The influence of subducting slab advance and erosion on overriding plate deformation in orogen syntaxes

Abstract. Focused, rapid exhumation of rocks is observed at some plate corners, but the driving mechanisms remain poorly understood and contested. In this study, we use a fully coupled thermo-mechanical numerical model to investigate the effect of slab advance and different erosion scenarios on overriding plate deformation. The subducting slab in the model is curved in 3D, analogous to the indenter geometry observed in seismic studies. We find that the amount of slab advance dramatically changes the orientation of major shear zones in the upper plate and the location of rock uplift zones. Shear along the subduction interface 5 facilitates the formation of a basal detachment situated above the indenter, causing localized rock uplift there. Switching from flat (total erosion) to more realistic fluvial erosion leads to variation of rock uplift on the catchment-scale. Here, deepest exhumation again occurred above the indenter apex. We conclude that the change in orientation and dip angle set by the indenter geometry facilitates creation of localized uplift regions as long as subduction of the down-going plate is active. Tectonic uplift is modulated on even smaller scales by lithostatic pressure from the overburden of the growing orogen, and 10 highest rock uplift can occur when a strong tectonic uplift field spatially coincides with large erosion potential. This implies that both the geometry of the subducting plate and the geomorphic and climatic conditions are important for the creation of focused, rapid exhumation.


Introduction
Plate corners connect long curved plate boundary segments (Figure 1) in subduction settings observed around the world.The deformation around plate plate corners, also referred to as orogen syntaxes, has been the subject of widespread attention over the last years due to the high, sustained, and spatially focused exhumation with rates in excess of 5 mm a −1 over million-year timescales.Classic examples of focused rapid exhumation at syntaxes have been documented for the Namche Barwa (e.g.Burg et al., 1998;Enkelmann et al., 2011;Stewart et al., 2008) and Nanga Parbat (Craw et al., 1994;Crowley et al., 2009) Himalayan Syntaxes, as well as for the Saint Elias Syntaxis (Berger et al., 2008;Enkelmann et al., 2010;Falkowski et al., 2014) in Alaska.
Despite nearly two decades of work, the tectonic-and climate-driven erosional mechanisms responsible for patterns and rates of upper plate deformation in these areas remain debated (Bendick and Ehlers, 2014;Lang et al., 2016;Wang et al., 2014;Zeitler et al., 2001).Plate corners are characterized by a unique 3D geometry of the subducting plate.The bending required by subduction on a sphere results in formation of long, slightly concave subduction segments connected by strongly bent, convex  (Mahadevan et al., 2010).These plate corner span a large variety of geodynamical and climatic settings and provide test cases to identify relevant mechanisms for rapid and focused exhumation.
arcs, the plate corners.These are flexurally stiffened by the bending and associated with the formation of orogen syntaxes (Mahadevan et al., 2010).This specific 3D shape can be clearly observed in seismic data from subduction zones, for example at the Cascadia subduction zone or the Alaskan plate corner (Hayes et al., 2012).
In many convergent plate settings, shortening is mostly accommodated by subduction of the down-going plate.However, an additional component of shortening can be taken up by slab advance, i.e. migration of the overriding plate toward the downgoing plate (Heuret and Lallemand, 2005).The Nazca-South America subduction zone (Russo and Silver, 1994;Schellart et al., 2007), and the early stages of the India-Eurasia collision (Capitanio et al., 2010), provide examples of shortening with large amounts of slab advance.In previous studies, numerical modeling has played a central role towards understanding focused exhumation in orogen syntaxes.Work by Koons et al. (2002Koons et al. ( , 2013) ) simulated focused exhumation in syntaxes as a function of focused (climate driven) denudation.In their approach, they approximate the subsurface to be of homogeneous composition and define a straight S-line boundary between the subducting and the overriding plate, thus not accounting for the 3D geometry of subducting plates observed in many syntaxes.Although the link between erosion and uplift through isostasy is well established (Molnar and England, 1990;Montgomery, 1994;Simpson, 2004), their hypothesis of additional positive feedback by thermal weakening of the crust, causing accelerated deformation beneath deeply incised valleys (see also Zeitler et al., 2001), is controversial.Following this, work by Bendick and Ehlers (2014) considered the effect of the 3D subducting plate geometry, but with simplified upper plate rheology and erosion.In both of these previous approaches the kinematic boundary conditions were fixed, and the effect of different slab advance scenarios not considered.
In this study, we complement previous work by investigating how slab advance and erosion impacts the pattern and rates of exhumation in a rheologically realistic upper plate and in a generalized plate corner setting featuring a 3D-shaped, flexurally stiffened down-going plate.Our first aim in this study is the characterization of upper-plate deformation in slab advance settings with a subducting 3D indenter geometry.Our second aim is to understand the effect of erosional efficiency on upper-plate deformation and exhumation by contrasting total erosion (i.e. a constant flat surface) with more realistic fluvial erosion.

