The Northridge earthquake

We demonstrate the method by simulating strong ground motions from several slip models for the 17 January 1994 Mw 6.7 Northridge earthquake. A model derived from near-source strong ground motions, P and SH teleseismic body waves, and GPS and spirit leveling geodetic data provides reference time histories and peak ground motions for comparison with the motions obtained from the source models inferred from uniform and variable-slip models derived from geodetic data only.

To assess the methodology, we compare the observed and modeled velocity waveforms at near-source stations surrounding the event. To account for site effects, we applied site amplification factors to the waveforms based on the upper 30 m shear wave velocity. Amplification factors are calculated by considering differences in impedances for the rock and soil layers (upper 30 m) for a normal incidence seismic wave.

For most stations and components, predicted strong ground velocities for the reference and variable-slip models are similar to the observed velocities, whereas the uniform-slip model tended to underpredict the velocities. Simulated peak ground velocity (PGV) ShakeMaps for the three models showed similar behavior with the PGV values derived from the uniform-slip model significantly less than the other two models, possibly because the model is too smooth and spread in time. The simulated PGV for reference and variable-slip models showed the roughly the same extent of the 10-20 cm/s contours, the level at which structural damage can occur (Figure 13.41).

Since our objective is to develop a method that is practical in a rapid (on the order of 30 minutes) post-earthquake time frame, we made several simplifying assumptions such as constant slip and rupture velocity. As more information becomes available following an earthquake it may be possible to invert for models to resolve these parameters, but such analysis is not applicable in an automated fashion. Nevertheless, to first order, assuming rupture velocity is constant performs well in describing the overall directivity effect and in simulating peak ground velocity for ShakeMap purposes. Although studies of many events show that the slip velocity is not spatially constant, the assumption is generally applicable given that many spatio-temporal slip models have longer duration for larger slip.

Future work will test this methodology for other large, well-studied events, such as 2003 San Simeon, 1992 Landers, and 2004 Parkfield, to optimize the rules for time variation of slip. Combining the optimized rules and near real-time measurement and inversion of GPS displacement vectors can reduce the estimation time of the extent of near-fault strong ground shaking, and thus could facilitate better emergency response activities.

Figure 13.41: (a) PGV ShakeMaps simulated for the Northridge earthquake for model UNI. Black contours of PGV represent the geometric mean of two horizontal PGV components. The minimum level of contour is 10 cm/s and the interval is 20 cm/s. The dotted line is the 20 cm/s contour, indicating the extent of significant structural damage. The black star indicates mainshock epicenter. The black dots are the grid locations where strong motions are derived for ShakeMap. Black bold numbers on the map indicate 15 observed PGV. (b) Same as (a) for modeL REF. (c) Same as (a) for model VAR. (d) The simulated PGV values at 15 stations as a function of real observations for models VAR (triangles), REF (squares), and UNI (inverted triangles) and the best fitting lines for models VAR (thin grey), REF (thick black), and UNI (dotted grey). (e) The best fitting lines assuming different stress drops for constant rupture velocity of 3.0 km/s. For the reference, the perfect agreement line is indicated with the dotted gray line. (f) The best fitting lines assuming different rupture velocities for constant stress drop of 100 bars.