Continuous GPS sites

Figure 1: Continuous GPS sites larger image [GIF, 36.1 kb, 497 X 396, notice]

Continuous GPS sites in southwestern BC and northwestern Washington State that are included in the daily analyses of continuous GPS data at the Geological Survey of Canada. Sites in Canada are part of the Western Canada Deformation Array (WCDA); sites in the U.S. are part of the Pacific Northwest Geodetic Array (PANGA). These sites were established expressly for the study of crustal deformation. Red stars show the epicentres for three large (M>7) historical crustal earthquakes; the blue star shows the epicentre of the recent Nisqually (in-slab) earthquake. Dashed lines show the nominal downdip extent of the seismo-genic zone on the subduction interface (green = locked zone extent; orange = transition zone extent) from Flueck et al. 1997.

Daily variations

Figure 2: ALBH-DRAO baseline variance larger image [GIF, 100.6 kb, 600 X 728, notice]

Example of daily variations in position (latitude, longitude, and height) of the site ALBH with respect to DRAO. From 1993 to 1998, the CGPS22 Analysis Software was used (Dragert et al., 1995) for the analysis of the daily data. Beginning in 1999, the Bernese Vs. 4.2 Software (Beutler et al., 2000) was adopted for routine automated analysis. The dramatic reduction in day-to-day scatter is primarily due to improved tropospheric modeling and the application of a regional filter computed from daily residuals averaged over all network sites.

Regional filters

Figure 3: Unfiltered vs filtered larger image [GIF, 47.0 kb, 600 X 395, notice]

Example of the effect of the regional filter at ALBH. The raw time series show the daily changes in position as computed by the automated Bernese V4.2 processing. To reduce further the day-to-day scatter common to all sites of the network, daily residuals in individual components are computed for each site by removing linear trends, annual signals, and steps, all estimated simultaneously by least-mean-squares regression. These residuals are averaged for all network sites and these averages are subtracted from the raw time series to generate the filtered time series. The application of this filter reduces the means of the daily (rms) scatter in the north, east, and up components from 1.3, 1.4, and 4.7 to 0.8, 0.8, and 3.1 mm respectively. This improved precision allows clear resolution of the brief reversal of motion beginning on Day 230 (Aug. 18), 1999, in both latitude and longitude.

Filtered time series

Figure 4: Filtered time series larger image [GIF, 68.9 kb, 648 X 446, notice]

Filtered time series of daily variations in relative longitude with respect to DRAO. Annual signals have also been removed. Red lines show the best-fitting linear trends which are assumed constant before and after the transient. These trends commonly form the basis for estimating long-term crustal motions. The vertical bars indicate the earliest date of detection (Day 230) of the transient. The time series are roughly positioned according to relative geographic locations of the sites and the delay of the signal in a northwest direction is apparent.

Timing and duration

Timing and duration of the transient displacements. Five-point triangular smoothing is applied to the filtered time series to resolve the timing of the transient. The onset and the duration of the transient displacements are estimated from the smoothed longitude series (see also Table 1).

Figure 5: Expanded view of transient event larger image [GIF, 41.6 kb, 600 X 752, notice]

Figure 6: Propagation of signal larger image [GIF, 32.4 kb, 600 X 369, notice]

When projected on a NW striking line (parallel to the strike of the subducting slab), the delay from SE to NW is clear. However, it is not resolved whether this move-out is step-wise (3 shaded regions), constant ( ~6km/day), or decelerates from SE to NW (curve).

Transient displacement

Figure 7: Displacements larger image [GIF, 57.1 kb, 600 X 479, notice]

Once the timing of the displacements had been established, the net offsets were estimated via least-mean-squares regression of the raw latitude and longitude time series assuming a single step at the mid-point of the transients and constant linear trends. Bold (red) arrows show displacements (with respect to DRAO). Error ellipses are double the 95% confidence limits derived from the formal regression errors of Table 1. Thin (black) arrows show 3 to 6 yr average GPS motions with respect to DRAO (Henton, 2000). The two dashed lines show the nominal downdip limits of the locked and transition zones from the model of Flück et al. (1997).

The key characteristics of the transient displacement are:

1. It shows no correlation with estimated tropospheric parameters;
2. It occurs at a contiguous subset of 7 of the 14 analyzed network sites;
3. Its direction is exactly opposite to longer-term margin deformation;
4. Its amplitude is larger at inner coastal than outer coastal sites and decays abruptly to the NE;
5. It lasts from 6 to 14 days at individual sites; and
6. It displays a delay in a NW direction, parallel to depth contours of the subducting slab.

