Two-dimensional thermal-mechanical models of thick-skinned, salt-based fold and thrust belts, such as the Zagros, SW Iran, are used to address: (1) the degree of deformation and decoupling between cover and basement rocks due to the presence of a weak salt detachment, (2) the reactivation potential of pre-existing basement normal faults due to brittle or ductile behavior of the lower crust (as related to cold or hot geothermal gradients), and (3) variations in deformation style and strain distribution. The geometry and kinematics of the orogenic wedge and the activity of pre-existing basement faults are strongly influenced by the geothermal gradient (defined by the Moho temperature, MT) and basement rheology. We infer that the MT plays a major role in how the lower and upper crust transfer deformation toward the foreland. In relatively hot geotherm models (MT = 600°C at 36 km depth), the lowermost basement deforms in a ductile fashion while the uppermost basement underlying the sedimentary cover deforms by folding, thrusting, and displacements along pre-existing basement faults. In these models, cover units above the salt detachment occur within a less deformed, wide plateau in the hinterland. In relatively cold geotherm models (MT = 400°C at 36 km depth), deformation is mainly restricted to basement imbricate thrusts that form within the orogenic hinterland. Detachment folding, thrusting, and gravity gliding occur within cover sediments above uplifted basement blocks. Gravity gliding contributes to a larger amount of shortening in the cover compared to the basement.
 Several studies have physically Graveleau et al., 2012] and numerically [Ellis et al., 2004; Stockmal et al., 2007; Yamato et al., 2011; Buiter, 2012 and references therein] simulated thin-skinned deformation at convergent plate boundaries and illustrated how different rheological and mechanical parameters change the mechanism and style of deformation of sedimentary cover rocks. Some studies have gone further and modeled the cover deformation above basement normal faults [e.g., Schedl and Wiltschko, 1987; Koyi, 1991, 1993; Hardy, 2011], and others have investigated the influence of cover strength on basement-involved fault propagation folding in cover sediments [e.g., Hardy and Finch, 2006]. A few studies have modeled thick-skinned deformation in which both brittle cover sediments and deeper brittle-ductile basement rocks are involved [e.g., Barr and Dahlen, 1989; Buiter and Torsvik, 2007]. However, scaled laboratory modeling of thick-skinned deformation is not easy, and more sophisticated materials and setups are required to handle temperature-dependent ductile rheologies at depth [Davy and Cobbold, 1991; Boutelier and Oncken, 2011]. Thermal-mechanical numerical experiments are therefore favored for modeling of thick-skinned deformation involving both brittle upper crust and ductile lower crustal rheologies [e.g., Bird, 1978; Buiter et al., 2009]. In thick-skinned deformation, weak zones such as pre-existing basement faults and their reactivation by inversion tectonics [Lacombe and Mouthereau, 2002], mechanically weak layers such as salt or shale within the cover sediments or above the basement acting as weak detachments [Yamato et al., 2011], and the depth- and temperature-dependent viscosity of the middle- to lower crust, are important parameters in crustal shortening studies [Buiter et al., 2009].
 Natural examples of basement-involved fold and thrust belts include the Andes [Kley et al., 1999; Cristallini and Ramos, 2000], Urals [Brown et al., 1999], Taiwan [Mouthereau and Lacombe, 2006], Rockies [Dechesne and Mountjoy, 1992], the Alps [Rotstein and Schaming, 2004], and the Zagros [Berberian, 1995; Molinaro et al., 2005; Mouthereau et al., 2007]. Some fold and thrust belts contain single or multiple weak detachments (salt or shale) such as the Zagros, SW Iran [Davis and Engelder, 1985; Talbot and Alavi, 1996; Bahroudi and Koyi, 2003; Sherkati et al., 2006], the Northern Apennines, Italy [Massoli et al., 2006], the southern Canadian Rocky Mountains [Stockmal et al., 2007], and the Jura of the Alps [Davis and Engelder, 1985]. Mechanically weak layers at the basement-cover interface or between stratigraphic levels of fold and thrust belts influence the deformation style and kinematics of an orogenic system [Davis and Engelder, 1985; Koyi, 1988; Cotton and Koyi, 2000; Costa and Vendeville, 2002; Nilforoushan et al., 2008; Nilfouroushan et al., 2012; Ruh et al., 2012]. In addition, pre-existing faults in the upper brittle crust, acting as weak zones in the basement, can be reactivated and change the localization and distribution of deformation in overlying sedimentary rocks. It is currently not clear how effectively ductile flow of lower crust can reactivate pre-existing basement faults in the upper brittle crust when there is a layer of salt between the cover and basement rocks.
