Acta Geophysica vol. 62, no. 6, Dec. 2014, pp. 1214-1245 DOI: 10.2478/s11600-013-0199-9 Study of Combined Effects of Sediment Rheology and Basement Focusing in an Unbounded Viscoelastic Medium Using P-SV-Wave Finite-Difference Modelling Jay P. NARAYAN and Vinay KUMAR Department of Earthquake Engineering, Indian Institute of Technology Roorkee, India e-mail: [email protected] (corresponding author) Abstract This paper consists of two parts. First, a fourth-order-accurate staggered-grid finite-difference (FD) program for simulation of P-SV-wave in viscoelastic medium is presented. The incorporated realistic damping is based on GMB-EK-model. The accuracy of program is validated by comparing computed phase-velocity and quality-factors with same based on GMB-EK-model and Futterman’s relations. The second part of paper presents the combined effects of sediment damping and synclinal basement focusing (SBT) on ground motion. The results reveal SBT focusing, mode conversion and diffraction of incident waves. The response of elastic SBT model reveals an increase of spectral amplification with increasing frequency. The viscoelastic response of SBT model reveals that a particular frequency may get largest amplification for a particular set of values for damping, focal-length and distance from tip of the SBT. This frequency-dependent amplification may explain mysterious damage reported in some past earthquakes if predominantly amplified frequency matches natural frequency of damaged structures. Key words: viscoelastodynamic P-SV-wave equations, finite difference program, fourth order spatial accuracy, basement topography effects, local site effects. © 2014 Institute of Geophysics, Polish Academy of Sciences Unauthenticated Download Date | 6/18/17 11:08 PM BASEMENT FOCUSING EFFECTS ON GROUND MOTION 1. 1215 INTRODUCTION A frequency-dependent damping in time-domain simulations of basin effects on the ground motion characteristics is essential to explain the very peculiar damage patterns and seismic phase generations in basins (Gao et al. 1996, Hartzell et al. 1997, Narayan 2005, 2012). The fourth-order-accurate staggered-grid finite-difference (FD) method, proposed by Madariaga (1976), is one of the most useful numerical methods to simulate the ground motion characteristics (Levander 1988, Moczo and Bard 2002, Narayan and Kumar 2008). The explicit staggered-grid FD has been used successfully to simulate the scattering and diffraction phenomenon at the basin edges, underground ridges and surface topography as well as the role of these phenomena in surface wave generation (Virieux 1986, Moczo and Bard 1993, Narayan and Prasad Rao 2003, Narayan 2003, Narayan and Ram 2006, Narayan and Richhariya 2008, Frehner et al. 2008, Narayan and Kumar 2010). In the past, simple approaches have been used to incorporate the damping in timedomain FD simulations due to non-availability of required computational memory and time (Narayan and Kumar 2008). Day and Minster (1984) attempted to incorporate the realistic viscoelastic damping into a 2D timedomain FD program for the first time based on the Padé approximation. Emmerich and Korn (1987) improved the incorporation of viscoelastic damping in time-domain simulations using a rheological model widely known as generalized Maxwell body – Emmerich and Korn (GMB-EK) model, in which n-Maxwell bodies and one Hooke element are connected in parallel (Moczo et al. 1997, Galis et al. 2008). A further improvement was made by Kristek and Moczo (2003) by introducing a material independent anelastic function, which is preferable in case of material discontinuities in the FD grid. Saenger et al. (2005) studied the coupling mechanism of fluidsolid interaction using viscoelastic damping based on the GMB-EK model in a rotated staggered FD grid. As a result, all the viscous parameters were located at the centre of a grid cell, which allowed saturating the rock models with realistic approximations of Newtonian fluids. On the other hand, Carcione et al. (1988) used the rheology of a generalized Zener body (GZB model), in which n-Zener bodies are connected in parallel. Sometimes, very peculiar damage patterns observed in a basin during an earthquake cannot be explained using simple soil amplification and resonance effects, as was for example reported in the Santa Monica area, Los Angeles basin, during the Northridge earthquake of 1994 (Gao et al. 1996, Hartzell et al. 1997, Alex and Olsen 1998, Davis et al. 2000). Based on the analysis of aftershock records, Gao et al. (1996) suggested that the amplification in Santa Monica was due to basement focusing from a lens shaped structure of the deep Los Angeles basin sediments. Alex and Olsen (1998) estimated the contribution of the proposed “Lens Effects” on the amplificaUnauthenticated Download Date | 6/18/17 11:08 PM 1216 J.P. NARAYAN and V. KUMAR tion in Santa Monica. They simulated SH-wave and P-SV-wave responses in the Santa Monica region and inferred that deep basement focusing can only account for around 50% amplification of ground motion observed in Santa Monica. Davis et al. (2000) carried out inversion of aftershock records using SH-wave FD simulations and inferred that the damage in Santa Monica occurred due to focusing caused by the presence of several underground acoustic lenses at depths of around 3 km in the Los Angeles basin. They also reported frequency-dependent amplification of ground motion due to basement focusing based on an analytical formulation. Basement focusing effects on the ground motion characteristics were also observed in the form of consistent anomalous damage to only unreinforced brick chimneys, as shown in Fig. 1, in west Seattle, Washington, during the 1949 Olympia earthquake of magnitude 7.1, the 1965 Tacoma earthquake of magnitude 6.5, and the 2001 Nisqually earthquake of magnitude 6.8 (Ihnen and Hadley 1986, Booth et al. 2004, Stephenson et al. 2006). Booth et al. (2004) carried out a detailed damage survey after the Nisqually earthquake of 2001 and reported severe damage to collapses of chimneys only in west Seattle. They also inferred that the damage pattern does not correspond to epicentral distance, ground motion amplification caused by soft soil deposit or surface topography. Based on SH-wave modelling, Stephenson et al. (2006) inferred that basement focusing of direct waves played a major role in the damage of chimneys in west Seattle during the Nisqually earthquake. Frankel et al. (2009) carried out 3D simulations for the Nisqually earthquake and inferred the focusing of S-waves in west Seattle that experienced increased chimney damage. The question remains why only unreinforced chimneys were damaged selectively. Such a selective damage cannot be explained by basement focusing alone, but an additional frequencydependent amplification is required. In order to answer the above question, a parametric study is carried out considering the seismic responses of simple synclinal basement topography (SBT) models with a varying rheology of the overlying sediments. The first aspect of this paper is to write an efficient second-order-accurate-in-time and fourth-order-accurate-in-space (2, 4) explicit staggered-grid finite-difference program based on the approximation of velocity-stress viscoelastic P-SV-wave equations for heterogeneous media. The validation of an implementation of frequency-dependent damping in time-domain simulations is done by comparing the numerically computed frequency-dependent quality factor and phase velocity of both P- and SV-waves with the same computed using the GMB-EK rheological model and the Futterman’s relation (Futterman 1962). The numerical grid dispersion and stability condition are also analyzed in details. The second aspect of this paper is to study the effects of basement focusing, focal length and sediment damping on ground motion Unauthenticated Download Date | 6/18/17 11:08 PM BASEMENT FOCUSING EFFECTS ON GROUND MOTION 1217 Fig. 1. Collapsed chimney during the Nisqually earthquake 2001, west Seattle, Washington (after Booth et al. 2004). characteristics. Seismic responses of an unbounded SBT model with different focal lengths on a vertical array along the focal length are simulated for different sediment damping to infer whether there is a particular frequency which gets maximum amplification for a particular set of values for sediment damping, SBT focal length and SBT depth. Snapshots in a rectangular area are also computed to identify the different seismic phases developed at the base of the SBT. 2. VISCOELASTIC P-SV-WAVE FINITE-DIFFERENCE PROGRAM In this subsection, salient features of the developed viscoelastic second-order accurate in time and fourth-order accurate is space P-SV-wave explicit staggered-grid finite-difference program based on the GMB-EK rheological model is described in brief (Emmerich and Korn 1987, Kristek and Moczo 2003). The heterogeneous viscoelastodynamic P-SV-wave equations based on the GMB-EK rheological model and using material-independent anelastic functions (Kristek and Moczo 2003) are given below ∂σ xx ⎛ ∂U = Ku ⎜ ∂t ⎝ ∂x ρ ∂U ∂σ xx ∂σ xz = + , ∂t ∂x ∂z (1) ρ ∂W ∂σ xz ∂σ zz = + , ∂t ∂x ∂z (2) ⎞ ⎛ ∂W ⎟ + λu ⎜ ∂Z ⎠ ⎝ ⎞ ⎟− ⎠ m ∑ ⎡⎣Y α ( χ ) + Y λ ( χ )⎤⎦ , l xx l l zz l l = 1, 2,..., m l =1 Unauthenticated Download Date | 6/18/17 11:08 PM (3) 1218 J.P. NARAYAN and V. KUMAR ∂σ zz ⎛ ∂W = Ku ⎜ ∂t ⎝ ∂z ⎞ ⎛ ∂U ⎟ + λu ⎜ ∂x ⎠ ⎝ ∂σ xz ⎛ ∂U = μu ⎜ ∂t ⎝ ∂z ⎞ ⎟− ⎠ ⎞ ⎛ ∂W ⎟ + ⎜ ∂x ⎠ ⎝ m ∑ ⎡⎣Y α ( χ ) + Y λ ( χ )⎤⎦ , zz l l l xx l l = 1, 2,..., m (4) l =1 ⎞ ⎟− ⎠ m ∑ ⎡⎣Y μ ( χ )⎤⎦ , l xz l l = 1, 2,..., m (5) l =1 where U and W are the particle velocity components in the x- and zdirections, respectively. σxx , σzz , and σxz are the stress components, and ρ is the density. Ku , λu , and μu are the modified unrelaxed elastic parameters and Yl α , Yl λ , and Yl μ are the modified anelastic coefficients. χlxx , χlzz , and χlxz are the anelastic functions and ∂/∂x, ∂/∂z, and ∂/∂t are the differential operators. Parameter m is the number of relaxation frequencies. Figure 2 shows the staggering technique, where normal stress components σxx and σzz , unrelaxed elastic parameters Ku and λu , anelastic coefficients Ylα and Ylλ and Fig. 2. FD staggering technique for P-SV-wave modeling with fourth order spatial accuracy, where dark circle denotes the defined position of normal stress components σxx and σzz, unrelaxed elasti parameters Ku and λu, anelastic coefficients Ylα and Ylλ , and anelastic functions χlxx and χlzz at the nodes. Dark triangle and rectangle denotes the position of density and particle velocity components U and W in the x- and z-directions, respectively. Open circle denotes the position of shear stress σxz component, unrelaxed modulus of rigidity μu, anelastic coefficient Yl β , and anelastic function χlxz . Unauthenticated Download Date | 6/18/17 11:08 PM BASEMENT FOCUSING EFFECTS ON GROUND MOTION 1219 anelastic functions χlxx and χlzz are defined at the nodes. The particle velocity components U and W in the x-direction and z-direction and density ρ are defined midway between two adjacent nodes and the shear stress σxz , unrelaxed modulus of rigidity μu, anelastic coefficient Yl μ , and anelastic function χlxz are defined at the centre of the grid cell. In the present FD formulation, material-independent anelastic functions (χl) have been computed at four relaxation frequencies (m = 4) using the following equation: ( χl )n + 1 2 = 1 2 − Δtωl 2Δtωl ε, ( χl )n − 2 + 2 − Δtωl 2 + Δtωl l = 1, 2,..., m , (6) where superscript n denotes the time index, ε is the strain , and Δt is the time increment. The details of computation of the modified elastic parameters ( Ku , μu , ( ) and λu ) and the anelastic parameters Yl α , Yl μ , and Yl λ are given by Kristek and Moczo (2003). The unrelaxed elastic parameters Ku and μu for P- and S-waves have been obtained using the phase velocity of P- (VP ,ωr ) and S-waves (VS ,ωr ), respectively, at a reference frequency (ω) and the following equations (Moczo et al. 1997). μu = ρVS2, wr Ku = ρVP2, wr ϑ1 = 1 − m ⎡ 1 ∑ ⎢⎢Y 1 + (ω l l =1 ⎣ r R + ϑ1 2R 2 R + ϑ1 2R2 where R = ϑ12 + ϑ22 , (7) where R = ϑ12 + ϑ22 , (8) ⎤ ⎥ , ϑ2 = 2 ωl ) ⎥⎦ m ⎡ l l =1 ⎣ ⎤ ⎥ , l = 1, 2,..., m . (9) 2 ωl ) ⎥⎦ ωr ωl ∑ ⎢⎢Y 1 + (ω r The effective value of the unrelaxed modulus of rigidity, μu, and density at the desired location are obtained using harmonic and arithmetic means, respectively (Moczo et al. 2000). The anelastic coefficients for P- and S-waves have been computed using Futterman’s equation (Futterman 1962) and a least square optimization technique. In Equations 1-5, the time derivative was replaced by a second-order accurate central FD operator and space derivatives were replaced by a fourthorder staggered grid FD operator (Levander 1988, Narayan and Kumar 2008). Both the sponge boundary condition of Israeli and Orszag (1981) and Unauthenticated Download Date | 6/18/17 11:08 PM 1220 J.P. NARAYAN and V. KUMAR absorbing boundary condition of Clayton and Engquist (1977) were implemented on all the four model edges to avoid edge reflections (Kumar and Narayan 2008). 3. VALIDATION OF PROGRAM To validate the accuracy of the developed explicit FD program for P-SVwave as well as the implementation of viscoelastic damping in time-domain simulations, seismic responses of unbounded homogeneous elastic and viscoelastic models were simulated using plane horizontal wave fronts of SVwave as well as P-wave separately. The simulated seismic responses were compared with the analytical solutions. 3.1 Numerical grid dispersion and stability To study the numerical grid dispersion, seismic responses of an unbounded homogeneous elastic half-space model with VS = 3200 m/s, VP = 5542.5 m/s, and density equals to 2800 kg/m3 were computed. A Ricker wavelet with 4.0 Hz dominant frequency with considerable spectral amplitude in a frequency range between 0.25 and 10.0 Hz was used as an excitation function for generation of a horizontal plane wave front using five point sources per dominant wavelength. The model was discretised using a square grid of 60 m side length. The computed seismic responses at two locations (60 m apart) on a vertical array at a distances of 1200 and 1260 m from the linear source were used for the numerical computation of phase velocity Vgrid(ω). First, Fourier transforms of the seismic responses at the two locations were used to determine the spectral phase difference Δϕ(ω), then the phase velocity was computed using the following equation: V grid (ω ) = ωΔH , Δϕ (ω ) (10) where ∆H is the distance between the two considered receiver points. The numerical grid dispersion curves for S- and P-waves are plotted as the normalized phase velocity (Vgrid/V) and the number of grid points per wavelength (λ/Δz). Parameter V is the S-/P-wave velocity assigned to each grid point. Analytical grid dispersion curves for S- and P-waves were computed using the same model parameters and the methodology given by Moczo et al. (2000) for a fourth-order spatial accuracy. Figure 3 shows a comparison of the numerical grid dispersion curves with the analytical one for S- and P-waves, respectively. A good agreement of numerical dispersion curves with the analytical dispersion curves reveals that the error in the