Subduction interface

Figure 8: Subduction Interface larger image [GIF, 68.7 kb, 750 X 705, notice]

3-D model of slip on the subduction interface. We use a three-dimensional model of faulting in an elastic half-space with the geometry of the subducting slab as compiled by Flueck et al. (1997). The rupture is allowed to propagate along the curved upper surface of the slab and the displacement at each GPS site is calculated by integrating the contribution from every point on the fault rupture surface using the Okada (1985) point-source solution. The relationship of the slip surface to the nominal locked and transition zones are shown in the schematic diagrams above. The downdip limit of the slip zone, kept constant at a depth of ~ 42 km, is relatively sharp, as required by the rapid landward decrease in the observed GPS site displacements, whereas the updip boundary requires a more gradual transition from full slip to no slip as it approaches the transition zone region.

Slip evolution

Figure 9: Slip evolution larger image [GIF, 70.0 kb, 1000 X 247, notice]

The evolution of the slip event is presented to the right in three snapshots. For each time step, the geometry of the slip area and the slip distribution, which are kept as simple as possible, are adjusted to fit the GPS observations. Slip direction is set constant at 235o, the direction of motion of the North America plate with respect to the Juan de Fuca plate. Dark shading indicates the plate interface area with full (2.1 cm) slip; lighter shading indicates area where slip tapers linearly from 2.1 cm to zero updip. Panels, marked by the day of year 1999, show the total area of slip on the interface in 3 time slices and the commensurate evolution of the surface displacement vectors (broad (yellow) = model; thin (blue) with error ellipses = observed). Day 240 is within the time interval of the GPS transient at PGC5 and ALBH, and their observed displacement vectors have been scaled, assuming a linear increase of the displace-ment with time.

The last panel shows the full area of slip. Assuming a rigidity of 40 GPa, the total cumulative moment of the rupture is 1.35 × 1019 N m, equivalent to an earthquake of moment magnitude Mw = 6.7. Also shown are the displacements at WHD1 and RPT1 (green arrows) calculated by simple averaging of positions before and after the slip event. These two USCG CORE sites are not included in our routine GPS network analyses, but their displacements are consistent with the slip model.

Spatial distribution

Figure 10: Spatial distribution larger image [GIF, 86.9 kb, 750 X 599, notice]

An examination of the regional seismicity failed to find any correlation between the slip event and seismic events before, during, or after the slip. The nominal spatial distribution of seismicity is indicated in the above figure which shows the epicentres for both crustal and subducting slab events for the period of April 1995 to March 2000 (orange symbols), as well as the epicentres for large (M > 6) historical crustal events (yellow stars) and JdF (Juan de Fuca) slab events (blue stars).

Temporal distribution

Figure 11: Occurence of earthquakes larger image [GIF, 30.4 kb, 600 X 397, notice]

Figure 12: Juan de Fuca slab earthquakes larger image [GIF, 47.1 kb, 600 X 556, notice]

The temporal distribution of seismicity for the period of June through October, 1999, projected onto a NW-striking line is shown in the figure on the left. Blue circles denote crustal events; orange circles denote JdF slab events. The red symbols denote the occurrence of the slip event as per Figure 6. During the period of the slip event, background seismicity was typical for both the overriding and the down-going plates, totaling about 80 regional events, none of which exceeded Ms = 3.0. As shown in the figure on the right, a moderate JdF slab earthquake occurred 103 km WSW of Seattle 45 days before the slip event (Satsup earthquake of Jul. 2, 1999; Ms = 5.5; depth ~41 km) and the long time separation makes a direct causal connection unlikely.

In the absence of any seismic trigger, we conclude that the slip event was associated with a sudden change in rheology or friction on the deep slab interface.

Figure 13: Latitude and longitude larger image [GIF, 60.6 kb, 550 X 682, notice]

It is possible that slip events have occurred in the past but have not been resolved due to the limited number of sites and the greater day-to-day scatter (2 to 4 mm) in the horizontal components. In retrospect, a similar transient event occurred at ALBH between Oct. 1 to Oct. 15, 1994, with an esti-mated displacement of 5.2 + 0.8 mm in a direction of N101.9oW + 6.4o. In response to this event, a local stability survey was carried out. The survey showed the monument to be stable at the 1mm level and the change remained unexplained. This isolated incident could not be corroborated in any way, but the similarity of the displacement to the more recent event points to aseismic slip as a likely cause.

