Measuring crustal motions in coastal British Columbia with continuous GPS
H. Dragert , M. Schmidt , J. Henton , Y. Lu
This article describes the Western Canada Deformation Array (WCDA), a network of automated continuous Global Positioning System (GPS) stations located in southwestern British Columbia. Started by the Geological Survey of Canada in 1991, the network has gradually expanded to serve as the northern portion of the Pacific Northwest Geodetic Array (PANGA). The objectives of the WCDA are to 1) provide high-quality GPS data for global geodynamic studies; 2) provide a precise, common reference frame for all deformation surveys carried out in this active seismic region; 3) serve as a strainmeter to map regional strain and monitor possible transient strain signals. This article describes how continuous GPS data is collected, verified, and archived, and how that data is used to identify transient signals, and estimate relative plate velocities with a precision of better than 1mm/yr. The data are also used to constrain tectonic dislocation models and plate trajectories, invaluable contributions to current studies of regional crustal dynamics.
The Western Canada Deformation Array (WCDA) consists of 8 continuous GPS tracking stations located in southwestern British Columbia. The purpose of the array is to monitor crustal motions in this most active seismic region of Canada and thus provide constraints for the modelling of plate interactions along the northern Cascadia margin and help in the estimation of current seismic hazard.
The first station, DRAO, was established in Feb. 1991, and one of the latest additions was WSLR which began continuous operation in Sept. 1996 (a new Site CHWK came online November, 1998). All sites use TurboRogue receivers with Dorne-Margolin choke-ring antennas.
The GPS antennas are mounted on forced-centre monuments embedded in concrete piers which are anchored into bedrock. Details of the antenna set-up are shown in the schematic diagram of the pier.
See also full description of the WCDA as well as all continuous GPS data
30-sec sample GPS data are collected automatically every 4 hours from all sites and binned into 24-hour data sets. Using precise IGS orbits and the station DRAO as a fixed reference site, daily solutions of relative positions of all network sites are computed at 2 min. epochs using the CGPS22 software developed by Jan Kouba, Geomatics Canada. This software employs double-difference, L3-phase solutions with orbits held fixed and ambiguities not fixed. One of the strengths of this analysis software is the use of interactive graphics to examine and edit the L3-phase data.
The changes in latitude, longitude, and height at the various sites relative to DRAO exhibit day-to-day variations with characteristic sigmas of 2 to 3 mm, 3 to 4 mm, and 6 to 9 mm respectively, dependent primarily on baseline length. The monthly scatter in each of the components (including baseline length) are shown for two stations (HOLB and WILL) for the most recent three years. These values provide a realistic estimate of the accuracy achievable for baselines measuring hundreds of kilometres.
Apparent in the plots of the daily solutions are distinct sudden offsets and periodic variations. Since longer period variations and step-functions can significantly bias linear trends, especially for data lengths <2 yrs, these effects were removed through regression techniques. It was assumed that the longer-period sinusoidal variations were predominantly annual, and step-functions were allowed for when antenna set-up changes or reference frame changes were known to have occurred. Note that the estimates of the step magnitudes are analysis dependent and cannot be calibrated in an absolute sense.
Using the data for HOLB as an example, the three types of regression parameter fits (i.e. step functions, annual variations , and linear trends) are illustrated for three components. Steps in the horizontal components for all sites were generally less than 3 mm although occasionally as large as 5 mm. Steps in the vertical ranged from 3 mm to as large as 19 mm showing the strong dependency of the vertical position of the antenna phase centre to the near-field EM environment. Annual signals in the horizontal components have amplitudes less than 1.6 mm while the vertical annual signal amplitudes range from 0.5 to 4.3 mm.
Least Squares Spectral Analysis (LSSA) confirms that periodic energy in the vertical component is primarily limited to the annual signal, although at HOLB a significant semi-annual signal is also apparent. The previously reported sharp Mf (period = 13.67 days) spectral peak has diminished and become more diffuse for the entire 6 year data set because over the past 2 years, this spectral component has effectively disappeared. It is significant that little energy is seen for periods exceeding a year indicating minimal, if any, 'random walk' behavior of monuments.