numerically computed grid dispersion curves are within the permissible limit when the number of grid points per shortest wavelength is more than six. Unauthenticated Download Date | 6/18/17 11:08 PM BASEMENT FOCUSING EFFECTS ON GROUND MOTION 1221 Fig. 3. A comparison of numerical and analytical grid dispersion curves for S- and P-waves with fourth order spatial accuracy. Special attention may be required to make the explicit time stepping scheme stable in case of simulating a viscoelastic model where velocity is frequency dependent, particularly when the considered frequency bandwidth is very large (Saenger et al. 2005). In case of the here developed code, the computed unrelaxed moduli based on Eqs. 7-9 and anelastic coefficients are used to find out the P-wave velocity for the purpose of finding the stability condition. The P-wave velocity computed using unrelaxed moduli gives the largest velocity for the considered model parameters and frequency bandwidth. This largest P-wave velocity can be conservatively used for computing the stability condition. The stability condition for the presented explicit FD program was obtained based on various iterative numerical experiments for a viscoelastic model i.e. using different time increments Δt for the simulation and then analyzing whether an error is arising (Δt was varied in a very small steps near the upper limit of the stability condition). It was finally concluded that the scheme is stable for both the homogeneous and the heterogeneous viscoelastic models if the following stability condition is locally satisfied (Moczo et al. 2000). VP Δt ≤ 0.7 , min ( Δx, Δz ) (11) where VP is the P-wave velocity computed using unrelaxed moduli, Δx and Δz are the grid size in x- and z-directions, respectively at a particular node. 3.2 Phase velocity and quality factor computation The accuracy of the presented FD program is also validated by comparing the numerically computed phase velocities VP(ω) and VS(ω) and quality factors QP(ω) and QS(ω) with the same obtained using Futterman’s relations and the GMB-EK model for an unbounded homogeneous viscoelastic medium. The responses of an unbounded model were computed at two locations on a vertical array for two rheologies of the medium (HVM1 and HVM2). The P- and S-wave velocities and quality factors at reference frequency (FR = 1 Hz), density and the computed unrelaxed moduli for HVM1 and Unauthenticated Download Date | 6/18/17 11:08 PM 1222 J.P. NARAYAN and V. KUMAR T ab l e 1 Rheological parameters for unbounded homogeneous viscoelastic models with different damping (HVM1-HVM2) Rheological models HVM1 HVM2 Velocity at FR [m/s] S-wave P-wave 3200 5542 3200 5542 Quality factor at FR S-wave 32 64 P-wave 55 110 Density [kg/m3] 2800 2800 Unrelaxed moduli [GPa] MU KU λU 31.455 90.732 27.821 30.021 88.323 28.280 HVM2 rheological models are given in Table 1. The reference frequency is chosen randomly for this theoretical study. If simulations had to be done for a specific site then the frequency at which quality factors and phase velocities are measured in the field has to be used as a reference frequency for the computation of anelastic coefficients. Four relaxation frequencies were taken as 0.02, 0.2, 2.0, and 20 Hz. To compute the phase velocity and quality factor, horizontal plane P- and SV-waves fronts were generated by applying only normal stress (σzz) and shear stress (σxz), respectively at the position of different point sources (Note: five point sources pre dominant wavelength was used to generate a line source). The Fourier transform of the seismic responses at two locations (∆H = 30 m apart) on a vertical array were used for the numerical computation of phase velocity using Eq. 10. The frequency-dependent quality factors were numerically computed using the phase velocity and the amplitude spectra of particle velocity V(H, ω) of the two traces and the following relationships: 1 QS (ω ) 1 QP (ω ) =− =− 2VS (ω ) ⎡ln V ( H + ΔH , ω ) − ln V ( H , ω ) ⎤ , ⎦ ωΔH ⎣ 2VP (ω ) ⎡ln V ( H + ΔH , ω ) − ln V ( H , ω ) ⎤ , ⎦ ωΔH ⎣ (12) (13) |V(H, ω)| is the mod of Fourier amplitude spectra of particle velocity at a distance H from the line source. The phase velocities for P- and S-waves were also computed analytically using the GMB-EK model (Moczo et al. 1997). First, relaxed moduli at the desired frequencies were computed using unrelaxed moduli given in Table 1 with the help of following equations and then relaxed moduli were used to compute phase velocity ⎧⎪ μ (ω ) = μu ⎨1 − ⎩⎪ m ∑ l =1 Yl μ ωl2 ⎫⎪ ⎬, ωl2 + ω 2 ⎭⎪ Unauthenticated Download Date | 6/18/17 11:08 PM (14) BASEMENT FOCUSING EFFECTS ON GROUND MOTION ⎧⎪ K (ω ) = Ku ⎨1 − ⎩⎪ m ∑ l =1 Ylα ωl2 ⎫⎪ ⎬, ωl2 + ω 2 ⎭⎪ 1223 (15) where μ(ω) and K(ω) are the relaxed moduli for S- and P-waves, respectively. The frequency dependent quality factors for P- and S-waves were Fig. 4. Comparison of numerically computed phase velocity and quality factor for S-wave in an unbounded homogeneous viscoelastic medium with the same computed analytically using Futterman’s relations and the GMB-EK rheological model. Fig. 5. Comparison of numerically computed phase velocity and quality factor for P-wave in an unbounded homogeneous viscoelastic medium with the same computed analytically using Futterman’s relations and the GMB-EK rheological model. Unauthenticated Download Date | 6/18/17 11:08 PM 1224 J.P. NARAYAN and V. KUMAR computed analytically using GMB-EK model with the help of following equations: 1 QS (ω ) 1 QP (ω ) m = l l =1 m = ωωl 2 2 l +ω ⎡ ⎢1 − ⎢⎣ ωωl ωl2 + ω 2 ⎡ ⎢1 − ⎣⎢ ∑Y β ω ∑ l =1 Ylα m ωl2 ⎤ ⎥, 2 2 l +ω ⎥ ⎦ (16) ωl2 ⎤ ⎥. + ω 2 ⎦⎥ (17) ∑Y β ω l l =1 m ∑Y α ω l l =1 2 l The phase velocities and frequency dependent quality factors for P- and S-waves were also computed using Futterman’s relationship (1962). Figure 4 shows a comparison of numerically computed phase velocity VS(ω) and quality factors QS(ω) with the same computed analytically using GMB-EK model and the Futterman’s relationship. Analysis of Fig. 4 reveals an excellent agreement between the numerically computed phase velocity and quality factor for S-waves and the same computed using Futterman’s relation and the GMB-EK model. Similarly, Fig. 5 shows a comparison of numerically and analytically computed P-wave phase velocity and quality factor. Analysis of Fig. 5 also reveals an excellent agreement between the numerically computed P-wave phase velocity and quality factor and the same computed using Futterman’s relation and the GMB-EK model. Figures 4 and 5 validate the accuracy of implementation of realistic viscoelastic damping in the developed time-domain P-SV-wave FD program. 