Numerical model details
Geodynamical simulations of plate corner subduction zones are performed with the program DOUAR (Braun et al., 2008;Thieulot et al., 2008), a fully coupled three-dimensional thermomechanical numerical modeling program designed to solve visco-plastic creeping flow equations at the lithospheric scale.Models in DOUAR are defined by a set of velocity boundary conditions, material properties, and model geometry defining the material domains.Bulk velocities are the result of solving the quasi-steady state force balance equations for nearly incompressible fluids with a finite element approach on an octree mesh.Pressure values are derived indirectly from the velocity solution by the penalty method (e.g.Bathe, 1982) and local smoothing is applied to avoid small-scale pressure oscillations.Finally, temperature is computed, incorporating the effects of material velocities and thermal properties.Material interfaces and bulk volume properties are stored on a self-adapting cloud of particles advected within the model domain.Additional details of the model and governing equations are given in the appendix.
For this study, we developed a new module that makes use of DOUAR's particle-in-cell approach and permits extraction of the pressure and temperature history (p-T-t paths) of particles exhumed at the free surface.Following the methods detailed in Braun (2003), Ehlers (2005), and Whipp et al. (2009), and using the parameters given therein, thermochronometric cooling ages can be calculated from these paths to quantify upper plate exhumation patterns and rates over time.

Geometry
The model domain is 800×800 km in planform and 81 km deep.

Boundary conditions
In the first part of this study, we investigate the effect of different slab advance scenarios.In the case of no slab advance, all shortening is taken up by subduction and accretion, i.e. material inflow into the domain only from left side (v sub = 30 mm a −1 , v adv = 0).Full slab advance corresponds to inflow only from the right side (v sub = 0, v adv = 30 mm a −1 ) and in the case of half advance both velocities are equal (v sub = v adv = 15 mm a −1 ).This means the S-line, intersection of overriding and down-going plate at the bottom boundary of the model domain (see Figure 2), is kept at a constant position.To allow better comparison between model scenarios, the overall shortening rate v short = v adv + v sub is kept constant at 30 mm a −1 , so that the model mass balance is the same for all cases.Additionally, outflux at the bottom of the lithospheric mantle increases from zero at the right boundary toward the center in order to compensate material influx from the right.This ensures all ratios of v sub and v adv effect the same mass added to the model domain.

Erosion
The second part of this study contrasts two erosion scenarios of flat and fluvial erosion.Flat (or total) erosion removes all material uplifted above the initial surface elevation instantly.In this special case, rock uplift rate is equal to the exhumation rate (England and Molnar, 1990).For more realistic fluvial and diffusive erosion the landscape evolution model FastScape (Braun and Willett, 2013) was coupled to DOUAR.
Fluvial erosion is computed on a regular grid of 0.78 km resolution with uniform precipitation of 1 mm a −1 .Erosion constants were 8.0•10 −5 m −2 for fluvial and 4.0 m 2 a −1 for hillslope erosion with a stream power exponent of 0.4.The edges at x = 0 km and x = 800 km are fixed to baselevel and local minima are filled so that each catchment drains to one of those edges.

