Bruce A. Bolt

The SMART I Accelerograph Array (1980‐1987): A Review

SMART 1 is the first large digital array of strong-motion seismographs specially designed for engineering and seismological studies of the generation and near-field properties of earthquakes. Since the array began operation in September 1980, it has recorded over 3000 accelerogram traces from 48 earthquakes ranging in local magnitude (M L ) from 3.6 to 7.0. Peak ground accelerations have been recorded up to 0.33g and 0.34g on the horizontal and vertical components, respectively. Epicentral distances have ranged from 3 km 200 km from the array center, and focal depths have ranged from shallow to 100 km. The recorded earthquakes had both reverse and strike-slip focal mechanisms associated with the subduction zone and transform faults. These high quality, digital, ground motions provide a varied resource for earthquake engineering research. Earthquake engineering studies of the SMART 1 ground motion data have led to advances in knowledge in several cases: for example, on frequency-dependent incoherency of free-surface ground motions over short distances, on response of linear systems to multiple support excitations, on attenuation of peak ground-motion parameters and response spectra, on site torsion and phasing effects, and on the identification of wave types. Accelerograms from individual strong-motion seismographs do not, in general, provide such information. This review describes the SMART 1 array and the recorded earthquakes with special engineering applications. Also, it tabulates the unfiltered peak array accelerations, displays some of the recorded ground motion time histories, and summarizes the main engineering research that has made use of SMART 1 data.

The density distribution near the base of the mantle and near the earth's center

A detailed study of times and amplitudes of P waves in the range 95° < Δ < 120° from two intermediate focus earthquakes has provided new information on the physical state of the transition shell D″ just above the core. P waves of period near 2 s are clearly observed out to 118° (Uinta Basin Observatory). For 105° < Δ < 115° the mean wave slowness is 4.55 ± 0.05 s/°. This value and empirical PcP times then yield, relative to the 1968 P travel-times, a least squares estimate for a core radius of 3475 ± 2 km; from 13.63 km/s at the top of D″, the P velocity drops to an average of 13.33 ± 0.15 km/s towards the mantle-core boundary. Short-period P waves in the core shadow do not attenuate like pure diffracted waves; the notation Pc rather than dif PcP is suggested. The P and S velocity decrease in D″ entails an upper limit in density of about 5.9 g/cm3 at the mantle-core boundary. Reflections of steeply incident PKiKP short-period waves from the inner core boundary provide an estimate of the density contrast there. They also entail that there must be a jump in incompressibility and/or density (and not just in rigidity). Three measurements of the amplitude ratio PKiKP/PcP give a weighted mean value for the minimum density ratio across the inner core boundary of 0.87 ± 0.04. The maximum central density is then about 14.10 g/cm3. A recent, well-constrained, P velocity model KOR5 for the central core is now available as a basis for inferences on physical properties. Computation of many central core models compatible with KOR5 then demonstrates that the central density probably lies between 13.0 and 14.0 g/cm3. A value near the lower limit is more compatible with shock wave experiments on iron. A model CAL 1G for the central core with g90 = 13.0 g/cm3 and which satisfies all available constraints has k0 = 15.05 × 1012 dyne/cm2, μ0 = 1.25 × 1012 dyne/cm2. The corresponding seismic velocities at the center are α0 = 11.35 and β0 = 3.10 km/s, respectively. A formula is derived which gives the central density directly from the gradient of the empirical P and S velocity distributions in the inner core.

Joint statistical determination of fault-plane parameters☆

Abstract Joint treatment of first-motion data for sets of earthquakes has great advantages over determinations of individual solutions. An algorithm is used to find maximum-likelihood estimates of the orientations of the P and T axes and of the fault planes (and their variances), together with two precision parameters: ρ i , that weights each event according to its agreement with the group solution, and α j , that weights each station according to its performance. The method has been applied to two sets of earthquakes having different characteristics. The first is formed by 19 earthquakes in Bear Valley, CA, with magnitudes of ∼3 and recording distances less than 200 km, and the second by 29 microearthquakes with magnitudes of ∼1.5 and recorded at distances less than 60 km in the Pyrenees. The data from Bear Valley form a single homogeneous group having a mechanism related to motion on the San Andreas fault. The data from the Pyrenees are divided into five groups having different mechanisms.

Balance of Risks and Benefits in Preparation for Earthquakes

Widespread proposals to benefit from lessons of the 17 October 1989 (Loma Prieta) earthquake dramatize the difficulties associated with reducing seismic risk. There are three main problems. First, the understanding of earthquake generation is far from complete. For example, the unanticipated source style of this earthquake raises vital questions; claims of predicting its occurrence are weak, and, for practical reasons, the detailed pattern of damaging strong ground shaking was not predicted. Second, although their interactions are not well understood, competing social forces continue to prevent the optimum growth and application of knowledge for earthquake hazard mitigation. Third, the recent use of the probabilities of seismic risk has had mixed results. Because of indecision between minimizing loss of life and maximizing broader benefits, general agreement on acceptable earthquake risk remains confused.

Finite-element computation of seismic anomalies for bodies of arbitrary shape

A method which uses observed frequency spectral ratios of seismic plane waves for exploration of ore bodies is now available. The new method is based on the numerical solution of the response of a two‐dimensional shallow structural anomaly to an upward‐moving seismic wave from a distant earthquake or explosion. Finite‐element analysis is used for both P- and S-waves. Solutions to the direct problem for bodies of arbitrary shape have not previously been available. Results in the time and frequency domains are discussed here for a salt ridge and for a massive sulfide body. For the inverse problem, interpretation using contours of spectral ratios along a surface profile is suggested.

