The Colorado Plateau (Figure 1) is a distinct tectonic and physiographic province in the southwestern United States characterized by high elevation and long-term tectonic integrity since the Proterozoic. Xenoliths, along with surface outcrops of mid-Proterozoic rocks, indicate that the CP lithosphere was formed in a series of continent building events in which Proterozoic terranes were accreted from northwest to southeast, culminating in the formation of the Rodinia supercontinent at about 1Ga [e.g., Karlstrom et al., 1999]. These accreted terranes consist of island and continental arcs, oceanic plateaus, and marginal basin units [Condie and Selverstone, 1999], and remained intact until the western margin of Rodinia rifted away at about 650 Ma [Karlstrom et al., 1999]. Although the CP did experience episodes of significant tectonism over the next 500 Ma, such as faulting related to the Ancestral Rockies formation, marine Paleozoic and Mesozoic units deposited on the CP indicate that it was relatively stable and at or near sea level during this time [Spencer, 1996].
 After this period of relative stability and deposition, the northwestern, western, and southwestern edges of the CP experienced intense Sevier orogeny thrust faulting beginning in the latest Jurassic or earliest Cretaceous (∼150 Ma), reaching a peak during the Late Cretaceous (∼75 Ma) [Willis, 1999]. The culmination of thrust faulting resulted in upper crustal thickening of ∼16 km, producing crustal thickness in excess of 50 km with elevations most likely greater than 3 km. Much of the western margin of the CP at this time may have been a high-elevation plateau with a rugged topographic front similar to the central Andean orogen [DeCelles and Coogan, 2006].
 During the Laramide (80–50 Ma) orogeny the region underwent further compression due to Farallon slab subduction along the North American western margin, with crustal deformation characterized by fault-bounded, basement-cored uplifts and an estimated mean lateral contraction of 5% [Spencer, 1996]. During this period the Farallon slab is widely interpreted to have subducted at a low angle [e.g., Dickinson and Snyder, 1978]. Between approximately 43 to 20 Ma, the slab likely detached, with the accompanying mantle upwelling being the controlling factor for the space-time evolution of extensive mid-Tertiary igneous activity across much of southwestern North America [Humphreys, 1995]. However, the central CP experienced relatively low levels of mid-Tertiary magmatism, and CP mantle xenoliths from this period suggest that at 30–20 Ma the Colorado Plateau had a cool root, somewhat analogous to those beneath Archean cratons, that extended to depths of up to 140 km [Riter and Smith, 1996] and had mid-Proterozoic U-Pb zircon ages [Smith et al., 2004]. The root persists today, as indicated by recent body wave [Sine et al., 2008] and surface wave inversion results [West et al., 2004] showing high velocity lithosphere beneath the Colorado Plateau to 140–160 km depths. This evidence of present day thick lithosphere of probable Proterozoic age, and lack of significant CP deformation [Spencer, 1996], rule out substantial Laramide-age mechanical erosion of the lithosphere by a shallowly subducting Farallon slab.
 Over the last 30 Ma, as the western U.S switched from a contractional to tensional tectonic regime, the formerly thickened crust surrounding the CP margins has drastically extended and thinned [Coogan and DeCelles, 1996]. At the eastern margin of the CP, extension led to the creation of the Rio Grande Rift with initial stages of rifting commencing at 30–35 Ma [Olsen et al., 1987]. It was at this time that the CP began to evolve as a distinct tectonic province; structurally differentiated from the surrounding region [Rowley et al., 1978], and began to assume its modern physiographic character as a high plateau while significant portions of the western U.S. extended and collapsed around its margins in what now is the Basin and Range and Rio Grande Rift.
 We examined the present lithospheric structure of the CP by analyzing P-to-S scattered energy, immediately following (several tens of seconds) the teleseismic (epicentral distances of 30–90 degrees) P-wave arrival. We use receiver function analysis which emphasizes P-to-S converted arrivals via signal deconvolution between horizontal and vertical component seismograms [e.g., Langston, 1977]. We applied the receiver function imaging methodology of Wilson and Aster  that uses both forward- and back-scattered P-to-S converted energy to produce a continuous seismic crust and upper mantle cross-section.
