J. A. Henton, H. Dragert, R. McCaffrey, K. Wang and R. D. Hyndman

This article compares results from Continuous GPS stations and campaign GPS surveys and finds that the horizontal velocities identified by each method are consistent. The horizontal deformation is consistent with strain accumulation across the subduction thrust, indicating that the northern Juan de Fuca plate is fully locked. The GPS observations fit an elastic dislocation model to 1st order. No clear deformation patterns, relating to the 1946 or 1918 crustal earthquakes, are resolved. Rates of shortening across the coastal margin from the elastic model are ~1.5 times larger than observed (likely due to the limitations of such models).

Recent densification of repeated GPS campaign measurements across Vancouver Island, British Columbia, is beginning to consolidate the pattern of crustal deformation for the northern Cascadia Subduction Zone (CSZ). Deformation is generally consistent with a fully locked subduction megathrust fault surface and no free-slip zones are resolved. As a result, the northern CSZ margin accumulates the full slip-deficit (i.e. at the rate of Juan de Fuca/North America relative plate convergence) along its length with no apparent abrupt changes in the width of the seismogenic zone. The observed horizontal velocity field fits the elastic dislocation model well to first order and shows no significant components of rigid block motion. The widths of the locked zones off central Vancouver Island and southern Vancouver Island are approximately 55 km and 85 km respectively, assuming equivalent widths for their respective transition zones. Directions of observed principal strain axes are well aligned with the direction of convergence but the magnitude of the observed rate of margin normal shortening is about 1.5 time smaller than predicted by the dislocation models; i.e. observed velocities for the inland GPS sites are larger than predicted by the model. This discrepancy can be minimized by introducing time-dependent viscous deformation within the mantle underlying the CSZ as demonstrated by the viscoelastic model developed by Wang and He. For northern Vancouver Island, the deformation measurements also provide evidence for crustal strain that is not fully accounted for by current elastic models of a locked subduction thrust fault. The northwesterly motion of the continuous GPS site HOLB, the reduced deformation motions in this region, and the small-scale margin parallel extension measured here mark a pronounced change in the character of the accumulating crustal strain in the margin overlying the Explorer Plate. This change in strain rates to the north of the locked subducting Juan de Fuca Plate may play a role in the origin of large crustal earthquakes on central Vancouver Island.

WCDA horizontal velocity field larger image [GIF, 52.0 kb, 600 X 442, notice]

The black arrows show the observed horizontal velocities at GPS sites relative to DRAO (Penticton, BC). The data are for periods of two years (WSLR) to six years (ALBH & HOLB) and were processed with CGPS22. Velocity vectors are shown with their 95% confidence ellipses determined from regression errors. Datum biases (i.e. steps typically associated with physical alterations affecting the antennae at tracker sites or reference frame changes) were removed by permitting step functions to be fit during the linear-regression process. We also simultaneously fit an annual (365.25-day) period sinusoid to the data during regression. Velocities have been corrected for differential North American plate motion between each tracker site and DRAO, the reference site.

GPS campaign sites of southwestern British Columbia larger image [GIF, 45.9 kb, 600 X 442, notice]

The networks comprising these sites are summarized in Table 1. The surveys were carried out by the Geodetic Survey of Canada. The PAL survey of 1999 was done in cooperation with Rob McCaffrey, RPI, supported by NSF funding. Each survey used 4 to 6 Ashtech receivers with each site being occupied a minimum of two days. All site positions were calculated with respect to DRAO by rigorously combining daily solutions into survey epoch averages.

Campaign velocities

The site displacements divided by the time between surveys determines site velocities relative to the reference site DRAO. The rate of differential, rigid-body North American plate motion between individual campaign GPS sites and DRAO is removed. Velocity vectors are plotted with estimated 95% confidence ellipses.

The vector directions are all nearly parallel within the estimated uncertainties and largely agree with the direction of Juan de Fuca Plate convergence. The displacements are consistent with the deformation signal expected from a locked CSZ subduction thrust surface. The largest displacements (~10 mm/yr) occur towards the coastal margin and decrease landward. For the central Vancouver Island region, the effect of deformation associated with the 1918 or 1946 crustal earthquakes is not evident in the data. The poorer CVI solutions (i.e. error ellipses greater by a factor up to ~2) are likely a consequence of GPS broadcast modifications and testing (i.e. anti-spoofing) during the period of the 1992 survey.

Campaign Strains

Principal horizontal strain rates averaged over an entire network are also computed from the observed velocities. The results for each campaign network are given in the inset boxes in the upper-left corner of the velocity field diagrams (error values are 1-σ). The principal shortening-axis rate (negative values indicate shortening) is ₂ which is oriented at an azimuth (clockwise from north) of θ₂ ; ₁ is the principal extension-axis rate. These strain rates are assumed to be uniform over the entire network (i.e., velocity gradients are assumed to be linear). Measurements are not sufficiently dense nor sufficiently precise to verify this assumption.

For the networks of southern Vancouver Island (i.e. Juan de Fuca & Port Alberni) the principal axes of shortening correspond well to the direction of relative Juan-de-Fuca/North-America plate convergence for northern Cascadia. Furthermore, these regions exhibit nearly uni-axial shortening:

{magnitude of ₂} >> {magnitude of ₁}

The region of central Vancouver Island is located near the northern edge of the of the subducting Juan de Fuca plate. This region is more tectonically complex and may be subject to interactions between the Juan de Fuca, North America, and Explorer plates. Additionally, two large crustal earthquakes (M7) have occurred in this area over the past century. For this network, the principal axis of shortening deviates significantly from the N62°E direction of JDF/NA plate convergence averaged for northern Cascadia. There is also increased margin-parallel extension, consistent with the north-northwest motion observed for the continuous GPS site HOLB on northern Vancouver Island. Nonetheless, shortening in the direction of plate convergence remain.

