Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.
APPENDIX A RESEARCH DIRECTIONS What kinds of global seismological studies are foreseen over the next decade? What could be the effect of a readily available, global, high-quality digital-data base? What research needs may still not be met by an integrated global system? This appendix attempts to answer these questions and to develop a rationale, based on research potential, for an optimum global data-acquisition and distribution system. Data from the integrated global network should be of sufficiently high quality to permit refined resolution of both global and regional earth structures. Well-calibrated instruments would enable the use of amplitudes of long- period waves and oscillations and of short-period pulses. The amplitude data are valuable for the study of both source mechanisms and structures and offer the possibility of improving our knowledge of seismic-energy dissipation. Further improvement in the identification of normal modes will be made possible by the expanding network of broad- band stations. The short-period data could be important in reducing the reading error in travel-time studies and in the identification of depth phases. High-quality data of wide dynamic range could be essential in the identifi- cation of multiple events accompanying large earthquakes and could lead to a better understanding of the propaga- tion of rupture along a fault. Data from the global network would be used primarily in studies of earthquakes from far-field displacements and in investigations of earth structure. In general, these two objectives cannot be separated, since our ability to infer properties of a source from seismic recordings is limited by our knowledge of the mechanical properties of the medium (earth structure). In recent years, significant progress in the theoretical and numerical aspect of 65
66 earthquake-mechanism studies has greatly increased our qualitative understanding of the physical properties involved in naturally occurring faulting. Concurrent advances in the calculation of synthetic motion by a given source excitation embedded in realistic earth models pro- vide a powerful tool to a fuller understanding of the considerable complications of actual seismograms and to separate complexities of the source from those arising from the propagation of the disturbance in a heterogeneous earth. In other words, seismologists want to explain observed motions on a global scale in terms of the source after separating out those effects associated with propaga- tion. The solution of this basic problem is probably essential to an understanding of local strong motions, which, in turn, is needed in testing the dynamic response of critical structures (e.g., nuclear reactors) and in the general estimation of ground motion of engineering interest. Secondly, a better appreciation of earthquake characteristics such as depth, dimension, and stresses is essential to a clearer understanding of the conditions that exist in fault zones and thus is an aid in earthquake prediction. If we assume that the ground motion at the recording site following an earthquake can be derived from a seismic recording by removing the effect of instrumental response, then the motion is determined by two factors: the seismic source and the mechanical structure of the earth through which the seismic waves pass. Obviously, to make infer- ences about the source we must have a model of the medium (the earth). Thus, studies of source properties cannot be detached from measurements of propagation effects. The next stage would be the use of inversion models of the source and the medium, using modern methods. We may then generate synthetic data and compare them with observations. The differences will suggest inadequacies in our model of the medium (for example, those due to the effect of lateral heterogeneities) or of the source. Phase-equalization techniques can be used to provide initial information about the source mechanism, which, in turn, can be used to improve our knowledge of the medium. With the improved model of the medium we may repeat inversion for the source mechanism. What has been described is actually an iterative, self- improving process of learning. Source physics is one of the least understood areas of seismology, and by this process we hope to gain further insight. The advent of digital computers has significantly broadened our ability to pull useful information out of
67 the data base used in seismological research. The first observations of free oscillations of the earth following the Chilean earthquake of May 22, l960, are a good example. To take advantage of this computer capability it is necessary, however, to represent the time series in numer- ical form. In a vast majority of cases, conversion from analog data to digital data has been achieved by manual operation. This process is time-consuming and not very accurate, and the small dynamic range of analog record- ings represents a serious limitation on the magnitude span of events that can be analyzed in this way. Nevertheless, application of numerical methods to time-series analysis has led to significant progress in our knowledge of long- period characteristics of seismic sources, average prop- erties of the earth, and regional variations in surface- wave dispersion. The advantages of using digital data in body-wave studies have been well illustrated by the results obtained from seismic arrays such as the Large Aperature Seismic Array (LASA) and the Norwegian Seismic Array (NORSAR). To achieve our goal of inferring quantitatively the spatial and temporal distribution of stress release at depth requires detailed modeling of broadband observations, recorded on a global scale because of the complexity of the problem. Observations obtained from the WWSSN have been used almost exclusively in such analyses in recent studies. The existence of a worldwide broadband network of seismological instruments with digital recording capa- bilities would mean that many new and currently available analytical methods could yield more accurate results than have been possible thus far and could be applied on a routine basis to a great many events. This does not mean that we underestimate the significance of the data recorded by analog systems. Much of the excellent seismological research of the last decade testifies to the success of the WWSSN, and its continued operation is considered es- sential. The following is a discussion of opportunities to apply digital data from a global network to studies of source mechanisms and of mechanical properties of the earth. For wavelengths greater than about six times the largest source dimension, the source can be treated as a point in space. A detailed treatment of the concept of representing the properties of a seismic source in terms of the moment tensor and a description of the phase-equalization tech- niques can be found in J. F. Gilbert and A. Dziewonski (Phil. Trans. Roy Soc. London A278, l87-269, l975). The