 Using a series of 2-D thermal-mechanical finite element models, we explore how thick-skinned deformation and the interactions between basement and cover rocks are influenced by different parameters such as the geothermal gradient, expressed as Moho temperature, and the presence of weak basal detachments like salt in the basal succession of the sediment cover, and weak zones in the basement (e.g., pre-existing normal faults). Owing to the inherent approximation of numerical modeling to the complex real Earth, we do not aim to model exactly the history and present-day deformation of any orogenic system. Rather, we relate our modeling approach to nature by considering the available thermal-mechanical parameters of the Zagros fold and thrust belt, as an opportunity to examine thermal-mechanical, rheological, and structural interactions with constraints from an active orogen. Although the experiments focus on the Zagros fold and thrust belt, the results can be applied to other fold and thrust belts with similar tectonic configurations.
1.1 The Zagros Fold and Thrust Belt
 The Zagros fold and thrust belt (Figure 1a) is the result of active convergence between the rifted continental margin of the Arabian plate and the Iranian continental blocks following the closure of the Neo-Tethys Ocean during the Late Eocene [Frizon de Lamotte et al., 2011]. The ongoing deformation of this convergence is not distributed homogenously in Iran and is mainly taken up by mountain belts like the Zagros. Central Iran is undergoing relatively little deformation and acts as the rigid backstop to the Zagros [Vernant and Chéry, 2006]. The present-day active deformation within the Zagros has been defined by several GPS measurement campaigns [e.g., Nilforoushan et al., 2003; Vernant et al., 2004; Hessami et al., 2006; Walpersdorf et al., 2006; Tavakoli et al., 2008]. These studies show that the cover sequence of the Zagros is shortening relatively slowly (5 ± 3 mm/yr) above a high-frictional basal detachment in the NW Zagros, whereas in the southeast of the Zagros, the shortening rate is higher (8 ± 3 mm/yr) (Figure 1). Total shortening since ~5 Ma [Blanc et al., 2003; Allen et al., 2004] in the Zagros is estimated to range between 45 and 65 km [Blanc et al., 2003; McQuarrie2004; Oveisi et al., 2007], consistent with GPS-derived shortening rates [Hessami et al., 2006; Walpersdorf et al., 2006; Tavakoli et al., 2008].
 Based on stratigraphic data, the total thickness of the cover sediments in the Zagros region varies between 7 km in the Fars area to 9 km in the Dezful area (Figure 1) [Alavi, 2007]. The southeastern cover sediments overlie a 1–2 km thick layer of weak Hormuz (Neoproterozoic-Cambrian) salt that partially decouples the Phanerozoic cover from its Precambrian crystalline basement [Koyi, 1988; Talbot and Alavi, 1996; Bahroudi and Koyi, 2003]. The thick Hormuz salt layer, not only acts as an efficient detachment but also by feeding salt diapirs, changes the deformation mechanism within the cover units mainly in the eastern Zagros. The numerous emerged or buried salt diapirs rising from this salt detachment have influenced the shape, localization, propagation, and orientation of the folds in the eastern Zagros [Talbot and Alavi, 1996; Jahani et al., 2009].