Central Vancouver Island GPS campaign velocity field larger image [GIF, 52.3 kb, 600 X 442, notice]

Port Alberni GPS campaign velocity field larger image [GIF, 46.8 kb, 600 X 442, notice]

Juan de Fuca GPS campaign velocity field larger image [GIF, 49.4 kb, 600 X 442, notice]

Cascadia elastic slip-dislocation model

The interplate surface may be characterized by four regimes. Proceeding landward or downdip, there is first a narrow region of stable sliding inferred to be limited to the frontal thrusts. From the deformation front extending downdip, the plates remain fully locked. Then a transition zone from fully locked to stable sliding occurs. Finally stable sliding resumes at greater depths. Together, the locked and transition zones form the zone of potential seismic rupture. Varying widths of these zones are modelled for comparison with observed surface deformation patterns and rates. The Cascadia model widths are based upon previous horizontal and vertical geodetic data as well as thermal data. The depth contour geometry of the megathrust fault surface was determined from a suite of data including offshore multichannel seismic reflection surveys, seismic refraction studies, Benioff-Wadati seismicity, seismic tomography, and teleseismic receiver waveform analyses.

Cascadia subduction thrust model: three-dimensional view of fault-model elements from the southeast larger image [GIF, 28.5 kb, 600 X 321, notice]

a sample cross-section reflecting realistic geometry for the subducting slab a half-space geometric approximation (i.e. a regional Lambert conformal conic projection) that corrects for surface topography larger image [GIF, 28.1 kb, 600 X 471, notice]

Comparison of model-predicted and observed velocities for GPS sites in southwestern B.C.

For this region the dislocation model predicts motions largely in the direction of plate convergence which is modelled at 42 mm/yr @ N62°E. This agrees well with the observed velocity field for the region. Although the magnitude of the velocity vectors from the model were comparable to the measurements near the seaward coast of Vancouver Island, the rates of the more landward model-predicted vectors are significantly smaller than the observed rates. As a result, the model predicts slightly more shortening should be accommodated across the island than recorded by the observations. However, overall the three-dimensional elastic dislocation model for a fully-locked subduction thrust fits the observations well.

The model predictions for central Vancouver Island also match the CVI GPS-determined velocity field rather well, although the GPS data for this region are quite noisy. The dominant signal observed appears to match the velocity field predicted at the northern terminus of the subducting Juan de Fuca plate.

Model-predicted strain rates

Model predicted strain rates larger image [GIF, 54.6 kb, 600 X 442, notice]

For the sites of the campaign networks, principal horizontal strain rates averaged over an entire network are also computed from the model-predicted velocities. For the regions of Juan de Fuca and Alberni Inlet, the model predicts nearly uni-axial shortening approximately in the direction of relative plate convergence. This agrees well with the principal strain rates calculated from the observed velocity field for these regions. However, the model predicts shortening rates that are approximately 1.5 times larger than measured. This was already evident from the comparison of the model velocity field to the observed velocity field.

For central Vancouver Island, the model also predicts strain rates larger than the observed but, with the increased noise in the data, they cannot be said to differ at the 95% confidence level. Nonetheless, similar to the results for southern Vancouver Island, there is a suggestion that the model-predicted principal rate of shortening is 50% greater than the observed rate. As opposed to the largely uniaxial character for the region of southern Vancouver Island, the three-dimensional nature of the strain field at the northern end of the Juan de Fuca plate system is evident in both the observations and the JDF plate model.

Comparison of viscoelastic model predictions to observed GPS velocities

Viscoelastic model velocities larger image [GIF, 48.5 kb, 600 X 480, notice]

Wang et al. (Earth Planets Space, in press, 2000) have used a viscoelastic finite-element model to calculate deformation velocities across the Cascadia margin. Their model adopts the geometry of the slip-dislocation model shown in Figure 6 and adds a viscoelastic substrate and a thin viscoelastic layer on the subduction interface downdip of the locked zone, both with a viscosity of 1019 Pa·s. Predicted deformation velocities from this model are in better agreement with the observed and better replicate the reduced shortening rates observed across the margin.

• Continuous GPS & campaign GPS horizontal velocities are consistent

• Horizontal deformation is consistent with strain accumulation across the subduction thrust:
  • the northern Juan de Fuca plate is fully locked
  • observations fit an elastic dislocation model to 1st order

• No clear deformation patterns relating to the 1946 or 1918 crustal earthquakes are resolved.

• Rates of shortening across the coastal margin from the elastic model are ~1.5 times larger than observed (likely due to limitations of such models)

The authors wish to thank Mike Schmidt and Yuan Lu for their important roles in data acquisition, validation, and computer support. Thanks to Alex Smith for his good-natured help and for his role in collecting additional data at Gabriola Island. Thanks to the Geodetic Survey of Canada for collecting the GPS campaign network data. Finally, many thanks to Richard Franklin for producing this poster. Many of the figures used in this poster were generated using the Generic Mapping Tools (GMT) of Wessel and Smith [1995].