SNARF Working Group - Report of the First SNARF Workshop
Summary Report on Intraplate Deformation, Eric Calais
Several lines of evidence suggest that both the central and eastern United States are part of the stable interior of the North American (NOAM) plate. Geodetic studies by Argus and Gordon (1996) and Dixon et al. (1996) that employ velocities from 9-20 sites in the NOAM plate interior suggest a 2 mm/yr upper bound on non-rigid behavior in these and other parts of the stable NOAM plate. A more recent study by Sella et al. (2002) based on 65 continuous GPS (CGPS) sites in the plate interior suggests a similar upper bound (1.7 mm/yr at the 95% level). Assuming these are accurate, geodetic observations presented in these regional scale studies are consistent with overall deformation ranging from 0-2 mm/yr. The general absence of active surface tectonic features in the stable interior argues against the uppermost rate, which would give rise to integrated deformation of 2 km over a 1 Myr period. However, a recent study by Gan and Prescott (2001) suggests residual rates of 1.7-0.9 mm/yr at several permanent GPS stations in the Mississippi embayment with respect to the NOAM plate. Their results translate into a significant rate of seismic moment accumulation rate in the New Madrid area, where earthquake risk is assumed to be high (Frankel et al., 1996). In addition to possible tectonic deformation, Glacia Isostatic Adjustment (GIA) may significantly contribute to intraplate deformation in stable North America (e.g., Peltier, 1995; 1996; 1998).
Intraplate deformation in stable North America and our ability to measure it with GPS have significant implications on the definition of a reference frame for mapping velocities in the western US. This question was approached during the workshop in terms of direct GPS measurements of plate rigidity (presentations by G. Sella and by E. Calais) and in terms of GIA models (presentation by M. Tamisiea).
Calais has processed CGPS data from the CORS network (~300 sites for the central and eastern US) for the 1996-2003 period using the GAMIT software. He has combined his results with independent solutions computed using GIPSY-PP at Univ. Wisconsin (C. DeMets), with the IGS combined solution, and with the ITRF2000 solution. He argues that the combination minimizes systematic errors associated with each individual processing strategy and therefore ensures a higher accuracy. Also, sites shared by several solutions can serve to cross-check the solutions and detect outliers. He is using the combination model of by Altamimi et al. (2002), implemented in the CATREF software. After removing a NOAM rigid rotation, he finds an average residual velocity of 0.2-0.9 mm/yr (1-sigma) for GPS sites in the central and eastern US.
None of the residual velocities are significant at the 95% confidence level, except for a few sites where anomalously high velocities have been traced back to poor geodetic monument design or location (fence posts, unconsolidated sediments, etc.). Residual velocities show a random pattern, except for the northeastern part of the US, where they seem to be systematically be pointing south (Michigan) to southeast (Ohio) to east (east coast). This may be a signature of GIA. But again, these velocities are less than 2 mm/yr and are not significant at the 95% confidence level. Calais showed that the choice of stations used to define the rigid North American plate has little influence on its rotation parameters and on site velocity estimates in the western US (less than 0.1 mm/yr). He also showed simulations where a GIA signal is added to north American site velocities in ITRF. This theoretical data set is then used to invert for a North America angular velocity. In that case, depending on the GIA model used, rotation parameters for North America can be significantly affected However, the impact on velocities in the western US remains small, less than 0.5 mm/yr. Calais insists in his conclusion on (1) the need to combine a larger number of independent GPS solutions to minimize systematic errors, (2) the importance of a careful and rigorous site selection procedure to define stable North America sites. He mentions that the relationship between geodetic monuments and local geologic setting and residual velocities should be carefully investigated. He raises the issue of the need for a non-linear time-depended frame definition to investigate transient motions in actively deforming areas in the western US and Alaska.
Sella has processed CGPS data from the CORS network in the central and eastern US, as well as campaign data from the Canadian Survey, using GIPSY-PP. He presented a strategy for choosing sites suitable for defining ``Stable North America'' based on a priori geologic criteria. He uses sites that are (1) located east of the Rockies and north of the central Gulf of Mexico, (2) not located in the Memphis or Charleston area because of large historical earthquakes, (3) located more that 1800 km from Hudson bay to avoid GIA effects. This leaves 83 suitable CGPS sites. Assuming a GPS noise model that accounts for white and colored noise, the reduced chi-squared of the fit of a rigid plate model to this 830-site subset is 1.08. The chi-squared is higher (1.33) when sites within 1800 km of Hudson Bay are used in the NOAM/ITRF angular velocity estimation. He finds a random pattern of residual velocities in the central and eastern US, and systematic patterns in Canada, probably reflecting GIA effects. He concludes that GIA significantly contributes to horizontal and vertical deformation within ~1800 km from Hudson Bay, but that the central and eastern US define a rigid block at the precision level of his GPS velocities. He argues that GIA models show large variations in predicted 3D velocities depending on the loading history and lower mantle viscosity used. Therefore, GPS results should first be used to help better constrain GIA models. Once better models are available, they can be used to remove the GIA contribution from GPS observables and hopefully improve our ability to derive a stable North America reference frame.
Tamisiea presents an overview of GIA models. He shows that, although the elastic earth model is well known (PREM), the viscosity model is vigorously debated, implying a wide range of surface deformation prediction according ot the (lower) mantle viscosity used. In addition, ice models are also debated and it is now recognized that the history of ice melting is complex in time and space, which also has significant implications on GIA-induced surface deformation. He explains that GIA models use the sea level equation, applying mass conservation and ocean and ice loading to determine surface deformations in 3 dimensions. Recently, a number of refinements have been added to teh sea level theory such as variable coastlines, water dumpling, and feedback from the rotation perturbation induced by the ice melting. However, there does not seem to be a ``best model'' at this point, in part because everyone uses a different data set and works in a different region to validate their models.
Tamisiea presents a series of models from Mitrovica et al. (GJI, 2001). Radial displacements in North America range from -6 to +8 mm/yr, with two maxima centered on Hudson Bay and westernm Canada. Horizontal displacements range from 2 to 0.5 mm/yr, with a radial pattern centered no north-central Canada. The horizontal velocity gradient is essentially north-south over most of north America. A discussion started on the reference frame used to expressed velocities in GIA models, in particular horizontal velocities. The frame for GIA calculations is chosen so that the degree one potential perturbation of the solid earth is always zero (Farrell, 1972). Vertical velocities are therefore expressed with respect to the Earth center of mass, but we were unclear on how to compare horizontal velocities from GIA models with GPS velocities in ITRF2000, for instance. This point will need to be clarified in coming meetings.
Tamisiea presented a series of GIA simulations from Mitrovica et al. (JGR, 1994) showing the influence of the ice model (ICE-1 vs. ICE-3G), the lithospheric thickness,and the upper and lower mantle viscosities on surface displacements in North America. The lower mantle viscosity has a particularly large effect on the peak amplitude of surface diaplcements (horizontal as well as vertical): the higher the lower mantle viscosity, the larger the north-south GIA velocity gradient in north America.
He concluded that the uncertainties in ice and Earth models are probably too large at this point to explicitely use GIA models in the definiton of a geodetic reference frame for north America, but that GIA models should benefit from GPS measurements of 3D surface displacements in north America.
Last modified: 2019-12-24 02:12:54 America/Denver