 By opening of the Neo-Tethys Ocean during Permian-Early Triassic [Berberian and King, 1981; Vergés et al., 2011], a passive margin and several extensional faults formed in the Arabian basin that were then covered by sediments [Berberian and King, 1981; Bahroudi and Talbot, 2003]. The exact locations of these pre-existing rift-related extensional faults which have been inverted during the collision stage are not clear but their existence and approximate locations have been inferred from analysis of surface geomorphologic features, topographical sections, and earthquake spatial distribution [Berberian, 1995; Mouthereau et al., 2006; Alavi, 2007]. The vertical distribution of earthquakes in the Zagros ranges from 4 to 30 km in depth with the majority between 7 and 20 km [Maggi et al., 2000; Tatar et al., 2004; Nissen et al., 2011; Yaminifard et al., 2012a, 2012b]. This indicates that faults are active in both the cover and basement sequence, suggesting that the shortening of the Zagros fold and thrust belt is not taken up only by the cover sediments but also by the basement [Jackson, 1980; Berberian, 1995; Hessami et al., 2001; Talebian and Jackson, 2004; Oveisi et al., 2009]. Further, the spatial distribution of earthquakes is limited to the upper 20 km of the crust and implies that the Brittle Ductile Transition (BDT) zone is located at a depth of around 20 km or deeper and faulting is limited to the upper 20 km crust of the Zagros (Figures 1 and 2) [Nissen et al., 2011].
 The Moho depth beneath the Zagros inferred from geophysical measurements using receiver functions is estimated to be between 40 and 50 km (Figure 1) [Hatzfeld et al., 2003; Paul et al., 2006, 2010; Manaman et al., 2011; Yaminifard et al., 2012a]. The hinterland-dipping Moho (β = 0.5°) makes the crust slightly thicker under the Sanandaj-Sirjan zone relative to the foreland in the simply folded zone (Figure 1).
 Geothermal gradient contours in Motiei  and Bordenave , both based on well data from Orbell , indicate a variation in geothermal gradient across and along the belt from 10 to 28°C/km. The increasing geothermal gradients across the belt from the High Zagros to Persian Gulf were partly assigned to tectonically thickened crust near the suture zone [Bird, 1978]. The anomalies in the contours of geothermal gradients along the belt were correlated with north-south trending reactivated old basement faults in the Zagros [Bahroudi and Talbot, 2003]. For geodynamic modeling of the Zagros, Mouthereau et al.  used a geothermal gradient of 10–15°C/km consistent with a Moho temperature (MT) of 450°C–675°C. The average surface heat flow of 40 mWm−2 used by Bird  for thermal-mechanical modeling of the Zagros also corresponds to a MT of about 500°C at a Moho depth of 40 km.
 Balanced cross sections [e.g., Sherkati and Letouzey, 2004] and analysis of topographic profiles across the Zagros [Mouthereau et al., 2006] also show that basement faulting is required to explain the present-day topography. However, in the Zagros, where a relatively thick salt layer covers active basement faults, seismic reflection data fail to image basement structures [Blanc et al., 2003; Alavi, 2004; McQuarrie, 2004; Sherkati and Letouzey, 2004; Paul et al., 2006] (Figure 1). Moreover, due to the presence of this thick salt detachment, the seismicity has a diffuse pattern (Figure 1) [Koyi et al., 2000; Nissen et al., 2011]; and therefore, locating seismogenic basement faults is elusive [Berberian, 1995].
2 Numerical Experiments
 In our experiments, we used a two-dimensional, thermal-mechanical finite element code (SOPALE) which can model high finite strain based on an Arbitrary Lagrangian Eulerian (ALE) method [Fullsack, 1995]. This code has been extensively used in a range of different geodynamic modeling applications [Pysklywec and Shahnas, 2003; Beaumont et al., 2004; Pysklywec and Cruden, 2004; Cruden et al., 2006; Buiter and Torsvik, 2007; Stockmal et al., 2007; Beaumont et al., 2009; Buiter et al., 2009; Gray and Pysklywec, 2010; Nilfouroushan et al., 2012].
 In the ALE numerical technique, SOPALE simultaneously uses Eulerian and Lagrangian grids: Finite element computations are performed on a Eulerian grid whose elements are only stretched vertically to accommodate the evolution of topography on the free upper surface; the fully deforming Lagrangian grid tracks the migrating interfaces and material properties. Each of these grids is made up of initially rectangular elements.
 The models assume incompressibility of materials. While studies show that compressibility of rock may have an effect for deep mantle convection [e.g., Jarvis and Mckenzie, 1980] at the crustal scale, the approximation of incompressibility will not influence the behavior of the models. The governing equations for the models are
where u, σij, ρ, g, cp, T, k, H, and t are velocity, stress tensor, density, gravitational acceleration, specific heat capacity, temperature, thermal conductivity, volumetric rate of internal heat production, and time, respectively. The other variables, α, ρ0, and T0 are thermal expansivity, reference material density, and reference temperature, respectively.
