Electronic Supplement to
Moment Tensor Inversion of Seismic Events Associated with the Sinkhole at Napoleonville Salt Dome, Louisiana

by Avinash Nayak and Douglas S. Dreger

To analyze the effects of different values of Q on synthetic waveforms, we computed Fourier amplitude spectra of synthetic radial displacements at stations LA02 (sediment) and LA03 (salt) for event TE1 using Green’s functions (GF) computed for different values of Q_P (= 2 × Q_S) and Q_S (Fig. S1). The amplitudes at frequencies of 0.1–0.2 Hz used for waveform inversion in this study are found to be insensitive to anelastic attenuation (Q_P and Q_S) down to Q_S = 25. Figure S2 shows the spatial distribution of moment tensor (MT) solutions represented by P-wave first motion mechanisms for 39 events detected from 00:00 to 19:00 hours on 2 August 2012. Figure S3 shows various MT solutions and locations from the sensitivity analyses of TE1 for comparison. We also estimate full MT solutions for TE1 for various four-station combinations (Fig. S4) using the hypocenter obtained from five-station GRiD MT results (Fig. 7b,c in the main article; Kawakatsu, 1998; Tsuruoka et al., 2009). The dominant isotropic volume-increase component persists in all cases, which suggests that it is not an artifact caused by data from one particular station. The layered synthetic and salt dome velocity models used to compute GF (shown in Fig. 5 in the main article) are provided in Table S1. Table S2 provides a catalog for all 62 events detected from 19:00 hours on 1 August 2012 to 19:00 hours on 2 August 2012, showing date, centroid time, centroid grid point hypocenter, full MT solution, and uncertainties in the MT elements.

MT Inversions in the 0.1–0.3 Hz Frequency Range

We observed that the preliminary velocity models shown in Figure 5 of the main article fail to fit displacement waveforms at frequencies higher than 0.1–0.2 Hz. Therefore, for TE1, keeping the hypocenter and full MT solution fixed, we iteratively perturb the 1D sediment and salt dome velocity models to improve the waveform fits at higher frequencies. The revised velocity models are shown in Figure S5. Compared to the preliminary models, the revised V_S models are slower in a general sense. The displacement waveforms in the 0.1–0.3 Hz frequency range are weakly sensitive to V_P models; hence, the modifications to V_P values are minor. Using these revised velocity models, we repeat the entire procedure of computing GF and applying GRiD MT to displacement waveforms filtered using a causal four-pole Butterworth band-pass filter with corners at 0.1 and 0.3 Hz and decimated to 0.25 s sampling interval. Displacement waveforms at each station were weighted with the reciprocal of their variance to account for large differences in amplitudes at different stations. The results for TE1 are shown in Figure S6. The full MT solution fits the farthest three stations (LA01, LA02, and LA03) quite well and possesses a dominant volume-increase component (isotropic [ISO] 64%) similar to the low-frequency MT solution estimated using the preliminary velocity models (Fig. 7a in main article). The revised centroid hypocenter is shallower with a minor change in epicenter (Fig. S3). We are unable to fit to waveforms of LA08 and LA09 as well, which appear to be more complex; however, the long-period signals at these stations are fit well. In Figure S6, waveforms at LA08 and LA01 show the 0.4 Hz weakly damped harmonic signal mentioned in the article.

Figure S7 shows MT solutions and waveform fits for TE1, assuming various source mechanisms: (a) shear-tensile and (b) pure crack in sediments (ν = 0.39), (c) pure isotropic explosion, and (d) pure doublecouple (DC).

---

Figures

Figure S1. Fourier amplitude spectra of synthetic radial displacements at stations LA02 (sediment) and LA03 (salt) for event TE1 using GF computed for different values of Q_P (= 2 × Q_S) and Q_S. R is the epicentral distance.

Figure S2. Mechanisms and locations of 39 events detected from 00:00 to 19:00 hours, 2 August 2012. Solid and dotted lines are the 1000 ft and 10,000 ft depth contours of NSD, respectively. The shaded polygon shows the approximate surface extent of the sinkhole in July 2013; black triangles are station locations; the black diamond is the Oxy Geismar 3 cavern.

Figure S3. Map showing various MT solutions from the sensitivity analyses of TE1. Full MT, the original full MT solution (Fig. 7a); Deviatoric MT, deviatoric MT solution (Fig. 10); Sediment, full MT solution using GF from the sediment velocity model (Fig. 11a); Salt, full MT solution using GF from the salt dome velocity model (Fig. 11b); and High Frequency, full MT solution using 0.1–0.3 Hz displacement waveforms and calibrated velocity models (Fig. S5). The meaning of other symbols is the same as in Figure S2.

Figure S4. Full MT four-station solutions and waveform fits (solid line, observed; dashed line, synthetic) for TE1, computed excluding stations (a) LA01, (b) LA02, (c) LA03, (d) LA08, and (e) LA09. Event depth and station-specific epicentral distance (R), azimuth (Az, in degrees), and maximum displacement amplitude at a station (Max Amp) are the same in all subplots. Components are double couple (DC), compensated linear vector dipole (CLVD), and isotropic (ISO), and VR is variance reduction.

Figure S5. Revised (solid lines) and preliminary (dashed lines) 1D velocity models. Gray and black lines represent salt and sediment velocities, respectively. Thick and thin lines represent V_P and V_S, respectively.

Figure S6. Results for TE1 using revised sediment and salt dome 1D velocity models (Fig. S5) and 0.1–0.3 Hz displacement waveforms (solid line, observed; dashed line, synthetic). Abbreviations are as defined for Figure S4.

Figure S7. MT solutions and waveform fits (solid line, observed; dashed line, synthetic) for TE1, assuming the source mechanism to be (a) shear-tensile (crack + DC) and (b) pure crack in sediments (ν = 0.39), (c) pure isotropic explosion, and (d) pure DC. Event depth and station-specific R, Az, and Max Amp are the same in all subplots; other abbreviations are as defined for Fig. S4.

---

Tables

Table S1. Velocity models used to compute GF.

Table S2. Catalog for all events detected from 19:00 hours on 1 August 2012 to 19:00 hours on 2 August 2012. Note that locations and depths correspond to grid points, and origin time (OT) is the centroid origin time.

---

[ Back ]