 Data were collected during the Colorado PLateau/Rio Grande RIft/Great Plains Seismic TRAnsect (LA RISTRA) experiments (Figure 1). RISTRA consisted of two stages, the first (1999–2001) occupied a 950 km near-great-circle transect with endpoints near Lake Powell, UT and Pecos, TX, and the second (2004–2006) occupied a 350 km continuation segment extending from Lake Powell, UT northwest to more than 100 km into the Great Basin (Figure 1). The overall northwest-to-southeast transect was oriented parallel to the azimuth of the very active South American and western Pacific seismogenic source zones. RISTRA instrumentation was unusually dense for a long transect (mean station spacing of around 19 km), which enables us to create a uniquely high resolution image of crustal-scale structure. Combining these two data collection phases, we apply a new receiver function velocity analysis technique, and construct an image of the crust and upper mantle seismic structure that is the first complete high-resolution seismic transect of the CP.
 Measuring the relative arrival times of P-to-S converted arrivals in receiver functions allows for the estimation of local depth and average Vp/Vs ratio between the surface and a strong impedance contrast such as the base of the crust. This is typically done by stacking receiver function amplitudes of direct and reverberated receiver function modes at a single station over a range of possible crustal thicknesses (H) and Vp/Vs ratios [Zhu and Kanamori, 2000]. This represents a mapping of these amplitudes to the H- Vp/Vs plane where the stacked trace amplitude will produce a local maximum at the best solution where the different modes add constructively. A major assumption in this method is that the velocity structure is composed of locally flat layers.
 In this study we use a novel variation on this receiver function velocity analysis technique. First, we create a multimode receiver function image [Wilson and Aster, 2005] using a reference P wave velocity model [e.g., Roller, 1965] and a starting Poisson solid Vp/Vs ratio of 1.73. For this starting model, we allow the lower crustal velocities to extend into the upper mantle, then apparent Moho depths can be picked from the local maxima on each of the three multimode images produced (Figure S1 of the auxiliary material). Using the initial velocity model, and these newly picked apparent Moho depths, we can calculate the actual crustal vertical P and S wave transit times (ts and tp) and the true crustal Vp/Vs ratio (Vp/Vs = ts/tp) at each imaging point (Figure 2). Finally, the images are recomputed using this new velocity model.
 By mapping receiver function amplitudes from all stations to 3-dimensional space (via multimode image processing) prior to velocity analysis, we reduce dipping layer effects that are a major source of error in traditional receiver function velocity analysis. Because the Vp/Vs estimates are based on picking peak amplitudes we can estimate the error based on wavelength picking error which translates into a maximum error in Vp/Vs ratio of about 0.035. Rather than producing a single 1-dimensional velocity model estimate beneath each station, we produce a velocity model estimate at each imaging point, thereby creating a 2-dimensional velocity model from which to compute a final receiver function image.
 The resulting multimode receiver function image is shown in Figure 2 along with the associated Vp/Vs analysis results and the transect elevation profile (the individual mode migrations are shown in the auxiliary material). Positive receiver function amplitudes (red) indicate a discontinuity that is high velocity and/or seismic impedance below and low velocity and/or impedance above the discontinuity.
 The most prominent discontinuity is the base of the crust (Moho) seen near 30 km depth in the Great Basin, thickening to 45–50 km thick beneath the central CP, then thinning again to 35 km in the Rio Grande rift. For the central CP, crustal Vp/Vs ratios average 1.71 to the northwest and 1.80 to the southeast of the Yavapai/Southern Yavapai province boundary. Crustal xenolith work near this boundary [Condie and Selverstone, 1999; Selverstone et al., 1999] indicates that the Yavapai terrane to the northwest was a continental land mass that the Southern Yavapai (composed of a continental-margin arc) was subducted beneath. Higher Vp/Vs ratios to the southeast of the boundary indicates that the Southern Yavapai crust may on average be slightly more mafic which is consistent with the lower concentrations of incompatible elements found to the southeast of this boundary [Condie and Selverstone, 1999].