| Date | Amp. (mm) | Error (mm) | Phase (days) | Error (days) |
|---|---|---|---|---|
| 1992-98 | 1.83 | 0.32 | 0.69 | 0.38 |
| 1992 | 6.28 | 2.53 | -1.57 | 0.88 |
| 1993 | 5.56 | 1.07 | -2.02 | 0.42 |
| 1994 | 4.41 | 0.71 | 3.81 | 0.35 |
| 1995 | 2.98 | 0.79 | 1.99 | 0.58 |
| 1996 | 4.71 | 0.76 | 1.01 | 0.35 |
| 1997 | 1.02 | 0.65 | -0.85 | 1.37 |
| 1998 | 0.10 | 0.78 | -5.20 | 16.67 |
| Date | Site | Description of changes |
|---|---|---|
| 94:041 | DRAO | New antenna; New antenna mount; |
| 94:104 | ALBH | New antenna; New antenna mount; |
| 94:124 | HOLB | New antenna; |
| 94:173 | WILL | Added acrylic dome & mounting ring; |
| 95:011 | ALBH | New antenna mount; |
| 95:103 | DRAO | New antenna mount; |
| 95:158 | ALBH | New antenna; |
| 95:189 | WILL | New antenna mount; Added RF screening skirt; |
| 95:202 | ALBH | Added RF screening skirt; |
| 95:223 | UCLU | Added RF screening skirt; |
| 96:010 | DRAO | Added RF screening skirt; |
| 96:044 | UCLU | New antenna mount; |
| 96:046 | NANO | Added RF screening skirt; |
| 96:089 | HOLB | New antenna; New antenna mount; Added skirt; |
| 97:033 | NEAH | New antenna; New Ashtech cone dome; |
| 97:330 | HOLB | Replaced torn screening skirt; |
In the regression, all parameters (steps, annual signals, and linear trends) are solved for simultaneously in each component. It is of course the linear trends that are interpreted in terms of tectonic motions. The linear trends for the north and east components are combined to produce estimates of horizontal site velocities with respect to DRAO which is assumed fixed on the North American plate. These vectors along with their 95% confidence ellipses are drawn in the map diagram. Also shown are velocities for the longer-running PANGA sites in northwestern Washington (from G. Khazaradze, Univ. Washington) and the velocities predicted by an elastic dislocation model of the locked Cascadia subduction thrust.
Surface crustal deformations are interpreted to be due to strain accumulating across a locked thrust fault at the interface of the subducting Juan de Fuca Plate and the Cascadia margin. This locked fault comprises the seismogenic zone which will generate the next great earthquake. To the first order, observed deformations can be accounted for by simple 2-D and 3-D elastic dislocation models which place the seismogenic zone offshore beneath the continental slope and vary the fault width between 40 to 100 km along the margin.
Refined observations of crustal motions obtained from continuous WCDA data are beginning to reveal some limitations of such simpler models. The most striking example is the margin-parallel motion observed for northern Vancouver Island which cannot be replicated by our current dislocation models. It is also noteworthy that the model calculations have been brought into better agreement with the observed velocities by using an azimuth of convergence of 62 degrees between the Juan de Fuca and the North American plates instead of the nominal 69 degrees.
The Western Canada Deformation Array (WCDA) is a network of automated continuous Global Positioning System (GPS) stations located in southwestern British Columbia. Started by the Geological Survey of Canada in 1991, the network has gradually expanded to the current 8 sites which now also serve as the northern portion of the Pacific Northwest Geodetic Array (PANGA). The initial objectives of this regional array were to 1) provide high-quality GPS data for global geodynamic studies; 2) provide a precise, common reference frame for all deformation surveys carried out in this active seismic region; and 3) serve as a strainmeter to map regional strain and monitor possible transient strain signals. As such, the WCDA was a new tool being adopted in a program of crustal deformation studies that had utilized tide gauge, levelling, precise gravity, and trilateration surveys.
The utility of this new tool has proven itself several times over: The infra-structure is in place and we currently collect, verify, distribute, and archive the continuous GPS data in near-real time with minimal manual intervention and therefore minimal manpower; continuous regional reference stations allow the integration of strain results from repeated campaigns, and horizontal strain tensors can be derived in a common regional framework; relative crustal motions have been resolved with a precision of better than 1 mm/yr within a time span of two to three years and these observed motions are helping to constrain models of the Cascadia Subduction Zone; seasonal and other transient signals have been observed emphasizing the usefulness of continuous GPS coverage; sudden shifts in positions have been found to be caused by near-field effects on the antenna phase-centre location and pointed out the critical importance of antenna set-up.
Although already invaluable to current studies of regional crustal dynamics, the sparseness and limited extent of this array present problems. The small number of continuous sites makes each site critical for the robust estimate of regional strain and great care must be exercised, beginning with the installation of stable monuments through to the elimination of non-tectonic effects in the analysis of data. This sparseness also prevents the resolution of variations in the strain field related to active structures of modest (~100 km) spatial scales and it prevents the clear identification of aseismic signals. With its current geographical coverage, the WCDA network limits deformation monitoring to the north Cascadia margin and leaves unmeasured the crustal motions associated with the extensive seismic activity along the Queen Charlotte Fault and regions further north.