 Similar to Mohr-Coulomb failure, the brittle deformation for frictional-plastic materials is specified by a pressure-dependent incompressible Drucker-Prager yield criterion:
where J′2 is the second invariant of deviatoric stress, P is the pressure, λ is the pore fluid factor, ϕ is the angle of internal friction, and C is the cohesion. The ϕ value can decrease linearly in a range of specified strain (frictional weakening).
 The viscous deformation of materials follows either linear (Newtonian) viscous behavior or power law creep where the effective viscosity, , is
where A is the material constant, n is the power law exponent, is second invariant of the deviatoric strain rate tensor, Q is the thermal activation energy, R is gas constant, and T is temperature.
2.1 Model Setup
 A series (Table 1) of 2-D thermal-mechanical shortening experiments was run to investigate the thick-skinned deformation of a salt-based fold and thrust belt with pre-existing basement faults. The model parameters (Table 2) were set based on the Zagros as a natural prototype. Our aim is not to simulate the full deformational history of the Zagros and its structure in detail. Instead, we use the best available observational constraints for parameters in the models and assess the interaction between the cover and basement, localization and distribution of deformation, and how variations in the geothermal gradient change the strain distribution when there is a weak salt detachment and pre-existing weak zones present in the basement.
 To set up our experiments, we started with a rectangular box of 300 by 36 km that consisted of 8 km sedimentary rocks overlying a stepped basement (inherited from a rifting episode) and separated by a 0.5–2.5 km thick salt layer (Figure 3a). The arbitrary three basement steps (with heights of 0.5, 1, and 1 km, respectively, from left to right in Figure 3) are not considered as faults, and they only introduce velocity discontinuities. Since the current Moho depth is ~40–50 km in the southeastern part of Zagros (Figure 1) [Hatzfeld et al., 2003; Paul et al., 2006, 2010; Yaminfard et al., 2012], we considered the initial Moho depth of 36 km [Vergés et al., 2011] to take into account crustal thickening after 50 km shortening in about 5–6 Ma [Blanc et al., 2003; Sherkati et al., 2006; Oveisi et al., 2007; Vergés et al., 2011]. As stated before, geophysical studies of the current geometry of the Moho under the Zagros [e.g., Hatzfeld et al., 2003; Paul et al., 2006] indicate a gentle hinterland-dipping geometry (β = 0.5°) across the belt (Figure 1). However, this is only an approximation of the current geometry of the Moho, and the initial Moho geometry at the onset of the Zagros shortening is still debated [Vergés et al., 2011]. Therefore, to simplify our models, we assumed a typical horizontal Moho (i.e., β = 0°). The models were shortened orthogonally and continuously from one side by pushing a strong indenter into the material box. The typical orthogonal indentor setup used in previous analytical, analog, and numerical wedge simulations [e.g., Davis et al., 1983; Buiter and Torsvik, 2007; Nilfouroushan et al., 2012; Ruh et al., 2012] simplifies the model setup and the interpretations of the results, and facilitates comparisons between different methods and results. All models were shortened at a rate of 8 mm/yr, a rate deduced from present-day GPS measurements (Figure 1) and consistent with geological rates [Hessami et al., 2006; Walpersdorf et al., 2006]. A high viscosity of the indenter (1030 Pa s) relative to the other materials in the box meant that the indenter did not deform as it was pushed into the solution space. The relative shortening between the indenter and the material in the box in our models is similar to the Central-Iranian block (Figure 1) as a rigid indenter pushing against the southeastern part of Zagros. Central Iran is presently undergoing relatively little deformation (±2 mm/yr) and can be considered a relatively rigid backstop to the Zagros [e.g., Vernant and Chéry, 2006].