68 source is characterized by the volume integral of the stress- release mechanism, the moment tensor, or the moment-rate tensor as a function of time. With MS > 1, ten global stations can be used to infer the moment-rate tensor by a linear inversion procedure. The moment-tensor approach is not limited to large earthquakes that generate free oscilla- tions. It can also be used in the traveling-wave formulation applied to analysis of any chosen portion of a seismogram. The relevant theory exists, and the numerical methods are rather straightforward. Let us define a seismic-moment-rate spectrum as one that is flat. In this case, the source-time function is a step function. There is considerable evidence from body waves, surface waves, and free oscillations that actual moment-rate spectra can exceed the seismic sprectra at either end of the frequency range. Some earthquakes are rich in high frequencies ("ultraseismic" events), and these need not be small earthquakes only. Others are rich in low frequencies ("infraseismic") and may approach being "aseismic"âthat is, they have so little high-frequency energy that they are un- observable on traditional seismographs. An extreme example of the latter is creep. Global data are essential to deter- mine whether any seismic activitiy exists at energy levels between those of creep and of infraseismic events, to ex- amine ultraseismic events more carefully, and to determine how earthquakes begin. The point source is the simplest possible representation of a seismic event, and it is a good approximation above a certain range of wavelengths, that is, at periods signifi- cantly greater than the rupture-propagation time. Also, approximation of the physical character of the earth by a radially symmetric average earth model is better at long periods. If we consider surface waves as an example, the great-circle-path phase velocities of Rayleigh waves differ by no more than l-2 percent in a period range of l50-300 sec. At a period of about 40 sec, the differences can be as great as l5 percent and the effect of regional variations cannot be discarded in an attempt to resolve source proper- ties. Studies of infraseismic events and surface waves re- quire the long-period instrumental response to be as high as possible. At wavelengths so short that the point-source approxi- mation is not valid, it is necessary to investigate the volume density of the dislocation mechanism. To do this, we must consider the moment-density tensor, which has dimensions of stress, as a function of time. Thus far, nearly all studies of the moment-density tensor have
69 depended on assumed characteristics of the source mechanism. It is almost a tradition to assume the form of the source- time behavior, constant rupture-propagation velocity, and fixed nodal planes. It is probably inevitable that some degree of a priori parameterization of the source-mechanism will be needed, and the advent of the global digital net- work will help in the refinement of such parameters as rupture-propagation velocity and source-time behavior. There is a need for more fundamental investigations of the nature of distributed sources. The solution of this basic problem, of explaining ob- served motions on a global scale in terms of the source and propagation effects, is probably essential to the under- standing of local strong motions and, hence, in the general estimation of ground motion of engineering interest. This is especially important to the USGS Earthquake Hazards Re- duction Program. Near-field measurements of earthquakes with sufficient strength to be recorded globally are es- sential to this approach. Teleseismic body-wave travel times, the periods of free oscillations, and the dispersion characteristics of very long-period surface waves have traditionally been interpreted in terms of a spherically symmetric model of the earth. Now, the differences among such models are quite small, except for the upper mantle. Actually, the concept of a spherically averaged upper mantle probably may be of limited geophysical interest, even though it is mathematically use- ful. More recently, considerable evidence has been gathered to support the idea that aspherical structures exist as deep as the core-mantle boundary. Nevertheless, a spherically symmetric earth model is a very good approximation, as such a model is capable of explaining some 99.9 percent of the values of observed normal modes. However, for a rotating, laterally heterogeneous earth, a normal mode of angular order, l, is split into 21+l spectral lines in the generally dis- tinct but only slightly different eigenfrequencies. The complex amplitude of each spectral line depends on the loca- tion of the source and receiver, the earth's heterogeneity, and the source mechanism. Such a line convolved with the resonance curve (determined by Q of the mode) represents a "singlet." What we observe in a spectrum of an individual seismogram is a "multiplet," the result of superposition of 2l + l singlets. Thus, the next step would be to resolve the fine structure of spectral peaks. It is a difficult problem, as a reasonable good starting model of a laterally heterogeneous earth is necessary so that phase-equalization procedures can be applied.