3.3 Response of bounded heterogeneous viscoelastic model To further validate the accuracy of implementation of realistic viscoelastic damping in a heterogeneous medium, seismic response of a bounded viscoelastic model containing a horizontal soil layer over the half-space was simulated at the free surface using S-wave horizontal line source. Rheology of soil was changed by taking the different values of the quality factor at a reference frequency and at the same time maintaining the other model parameters constant. The velocities and quality factors for P- and SV-waves at reference frequency (Fr), soil thickness, density, and unrelaxed moduli for the soil layer corresponding to different considered models (SRM1-SRM3) are given in Table 2. The velocities and quality factors for P- and SV-waves at reference frequency (Fr), density and unrelaxed moduli for basement were taken as that of the HVM1 model. Figure 6a depicts the simulated seismic responses of half-space (without soil layer) and various heterogeneous viscoelastic models (SRM1-SRM3). The spectral amplifications were computed by taking the ratio of spectra of seismic responses with and without soil layer. Analysis of Fig. 6b reveals a decrease of spectral amplification with an increase of sediment damping. This figure also depicts an increase of Unauthenticated Download Date | 6/18/17 11:08 PM BASEMENT FOCUSING EFFECTS ON GROUND MOTION 1225 T ab l e 2 Rheological parameters of soil considered in various inhomogeneous viscoelastic models (SRM1-SRM3) and a comparison of obtained fundamental frequency (F0) and spectral amplification at F0 (SAF) with the analytical one Model parameters Velocity at FR [m/s] Densiy [kg/m3] Quality factors at FR Unrelaxed moduli Fundamental frequency of S-wave [Hz] Amplification of S-wave at fundamental frequency S-wave P-wave S-wave P-wave [GPa] Numerical Analytical Numerical Analytical SRM1 525.00 909.00 2000.00 10.00 17.00 7.476 19.725 4.762 1.65 1.66 5.20 5.10 SRM2 525.00 909.00 2000.00 20.00 34.00 6.401 18.030 5.227 1.64 1.65 6.40 6.38 SRM3 525.00 909.00 2000.00 50.00 85.00 5.847 17.106 5.411 1.64 1.64 7.50 7.53 Fig. 6. Unbounded (with no free surface) synclinal basement topography (SBT) model with horizontal and vertical arrays. The distance of horizontal array from the flat part of SBT is variable. The vertical array is passing along the focal length of SBT. Note: on the vertical array, the distances of receiver’s points from the tip of SBT are normalized with the focal length of SBT – mentioned as NDTSBT in the manuscript. Unauthenticated Download Date | 6/18/17 11:08 PM 1226 J.P. NARAYAN and V. KUMAR fundamental and higher mode frequencies with an increase of sediment damping. This finding corroborates with the GMB-EK rheology, where there is an increase of phase velocity with an increase of sediment damping. The spectral amplifications at fundamental frequency (SAF) were also computed analytically using the phase velocity VS(ω) and quality factor QS(ω) corresponding to GBM-EK model and the following relationship (Bard and RieplThomas 2000) ⎛ πIC ⎞ SAF = IC ⎜1 + , ⎜ 4Q (ω ) ⎟⎟ ⎝ ⎠ (18) where IC is impedance contrast between soil layer and the half-space. A comparison of the numerically obtained fundamental frequency (F0) and the spectral amplification at F0 (SAF) with analytical solutions reveals an excellent matching (Table 2). 4. EFFECTS OF RHEOLOGY AND SYNCLINAL BASEMENT TOPOGRAPHY To study the combined effects of rheology of sedimentary deposits and synclinal basement topography (SBT) on the ground motion characteristics, seismic responses of an unbounded 2D SBT model consisting of a single sedimentary layer with different rheology overlying the SBT are simulated using a horizontal plane SV-wave front. Figure 7 shows a 2D cross-section (XZ plane) of a 3D SBT, which is extending infinitely perpendicular to the considered cross-section. The shape of the SBT is semi-circular and numerically discretized in the form of a staircase since square FD grids are used. In order to get the shape of the synform as close as possible to a semi-circular SBT, finer grids (∆x = ∆z = 10 m) are used as compared to the required grids per wavelength to avoid grid dispersion. In the considered XZ plane, positive Z-coordinates are pointing upward from the tip of the SBT. All the distances were measured with respect to the tip of the SBT. The radius of curvature of the SBT is 3000 m. The P- and S-waves velocities and quality factors at reference frequency, density and unrelaxed moduli of the sedimentary deposit and basement rock are given in Table 3 for different rheological models for sediment deposit (BTM1-BTM4). The BTM1 model is elastic and the other models are viscoelastic. A plane horizontal SV-wave front was generated at a depth of 1000 m below the tip of the SBT (i.e., at a depth of 4000 m below the top-flat part of the SBT; Fig. 7). To avoid edge reflections from the four edges of the model, both the sponge boundary condition (Israeli and Orszag 1981) and absorbing boundary conditions of Clayton and Engquist (1977) were implemented on all the model edges. The time increment was taken as 0.001 s, which fulfills the stability criterion (Eq. 18). Unauthenticated Download Date | 6/18/17 11:08 PM BASEMENT FOCUSING EFFECTS ON GROUND MOTION 1227 Fig. 7. Seismic responses of BTM1-BTM4 models, for an incident horizontal SVwave front, on a vertical array. Note: the amplitude scale used for normalizing the vertical component is 104 times larger. Basement rock Sediment deposits T ab l e 3 Rheological parameters for sediment with different damping and basement for various basement topography models (BTM1-BTM4) SBT models Velocity at FR [m/s] S-wave P-wave BTM1 1750 3031 BTM2 1750 BTM3 1750 BTM4 Density [kg/m3] Quality factor at FR Unrelaxed moduli [GPa] S-wave P-wave MU KU λU 2200 – – 6.737 20.212 6.737 3031 2200 175 303 6.851 20.407 6.704 3031 2200 131 227 6.889 20.473 6.693 1750 3031 2200 85 151 6.967 20.605 6.670 BTM1 3200 5542 2800 – – 28.672 86.016 28.672 BTM2 3200 5542 2800 320 554 28.935 86.453 28.582 BTM3 3200 5542 2800 320 554 28.935 86.453 28.582 BTM4 3200 5542 2800 320 554 28.935 86.453 28.582 Both the elastic (BTM1) and viscoelastic (BTM2-BTM4) seismic responses of the SBT models were computed for the quantification of the combined effects of sediment rheology and SBT on ground motion characUnauthenticated Download Date | 6/18/17 11:08 PM 1228 J.P. NARAYAN and V. KUMAR teristics. Seismic responses without SBT were also computed taking the same model parameters and the source-receiver configuration but using a horizontal interface between the sedimentary deposit and the basement rock at the depth of the tip of the SBT. 4.1 Seismic response of SBT models Seismic responses of SBT models were computed on both the horizontal and vertical arrays, as shown in Fig. 7. The vertical array along the focal length of the SBT extends from 450 m below the tip of the SBT to 6550 m above the tip of the SBT with 15 equidistant (500 m apart) receiver points. On the other hand, the horizontal array extends from 5000 m left to 5000 m right of the SBT axis with 21 equidistant (500 m apart) receiver points. Further, the position of the horizontal array is varied in vertical direction. Response on vertical array First, the seismic response of the elastic BTM1 model was simulated on the vertical array along the focal length of the SBT. Figure 8a shows the Fig. 8. Continued on next page. Unauthenticated Download Date | 6/18/17 11:08 PM BASEMENT FOCUSING EFFECTS ON GROUND MOTION 1229 Fig. 8. Horizontal and vertical components of seismic responses of BTM1 model on different horizontal arrays at a distance of 6750, 5750, and 4750 m from the tip of SBT, respectively. horizontal and vertical components of ground motion. The amplitude in the vertical component is almost negligible as compared to the horizontal component (Note: amplitude scale used to normalise the traces in vertical component is 104 times larger). The normalised distance of receiver points from the tip of the SBT with respect to the focal length (FL) of the SBT are given in the brackets. The focal length (FL) of the SBT is 6620.7 m and was obtained using the following equation: FL = r , 1 −η (19) where r is the radius of the semi-circular SBT (3000 m) and η is the ratio of the S-wave velocity in the sediment to that of the basement rock. Figure 8a depicts that there is a tremendous increase of amplitude of transmitted SVwaves in the horizontal component towards the focus of the SBT, which may be due to SBT-focusing effects. Diffracted body waves from the top corners