 We used a rectangular node resolution of 601 × 121 (equal to 0.5 km for horizontal and 0.3 km for vertical resolution) for the Eulerian and 1801 × 361 for the Lagrangian grids. The upper surface is a free surface, the sidewalls are free-slip, and the bottom surface, the Moho, is a no-slip boundary (Figure 2). A no-slip basal boundary assumption implies no horizontal movements occur right at the bottom surface (Moho). Due to the temperature-dependent rheology used to model lower crustal rocks, the ductile behavior of the lower crust permits the materials just above the Moho to deform readily, thereby minimizing the effect of the no-slip lower boundary. A similar no-slip assumption for the Moho boundary has been used in other modeling studies [e.g., Mouthereau et al., 2006; Buiter and Torsvik, 2007].
 We set the angle of internal friction to 15° for the overburden sediments and 20° for the basement rocks and let the frictional strength of the brittle crust decrease linearly by 50% across a strain range of 0.5 to 1.5 to approximate material weakening due to, for example, an increase in pore fluid pressure in nature [e.g., Buiter and Torsvik, 2007; Gray and Pysklywec, 2012a]. Similar to previous modeling studies [Pysklywec and Beaumont, 2004; Gray and Pysklywec, 2012a], we use an “effective angle of internal friction” for φ with pore fluid factor λ = 0. This factors in the pore fluid pressure implicitly with the assumed (lower) angle of internal friction of 15°–7.5° for overburden sediments (Table 2).
 Following previous studies, we model the weak Hormuz salt in Zagros by a Newtonian viscous rheology with an effective viscosity of 1018 Pa s and density of 2200 kg m−3 (Table 2) [e.g., Mouthereau et al., 2006; Yamato et al., 2011]. The extension of the salt detachment is varied in our models to study its effect on orogenic wedge deformation (Table 1). However, we limit the frontal extension of the salt detachment to a distance of 240 km (in initial setups) from the indenter (Figure 2) to avoid any frontal boundary effect on cover deformation. We used density values of 2600 kg m−3 for the cover sediments and 2900 kg m−3 for the crystalline crust [Snyder and Barazangi, 1986; Paul et al., 2006].
 The composition of the basement of the Zagros is poorly known [e.g., Bahroudi and Talbot, 2003]. The only observed basement rocks are blocks of orthogneiss, metasediments, amphibolites, and serpentinites intruded by granite, gabbro, and basalt brought to the surface in salt diapirs [Haynes and McQuillan, 1974; Kent, 1979]. Following Mouthereau et al. , we assumed quartz diorite and diabase to be suitable compositions for the basement rocks of the Zagros basement. Hence, we evaluated the available temperature-dependent power law creep laws for these compositions in our numerical experiments (Tables 1 and 2). As stated above, earthquakes in the southeastern Zagros occur at depths between 4 and 30 km with the majority between 7 and 20 km [Maggi et al., 2000; Talebian and Jackson, 2004, Nissen et al., 2011; Yaminifard et al., 2012a, 2012b], which suggests that ductile deformation should occur below ~20 km. To demonstrate the strength profiles that result from using the flow laws for diabase and quartz diorite, we used the same flow laws as Mouthereau et al.  (Figure 4). The input parameters for the brittle and ductile deformation of the model crust are given in Table 2, using parameters suggested by Vernant and Chéry , Mouthereau et al. , and references therein [i.e., Goetze, 1978; Hansen and Carter, 1982; Wilks and Carter, 1990]. The selection of diabase and quartz diorite flow laws to represent the rheology of the ductile crust is reasonable as their strength profiles result in a brittle ductile transition (BDT) depth of ≥ 20 km depth for MTs of 400–600°C (Figure 4).
 The number, location, and geometry of the basement faults in the Zagros are poorly known; hence, we introduce them arbitrarily in our models. In experiments with pre-existing basement faults, we consider the faults to be 1500 m wide and dipping at 60° (as such, the faults are resolved by three Eulerian elements, 3 × 500 = 1500 m). Following Buiter and Torsvik , the faults are filled with a Newtonian material with a viscosity of 1020 Pa s to mimic weak inherited normal faults in the basement. The faults extend down to 20 km depth where the brittle deformation depth has been inferred from earthquakes studies [Tatar et al., 2004; Yaminifard et al., 2012a, 2012b].
