Interpretation of Near-Source Ground Motion and Implications

• Main Author: Liu, Hsui-Lin
• Format: Others
• Language: en
• Published: 1983
• Online Access: https://thesis.library.caltech.edu/3374/1/Liu_hl_1983.pdf; Liu, Hsui-Lin (1983) Interpretation of Near-Source Ground Motion and Implications. Dissertation (Ph.D.), California Institute of Technology. doi:10.7907/5hbj-fd95. https://resolver.caltech.edu/CaltechETD:etd-09072006-111327 <https://resolver.caltech.edu/CaltechETD:etd-09072006-111327>

<p>This thesis presents some deterministic modeling and interpretation of various aspects of observed near-source ground motions.</p> <p>In Chapter 1, finite source parameters determined from waveform modeling studies are presented for two California earthquakes; the 1979 Coyote Lake event and the 1966 Parkfield event. These events were recorded by strong motion arrays with similar station to fault rupture geometries. Thus it is possible to demonstrate that differences in the ground motions recorded within 30 km of the epicenter are indeed due to the differences in rupture fault length and dislocation distribution.</p> <p>Details of the waveform modeling for the August 6, 1979 Coyote Lake earthquake are described in part 1-A. A finite fault striking N24°W and extending to a depth of 10 km is proposed to model the strong-ground motion data. The source model suggests that right-lateral faulting initiated at a depth of 8 km and ruptured towards the south with a velocity of 2.8 km/sec. This unilateral rupture can explain the large displacements recorded south and southwest of the epicenter. However, the waveform coherency observed across an array south and southwest of the epicenter suggests that the rupture length is less than 6 km. The maximum dislocation is about 120 cm in a small area near the hypocenter and the total moment is estimated to be 3.5 x 10²⁴ dyne-cm. An abrupt stopping phase, which corresponds to a cessation of right-lateral motion, can explain the high peak acceleration recorded at array station 6. The stress drop in the hypocentral area is about 140 bars; although the average stress drop over the entire rupture surface is 30 bars. This preferred finite source model can predict observed P<i>_ni</i> waveforms as well as the beginning features of teleseismic body waves.</p> <p>In part 1-B, a similar waveform modeling technique is used to interpret the ground motions recorded during the June 28, 1966 Parkfield earthquake. The preferred model suggests that the earthquake involved two fault segments; one is the NE branch which extends 22 km southward from epicenter and has an average slip of 45 cm, another is the SW branch which ruptured less than 10 km and has an average slip of about 22 cm. The total moment indicated by this model is 1.25 x 10²⁵ dyne-cm. The anomalous large amplitude ground displacement seen at station Cholame No. 2 is modeled as a local amplification effect rather than a source effect due to significant dislocation near this station.</p> <p>Direct waveform comparisons between recordings of the Parkfield event and the Coyote Lake event also support the conclusion that the rupture length of the Coyote Lake earthquake is much shorter than that of the Parkfield event. The waveform modeling also emphasizes the importance of using array data to constrain source parameters. The solution derived from a single station's recording, which in many cases is the only available information, may often produce misleading results.</p> <p>In Chapter 2, high-frequency ground motions (ground velocity and acceleration) recorded at less than 30 km epicentral distances are studied for two aftershocks of the 1979 Imperial Valley, California earthquake. In the past, little has been done to understand these high frequency waves through a deterministic modeling approach. The waveform modeling technique and the source mechanisms of these two aftershocks are described in sections 2-A and 2-B.</p> <p>An important feature of the ground motions recorded during the October 15, 1979 Imperial Valley earthquake sequence is the strong high frequency waves observed on the vertical components. This feature is also seen in recordings of the aftershock of October 16, 23:16, 1979, which is described in section 2-A. This polarization feature is easily explained by the basin velocity structure which bends rays towards the vertical at the free surface. Short S-P times are observed at the three closest stations (epicentral distances of 3 km to 5 km) suggesting that this aftershock occurred at a very shallow depth of about 2 km. A fault plane orientation (strike = N20°E, dip = 30°SE, and rake = -80°) obtained from a first P-motion study, generates synthetic waveforms of the strong ground velocities which are similar to those observed at three closest stations. The source time duration is determined to be 1.0 second and the moment is 1.6x10²³ dyne-cm. Synthetics for a number of line source models are compared with the observations. These comparisons lead to two basic mechanisms that are necessary to explain the frequency content of the observed P- and S- waves. One is that the source process is characterized by irregular rupture. It is postulated that the heterogeneous stiffness in the layered medium is the basic cause of the irregular rupture. Heterogeneous rupture generates both high-frequency P- and S-waves. In order to explain the contrast in observed frequency content it is also necessary that there is a mechanism for attenuating S-waves much stronger than P-waves.