Thermochronometric cooling ages
Thermochronometric cooling ages are calculated using tracer particles within the domain, and assuming no deformation for 30 Myr prior to the model start.Each predicted age represents an integration of deformation from its respective partial annealing or retention zone to the surface (Dodson, 1973;Reiners et al., 2005).Consequently, exhumation of material from the respective zone to the surface is required before the predicted cooling age can be interpreted in a meaningful way.As thermochronometric ages are sensitive to both changes in particle trajectory and thermal gradients, thermochronometric ages will continue to adjust to the evolving geodynamic conditions after the initial phase.

Results
We investigate the effects of crustal shortening by both subduction and frontal accretion (v sub > 0) and slab advance (v adv > 0) on upper crust deformation in a subduction zone with a rigid indenter in the central part of the model domain.We evaluate those models with respect to the resulting rock uplift and strain rates due to the variation in boundary inflow velocities.Additionally, we discuss the models' temporal evolution and the predicted thermochronometric ages.In total, we present four models.The first three (no, half, and full advance) are run with flat erosion, the fourth (fluvial erosion) uses the half advance velocity boundary conditions with fluvial instead of flat erosion.An overview is presented in table 2. Different amounts of slab advance change the distribution of strain rates significantly, as can be seen from Figure 4.In the no slab advance (Fig. 4a), the structures to the left of the S-line, the pro-shear band and the basal detachment, are expressed even more strongly than in the half advancing scenario, whereas the retro-shear band is weaker.The case of full slab advance (Fig. 4b) exhibits the opposite behavior: The retro-shear zone to the right of the S-line is more active, while the structures on the left are expressed more weakly.In both end-member cases, a shallow basal detachment forms in the upper crust above the more active of the steeply dipping shear bands, which is not observed in the half advance case.

Effect of slab advance on the spatial distribution of rock uplift
Figure 5 shows two horizontal slices of the rock uplift rate at the surface and 10 km depth for the three scenarios.In general, three zones of rock uplift, corresponding to the detachments described in the previous section, can be identified.Two bands of uplift to the right and left of the S-line, associated to the steeply dipping pro-and retro-shear zones, respectively, and a curved ellipsis located above the indenter basal detachment.In all scenarios, the mean uplift rate is about 1.3 mm a −1 , but the distribution strongly depends on slab advance velocity.For no slab advance (v sub = 100 %, v adv = 0 %), the pro-side shear zones, i.e. those above the background slab and the indenter ellipsis, show the highest rock uplift rates (6.4 and 8.3 mm a −1 ), respectively).In the half advance scenario (v sub = 50 %, v adv = 0 %), rock uplift is concentrated mostly in the indenter ellipsis (8.0 mm a −1 ).Finally, the full advance scenario (v sub = 0 %, v adv = 100 %) exhibits only a comparatively small increase of rock uplift above the indenter (5.4 mm a −1 ), whereas the right-side uplift band is well pronounced with rates reaching 6.3 mm a −1 .
For all scenarios, the horizontal distribution of rock uplift rates at 10 km depth mimics the pattern at the surface, but uplift rates are attenuated by roughly 50 %, indicating that upward motion is concentrated at shallow depths.

Relative slab motion effects on temporal evolution
As simulation time progresses, the deformation patterns and rock uplift change differently for the three scenarios.The central region depicted in Figure 6 shows that both for the no and half advancing slab scenario, the uplift peak above the indenter increases and reaches a quasi-steady state.In case of full slab advance, it decreases after reaching a maximum at ∼4 Myr.
Here, the strong and lasting increase can be observed in the right uplift band instead.These changes in uplift are also reflected in the position and modes of shear zones (compare figs. 3 and 4 showing strain rates at 4 Myr modeling time with strain rates at 8 Myr in Fig. 6).Without slab advance, the indenter basal detachment slowly shifts to the right over time, while keeping its general shape.In contrast, additional shear zones develop alongside the basal detachment for the half advance scenario.In the case of full slab advance, however, motion across the basal detachment almost ceases in favor of a newly formed normal fault.