Unauthenticated Download Date | 6/18/17 11:08 PM 1230 J.P. NARAYAN and V. KUMAR of the SBT are clearly visible on receivers R2-R10 in the horizontal component. Their amplitude is highly variable due to divergence, damping and interference effects. For example, maximum amplitude seems to be reached at receiver R4 in the horizontal component. Diffracted waves are merged with the transmitted SV-wave on receivers R11-R15. A decrease of amplitude of horizontal components of SV-waves at R15 very near the focus can be inferred. This may be due to diffracted waves being out of phase with the transmitted SV-waves. Maximum amplitude of SV-waves in horizontal component is obtained at receiver R14 at a distance of 6050 m from the tip of the SBT. The normal incidence of SV-wave on the SBT and the recording on the vertical array depicts that the mode converted P-wave should not be present in any component and transmitted SV-wave should not be present in the vertical component. This means that the seismic phases visible in the vertical component with negligible amplitude may be edge-reflections. However, the maximum amplitude in the vertical component is of the order of 104-105 times smaller than that in the horizontal component and these edgereflections are therefore negligible. Further, the polarity of mode converted P-waves, SV-waves, and the diffracted P- and SV-waves in the vertical component on the left side of the SBT axis is opposite to that on the right side (see in Fig. 9) and cancel each other. On the other hand, these phases have the same polarity in the horizontal component. Therefore, the diffracted Pand SV-waves cannot be present in the vertical component, even after the SBT focusing. The presence of a direct P-wave in the horizontal component first arrival (first arrival near the focus) may be attributed to the focusing of the mode converted P-wave. Another reason may be the position of the horizontal component in the staggered grid, which is at an offset of half grid spacing from the SBT axis. The receiver R1 located in the basement rock shows both the up-going SV-wave and the reflected SV-wave from the SBT. Similarly, Fig. 8b-d shows the horizontal and vertical components of the seismic responses of the viscoelastic BTM2-BTM4 models, respectively. It is surprising to notice the presence of the mode-converted P-wave, transmitted SV-wave, and diffracted P- and SV-waves in the vertical components, as well as an increase of amplitude of the diffracted waves with the increase of sediment damping. An increase of amplitude of the suspected modeconverted P-wave in the vertical component towards the focus point of the SBT may be due to the interference with the suspected diffracted waves. Further, an increase of amplitude of noise (edge-reflections) can also be noticed with the increase of sediment damping. The only possible reason for the presence of P- and SV-waves and an increase of amplitude of diffracted waves in the vertical component is due to the somewhat poor performance of the second-order accurate anelastic function at the sediment-basement interface. A decrease of amplitude of transmitted SV-waves in horizontal Unauthenticated Download Date | 6/18/17 11:08 PM BASEMENT FOCUSING EFFECTS ON GROUND MOTION 1231 Fig. 9. Snapshots of horizontal component in a rectangular area (10 000 and 7480 m in the horizontal and vertical directions, respectively) at different times for incident SV-wave. components with the increase of sediment damping can be inferred. Generally, the analysis of Fig. 8 reveals tremendous effects of SBT focusing and rheology of sediments on the ground motion characteristics in the horizontal components. Response on horizontal array To further study the effects of the SBT on the mode conversion of the incident SV-waves, seismic responses of the elastic BTM1 model were computUnauthenticated Download Date | 6/18/17 11:08 PM 1232 J.P. NARAYAN and V. KUMAR ed at a distance of 6750, 5750, and 4750 m from the tip of the SBT on a horizontal array (Fig. 7). Receivers HR5 and HR17 are located vertically above the left and right top corners of the SBT. Figure 9 depicts the horizontal and vertical components of the seismic responses. Analysis of Fig. 9 reveals a strong mode conversion of SV-waves, particularly from the upper part of the SBT, where the angle of incidence is larger. For example, the recorded mode converted P-waves in the vertical component on receivers HR6-HR9 and HR13-HR16 (Fig. 9a) have larger amplitudes than those of the transmitted SV-wave in the horizontal component on the same receiver points. In other words, at certain locations, maximum amplitude may be associated with Pwaves for an incident SV-wave due to mode conversion. Diffracted P-waves are also visible in these seismograms (HR1-HR4 and HR18-HR21). The analysis of Fig. 9 also shows that depending on the vertical distance of the horizontal array to the focus point, there is a lot of spatial variation in ground motion characteristics (particularly in the vertical component). So, depending on the depth of the basement, spatial variation in ground motion may occur due to intense mode conversion and diffraction phenomena. In the vertical component, polarity of transmitted, mode converted and diffracted waves on the right of the SBT-axis have opposite polarity to that on the left of the SBT-axis. So, it can be concluded that basement focusing effects may not occur in the vertical component for incident SV-wave. However, the spatial variability in the vertical component of ground motion may occur due to the diffracted and the mode converted waves. Snapshots To further demonstrate the SBT-focusing effects, mode conversion and development of diffractions from the top corners of the SBT, snapshots for both the horizontal and vertical components were computed in a rectangular area at different times. Snapshots were computed in an area extending from 730 m below to 6750 m above the tip of the SBT and 5000 m left to 5000 m right of the tip of the SBT. The snapshots at different times for the horizontal and the vertical components are shown in Figs. 10 and 11, respectively. The snapshots at time 0.1 s, shown in Figs. 10a and 11a, depicts that the incoming waves are below the considered area. Just entered SV-wave in the considered rectangular area can be seen only in the horizontal component of snapshot at time 0.5 s (Fig. 10b). Figure 10c shows the SV-wave entered into the synclinal part of the SBT and the reflected SV-wave from the bottom of the SBT. On the other hand, the vertical component at time 1.0 s (Fig. 11c) depicts the transmitted mode-converted P-wave and the transmitted SV-wave in the basin as well as the reflected mode-converted P-wave in the basement rock, moving downward. Further, the polarity of the mode converted P-wave and the transmitted SV-wave towards the right of the SBT-axis is opposite to Unauthenticated Download Date | 6/18/17 11:08 PM BASEMENT FOCUSING EFFECTS ON GROUND MOTION 1233 Fig. 10. Snapshots of vertical component in a rectangular area (10 000 and 7480 m in the horizontal and vertical directions, respectively) at different times for incident SV-wave. that towards the left of the SBT-axis, as was inferred from the records on the horizontal arrays (Fig. 9). Similar effects can be seen in Figs. 10d and 11d (snapshot time 1.5 s) except that the amplitude of the transmitted SV-wave in Unauthenticated Download Date | 6/18/17 11:08 PM 1234 J.P. NARAYAN and V. KUMAR Fig. 11. Spectral amplification factors for horizontal component of SV-wave for BTM1-BTM4 models. Unauthenticated Download Date | 6/18/17 11:08 PM BASEMENT