</p> <p>The aftershock that occurred about 3 minutes after the mainshock, at 23:19 October 15, 1979 is presented in section 2-B. This aftershock was located on the Imperial fault near Highway 8 and close to the zone of high frequency energy release of the main event. The impulsive seismograms for 16 array stations, ranging from 8 km to 26 km in epicentral distance, are well suited for source parameter inversion studies to obtain an optimal solution for ground velocity and acceleration. The earthquake source is approximated by a model consisting of several point dislocation sources separated in space and time and having different dislocation orientations and moments. These source parameters were deduced by trial and error modeling as well as by applying inversion procedures. The waveforms and amplitudes of horizontal ground velocities are well modeled by two predominantly strike-slip point sources; the first source (strike = N41°W, dip = 42°NE and rake = 174°) has a moment of 0.7 x 10²⁴ dyne-cm, the second source (strike = N36°W, dip = 82°SW and rake = 181°) lies about 1 km to the north of the first and has a seismic moment of about twice that of the first source. It is suggested that the higher-frequency ground motions, such as accelerations, can be derived from very irregular source processes, whereas the longer-period ground motions, such as ground displacements, can be well modeled by simpler planar source.</p> <p>A Futterman attenuation operator with a t^*_β of about 0.08 to 0.1 and a t^*_α of about 0.001 in the sedimentary region produces longer period S waves and the proper amplitude ratio between P and S waves.</p> <p>In Chapter 3, the ground motion data from the 1971 San Fernando earthquake recorded at epicentral distances of less than 100 km are presented. Three long profiles (&gt; 50 km ) and three short profiles (&lt; 2 km) of ground velocity and acceleration, displayed as a function of epicentral distance are analyzed.</p> <p>Although there is considerable variation in waveforms and peak amplitudes observed along the long profiles, there are also many examples of coherent phases seen at adjacent stations. Ground velocity profiles show striking differences in amplitude and duration between stations located on hard rock sites and stations located within the sedimentary basins. The San Fernando basin, in which the source is located, seems to respond quite differently from the Los Angeles basin which is about 30 km from the earthquake source area. Ground acceleration profiles show that there is little change in the duration of high-frequency shaking along the long profiles.</p> <p>The three short profiles, which are all located within the Los Angeles basin, demonstrate that ground velocity waveforms are nearly identical along these profiles. Although greater variation of waveforms and amplitudes are seen for ground acceleration along these short profiles, strong phase coherence is still observed.</p> <p>The 2D acoustical finite difference method is used to compute the effects on SH-waves of irregular velocity structures believed to exist along Profile I and Profile II. Profile I extends 65 km southward from the epicenter across the San Fernando and Los Angeles basins to a station on the Palos Verdes Peninsula. Profile II extends 95 km S 40° E along the front of the San Gabriel mountains and across the San Gabriel and Los Angeles basins. These numerical models consist of low-velocity sedimentary basins (β = 2.1 km/sec) of irregular shape which are imbedded in high-velocity basement rock (β = 3.5 km/sec). Heaton's (1982) finite source model derived from modeling the five nearest stations for the San Fernando event, is also incorporated in the interpretation. The resulting simulation suggests that the smaller S! phases in both Profile I and Profile II are direct S waves from the deep source region (13 km). The shallow source region (at 1 km) dominates high amplitude later arrived phases observed along Profile I and are due to the complicated basin path along this profile. The shallower source region, however, contributes little to the ground motions along Profile II due to the lack of thick sediments near the source region along this azimuth.</p>