Prediction of thermochronometric ages
Figure 7 shows thermochronometric age predictions and physical parameters of exhumed material for the half advancing slab after 8 Myr modeling time and illustrate the possibilities of the newly added particle tracing module.The relative distribution of predicted cooling ages is similar for all systems and the relative increase in age corresponds to the respective increase in closure temperature.Likewise, all physical parameters show the same pattern, which resembles that of rock uplift rates at the 5 surface (cf.Fig. 5).Although of similar range, cooling ages above the indenter are systematically younger in the area above the indenter compared to those from the uplift band at the model sides.Figure 8 shows a comparison of resulting zircon (U-Th)/He cooling ages and rock uplift rates for the three flat erosion scenarios.In all cases, the distribution of cooling ages again matches that of rock uplift rates.The shift in deformation towards the direction of inflow is clearly visible.Additionally, the concentration of young ages above the indenter can be seen in all scenarios.Even in the case of a fully advancing slab (Fig. 8c), where most deformation is accommodated by the right steeply dipping shear zone, this increase towards the center is visible in the zircon (U-Th)/He ages.5

Effect of variable erosion on exhumation
For the half advance velocity boundary conditions, an additional simulation coupled to a landscape evolution model (FastScape, Braun and Willett ( 2013)) was performed.The distribution of strain rates and large-scale particle trajectories, as described in local maxima (up to 2.9 mm a −1 compared to an average 1.0 mm a −1 ) correspondingly show deeper exhumation, as can be seen from panel c.The two uplift peaks in the marked catchments show the deepest exhumation (∼8 km).They are situated above the indenter and within the region of deepest exhumation in the flat erosion scenario, shown in panel e for comparison.
They also comprise locations where the youngest thermochronometric ages are predicted (not shown).In summary, fluvial erosion leads to variation of rock uplift and exhumation on the catchment scale and maxima in rock uplift roughly 60 km in 5 diameter, while regions of high uplift in the flat erosion scenarios extend ∼250 km along strike.These zones of deepest and fastest exhumation are situated above the indenter apex.the pro-band diminishes towards the model center.There, above the indenter apex, a basal detachment formes and creates a spatially focused region of uplift.
The two main factors for the differences between the indenter and background sections of the subducting slab can be attributed to the slab geometry.The shallower slab and lower dip angle of the indenter exerts a stronger traction on the overriding plate, both forward and downward, that furthers the development of the observed basal detachments.Additionally, the steeper dip of the indenter at depth provides better conditions for strain transfer along the interface than the background slab, so that no additional shear zone is needed to accommodate shortening in this region.It important to note that the different depths to the subducting slab may contribute to those effects.Nevertheless, the variation of slab dip angles conforms to natural conditions.
A prime example are the Olympic mountains, where the accretionary wedge is exposed on land only above the observed slab bulge (Brandon et al., 1998), which illustrates the substantial effects changes in the subducting slab geometry can cause.

Effect of slab advance on strain rates
The slab advance velocity has a first-order influence on the orogen dynamics, as the steeply dipping shear zone oriented toward the primary influx direction is expressed more strongly.Contrary to two other scenarios, the basal detachment in the full advance case is only of transient nature (cf.Fig. 6), and rock uplift rates above this feature reach barely half those seen in the other scenarios.As much of the overall shortening in this case is taken up by the retro-shear band, even increasing by time through thermal and strain weakening, strain transfer along the weak subduction interface and thus uplift on this side is strongly reduced.The half and no advance scenarios, however, reach a dynamic steady state at ∼4 Myr.Due to no opposing influx, the peak position in the no slab advance scenario continues to move 50 km forward in 8 Myr of modeling time, but the general structure of shear zones and strain distribution stays the same.The half advance scenario is the intermediate case and shows features of both other scenarios, with a stronger retro-shear zone and a persistent, but spatially stable, basal detachment above the indenter.In the no and half advance scenarios, rock uplift rates above the indenter reach 7.5 mm a −1 at 3.5-4 Myr, but only 4 mm a −1 in the fully advancing slab scenario, decreasing again after 4 Myr.