FOCUSING EFFECTS ON GROUND MOTION 1235 the horizontal component above the horizontal part of the SBT is larger than that above the synclinal part of the SBT. This may be due to the intense mode conversion of SV-wave to P-wave along the upper curved part of the SBT. The transmitted SV-wave and reflected SV-wave from the flat part of the SBT along with SBT-focusing is more pronounced in the horizontal component at times 2.0 and 2.5 s. Diffracted waves are not well separated from the transmitted SV-wave in Fig. 10e-f. On the other hand, the transmitted modeconverted P-wave and the transmitted SV-wave along with diffracted P- and SV-waves are very clearly visible in Fig. 11e-f at snapshot times 2.0 and 2.5 s, respectively. The transmitted and diffracted SV-waves are only visible in horizontal and vertical components of snapshots at times 3.0 and 3.5 s, as shown in Figs. 10g-h and 11g-h, respectively. The focusing of the horizontal component of the transmitted SV-wave towards the focus point can be inferred in snapshots at times 2.0, 2.5, 3.0, and 3.5 s in Fig. 10e-h, respectively. The maximum focusing effects is visible in the horizontal component of snapshot at time 3.5 s. 4.2 Amplification factors To assess the combined effects of sediment rheology and SBT-focusing quantitatively, spectral amplification of the SV-wave in each trace above the SBT is computed with respect to a reference trace (response at the R2 receiver point of a model with horizontal interface between sediment and basement rock passing through the tip of the SBT is used as reference trace). The spectral amplifications at different normalized distances from the SBTtip (NDTSBT) are shown in Fig. 12. Diffracted P- and SV-waves were removed manually from the traces R2-R7 and R2-R10, respectively, before computation of the amplitude spectra. Diffracted waves were removed to infer whether focusing is frequency dependent. Figure 12a shows the spectral amplification for the elastic BTM1 model. This figure depicts an increase of spectral amplification with increasing NDTSBT value. It can also be inferred that spectral amplification is increasing with increasing frequency. Further, the rate of increase of spectral amplification with frequency is increasing with increasing NDTSBT value. However, spectral amplification is highly affected by the presence of diffracted waves where manual removal of diffracted waves was not possible. The effect of diffraction can be seen in the form of ups and downs in the spectral amplification instead of a steady increase of spectral amplification with frequency in the considered bandwidth. Effect of diffraction is not visible in the traces with NDTSBT values less than 0.53 since it has been removed from the traces before Fourier transform. Maximum effect of diffraction in the form of ups and downs of spectral amplification can be seen in the record at NDTSBT value equal to 0.76. Unauthenticated Download Date | 6/18/17 11:08 PM 1236 J.P. NARAYAN and V. KUMAR The spectral amplification at a NDTSBT value equal to 0.91 reveals a maximum value of 5.15 at 10.0 Hz. This finding suggests that higher frequencies would be amplified even more if higher frequencies were used in the simulation. This corroborates with the finding of Davis et al. (2000), who predicted elastic amplifications analytically, without considering diffractions. The combined effects of SBT-focusing and the sediment rheology on the spectral amplifications are shown in Fig. 12b-d. This figure depicts a decrease of spectral amplification with increasing sediment damping. Further, higher frequencies are damped more compared to lower frequencies, even though higher frequencies were amplified more due to the SBT-focusing as was inferred from the elastic BTM1-model. The highest frequency, up to which the spectral amplification increased with frequency, depends on the Fig. 12. Continued on next page. Unauthenticated Download Date | 6/18/17 11:08 PM BASEMENT FOCUSING EFFECTS ON GROUND MOTION 1237 Fig. 12. Effect of sediment damping on amplitude amplification of horizontal component of SV-wave in time domain and average spectral amplification factors, respectively. NDTSBT value and sediment damping. For example, for a constant NDTSBT value of 0.91 the spectral amplification is increasing with frequency up to 6.32, 5.51, and 4.29 Hz in BTM2-BTM4 models, respectively. Even the amplification of certain higher frequencies and up to certain NDTSBT values can be inferred in case of models having higher damping. The computed amplification in time domain at different NDTSBT positions is shown in Fig. 13a. It was computed by taking the ratio of maximum amplitude at different NDTSBT positions and the maximum amplitude in the second trace of the model without SBT. There is an increase of amplification Unauthenticated Download Date | 6/18/17 11:08 PM 1238 J.P. NARAYAN and V. KUMAR Fig. 13. Effects of focal length on amplitude amplification in time domain and average spectral amplitude amplification factors, respectively. Note: focal length is varied with varying sediment velocity. in time domain with increasing NDTSBT value. This increase is to some extent linear up to an NDTSBT value of 0.76, thereafter the increase of amplification with NDTSBT value is tremendous (non-linear). The maximum amplification is obtained at an NDTSBT value of around 0.91. The decrease of amplification near the focus point may be attributed to the diffracted waves. The average spectral amplifications in the considered frequency range were also computed at different NDTSBT positions (Fig. 13b). There is also an increase of average spectral amplification with increasing NDTSBT value and decrease of sediment damping. It is also to some extent linear up to a NDTSBT value of 0.76; thereafter, the rate of increase of the average spectral amplification with NDTSBT value is very large. The maximum average spectral amplification is also obtained at NDTSBT value of 0.91. A comparison of panels (a) and (b) in Fig. 13 shows that the amplification in time domain is very similar to the average spectral amplification. This may be due to the presence of diffracted waves and spectral shape of the considered Ricker wavelet. 5. EFFECT OF SBT FOCAL LENGTH In this subsection, the effects of SBT focal length on ground motion characteristics are described. The SBT focal length is varied by varying both the sediment velocity and the radius of curvature of the SBT for a fixed aperture length (6000 m). Further, the damping of the sediment is kept constant for the different considered models. 