Dynamic topography during flat-slab subduction: A first approach in the south-Central Andes

DÁVILA, F.M., LITHGOW-BERTELLONI, C

Niza

Simposio; International Symposium of Andean Geodynamics; 2008

In the flat-slab segment of the south-Central Andes, recent stratigraphic reconstructions have proposed that long-wavelength and high-amplitude accommodation spaces have controlled the alluvial sedimentation in the foreland during Miocene to Present (Dávila et al., 2005, 2007). Given that the tectonic and sedimentary loads do not yield the large magnitude of these spaces by conventional flexural models (Dávila et al., 2005), alternative explanations are needed. Eastward of the Andean foreland system, within the pericratonic Pampean Plains (to >700 km from the Chilean trench, where the slab descends again at high angles; Booker et al., 2005), Cenozoic subsurface sequences show thicknesses of >0.5 km. But this region is even more distant from the High Andes loads. The closest topography is the modest Sierras de Cordoba range (easternmost range of the Sierras Pampeanas), which only records a shortening <5%, i.e. ~5 km within the ~100-km orogen wide. This undersized tectonic load can explain <<50% of the accommodation spaces in the Plains (see Dávila, 2008). Dávila et al. (2005, 2007) suggested “hidden sub-lithospheric loads” (like dynamic topography, lithospheric mantle densification or eclogitization of the lower crust) might overlap the tectonic and sedimentary loads and also explain the recorded load deficits. These mechanisms, likewise, would provide reconciliation with other geophysical and geomorphological features of the foreland (see Dávila et al., 2005). But a question arises: which of these controls occur in the Andes and what triggered it? With the onset of subduction, Earth’s surface deforms by vertical stresses induced by mantle flow. This deformation is called “dynamic topography” because the buoyancy forces driving the surface deflections are actively moving. Thus, whenever subduction occurs, dynamic topography will be present in a foreland. However, at >700 km from the trench, a second question arises: what is the magnitude and importance of the dynamic topography? We test and quantify the degree to which dynamic topography explains how much of the total subsidence (negative vertical deflection) in the south-Central Andean foreland is controlled by non-isostatic loads originated in the astenospheric mantle during flat subduction. We specifically focus on dynamic topography during flat subduction, as it seems clear that the synorogenic accommodations began in the Pampean regions in the last 10-7 my, coevally with the arrival of the Juan Fernandez Ridge to these latitudes, coincident with the initiation of slab flattening and basement thrusting in the Sierras Pampeanas (or broken foreland). We compare predicted values of dynamic topography with maximum deflections and maximum accommodation spaces estimated by flexural analysis and stratigraphic approaches, respectively. We calculated dynamic topography based on mantle flow models that assume a three-dimensional density distribution inside the Earth and solve the viscous flow induced by such density heterogeneity. The model is based on the history of 120 m.y. of subduction (Ricard et al., 1993) and assumes that cold subducted slabs are the main source of thermal buoyancy in the mantle, and therefore of mantle density heterogeneity. This is a reasonable approximation for a mantle largely heated internally, although it neglects the role of active upwellings. The model calculates predicted geoids, which can be compared with observed geoids. Both geoid and dynamic topography are sensitive to the mantle’s convective pattern and therefore to the mantle’s viscosity structure. But, the model assumes subvertical subduction from the trenches. Therefore, we had to modify the slab geometry in our region to simulate flat subduction from the trench to ~64° WL, where the slab submerges again vertically (Booker et al. 2005). As expected, the wavelength of dynamic topography is long and smooth at low (10-30) spherical harmonic degrees (Fig. 1a). However, to analyze a 1000 - 500 km length region, higher degrees (>30) were required. Dynamic topography in these cases is more complex and tends to adjust to the geometry of the subducting slab (Fig. 1b). According to previous calculations (see Lithgow-Bertelloni and Richards, 1998 and references therein), a lithosphere that is 10 times more viscous than the upper mantle and a lower mantle that is 50 times more viscous than the upper mantle can account for >80% of the present-day geoid. Testing different viscosity contrasts (lithospheric mantle / upper mantle / transitional zone / lower mantle), good correlations between predicted geoids and observed geoids (coefficient correlations of 0.65-0.90) were obtained. For a lithospheric mantle more viscous than the lower mantle (1000 / 1 / 1 / 150) dynamic topography above the subducting cells is -731.2 m. Keeping similar viscosity constrasts for the lithosphere and lower mantle (50 / 1 / 1 / 50), dynamic topography above the subducting cells is -1135.9 m. When lower mantle is much more viscous than lithosphere (1 / 1 / 1 / 50), dynamic topographic is  -580.2 m. During flat subduction and with variations in viscosity in the mantle wedge, dynamic topography increases and varies on much shorter wavelengths (Burgess et al., 1997, Billen et al. 2001) The accommodation spaces, estimated by negative values of dynamic topography, are in the order of hundreds of meters in amplitude (between 500-1100 m) and of thousands of kilometers in wavelength (between 700 km for 1 / 1 / 1 / 50 and 3500 km for 1000 / 1 / 1 / 100). This indicates mantle-driven forces during subduction are large enough to reproduce the subsidence in foreland and pericratonic areas even using higher spherical harmonic degrees, and other loads would not be needed. However, comparing these results with flexural models and stratigraphic studies (e.g., Dávila et al. 2005, 2007), nearly all predictions overestimate the magnitude and wavelength of the subsidence. A rigid lithosphere and a viscous lower mantle (e.g., 1 / 1 / 1 / 50) reproduce nearly proper wavelengths and amplitudes (~700 km and ~500 m, respectively). This result is consistent with geological studies and with low viscosities in the astenospheric wedge during slab flattening (due to low concentrations of water in the wedge, Billen and Gurnis, 2001). However, and contrary to previously proposed (Dávila et al., 2005), the modification of the slab geometry, to simulate flat subduction, does not reproduce negative dynamic topographies in the flattest part of the slab. Instead, it favors the generation of positive “relieves”. Thus, when the viscous shear generated by the subducting slab is cancelled laterally, upwarping surfaces develop close to the trench (Fig. 1c). This is in apparent contradiction with previous results (e.g., Burgess et al., 1997) that suggested subsidence even along the flattest part of the slab. However, taking into account the dynamic subsidence is triggered by mantle flows, it is reasonable that a reduction in the astenospheric wedge, by slab flattening, produces a decrease in the negative values of dynamic topography. Assuming flat subduction is a result of ridge interaction in the Andes and the slab flattening is not static and shifted southward during the Miocene (Yañez et al., 2001), subsidence driven by dynamic topography should have also migrated in this direction tracking the leading edge of the Juan Fernandez Ridge. If flat subduction favors surface uplift, the exhumation and denudation of the northern Sierras Pampeanas broken foreland may have occurred when the slab flattening passed across this area in the middle-late Miocene. The location of the flat slab at ~31° SL since the late Miocene-Pliocene (Kay and Mpodozis, 2002) would allow to suggest exhumation/denudation of the western broken foreland and the subsidence cratonward in pericratonic areas, out of the influences of tectonic topographies, are mainly a result of the sublithospheric processes. Based on these new results, some aspects might be re-evaluated, like the origin of “subduction erosion” in flat-slab segments and of the crustal overcompensations in the Sierras Pampeanas evidenced by seimic velocities studies (Fromm et al., 2004).