Sensitivity of exhumation to different erosion parameters
On large scales, the relative distribution of rock uplift is similar for both perfect and fluvial erosion model runs.Nevertheless, the growing topography produces an increased lithostatic overburden, causing both a general reduction in rate and spatial variations in rock uplift on the catchment scale.Higher uplift rates are situated in valleys, and the deepest and fastest exhumation is observed above the indenter apex, where strong tectonic forcing coincides with a high erosion potential, in this case through the large upstream catchment area.
From the previous results, we hypothesize that both focused deformation and focused erosion are necessary in order to form a region of localized and rapid uplift.The curved 3D-geometry of the subducting plate at plate corners sets the stage, as proposed by Bendick and Ehlers (2014) but only when this concurs with a sufficiently high erosion potential, a spatially limited area of intense uplift can form (cf. Zeitler et al., 2001).
A prime example for this are the Himalayan syntaxes, where the Yarlung and Ganges river, respectively, cut across the Himalayan range and steepen rapidly (Finlayson et al., 2002).In combination with active crustal-scale structures (Burg et al., 1998;Schneider et al., 1999), this has given rise to spatially limited, rapid exhumation as indicated by very young thermochronometric cooling ages (Zeitler et al., 1993;Winslow et al., 1996;Burg et al., 1998;Malloy, 2004;Zeitler et al., 2014)

Comparison to previous work
This work combines and builds upon previous work.Koons et al. (2002) showed that locally enhanced erosion rates and preweakened crust can lead to focused exhumation.However, they used a uniform visco-plastic material for both overriding and down-going plate and a straight S-line and material influx and basal drag only on one side, corresponding to our no advance scenario.Flat erosion was applied in only a limited region, while surrounding areas were not eroded, resulting in highest rock uplift rates in this very region.In contrast, Bendick and Ehlers (2014)

Model caveats and limitations
Our model setup is generic and includes first-order features of plate corner settings, but is not designed reproduce a specific region.Numeric resolution and large uncertainties on rheologic and thermal parameters do not permit this.Moreover, rock uplift rates in our model are overestimated for flat erosion simulations (section 3. 1-4.3; figs. 3-6), which inhibits the creation of topography that would exert an isostatic counter-force to material uplift.With these caveats in mind, rock uplift rates in the flat erosion models should be seen as upper bound in cases of extremely high erosion.Corresponding to the high rock uplift rates, our predicted cooling ages are very young, but nevertheless consistent with the range of exhumation rates reported at syntaxes: ∼1 mm a −1 in the Olympic mountains (Brandon et al., 1998), 3-5 mm a −1 in SE Alaska St. Elias range (Enkelmann et al., 2016), and 5-9 mm a −1 in the eastern Himalayan syntaxis (Enkelmann et al., 2011;Lang et al., 2016).
It is also worth noting that in the fluvial erosion simulations, uniform precipitation was used throughout the model run.
While this is justifiable for our study of contrasting erosional efficiency, it has been shown that changes in precipitation by orographic effects strongly influence the distribution of rock uplift and consequently orogen dynamics (Willett, 1999).If wind comes from the pro-side, the increased erosion might increase the focusing effect observed above the indenter apex, whereas we expect weaker erosion maxima with wind from the retro-side.