70 For a phase-equalization procedure to be successful, it is necessary to predict for each receiver the amplitude of the mode to be extracted, and this amplitude depends also on the earth's heterogeneity. Thus, global studies of lateral heterogeneities using surface waves and body waves are prerequisite to superresolutions (extraction of singlets from multiplets) of the normal-mode spectrum. The rewards, however, would be significant. If we could isolate singlets, we could measure the anelastic attenu- ation, separated from the effects of elastic scattering. The eigenfrequencies of singlets would represent a new class of data that could be used in construction of three- dimensional earth models. Despite some current difficulties caused by splitting, observations of free oscillations are the best tool for estimating source properties at very long periods. The distinction between surface waves, or even body waves, and normal modes is a matter of practice rather than theory. At sufficiently short wavelengths, the ray- theory approximation (in which the propagation effects are assumed to be associated with a particular ray path rather than with global properties of the earth) comes close enough to reality that assumption of its validity might lead to useful starting solutions. Much of what is assumed about the differences between upper-mantle structures under oceans and under continents comes from the "pure path" dis- persion studies. If reliable phase-velocity data should become available for "fractional" paths (R-[ or #2) , an inversion for the global three-dimensional structure of the upper mantle could be attempted. Knowledge of the lateral differences in phase velocities could be used to improve the precision with which source parameters are determined. At short periods (T < l00 sec), the effect of lateral variations in structure becomes very pronounced; also, the effect of multipathing can be significant, particularly in amplitude measurements. In recent years, numerous attempts have been made to measure "single-station" phase velocities, estimating the initial (source) phase shift from the fault-plane solution. It is possible to supple- ment this method by determining the source geometry from the long-period data, as previously discussed, and estab- lishing a library of "master events" with known phase- delay characteristics; these could then be used to deter- mine the source parameters of small-magnitude events that have not generated long-period surface waves of sufficient amplitude. Perhaps the iterative nature of the studies of
7l source mechanism and earth structure may find its most extensive application in studies of the dispersion charac- teristics and excitation of short- and intermediate-period surface waves. Long-period body waves can be used in the source- mechanism studies in much the same way as surface waves or free-oscillation data. It is possible to generate synthetic seismograms corresponding to each of the six elements of the moment tensor and then determine their components by in- version. The advantage of this procedure may be that in a given period rangeâsay, 30-50 secâthe body waves will show much less regional variation. Combining the normal- mode approach with one of several generalized ray methods may be optimal, as these two different approaches are most efficient in different frequency ranges. Body-wave data are important for accurate hypocenter location. Among the first-order source parameters, the least well determined is focal depth. More than three quarters of all earthquakes occur at depths of less than 70 km; for more than half of these shallow earthquakes, the depth can be determined with a minimum error of ±25 km. This large imprecision results mainly from the tradeoff between origin time and focal depth that occurs when the hypocenter is determined from teleseismic travel times. If the origin time were accurately fixed, the depth could be determined to approximately l sec of P-wave travel time at crustal velocities and vice versa. To fix the origin time, observations of S-P for distances less than about l0° of arc are required. Even with the most advantageous place- ment of, for example, a 30-station network, a great number of earthquakes would still lie beyond the required range. Hence, precise determination of focal depth will depend on the observation of near-source surface reflections, pP, sP, pPcP, sPcP, etc. Direct observation of the delay times of these phases can fix the depth to within several kilo- meters. Since the delay time for near-source reflections is less than about 20 sec for earthquakes occurring in the crust, the arrival time and even the presence of the phase itself will often be obscured by the waveform of the primary phase. Recent work on improvements of homomorphic decon- volution and cepstral analysis might be useful in routine estimation of focal depth under those circumstances. Other innovations include making use of maximum-entropy spectral analysis of the log spectrum and spatial and temporal averaging of spectra to enhance depth-phase peaks. Low- pass filtering of short-period seismograms before analysis should allow depth-phase detection where the roughest topography exists.
72 Another important area of investigation is that of coda waves; these waves include information on both the earth- quake process and earth structure. Separation and resolu- tion of these two effects require the availability of a vast amount of data for different sources and stations. Finally, there are studies that require strategic instru- mental coverage and therefore depend on the favorable occurrence of earthquakes from certain sources. Examples include studies of aftershocks; upper-mantle, core-mantle, and inner-core structures; absorption; diffracted waves; and geothermal research. Some operational aspects of the global seismograph net- work are relevant to research opportunities described above. If WWSSN data are used to infer the moment rate tensor by linear inversion, the errors are about l0 per- cent or worse. A global digital network should permit the inference of source mechanisms for smaller value of m, b as the number of stations increases and as the signal-to- noise ratio (S/N) improves. For WWSSN instruments, the S/N is rarely more than 20 dB. Digitization noise, caused by the finite and sometimes large width of the signal trace, and signal-generated noise, caused by as yet unknown in- strumental nonlinearities, are mainly responsible. The newer digital instruments appear to have a S/N of over l00 dB for events with MS > 7, but there remains considerable work to be done before their calibrations are well under- stood. The average noise level of a single instrument should not be necessarily taken as a cutoff level. Availability of a large network of instruments and application of phase- equalization techniques can result in a significant im- provement of the network S/N. Studies of infraseismic events also require the long-period response to be as high as possible. Perhaps the word "global" with respect to seismic instrumentation should be understood not only in the sense of geographical distribution but also in the sense of the frequency band of measurements. We have mentioned earlier the "infraseismic" events, those that are characterized by anomalously high spectral values of the moment rate tensor at very long periods. This is related to the difficult but important problem in geodynamics of whether plate slip is continuous or discontinuous. Although it is likely that the currently deployed, or being deployed, net- work of instruments may be material in the identification of infraseismic events, there is an urgent need for con- struction and worldwide application of instruments with
73 the necessary zero-frequency response. Measurements of this type are currently planned by NASA, but it seems likely that they should be supplemented by a network of high-class strainmeters that would provide continuous monitoring of the state of stress in the crust. It is important that these instruments be standardized, to make possible correlation of the data from various sites.