Added deformation from 'Impeded' slip

Figure 14: 'Impeded' slip larger image [GIF, 41.0 kb, 550 X 766, notice]

If slip on the deep interface occurs as a result of a sudden change in rheology or friction, there must be periods of time (possibly one to two years) when greater coupling of plates across the deeper subduction interface leads to stress accumulation in a region previously thought to be slipping freely. This transient stress accumulation should augment the crustal deformation velocity due to the locked thrust zone located offshore, especially at sites overlying the region of "impeded" slip. The time series at ALBH show this to be the case. If linear trends in the horizontal components are estimated without removing the displacement due to the slip event (Panel A), the resultant deformation velocity is more consistent with the longer-term (6 yr) averages shown in Figure 2. The velocity based on trend estimates on either side of the slip (Panel B) is significantly higher, consistent with contributions from coupling across the deeper interface (and a current build-up to another slip event).

Stress loading from deep-interface slip

Figure 15: Stress loading larger image [GIF, 84.7 kb, 550 X 766, notice]

Episodic deep-slip events result in a transfer of stress to adjacent regions in discrete pulses. For the shallow-dipping megathrust zone located updip from the slip zone, each slip event brings the locked zone closer to failure (Panel C1). It is conceivable that one of these events may keep propagating updip and evolve into a trigger mechanism for a great subduction thrust earthquake. The process of deep slip leading to a thrust earthquake is considered responsible for the 1960 (Mw = 9.5) Chilean earthquake (Linde and Silver, 1989)and the 1944/46 (both Mw = 8.2) Nankai Trough earthquakes (Linde et al. 1998).

For the in-slab, normal earthquakes occurring downdip from the slip zone, each slip event contributes to a stress shadow with reduced Coulomb stress (Panel C2). The fact that the largest in-slab earthquakes occur to the south and south-east of the slip zone (Figure 9, Panel 3) leads to the speculation of possible anti-correlation, i.e. the absence of episodic slip to the south of Seattle enhances the likelihood of larger in-slab events in this region.

• Improved spatial coverage & improved GPS data analysis have resolved transient displacements at 7 contiguous continuous-GPS sites along the northern Cascadia margin that constitute a brief, almost exact reversal of the longer-term crustal motions.
• The displacements are largest at sites on the inner margin and the time of occurrence of the transient is sys-tematically delayed in a NW direction.
• The transient surface displacements are best modeled by ~2cm slip on the deeper subduction interface that begins just south of Seattle and propagates NW, parallel to the subducting slab depth contours. Slip occurs over an area of about 50 x 300 km on the subduction interface between the 25 and 42 km depths. Cumulative moment is equivalent to a M=6.7 earthquake.
• The absence of seismic triggers, the evidence for a previous slip event, and the modulation of longer-term deformation velocities suggests this deep slip behavior is episodic and triggered by rheological instabilities.
• Impacts of this newly observed dynamic behavior on the deeper Cascadia subduction interface are significant:
  • it must be taken into account in "budgeting" observed velocities;
  • it may provide a physical basis for "transition zone" behavior;
  • it loads the updip locked thrust zone and could trigger rupture;
  • it decreases tension downdip within the subducting slab;
  • it may lead to real-time monitoring of the seismic potential of the subduction megathrust.

G. Beutler et al., Bernese GPS Software Version 4.2 (Astronomical Inst., University of Berne, 2000).

H. Dragert, X. Chen, J. Kouba, Geomatica, 49, 301, (1995).

P. Flueck, R.D. Hyndman, K. Wang, J. Geophys. Res. 102, 20539 (1997).

J. A. Henton, thesis, University of Victoria, British Columbia, (2000).

A.T. Linde, P.G. Silver, Geophys. Res. Lett. 16, 1305 (1989).

A.T. Linde, I.S. Sacks, M.T. Gladwin, M.J.S. Johnston, P.G. Silver, Eos Trans. AGU 79, F600, (1998).

Y. Okada, Bull. Seismol. Soc. Am., 75, 1135 (1985).

We thank M. Schmidt and Y. Lu for their support in GPS network operations, and the University of Washington and Central Washington University for providing GPS data for the PANGA sites. USGS NEHRP research grant 00HQGR0061.