 Another feature of central CP receiver function images is a weak Moho receiver function signal (Figure 2) indicating a reduced impedance contrast across the Moho in the central CP. It is unlikely that high impedance lower crust or low impedance just below the Moho are the result of Laramide-to-present tectonics because of the lack of significant deformation [Spencer, 1996] and lack of penetration of magmatism since the Laramide [Condie and Selverstone, 1999; Wendlandt et al., 1993]. A large number of lower crustal xenoliths from the central CP are Proterozoic, high-impedance eclogites [Wendlandt et al., 1993] which could act to produce a gradational Moho. This indicates that this low Moho impedance contrast is an ancient feature, and has not contributed to CP uplift since the Late Cretaceous.
 Thick, low-density crust can provide lithospheric buoyancy resulting in high elevations. If the current average central CP crustal thickness (46.7 km from Figure 2) reflects an estimated 5% of contraction during the Sevier-Laramide orogenies [Spencer, 1996] then the pre-Laramide crustal thickness would have been 44.3 km. This 2.4 km of increased crustal thickness would result in an isostatic elevation gain of only 300 m relative to pre-Laramide elevations (assuming crustal density of 2.76 g/cm3, mantle lithospheric density 3.36 g/cm3, and asthenospheric density of 3.18 g/cm3).
 Spencer  found that differing rates of erosion and sedimentation on the CP compared to the mid-continent could account for on average 650m of CP uplift. Since the average central CP elevation is 1940 m along our transect, removing the 300m and 650m buoyant contribution from crustal thickening and erosion/sedimentation respectively leaves 990 m of unexplained elevation gain in the central CP which must be supported by increased mantle buoyancy relative to the pre-Laramide CP.
 The buoyancy structure of the CP lithosphere was likely modified by metasomatism arising from the introduction of Farallon slab volatiles [Smith et al., 2004; Lee, 2005]. Geochemical analysis of the mafic, ultra-potassic rocks of the Navajo volcanic field in the central CP indicate that the magma source region was deep (140 to 200 km) and experienced metasomatism prior to eruption [Laughlin et al., 1986]. Metasomatism is a positive volume change process and could thus have provided additional buoyancy for the CP [Laughlin et al., 1986].
 Mantle upwelling following the Laramide slab detachment and sinking would have introduced more heat at the base of the CP lithosphere, and through thermal expansion due to conductive and advective heat transfer into the CP lithosphere, provided additional buoyancy to the CP [Roy et al., 2009]. This upwelling may also provide added elevation via dynamic support. Moucha et al.  calculated the viscous flow beneath North America as inferred from joint seismic-geodynamic modelling, and found that significant dynamic topography may be provided by mantle upwelling beneath the western U.S including the CP.
 During Farallon plate subduction, some erosion of the CP lithosphere may have also taken place [e.g., Spencer, 1996]. However, a regional earthquake study [Beghoul and Barazangi, 1989] found high Pn velocities beneath the CP. This argues against significant CP lithosphere thinning which would yield a much lower Pn velocity (7.8 to 7.9 km/s).
 Such arguments make a strong case for post-Laramide sub-crustal buoyancy changes. However the relative contributions from each of the processes above are poorly constrained. Some constraints are provided by analysis of CP geoid data [Chase et al., 2002] which show that the CP exhibits a relatively small geoid anomaly. This argues against the primary buoyancy changes (relative to pre-Laramide) being at the base of the lithosphere or deeper, which would predict a geoid anomaly that is too high. We thus support the contention [Roy et al., 2009] that the majority of remaining 990m of central CP elevation is from thermal expansion of the CP lithospheric root due to conductive heat transfer to relatively shallow levels. This, along with secondary contributions from dynamic support, metasomatism and minor thinning of the CP lithospheric root can easily balance the remaining central CP elevation budget (Table 1).
