Abstract
We use first-arrival wide-angle seismic refraction data recorded during the 1998 Seismic Hazards Investigation in Puget Sound (SHIPS) experiment to derive a 3-D model of the upper crustal velocity structure beneath the Strait of Georgia in southwestern British Columbia, Canada. This region, which marks the boundary between the Coast Belt to the east, and the Insular Belt to the west, includes the site of the shallow (3-5 km) 1997 M=4.6 and 1975 M=5 earthquakes. Twenty-five temporary land based stations and five ocean bottom seismometers recorded airgun blasts from within the Strait of Georgia. The dataset comprises approximately 100,000 first arrival traveltime picks of refractions through the upper crust. The onshore-offshore recording geometry in the narrow (210 X 60 km) study region results in anisotropic 3-D ray coverage, and limits the long offset, deeper penetrating arrivals to raypaths that travel along, as opposed to across, the strait. Nevertheless, to fully exploit these data a full 3-D analysis was required. We used a regularized inversion algorithm to provide the smoothest model appropriate for the data errors. Crustal structure is constrained to a maximum depth of 10 km. The most significant feature of the final model is a wedge-shaped region in the southeast with relatively low velocities (2-5 km/s), representing Georgia Basin sedimentary rocks of the Cretaceous and Tertiary Nanaimo Group and overlying younger sediments. The underlying basement rocks, possibly Wrangellia or Jura-Cretaceous intrusive rocks of the Coast Belt (or both), have velocities of 6-6.5 km/s with significant lateral variation. In cross-section, the base of the basin dips at about 4 degrees to the southeast to at least 7 km depth; however, the maximum thickness may be attained further to the southeast, outside our region of constraints. Depths for the 1997 and 1975 earthquakes estimated from the regional network places their hypocenters at the base of the dipping interface. The reliability of the final model is addressed with a suite of checkerboard resolution tests using different anomaly sizes to estimate lateral resolution at each point in the model.
Figures
Figure 1. Rectangle donotes surface location of 3-D study area. Heavy broken line is the
Insular-Coast Belt boundary. White dots are receivers (Refteks and ocean bottom
seismometers (OBSs)) used in this study. Light and dark grey dots are receivers not used
in this study. Dotted lines within waterways denote the shiptrack (shot point locations).
Star denotes the epicenters of two earthquakes: ML=4.6 (1997) and ML=4.9 (1975). VA,
Vancouver; VI, Victoria. Geology based on Monger (1990). [Excuse the "beta" symbols - they should be degree symbols...]
Figure 2. Receivers (numbered circles) and shot point locations (blue and red dots) along
shot lines 5 and 6 used in 3-D study. The circles are colored according to the percentage
of traces in each receiver gather that could be picked for first arrivals. Receivers
numbered 1-5 are OBSs. Only every tenth shot point is shown. Arrows depict the
coordinate system used in the modeling.
Figure 3. Record sections for shots from line 5 into receiver 50. Data have been bandpass filtered between 4-10
Hz and are reduced at 6.5 km/s. Every second trace is plotted.
Figure 4. (a) First arrival times (reduced at 8 km/s) of the 17,832 picks (red dots) used in
the modeling plotted against shot-receiver offset without regard to position or azimuth.
Solid line represents average time in 2-km-wide offset bins used to construct
starting 1-D velocity model. (b) Smoothed 1-D starting model derived from solid line in (a).
Figure 5. Traveltime residuals for 1-D starting model (top)
and final model (bottom) plotted against source-receiver offset. The bias near zero km
offset results from misfitting the near-receiver data at the 5 OBSs.
Figure 6. Perturbation (km/s) from the starting 1-D velocity model (Fig. 4b) at depths
between 1-12 km. Locations of receivers (dots) and shot points (heavy black line) are
shown on the z=1 km panel. Regions not sampled by ray paths are shown in grey.
Contour interval is 0.2 km/s.
Figure 7. Horizontal slices through the final velocity model at depths between 1-12 km.
Locations of receivers (dots) and shot points (heavy black line) are shown on the z=1 km
panel. Star on the z=3 km panel denotes approximate hypocentral location of the 1997
ML=4.6, and 1975 ML=4.9 earthquakes (Cassidy et al., 1999). Regions not sampled by
ray paths are shown in grey. Contour interval is 0.2 km/s. Thick black contour is 6.0
km/s; thick white contour is 6.4 km/s. Note that the color scales are not linear and
different scales are used to accentuate velocity variations at different depths. "A" on the
z=2 and 3 km panels is a high velocity anomaly that correlates with a high Bouguer gravity anomaly (Fig. 12).
Figure 8. Vertical slices through the final velocity model along the x axis at y=20, 30 and
40 km. Star on the y=30 km panel denotes approximate hypocentral location of the 1997
ML=4.6, and 1975 ML=4.9 earthquakes (Cassidy et al., 1999). Thick black line is the 6.0
km/s contour. Contour interval is 0.5 km/s for velocities <6 km/s and 0.1 km/s for
velocities >6 km/s (white contours). Note that the color scale is not linear. Regions not
sampled by ray paths are shown in grey.
Figure 9. (a) Ray coverage measured as the number of rays penetrating each cell of the
final model at depths between 1-12 km. Locations of receivers (dots) and shot points (red
line) are shown on the z=1 km panel. Contours are drawn at 1, 5, 10, 25, 50, 100, 250.
Thick black line is the 25 contour. Regions not sampled by ray paths are light blue. (b)
Ray diagram showing all rays in the final model projected onto the x-y plane. (c) Ray
diagram showing all rays in the final model projected onto the x-z plane.
Figure 10. Results of checkerboard tests using checkerboard cell sizes of 10 (left panels),
20 (center panels) and 30 km (right panels). Relative velocity perturbations with respect
to the starting model are shown for depths of 2, 4, 6 and 8 km for each checkerboard cell
size. Boundaries of true checkerboard pattern are indicated by grid lines. True
checkerboard pattern is ±10 per cent velocity variation.
Figure 11. Estimated lateral velocity resolution at depths between 1-12 km using the
method of Zelt (1998). Contour interval is 5 km. White line is the 20 km interval. Black
regions have better than 5 km resolution; white regions have worse than 30 km resolution
or are unsampled.
Figure 12. Bouguer gravity anomaly in study area. White lines are contours with interval
of 10 mGal. Relative gravity high "A" corresponds to high velocity anomaly "A" on Figs.
7 & 8.
