GPS-determination of along-strike variation in Cascadia margin kinematics: Implications for relative plate motion, subduction zone coupling, and permanent deformation M. Meghan Miller1, Daniel J. Johnson1, Charles M. Rubin1, Herb Dragert2, Kelin Wang2, Anthony Qamar3, and Chris Goldfinger4 earthquakes along the Cascadia subduction zone to- gether with frequent, smaller earthquakes within the Abstract. High-precision GPS geodesy in the Pacific Northwest provides the first synoptic view of the North America and downgoing Juan de Fuca plates char- along-strike variation in Cascadia margin kinematics. acterize seismic risk in the Pacific Northwest. The These results constrain interfering deformation fields in kinematics and dynamics of the Cascadia subduction a region where typical earthquake recurrence intervals zone have proved particularly elusive to seismologic are one or more orders of magnitude longer than the studies, owing to relatively low Benioff-Wadati zone decades-long history of seismic monitoring and where seismicity and short historical earthquake records geologic studies are sparse. Interseismic strain accumu- [Heaton and Kanamori, 1984]. lation contributes greatly to GPS station velocities During the last decade, however, new tools in pa- along the coast. After correction for a simple elastic leoseismology have revealed a rich Holocene seismic dislocation model, important residual motions remain, record including evidence for recurring catastrophic especially south of the international border. The mag- coastal subsidence, tsunamis, and ground shaking nitude of northward forearc motion increases southward from western Washington (3-7 mm/yr) to northern and [Atwater, 1987; Adams, 1990; Atwater et al., 1995; central Oregon (~9 mm/yr), consistent with oblique Darienzo and Peterson, 1995; Satake et al., 1996; convergence and geologic constraints on permanent Atwater and Hemphill-Haley, 1997]. Taken together deformation. The margin-parallel strain gradient, con- with evidence for a circum-Pacific tsunami [Satake et centrated in western Washington across the populated al., 1996] and local tree ring histories [Atwater and Puget Lowlands, compares in magnitude to shortening Yamaguchi, 1991; Yamaguchi et al., 1997], these data across the Los Angeles Basin. Thus crustal faulting point to a major earthquake on the Cascadia subduction also contributes to seismic hazard. Farther south in zone late in January, 1700 A.D. On the basis of the southern Oregon, northwestward velocities reflect the size of the tsunami in Japan, the entire length of the influence of Pacific-North America motion and im- Cascadia margin likely broke in a single event of pingement of the Sierra Nevada block on the Pacific Mw~9.0 at that time [Satake et al., 1996]. Whether Northwest. In contrast to previous notions, some de- formation related to the Eastern California shear zone this size of rupture also typifies earlier events has been crosses northernmost California in the vicinity of the considered. The simultaneity of turbidite triggering in Klamath Mountains and feeds out to the Gorda plate tributary and trunk submarine canyons and channels margin. along the length of the Cascadia margin during the Holocene supports the case for large (Mw~9) earth- 1. Introduction quakes that rupture the length of the convergent margin rather than smaller events along a seismically seg- The Cascadia convergent margin, located between mented subduction zone [Adams, 1990; Nelson et al., two migrating triple junctions, deforms in response to 2000]. The turbidite data further suggest recurrence for superimposed forces of North America, Juan de Fuca, such events on average every ~600 years. Nevertheless, and Pacific boundary interactions (Figure 1), including sparse empirical constraints on the interplay between oblique subduction and entrainment of the Sierra Nevada Juan de Fuca-North America convergence and crustal and Oregon Coast Range blocks. Infrequent great faulting within the North America plate have prevented characterization of the processes that drive deforma- 1Department of Geological Sciences, Central Washington tion. University, Ellensburg, Washington. 2Geological Survey of Canada, Pacific Geoscience Center, High-precision geodesy samples the deformation Sidney, British Columbia, Canada. field over timescales of years or decades. Separation of 3Geophysics Program, University of Washington, Seattle, mutually interfering deformation fields from plate Washington. 