Observations of pseudo-aftershocks from underground nuclear explosions

Recent large underground nuclear detonations in Novaya Zemlya have provided crucial new evidence of the fine structure of the Earths upper mantle and outer core. Observations of the prescursors to P′P′ waves at Jamestown and Priest from explosions in 1966, 1970 and 1971 entail a sharp reflecting interface at 650 km depth. A spike doublet occurs, in agreement with predicted travel-times for a triparite core. Additional precursor waves also were detected at least 9 s before the first spike. These waves may be reflections from minor discontinuities in a 40 km thick transition shell below 650 km. The difference between the observed waves PmMP (m = 4 and 7), reflected m−1 times within the core, exceeds by 1.4 s the difference predicted by the 1968 PcP and PKP tables at Δ = 69°.

Short-term Exciting, Long-term Correcting Models for Earthquake Catalogs

Short-term Exciting, Long-term Correcting Models for Earthquake Catalogs Frederic Schoenberg Bruce Bolt Dept. of Statistics Dept. of Geology and Geophysics University of California, Los Angeles University of California, Berkeley Abstract A class of probability models for earthquake occurrences, called Short-term Exciting Long-term Cor- recting (SELC) models, is presented. This class encompasses features of different models presently used in hazard analysis to characterize earthquake catalogs, such as that of F. Omori. It offers the potential for a unified approach to the analysis and description of different types of earthquake catalogs. Maximum likelihood estimation methods for the seismicity model parameters and standard errors are presented. Sample SELC models are shown to provide satisfactory fit to a seven-year catalog of microearthquakes occurring in Parkfield, California and a longer seismicity sequence from the San Andreas fault zone in Central California. Inferences on seismicity patterns and mechanisms are discussed. Both significant clustering and strain release are detected. Introduction Two widely noted features of earthquake catalogs are the following: 1) Earthquakes tend to occur in clusters. This clustering is both spatial and temporal, and is some- times referred to with terms such as swarms , foreshock activity , and aftershock and Adamopoulos, 1973; Kagan and Knopoff, activity (Hawkes 1984; Ogata and Tanemura, 1984; Bullen and Bolt, Ogata and Katsura, 1988). 2) The fault ruptures that generate earthquakes decrease the amount of strain present at the locations along the fault where rupture occurs. This tectonic strain is thought to rebuild gradually over time, eventually achieving a critical level at which time another earthquake, or sequence of earthquakes, is generated (Reid, 1911; Ogata and Vere-Jones, 1984; Wang et al., 1991; Ogata, 1994). Several probability models reflect efforts at modeling this first type of behavior. Earthquake catalogs are modeled as realizations of triggering, branching or epidemic-type point processes, and all the referenced models have the feature that the instance of an earthquake at point (x,ti) the likelihood of an earthquake at point (y,?2) m in space and time increases space and time, where t\ < The likelihood of a particular realization may be given by the conditional rate of the point process; these models prescribe that the conditional rate of the earthquake process increases if more earthquakes have occurred. The amount which the conditional rate increases as a result of one previous earthquake is generally assumed to taper off as both time and distance from the previous earthquake increase. Numerous researchers have instead focused on the second facet of earthquake behavior listed above. In some cases small events and/or aftershocks are removed from earthquake catalogs under consideration (e.g.

Summary value smoothing of physical time series with unequal intervals

Abstract An algorithm SUMMOOTH for smoothing raw observations that are unequally spaced is explained. The process generalizes the method of summary values introduced into geophysics by H. Jeffreys. Smoothed data points, defined as the intersection of the local linear and parabolic least-squares fits, are computed for overlapping intervals rather than fixed sequential intervals as in previous work. A new feature is the parallel computation of (smoothed) summary gradients, essential for the Herglotz velocity integration. The interval selection is objective because the position of the smoothed values at each stage depends only on the spacing of the raw sample points. Selection of the starting interval can be made objective by addition of a principle, such as symmetry, or a rule, such as the fitted local curvature must never exceed a fixed value. In practice, selection should employ a trade-off curve between resolution and variance. An advantage of the process is that uncertainties at the summary points are independent. A comparison is given, for scarce data of Mossbauer spectra, with the smoothing method of Talmi and Gilat. Application to ragged time series for v p v s observations in earthquake prediction studies and to the construction of seismological travel-time curves illustrates the value of the method in geophysics.

Incomplete formulations of the regression of earthquake magnitude with surface fault rupture length

Recently it has been correctly pointed out that because linear regression of two variables is not symmetric, regression of log L on M cannot be used to estimate magnitude ( M ) from fault rupture length ( L ). It must be stated further, however, that least-squares regression treating L as error free is also incorrect. Indeed, to date no completely correct regression analysis of M or log L has been published. Full treatment requires geologists to assess uncertainties and to assign weights to field observations of L .

Spectral Comparison of Vertical and Horizontal Seismic Strong Ground Motions in Alluvial Basins

We analyze observations from the SMART2 array and the 1994 Northridge, California earthquake of spectral differences between vertical and horizontal strong seismic motions in alluvial basins. Our explanation is that the most energetic of such high-frequency vertical ground accelerations are generated by S-to-P seismic wave conversion within the transition zone between the underlying bedrock and the overlying sedimentary layers. The differences in combined scattering and anelastic attenuation for P and S waves lead to the observed spectral differences of the vertical motions between rock and deep alluvium sites. This model also accounts for the frequency content differences between the vertical and horizontal motions at sites in alluvial basins than at rock sites at similar distance ranges. The high-frequency cutoff of the acceleration power spectrum, f max , is a useful comparison parameter. The results help in computing matched sets of synthetic ground motions above 2 Hz at alluvial sites.