5.1 Variation of sediment velocity First, the focal length of the SBT is varied by changing the sediment velocity. For this purpose, four SBT models (BTVM1-BTVM4) with different sediment velocities and densities but with a fixed SBT radius of 3000 m are Unauthenticated Download Date | 6/18/17 11:08 PM BASEMENT FOCUSING EFFECTS ON GROUND MOTION 1239 considered. The rheological parameters, namely velocity and quality factor at reference frequency, density and unrelaxed moduli for the different types of sediment and the basement are given in Table 4. Focal lengths for BTVM1-BTVM4 models were obtained as 4363, 4923, 5647, and 6620 m, respectively. The computed amplification in time domain and the average spectral amplification at different NDTSBT positions are shown in Fig. 14a-b, respectively. Both the amplification in time domain and the average spectral amplification at a particular NDTSBT value are decreasing with decreasing velocity in the sediment (focal length). This may be due to an increase of mode conversion along the upper part of the SBT with increasing impedance contrast. Further, amplification factors are highly variable near the focus point. This may be due to different NDTSBT values available in the different models at receiver points. A higher rate of decrease of ampliT ab l e 4 Rheological parameters of different sediment velocity considered for various basement topography models (BTVM1-BTVM4) SBT models Sediment BTVM1 deposits BTVM2 Basement Velocity at FR [m/s] S-wave P-wave Density [kg/m3] Q at FR S-wave P-wave Unrelaxed moduli [GPa] MU KU λU 1750 3031 2200 175 303 6.851 20.407 6.704 1500 2598 2050 175 303 4.690 13.970 4.590 BTVM3 1250 2165 1950 175 303 3.098 9.228 3.032 BTVM4 1000 1732 1800 175 303 1.830 5.452 1.791 3200 5542 2800 320 554 28.935 86.453 28.58 Fig. 14. Effects of focal length on amplitude amplification in time domain and average spectral amplification, respectively. Note: focal length is varied with varying the sediment velocity. Unauthenticated Download Date | 6/18/17 11:08 PM 1240 J.P. NARAYAN and V. KUMAR fication with increasing distance on the far side of the focus point can be observed compared to the near side of the focus point. This may be due to the combined effects of sediment damping and SBT de-focusing. 5.2 Variation of radius of curvature Second, the focal length was varied by taking different radii of curvature for the SBT models. The radius of curvature was taken as 3000, 3250, 3500, and 3750 m for BTRM1- BTRM4 models but with a fixed aperture length of 6000 m (maximum width of SBT). The rheological parameters, namely velocity and quality factor at reference frequency, density and unrelaxed moduli for the sediment are given in Table 5. Focal lengths for BTRM1-BTRM4 models were obtained as 4363, 4727, 5090, and 5454 m, respectively. The rheological parameters for the basement are the same as in the previous case. The computed amplification in time domain and the average spectral amplifications are shown in Fig. 15a-b, respectively. Figure 15 depicts a decrease of amplification factors with increasing focal length at a particular NDTSBT value due to the larger distance travelled. Again, a much higher rate of decrease of amplification on the far side of the focus point may be due to the T ab l e 5 Rheological parameters of sediments for BTRM1-BTRM4 basement topography models having different radii of curvature BTRM1BTRM4 SBT models Velocity at FR [m/s] S-wave P-wave 1000 1732 Density [kg/m3] 2200 Q at FR Unrelaxed moduli [GPa] S-wave P-wave MU KU MU 100 173 1.853 5.491 1.784 Fig. 15. Effects of focal length on amplitude amplification in time domain and average spectral amplification, respectively. Note: focal length is varied with varying the radius of curvature of the SBT for a fixed value of aperture length. Unauthenticated Download Date | 6/18/17 11:08 PM BASEMENT FOCUSING EFFECTS ON GROUND MOTION 1241 combined effects of sediment damping and SBT de-focusing. Finally, it can be concluded that basement focusing effects not only depend on NDTSBT value but also depends on focal length for a fixed value of sediment damping. 6. DISCUSSION This research work was motivated by the selective damage of chimneys in west Seattle, Washington, consistently during the past three earthquakes (Booth et al. 2004) as well as very peculiar damage pattern observed in the Santa Monica area, Los Angeles basin, during the 1994 Northridge earthquake (Gao et al. 1996, Davis et al. 2000). Based on the simulated and recorded seismograms, Stephenson et al. (2006) inferred that focusing of direct waves played a major role in the damage of chimneys in west Seattle during the 2001 Nisqually earthquake. Frankel et al. (2009) also carried out 3D simulations for the 2001 Nisqually earthquake and inferred focusing of S-waves in west Seattle. However, the above works cannot explain the selective damage of chimneys only. The main conclusion of the presented research work is that a particular frequency may be maximally amplified for a particular value of NDTSBT value and sediment damping. This maximally amplified frequency may cause selective damage if it matches with the natural frequency of a structure. This may be the reason behind the reported mysterious damage to only unreinforced brick chimneys in west Seattle during the 2001 Nisqually earthquake. Amplification factors caused by SBT focusing are highly affected by the presence of diffracted waves from the very sharp edges at the transition from the synform to the horizontal part of the SBT. In case of real geological models there may not be such very sharp edges due to the erosion prior to sedimentation. Further, SBT focusing may also be highly affected by variations of velocity in different sediment layers overlying the basement and seismic anisotropy of the sediment. The considered 2D SBT models are applicable to simulate the response of elongated synclinal basement topography in a direction perpendicular to the plane of simulation, for example, an elongated buried horst and graben structure below the sediment. However, the considered 2D SBT models are not applicable to simulate the response of a buried depression in the basin, which requires 3D models. The effect of SBT focusing only on incident S-waves is studied and not on incident P-waves since earthquake ground motion is dominated by S-waves and the damage potential of P-wave is much lower than that of S-wave. Unauthenticated Download Date | 6/18/17 11:08 PM 1242 7. J.P. NARAYAN and V. KUMAR CONCLUSIONS A second-order-accurate-in-time and fourth-order-accurate-in-space P-SVwave staggered-grid viscoelastic finite-difference (FD) program is written. Frequency-dependent damping in the time-domain FD simulations is incorporated based on the GMB-EK rheological model (Emmerich and Korn 1987, Kristek and Moczo 2003). An excellent match of numerically computed phase velocity and quality factors for both the SV- and the P-waves for an unbounded viscoelastic homogeneous medium with the same computed analytically using Futterman’s relation (Futterman 1962) and the GMB-EK rheological model validates the accuracy of implementation of realistic damping in the time-domain FD program. Grid dispersion and stability analysis reveals that the requirement for the number of grid point per wavelength and stability condition are the same as reported by others (Moczo et al. 2000). The snapshots along with the simulated responses revealed that SBT focusing effects depend on NDTSBT value and sediment damping. An increase of amplification with both increasing frequency and distance from the tip of the SBT was obtained in case of an elastic model. A similar conclusion was also drawn by Davis et al. (2000) using analytical response of an elastic SBT model without considering diffractions. In addition, an increase of the rate of amplification with frequency was inferred towards the focus point of the SBT. However, the sediment damping masks the increase of amplification with frequency. For example, the largest amplification of 3.37, 3.05, and 2.59 was obtained at a NDTSBT value of 0.91 for 6.32, 5.51, and 4.29 Hz in BTM2-BTM4 models, respectively. So, it is concluded that a particular frequency may be maximally amplified for a particular value of the SBT focal length, distance of a site from the tip of SBT and sediment damping. It is also concluded that focusing of mode converted waves may not occur, but amplitudes of mode converted waves may be larger than the amplitudes of transmitted waves at certain locations on the free surface. A c k n o w l e d g e m e n t . Authors are grateful to Dr. Marcel Frehner, ETH Zurich, and an anonymous reviewer for their insightful