Summary and conclusion
In a generalized 3D curved slab subduction setting, the effect of velocity boundary conditions and erosion on the deformation in the overriding plate is investigated, with focus on the emergence and properties of localized rock uplift.Particle tracking and thermochronometric age prediction has been added to and existing model (DOUAR) to allow for direct and more tangible evaluation of crustal deformation.
The common point of all models is the localization of deformation in two major systems.One is an pop-up structure composed of two steeply dipping shear bands rooting to the S-line and cutting across the entire lithospheric mantle and crust.The shear bands are expressed more strongly towards the model sides and toward the primary direction of inflow.For the no slab advance scenario, the retro-side shear band almost disappears, and in the case of a fully advancing slab, the pro-side shear band is much weaker.The second major shear system is a basal detachment situated above the indenter apex.This detachment is strongly expressed only for a not or half advancing slab and gives rise to a zone of localized rock uplift that reaches a quasisteady state with uplift rates of 7.5 mm a −1 at ∼4 Myr.In case of a fully advancing slab, uplift rates reach only 4 mm a −1 and decreases after a peak at 3.5 Myr, when larger amounts of shortening are accommodated by the retro-shear zone.
Applying a landscape evolution model and the consequent creation of topography causes rock uplift to vary on smaller

1Figure 1 .
Figure1.World map of plate corners.In these regions (red circles), changes in subduction orientation cause flexural buckling of the downgoing plate(Mahadevan et al., 2010).These plate corner span a large variety of geodynamical and climatic settings and provide test cases to

Figure 2 .
Figure 2. Model setup and properties.a) Cut-out oblique view of the model domain illustrating the geometry and material layers: The overriding plate is divided into an upper and lower crust and lithospheric mantle.The down-going plate is a uniform viscous material.b) Velocity boundary conditions.Downward rotation of the subducting plate and accretion at x = 0 km with vsub, as well as influx from x = 800 km with vadv is compensated up to the Moho depth by outflux at the bottom.c) Vertical profiles of the overriding plate effective viscosity and temperature initial conditions.Material parameters given in table 1.

Figure 3
Figure 3 illustrates the key features of the half advance model in a snapshot taken after 4 Myr modeling time.Strain rates (background colors in panel a) indicate two main systems of localized deformation (strain rates above 5.0•10 −15 a −1 ).In the first system, two broad, steeply dipping shear bands rooting to the S-line at an angle of ∼60 • form a lithospheric scale popup structure.Due to the mostly viscous deformation (cf.Fig. 2), the shear bands are poorly localized in the lower crust and lithospheric mantle.Laterally towards the model center and the indenter (y = 400 km), those shear bands attenuate.Here, deformation is accommodated by the second type of structure, a basal detachment located above the indenter.The particle

Figure 3 .
Figure 3. Vertical slices along dip after 4.0 Myr modeling time showing the second invariant of the strain rate tensor, particle trajectories, and the temperature distribution for the half advance scenario.Outlines of the Moho (light green), subducting plate interface (dark green), and S-line (magenta) are added for orientation.Note that results and slices are symmetrical about the y = 400 km plane.a) The strain rate outlines the major shear zones: Two broad, poorly localized shear zones nucleate at the S-line, while the pro-shear zone weakens towards the model center.This shear is accommodated by a basal detachment that is situated on top of the indenter.b) Temperatures follow crustal deformation.Significant upward heat advection can be seen in the vicinity of the basal detachment, while Moho deformation dominates the temperature anomalies towards the model sides.

SolidFigure 4 .
Figure 4. Vertical slices after 4.0 Myr modeling time showing the second invariant of the strain rate tensor and particle trajectories for a) no and b) full slab advance scenarios.Outlines of the Moho (light green), subducting plate interface (dark green), and S-line (magenta) are added for orientation.The main shear zone rooting to the S-line is oriented towards the direction of influx.At shallow depths, that shear zone forms a basal detachment in the respective orientation.The basal detachment above the indenter exists for both cases, but is pronounced more strongly in the no slab advance scenario.

