´s original definition their magnitude values become negative. Via empirical energy-magnitude-relationships the seismic energy Es irradiated by the seismic source as seismic waves can be estimated. Common relationships are those given by Gutenberg and Richter (1954, 1956) between Es and the so-called surface wave magnitude Ms and the body-wave magnitude mB, log Es = 11.8 + 1.5 Ms and log Es = 5.8 + 2.4 mB, respectively, with Es in erg (1 erg = 10-7 J). According to the first relationship, a change of M by two units corresponds to a change in Es by a factor of 1000. Nowadays, based on the analysis of digital recordings, there exist also direct procedures to estimate Es (e.g. Purcaru and Berckhemer 1978; Boatwright and Choy 1986; Kanamori et al. 1993, Choy and Boatwright 1995) and to define an "energy magnitude" ME (see section 4: Magnitude of seismic events, below). Since most of the seismic energy is concentrated in the higher frequency part around the corner frequency of the spectrum ME is a suitable measure of the earthquakes potential for damage. Contrary to this, the seismic moment (see below) is related to the final static displacement after an earthquake and consequently, the moment magnitude Mw to the tectonic effects of an earthquake.
Another quantitative measure of the size and strength of a seismic shear source is the so-called scalar static seismic moment:
with m - Young or shear modulus of the medium, `D - average final static displacement after the rupture earthquake, A - the surface area of the rupture. Mo is a measure of the irreversible inelastic deformation in the rupture area. This inelastic strain is described in (1) by the factor`D A. On the basis of reasonable average assumptions about the Young modul m and the stress drop Ds and assuming Ds/m = constant Kanamori (1977) derives the relationship Es = 5¥10-5 Mo. More information about the deformation in the source is described by the seismic moment tensor. Its determination is nowadays a standard routine in the analysis of strong earthquakes by means of waveform inversion of digital broadband records (cf. section 3: Source parameters and moment tensor solutions).
Mo can be determined from the spectra of seismic waves observed at the Earth surface by using the relationship:
with: d - hypocentral distance between the event and the seismic station; r - average density of the rock and Vp/s - velocity of the P- or S-waves around the source; Rqf - a factor correcting the observed seismic amplitudes for the influence of the radiation pattern of the seismic source (cf. Fig.3 above and Fig. 11 and 12 in the section Determination of fault plane solutions), uo - the low-frequency amplitude level as derived from the seismic spectrum of P- or S-waves, corrected for the instrument response and wave propagation effects (geometrical spreading, wave attenuation, boundaries, surface amplification). For more details see the Worksheet Exercise on the determination of source parameters derived from seismic spectra.
A simple seismic shear source with linear rupture propagation according to Aki (1967) shows in the far-field smooth displacement and velocity spectra. When corrected for the effects of geometrical spreading and attenuation we get so-called "source spectra" similar to the generalised ones shown in Figure 5.
Fig.5: "Source spectra" of ground displacement (left) and velocity (right) for a seismic shear source. The broken line shows the increase of corner frequency with decreasing seismic moment of the event while the dotted line gives the approximate source spectrum for a well contained underground nuclear explosion of an equivalent yield of 1 kt TNT.
In Figure 5 the low-frequency values have been scaled to the scalar seismic moment Mo (left) and moment rate dMo/dt (right), respectively. The given magnitude values Ms correspond to the related linear Ms-logMo relationship given by Kanamori (1977). Note, that there exist other, non-linear empirical Ms-logMo relationships (e.g. Geller, 1976; Purcaru and Berckhemer, 1978). They correspond to even stronger saturation effects than the one shown in Fig. 5. According to them the curve shown in Fig. 5 for Ms = 8 would correspond to Ms ª 8.5, the largest surface wave magnitude ever determined, e.g. for the Chile earthquake 1960. The latter had a seismic moment Mo of about 2 - 5¥1023 Nm.