4College of Oceanic and Atmospheric Sciences, Oregon boundary and crustal faults hinges on our ability to ac- State University, Corvallis, Oregon. curately model or measure processes that independently Tectonics – expected publication April, 2001 contribute to the integrated, instantaneous velocity respect to motion of other stations on stable North field. These processes include (1) convergence between America (Figures 2 and 3). the Juan de Fuca and North American plates (Figure 2 and Table 1) [Riddihough, 1984; Wilson, 1993; De- 2.1. GPS Observations and Uncertainties Mets and Dixon, 1999; this study], (2) elastic and vis- GPS phase data were analyzed using GPS Inferred coelastic strain that accumulates in the forearc of the Positioning System/Orbit Analysis and Simulation subduction zone in response to seismogenic coupling Software (GIPSY/OASIS II), developed by the Jet Pro- between the young and buoyant Juan de Fuca plate and pulsion Laboratory (JPL) for NASA, using the precise the Cascadia forearc [Dragert et al., 1994; Heaton and point-positioning technique and JPL-generated fiducial- Kanamori, 1984], (3) partitioned deformation within free orbit solutions [Zumberge et al., 1997]. Daily so- North America due to oblique convergence, (4) distribu- lutions for individual station position and correspond- tion of primarily northwest directed dextral shear re- ing matrices of the covariance among the three posi- lated to Pacific-North America plate motion that ulti- tion components were determined within the Interna- mately penetrates the continent in southern California tional Terrestrial Reference Frame (ITRF 97) [Boucher and feeds into the Cascadia arc and back arc region et al., 1999] without ambiguity resolution. For this [Pezzopane and Weldon, 1993] and into the forearc study, daily and seasonal irregularities in the global (this study), and (5) impingement of the Basin and reference frame were minimized using a regional stabi- Range on the back arc of southeastern Washington and lization based on subtracting average common-mode its northward transition to back arc contraction. Reso- position variations at six sites (ALBH, CABL, DRAO, lution of contributions from these processes require ob- GOBS, PABH, and QUIN, Plate 1). This technique servations of regional deformation in an internally [Bock et al., 1997; Wdowinski et al., 1997] is particu- consistent, robust, and stable realization of the North larly useful in cases where the length of observations is America reference frame as well as a model that corrects short, and hence the relative effect of uncorrected sea- for the first-order effects of interseismic strain accumu- sonal noise is potentially large. Using GIPSY/OASIS lation. Continuous Global Positioning System (GPS) II, daily positions accumulated over a 2–6 year period results from the Pacific Northwest Geodetic Array yield station velocities by least squares fit to the indi- (PANGA) provide the first synoptic view of Pacific vidual station time series, which take into account co- Northwest deformation along the Cascadia margin and variance among the three position components (Table place first-order constraints on processes that contrib- 2). ute to continental deformation. Both white noise and time-correlated noise charac- These geodetic results bear on convergent margins terize GPS velocity estimates. Time-correlated noise around the world. Careful differentiation of the in- effects include potential monument motion unrelated to terseismic strain that results from subduction zone cou- crustal deformation [Langbein and Johnson, 1997], un- pling and secular deformation on crustal faults illumi- certainty in the satellite orbit parameters, and atmos- nates the processes that drive deformation within plate pheric and local environmental effects [Mao et al., margins and constrains seismic risk. These results con- 1999]. Pure white noise models underestimate uncer- firm that GPS geodesy provides a critical tool for rap- tainty [Johnson and Agnew, 1995]; however, time- idly characterizing Earth deformation processes and correlated noise estimates, to date, largely rely on a constraining seismic potential. priori assumptions about monument behaviors that are not well determined for a wide variety of geodetic 2. Methods and Results monuments in realistic settings. PANGA uses a variety of regionally available GPS stations that include differ- PANGA, a network of continuously operating GPS ent styles of monumentation in a variety of geologic receivers distributed throughout the Pacific Northwest settings; in addition, different atmospheric effects are in the United States and Canada, is deployed and sup- likely across the network that spans 10° in latitude as ported by an international consortium of institutions well as a longitudinal maritime zone and mountain and agencies [Miller et al., 1998]. Within the Western range. Canada Deformation Array [Dragert and Hyndman, We adopted a simple empirical model for estimating 1995], the oldest component of PANGA, deformation the GPS rate error σ