SolidFigure 5 .Figure 6 .Figure 7 .
Figure 5. Rock velocity field and uplift rates after 2.5 Myr modeling time.Each panel shows two horizontal slices at the surface and 10 km depth for the three scenarios discussed, i.e. a) no, b) half, and c) full advancing slab.Background colors show rock uplift rate, with red tones indicating uplift and blues subsidence.Arrows illustrate the full 3-dimensional velocity field and are colored by total velocity.Rock uplift occurs in two bands along the S-line and in an ellipsoid region above the indenter.The relative uplift rate in the two bands depends on the inflow velocity ratio.For both no and half advancing slab, uplift above the indenter is equally strong, but more distinguished in the half advancing scenario.

Figure 8 .
Figure 8.Comparison of zircon-(U-Th)/He cooling age for three slab advance scenarios.In all cases, cooling ages match well with the distribution of rock uplift.The shift in deformation focus in direction of the main material influx is clearly visible.

SolidFigure 9 .
Figure 9.Comparison of rock uplift and exhumation depths with fluvial versus total erosion.a) Surface elevations, b) rock uplift rates, and c) exhumation depth for the half advance scenario coupled to a landscape evolution model (FastScape (Braun and Willett, 2013)) after 6.0 Myr modeling time.Uplift is focused along the flanks of the orogen, with exhumation maxima situated in river valleys of high erosion potential.Deepest exhumation occurs in the two uplift foci situated above the indenter, the catchments outlined in blue.Predicted thermochronometric ages are: apatite (U-Th)/He: 1.3 and 1.1 Ma; apatite fission track: 1.9 and 1.65 Ma; and zircon (U-Th)/He 4.2 and 4.1 Ma for the upper (light blue catchment) and lower (dark blue) hotspot respectively.d) and e) show the flat erosion results for rock uplift rates and exhumation depth for comparison.Note the range is increased by factor 2 for those two plots (gray labels).

4. 3
Effect of slab advance on exhumation and cooling ages 5 Exhumation and rock uplift, which are the same in case of flat erosion, are defined by the orogen-scale dynamics discussed in the previous section.Deformation and uplift are strongest in a broad region along the S-line, with a localized region of uplift created by the basal detachment, if active.The amount of slab advance influences rock uplift through the relative activity (or Solid Earth Discuss., https://doi.org/10.5194/se-2018-14Manuscript under review for journal Solid Earth Discussion started: 21 March 2018 c Author(s) 2018.CC BY 4.0 License.rate of deformation) of shear bands as discussed in the previous section.Higher slab advance velocities cause faster rock uplift along the right uplift band, whereas more frontal accretion strengthens the left band and centers the uplift above the indenter.

Table 1 .
The element size is 6.25×6.25×1.56km, corresponding to 128×128×52 elements for the entire domain.The direction of subduction is parallel to the x axis and the entire setup is Solid Earth Discuss., https://doi.org/10.5194/se-2018-14Manuscript under review for journal Solid Earth Discussion started: 21 March 2018 c Author(s) 2018.CC BY 4.0 License.Overriding pate material parameters

Table 2 .
Controlling Parameters Overview used an indenter-type geometry, but temperaturedependent viscous rheology for the overriding plate and uniform, flat erosion.Due to the viscous rheology, no strain localization was observed and rock uplift rates increased gradually from the edges toward the center, forming an uplift region of several 100 km extent.While they explored effects of viscosity and indenter geometry, only full slab advance velocity boundary conditions were used.The addition of frictional plasticity in this study naturally allows for strain localization.In combination with fluvial erosion, this illustrates possible mechanisms that create localized regions of uplift without relying on a priori defined structures.Solid Earth Discuss., https://doi.org/10.5194/se-2018-14Manuscript under review for journal Solid Earth Discussion started: 21 March 2018 c Author(s) 2018.CC BY 4.0 License.