|GB/CP Transition||Central CP||CP/RGR Transition|
|Mean crustal thickness||39.2||46.7 km||41.4|
|Mean elevation||2180 m||1940 m||1920|
|Elevation from crustal thickening (I)||−570 m||300 m||−230 m|
|Elevation from erosion and sedimentation [Spencer, 1996] (IIa)||650 m||650 m||650 m|
|Elevation from lithospheric thermal expansion and metasomatism and dynamic support from mantle upwelling (IIb)||990 m||990 m||990 m|
|Elevation from small-scale edge convection (III)||1110 m||0 m||510 m|
 While this combination of buoyant sources explains central CP elevations well, additional support is required at the CP margins. Average crustal thickness at the CP margins (transition zones (Figure 2)) is 39.2 km at the GB/CP edge and 41.4 km at the CP/RGR edge. Using the crust and mantle densities considered above, the thinner crust at the CP margins would predict edge elevations 870 m and 530 m below that of the central CP for the GB/CP and CP/RGR margins respectively (Table 1). However, quite dramatically, the opposite is true; the average central CP elevation is 1940 m while the average elevations at the GB/CP and CP/RGR margins are 2170 m and 1920 m respectively. This indicates that the mantle provides even greater support to high elevations at the CP margins than at the center of the plateau.
 We modelled the elevations predicted from the crustal profile (Figure 2) using an isostatic flexural response for the lithosphere (using an elastic thickness of 25 km). The starting reference lithospheric density model is derived from the densities given above and a 44.3 km reference crustal thickness (see above). Results (Figure 2) show that the high elevations found at the CP margins cannot be supported by the flexural strength of the lithosphere and requires additional sub-crustal buoyancy.
 Additional CP margin buoyancy may be explained by small-scale convection due to extension-driven asthenospheric upwelling around the CP edges [Sine et al., 2008; van Wijk et al., 2008]. Edge convection may also result in lithospheric erosion and dissociation at the CP margins [van Wijk et al., 2010]. This upwelling at the CP margins would also facilitate conductive and advective heat transfer into the CP lithosphere [Roy et al., 2009] creating additional buoyancy. The amount of additional buoyant uplift that this edge convection may provide (after subtracting other sources) is up to 1110 m on the GB/CP margin and 510 m on the CP/RGR margin (Table 1). Strong negative Bouger gravity anomalies are present on the CP margins and have been modelled [Roy et al., 2005] on the eastern margin as low-density anomalies on the order of −30 to −50 kg/m3 from the base of the crust to 100 km depth. Density anomalies of this magnitude, most likely associated with edge convection and conductive heat transfer, when integrated over tens of km can easily provide the remaining buoyant uplift needed to explain the topography at the CP margins.
 Similarly, sub-crustal high density anomalies in the central CP [Sine et al., 2008] could be responsible for the lower than predicted elevations where the imaging transect crosses the Colorado River (near −110.5 longitude). We have included these buoyant effects in our modelling (Figure 2) using the velocity models of Sine et al.  and velocity/density derivatives of Cammarano et al. . Flow caused by these anomalies [van Wijk et al., 2010] would further modify the predicted elevations, providing additional Great basin support, and down welling in the central CP.
 The lithospheric architecture of the CP is characterized by regionally high and highly variable crustal thickness and Moho impedance contrast, for which we have obtained new measurements. These results indicate that the crust provides only a small (<20%) component of isostatic CP compensation. Prior to the Laramide, thick crust and low density uppermost lithospheric mantle were balanced by a thick, dense Proterozoic lithospheric root that kept the CP near sea level. Laramide/post-Laramide tectonics altered this root through a combination of metasomatism, post-Laramide thermal conduction and expansion, and perhaps minor tectonic erosion. This, coupled with dynamic topography due to mantle upwelling, resulted in buoyant CP uplift to the present high elevations. These buoyant effects are complemented significantly by edge effects where CP lithosphere is undergoing mechanical forcing and tectonic erosion arising from small-scale convection in the asthenospheric mantle of the adjoining Great Basin and Rio Grande Rift.