The following general features are obvious from Figure 5:
The main causes for this difference in Es and high-frequency content between UNE and earthquakes are:
Note: Details of calculated "source spectra" depend on the model assumptions of the rupture process. E.g., when the rupture is propagating not one- but bi-directional and the spectrum of the source-time function is - more realistically - proportional to f-2, then the high-frequency decay for f >> fc is proportional to f-3. On the other hand, when the linear dimensions of the fault rupture differ in length and width then two corner frequencies will occur. A third one is related to the specifics of the source time function. Whether the two or three corner frequencies are separable will depend on their difference and, in case of real spectra derived from data limited in both time and frequency domain, mainly on the signal-to-noise ratio. Normally, real data are too noisy as to allow the discrimination between different types of rupture propagation and geometry.
The general shape of the seismic source spectra can be understood as follows: We know from optics that under a microscope no objects can be resolved and details of it seen anymore when its size becomes smaller than the wavelength l of the light with which it is observed. In this case it appears just as a blurred point or dot. In order to resolve more details, electron microscopes are used which operate with much smaller equivalent wavelength. Similarly, the reverse is true in seismology. When observing at some distance a seismic source of radius r with wavelengths l >> r then these do not carry any information about the details of the source process but only of the overall (integral) source process. They all "see" the source only as a point source. Accordingly, their spectral amplitudes are all constant forming the spectral plateau. On the other hand, wavelength l << r can resolve internal details of the rupture process. In case of an earthquake they correspond to smaller and smaller elements of the rupture processes or of the fault roughness (asperities and barriers). Accordingly, their spectral amplitudes decay rapidly with higher frequencies.Thus, the corner frequency fc marks a critical position in the spectrum which is obviously related to the size of the source. According to Brune (1970) and Maderiaga (1976), who assumed both a circular fault model, the corner frequency in the P- or S-wave spectrum, respectively is fc p/s = cm Vp,s / p r while according to Haskell (1964), who assumed a rectangular fault, fc p/s = cm Vp,s / (L¥W)1/2 with L the length and W the width of the fault. The values cm are model dependent constants. Accordingly, the critical wavelength l c = V/ fc, beyond which the source can be realised as a point source only, is l c = cm p r or l c = cm (L¥W)1/2, respectively.
Thus, when being able to determine both the source area (based on model assumption of the shape of the rupture!) and the seismic moment from seismic spectra, one can estimate from (1) the average total displacement `D. Knowing it other parameters such a the stress drop in the source area due to the faulting can be inferred too. More details are given in the chapter Physical and mathematical models of seismic source processes and in the related exercise Determination of source parameters derived from seismic spectra.
Assuming that the earthquake rupture occurs along a plane fault surface the orientation of this plane in space can be described by the angles of its strike f (against north) and dip d (against the horizontal) and additionally the direction of slip on the fault by the rake angle l. Fig. 16 and 17 in section 2 below on Determination of fault plane solutions define these angles and how to determine them from a stereographic (Wulff net) or equal area (Lambert-Schmidt net) projection of first motion polarity observations. It can be shown that a rupture along a plane perpendicular to the above mentioned fault plane with a slip vector perpendicular to the slip on the first plane causes an identical angular distribution of first motions. Therefore, on the bases of first motion analysis alone one cannot decide which of the two planes is the true acting plane.
Note that the fault plane solution (i.e. the information about the orientation of the fault plane and of the fault slip in space) forms, together with the information about the static seismic moment Mo , the seismic moment tensor. Its principal axes coincide with the direction of the so-called pressure axis P and the tension axis T given with fault plane solutions. They should not be mistaken for the principal axes s1, s2 and s3 (with s1 > s2 > s3) of the acting stress field in the earth which is described by the stress tensor. Only in case of a fresh crack in a homogeneous isotropic medium in a full space with no pre-existing faults and vanishing internal friction P is in the direction of s1 and T opposite to s3. P and T are perpendicular to each other and form, under the above conditions, an angle of 45° with the two possible conjugate fault planes (45°-hypothesis) which are in this case perpendicular to each other (cf. Figures 10 and 19 in section Determination of fault-plane solutions). The orientation of P and T is also defined by two angles each, the azimuth and the plunge. They can be determined by knowing the respective angles of the fault plane (cf. Exercise on fault plane solutions). If the above model assumptions hold true, one can, knowing the orientation of P and T in space determine that of s1 and s3. Most of the data used for compiling the global stress map (Zoback 1992) come from earthquake fault-plane solutions calculated under these assumptions.