comments and suggestions which significantly improved the revised manuscript. The first author is also thankful to the Ministry of Earth Sciences, New Delhi, and Council of Scientific and Industrial Research, New Delhi, for financial assistance through Grant Nos. MES-484-EQD and CSR-569-EQD, respectively. Unauthenticated Download Date | 6/18/17 11:08 PM BASEMENT FOCUSING EFFECTS ON GROUND MOTION 1243 References Alex, C.M., and K.B. Olsen (1998), Lens-effect in Santa Monica?, Geophys. Res. Lett. 25, 18, 3441-3444, DOI: 10.1029/98GL52668. Bard, P.Y., and J. Riepl-Thomas (2000), Wave propagation in complex geological structures and their effects on strong ground motion. In: E. Kausel and G. Manolis (eds.), Wave Motion in Earthquake Engineering, International Series on Advances in Earthquake Engineering, WIT Press, Southampton, 37-95. Booth, D.B., R.E. Wells, and R.W. Givler (2004), Chimney damage in the greater Seattle area from the Nisqually earthquake of 28 February 2001, Bull. Seismol. Soc. Am. 94, 3, 1143-1158, DOI: 10.1785/0120030102. Carcione, J.M., D. Kosloff, and R. Kosloff (1988), Wave propagation simulation in a linear viscoacoustic medium, Geophys. J. Int. 93, 2, 393-407, DOI: 10.1111/j.1365-246X.1988.tb02010.x. Clayton, R.W., and B. Engquist (1977), Absorbing boundary conditions for acoustic and elastic wave equations, Bull. Seismol. Soc. Am. 67, 6, 1529-1540. Day, S.M., and J. B. Minster (1984), Numerical simulation of attenuated wavefields using a Padé approximant method, Geophys. J. Roy. Astr. Soc. 78, 1, 105118, DOI: 10.1111/j.1365-246X.1984.tb06474.x. Davis, P.M., L.J. Rubinstein, K.H. Liu, S.S. Gao, and L. Knopoff (2000), Northridge earthquake damage caused by geologic focusing of seismic waves, Science 289, 5485, 1746-1750, DOI: 10.1126/science. 289.5485.1746. Emmerich, H., and M. Korn (1987), Incorporation of attenuation into time-domain computations of seismic wave fields, Geophysics 52, 9, 1252-1264, DOI: 10.1190/1.1442386. Futterman, W.I. (1962), Dispersive body waves, J. Geophys. Res. 67, 13, 52795291, DOI: 10.1029/JZ067i013p05279. Frankel, A., W. Stephenson, and D. Carver (2009), Sedimentary basin effects in Seattle, Washington: Ground-motion observations and 3D simulations, Bull. Seismol. Soc. Am. 99, 3, 1579-1611, DOI: 10.1785/0120080203. Frehner, M., S.M. Schmalholz, E.H. Saenger, and H. Steeb (2008), Comparison of finite difference and finite element methods for simulating two-dimensional scattering of elastic waves, Phys. Earth Planet. Int. 171, 1-4, 112-121, DOI: 10.1016/j.pepi.2008.07.003. Galis, M., P. Moczo, and J. Kristek (2008), A 3-D hybrid finite-difference-finiteelement viscoelastic modeling of seismic wave motion, Geophys. J. Int. 175, 1, 153-184, DOI: 10.1111/j.1365-246X.2008.03866.x. Gao, S., H. Liu, P.M. Davis, and G.L. Knopoff (1996), Localized amplification of seismic waves and correlation with damage due to the Northridge earthquake: Ewidence of focusing in Santa Monica, Bull. Seismol. Soc. Am. 86, 1B, S209-S230. Unauthenticated Download Date | 6/18/17 11:08 PM 1244 J.P. NARAYAN and V. KUMAR Hartzell, S., E. Cranswick, A. Frankel, D. Carver, and M. Meremonte (1997), Variability of site response in the Los Angeles urban area, Bull. Seismol. Soc. Am. 87, 6, 1377-1400. Ihnen, S.M., and D.M. Hadley (1986), Prediction of strong ground motion in the Puget Sound region: the 1965 Seattle earthquake, Bull. Seismol. Soc. Am. 76, 4, 905-922. Israeli, M., and S.A. Orszag (1981), Approximation of radiation boundary conditions, J. Comp. Phys. 41, 1, 115-135, DOI: 10.1016/0021-9991(81)90082-6. Kristek, J., and P. Moczo (2003), Seismic-wave propagation in viscoelastic media with material discontinuities: A 3D fourth-order staggered-grid finite difference modeling, Bull. Seismol. Soc. Am. 93, 5, 2273-2280, DOI: 10.1785/0120030023. Kumar, S., and J.P., Narayan (2008), Absorbing boundary conditions in a 4th order accurate SH-wave staggered grid finite difference program with variable grid size, Acta Geophys. 56, 4, 1090-1108, DOI: 10.2478/s11600-0080043-9. Levander, A.R. (1988), Fourth-order finite-difference P-SV seismograms, Geophysics 53, 11, 1425-1436, DOI: 10.1190/1.1442422. Madariaga, R. (1976), Dynamics of an expanding circular fault, Bull. Seismol. Soc. Am. 66, 3, 639-666. Moczo, P., and P.-Y Bard (1993), Wave diffraction, amplification and differential motion near strong lateral discontinuities, Bull. Seismol. Soc. Am. 83, 1, 85106. Moczo, P., E. Bystricky, J. Kristek, J.M. Carcione, and M. Bouchon (1997), Hybrid modelling of P-SV seismic motion at inhomogeneous viscoelastic topographic structures, Bull. Seismol. Soc. Am. 87, 5, 1305-1323. Moczo, P., J. Kristek, and E. Bystricky (2000), Stability and grid dispersion of the P-SV 4th-order staggered-grid finite-difference schemes, Stud. Geophys. Geod. 44, 3, 381-402, DOI: 10.1023/A:1022112620994. Moczo, P., J. Kristek, V. Vavryčuk, R.J. Archuleta, and L. Halada (2002), 3D heterogeneous staggered-grid finite-difference modelling of seismic motion with volume harmonic and arithmetic averaging of elastic moduli and densities, Bull. Seismol. Soc. Am. 92, 8, 3042-3066, DOI: 10.1785/0120010167. Narayan, J.P. (2003), Simulation of ridge-weathering effects on the ground motion characteristics, J. Earthq. Eng. 7, 3, 447-461, DOI: 10.1080/ 13632460309350458. Narayan, J.P. (2005), Study of basin-edge effects on the ground motion characteristics using 2.5-D modeling, Pure Appl. Geophys. 162, 2, 273-289, DOI: 10.1007/s00024-004-2600-8. Narayan, J.P. (2012), Effects of P-wave and S-wave impedance contrast on the characteristics of basin transduced Rayleigh waves, Pure Appl. Geophys. 169, 4, 693-709, DOI: 10.1007/s00024-011-0338-7. Unauthenticated Download Date | 6/18/17 11:08 PM BASEMENT FOCUSING EFFECTS ON GROUND MOTION 1245 Narayan, J.P., and S. Kumar (2008), A fourth order accurate SH-wave staggered grid finite-difference algorithm with variable grid size and VGR-stress imaging technique, Pure Appl. Geophys. 165, 2, 271-294, DOI: 10.1007/ s00024-008-0298-8. Narayan, J.P., and S. Kumar (2010), A 4th order accurate P-SV wave staggered grid finite difference algorithm with variable grid size and VGR-stress imaging technique, Geofizika 27, 1, 45-68. Narayan, J.P., and P.V. Prasad Rao (2003), Two and half dimensional simulation of ridge effects on the ground motion characteristics, Pure Appl. Geophys. 160, 8, 1557-1571, DOI: 10.1007/s00024-003-2360-x. Narayan, J.P., and A. Ram (2006), Numerical modelling of the effects of an underground ridge on earthquake-induced 0.5-2.5 Hz ground motion, Geophys. J. Int. 165, 1, 180-196, DOI: 10.1111/j.1365-246X.2006.02874.x. Narayan, J.P., and A.A. Richharia (2008), Effects of strong lateral discontinuity on ground motion characteristics and aggravation factor, J. Seismol. 12, 4, 557-573, DOI: 10.1007/s10950-008-9109-z. Saenger, E.H., S.A Shapiro, and Y. Keehm (2005), Seismic effects of viscous Biotcoupling: Finite difference simulations on micro-scale, Geophys. Res. Lett. 32, 14, L14310, DOI: 10.1029/2005GL023222. Stephenson, W.J., A.D. Frankel, J.K. Odum, R.A. Williams, and T.L. Pratt (2006), Towards resolving an earthquake ground motion mystery in west Seattle, Washington State: Shallow seismic focusing may cause anomalous chimney damage, Geophys. Res. Lett. 33, 6, L06316, DOI: 10.1029/ 2005GL025037. Virieux, J. (1986), P-SV wave propagation in heterogeneous media; velocity-stress finte-diffeence method, Geophysics 51, 4, 889-901, DOI: 10.1190/ 1.1442147. Received 14 November 2012 Received in revised form 9 May 2013 Accepted 14 May 2013 Unauthenticated Download Date | 6/18/17 11:08 PM
© Copyright 2026 Paperzz