In reality, the internal friction of rocks is not zero. This results, according to Andersons´s theory of faulting (1951) for most rocks in conjugate pairs of faults which are oriented at about ± 30° to s1. In this case, the direction of P and T, as derived from fault plane solutions, will not coincide with the principal stress directions. Near to the surface of the earth one of the principal stresses is almost always vertical. In case of a compressive regime, the minimum stress s3 is vertical while s1 is horizontal This results, when fresh faults are formed in unbroken rock, in about 30° dipping thrust faults striking parallel or anti-parallel to s2. In an extensional environment, s1 is vertical and the resulting dip of fresh normal faults is about 60°. When both s1 and s3 are horizontal, vertical strike-slip faults will develop, striking with ± 30° to s1. But most earthquakes are not due to fresh faulting but associated with the reactivation of of pre-existing faults. Since the frictional strength of faults is generally less than that of unbroken rock, faults may be reactivated also at angles between s1 and fault strike different from 30° In a pre-faulted medium this mostly prevents failure on a new fault. Accordingly, there is no straight forward way to infer from the P and T directions determined for an individual earthquake the directions of the acting principal stress. On the other hand it is possible to infer the regional stress based on the analysis of many earthquakes in that region since the possible suite of rupture mechanisms activated by a given stress regime is constrained. This method aims at finding an orientation for s1 and s3 which is consistent with as many of the actually observed fault plane solutions as possible (e.g. Gephart and Forsyth 1984; Reches 1987; Rivera 1989).
Above we have considered suitable parameters for describing the size and faulting parameters of earthquakes and to some extent also of explosions. As a matter of fact, earth ruptures in real nature, as geological faults in general, are no planes, neither circular nor rectangular, neither homogeneously nor unilaterally slipping etc. All these are simplifying zero or at best first order model approximations to the truth in order to make the problem with limited data tractable at all. Real faults show jogs, steps, branching, splays etc. both in their horizontal and vertical extent (Fig. 6). Such jogs and steps, depending on their severity, are impediments to rupture, so-called asperities or barriers, as are bumps or roughness features along the contacting fault surfaces. More illustrating examples can be found in Scholz (1990). Since these features exist at all scales, which implies the self-similarity of fracture and faulting processes and their fractal nature, this will necessarily result in heterogeneous dynamic rupturing and finally also in rupture termination.
Fig. 6: Several fault zones mapped at different scales and viewed approximately normal to slip (taken from Scholz, 1990, with permission of Cambridge University Press).
According to Figure 7 the complexity of the rupture process in time is very common to earthquakes, i.e. they are mostly multiple ruptures. This holds not only for very large earthquakes but is often observed even for small ones (Kikuchi and Ishida, 1993). And obviously, each event has its own "moment rate fingerprint".
Only in a few cases dense seismic strong-motion networks exist in the very source region of strong earthquakes which permit a detailed analysis of the rupture history in space and time described by the moment-rate density. As an example, Fig. 8 depicts data by Mendez and Anderson (1991) on the rupture process of the 1985 Michoacán Mexico earthquake. Shown are snapshots, 4 s apart from each other, of the dip-slip velocity field. One recognises two main clusters of maximum slip velocity being about 120 km and 30 s apart from each other. The related maximum cumulative displacement was more than 3 m in the first and more than 4 m in the second main cluster at about 55 km and 40 km depth, respectively. About 90% of the total seismic moment was released within these two main clusters which ruptured each within 8 s only while the total rupture lasted for about 56 s (Mendez and Anderson, 1991).
Fig. 7: Moment-rate (velocity source time) functions for the largest earthquakes in the 1960s and 1970s as obtained by Kikuchi and Fukao (1987). (Reproduced, with slight modification, from Kikuchi and Ishida, 1993, with permission of the Seismological Society of America).
Fig. 8: Snapshots of the development in space and time of the inferred rupture process of the Michoacán earthquake in 1985. The contours represent dip-slip velocity at 5 cm/s interval, the cross denotes the NEIC hypocenter. Three consecutively darker shadings are used to depict areas with dip slip velocities in the range: 12 to 22, 22 to 32, and greater than 32 cm/s, respectively. t - snapshot time after the origin time of the event, h - depth, D - distance in strike direction of the fault (redrawn from Mendez and Anderson, 1991).
This sequential rupturing of local asperities produces most of the high-frequency content of earthquakes. Accordingly, they contribute more to the cumulative seismic energy release than to the moment release. This is particularly important for engineering seismological assessments of expected earthquake effects. Damage to (the multitude of low-rise) structures is mainly due to frequencies > 2 Hz. They are grossly underestimated when analysing strong earthquakes only in teleseismic records on the basis of medium and long-period seismic records or when calculating model spectra assuming smooth rupturing along big faults of large earthquakes. This problem can be overcome only with detailed strong-motion networks in source areas of potentially large earthquakes and by complementary field investigations and related modelling of the detailed rupture process in case of clear surface expressions of the earthquake fault. But this is beyond the scope of seismological observatory practice. But observatory seismologists need to be aware of these problems and the related limitations of their simplified standard procedures. Nevertheless, they have a value on their own by allowing a rough first order analysis of the dominating types and orientation of earthquake faulting in a given region and their relationship to regional tectonics and stress field. The latter can also been inferred from other kind of data such as overcoring experiments, geodetic data or field geological evidence. Their comparison with independent seismological data, which are mainly controlled by conditions at larger depth, may provide a deeper insight into the nature of the fields observed.
The detailed understanding and quantification of the physical processes and geometry of seismic sources is one of the ultimate goals of seismology, be it in relation to tectonics, the improved assessment of seismic hazard or with the aim of identifying the type of source and discriminating between natural and man-made events. Earthquakes can be quantified with respect to various geometrical and physical parameters such as time and location of the (initial) rupture and orientation of the fault plane and slip, earthquake size with respect to fault length, rupture area, amount of slip, magnitude, seismic moment, stress drop and radiated energy, duration and time-history (complexity) of faulting, particle velocity and acceleration of fault motion etc. It is impossible, to represent this complexity in just a single number or a few parameters.
There are different approaches to tackle the problem. One aims at the detailed analysis of a given event, both in the near- and far-field, analysing waveforms and spectra of various kinds of seismic waves in a broad frequency range up to the static displacement field as well as looking into macroseismic data. Such a detailed and complex investigation requires a lot of time and efforts. It is affordable only for selected important events. The second simplified approach describes the seismic source only by a limited number of parameters such as the origin time and (initial rupture) location, magnitude, intensity or acceleration of observed/ measured groundshaking, sometime also the fault plane solution. They can more easily be obtained and have the advantage that rough but quick information can be given to the public and concerned authorities. Furthermore, this approach provides standardised mass data for comprehensive earthquake catalogues which are fundamental for other kinds of research such as earthquake statistics and seismic hazard assessment. But we need to be aware, that these simplified, often purely empirical parameters can not give a full description of the true nature and geometry, time history and energy release of a seismic source. In the following we will describe only the most common procedures in seismological (more or less routine) practice for single or more complex parameter determinations.
Date created: September 5, 2000
Last modified: September 14,
2000
Copyright © 2000, Global Seismological
Services
Maintained by: Eric Bergman bergman@seismo.com