 We ran 19 shortening experiments (Table 2) with or without basement faults, changing the basement rheology, salt distribution, and increasing Moho temperature (MT) systematically from 400°C to 500°C and 600°C (assuming geothermal gradient of 11–17°C/km for a 36 km thick crust) which covers the reported MT range in the previous studies [Bird, 1978; Mouthereau et al., 2006; Vernant and Chéry, 2006].
 We simplified our models by ignoring the effect of isostatic adjustment and thermal subsidence of the underlying lithosphere; also, erosional and depositional processes were not included, although these can have an influence on the behavior of such models [Pysklywec, 2006; Gray and Pysklywec, 2012b]. As described in Buiter and Torsvik , in orogenic wedge models with this type of configuration, the ductile/viscous lower crust in the models allows a “simple form” of effective isostatic compensation in the crust to occur. With a deeper (i.e., mantle) isostatic compensation, there may be some modification to details of the structural geometry of the crust, but this is beyond the scope of the modeling code at this scale of lithospheric investigation.
3.1 Rheology of Basement Rocks
 Using a diabase composition and rheology results in a higher effective viscosity lower crust compared to quartz diorite (Figure 4). The BDT depths are 24, 20, and 16 km for quartz diorite and 31, 26, and 22 km for diabase, for MTs of 400°C, 500°C, and 600°C, respectively. This means that diabase has a deeper BDT than quartz diorite under the same thermal and mechanical conditions.
 We tested basement comprising both diabase and quartz diorite in our models and observe composition has a strong influence on deformation behavior (Figure 5). In the early stages of shortening, deformation in the cover units starts by formation of shear zones in the frontal part of the cover where the salt is pinched out. With further shortening, deformation in the cover units propagates backward (Figure 5). Due to the shallower BDT depth in experiments using quartz diorite (Figure 4), flow in the more ductile lower crust suppresses significant deformation in the cover above the salt detachment (Figure 5b–5d). In all three models using quartz diorite with different MTs, the upper crust deforms similarly and no shear zones (except one near the indentor) are localized in the 200 km long sedimentary cover (Figure 5b–5d). However, the basement is folded in the case of all MTs. In the hottest model, MT = 600°C, the basement is also affected by the load of the especially thickened cover at the second basement step location. Due to the flow of hot ductile lower crust in this model, the upper brittle crust including the cover subside significantly at this location (Figure 5d).
 The cover units in models with diabase deformed more, with faulting and development of pop-up structures especially in the coldest model (MT = 400°C) (Figure 5e). In the case of the Zagros, with reported earthquakes depths to around 20 km, diabase seems to be a better choice for the composition and rheology of the basement rocks. This is in agreement with a recent study on lithospheric strength of the Zagros [Nankali, 2012] that suggests a relatively cold geothermal gradient and a diabase or granulitic composition with a BDT located around 21–28 km. We therefore focus our modeling using diabase as the preferred composition and rheology for the basement rocks.
3.2 Salt Detachment
 The viscous “salt” layer in our experiments decouples the cover sediments from the basement and causes rapid propagation of deformation in the cover toward the distal pinch out of the salt layer in the foreland during early stages of deformation (Figures 5a–5g). To better illustrate this decoupling effect and how its spatial distribution above the basement can influence wedge deformation, we ran three more models with three different MTs but the same diabase rheology and removed the salt between the cover and basement in the hinterland near the indenter (hereafter called partial-salt models, Figures 3b and 6a–6d). After 50 km shortening, wedge deformation, topography, and the localization of shear zones are very different in the hinterland in these partial-salt models compared to salt models, especially for MT = 400°C (cf. Figures 5e–5g and 6b–6d). For example, in salt models with MT = 400°C, the basement is highly deformed into stacks of thrust sheets and the cover sediments are extended and thinned in the hinterland (Figure 5e). In contrast, for the partial-salt model with the same MT, the cover sediments do not localize many shear zones and are less uplifted in the hinterland near the indenter (Figure 6b). This indicates that the basement and cover deformation is strongly affected by the salt detachment near the indenter in these “cold” models (MT = 400°C). However, the foreland deformation in both the cover sediments and basement is very similar in both salt and partial-salt models with equal MTs, where pop-up structures above the salt layer are similarly developed (Figures 5e–5g and 6b–6d). By increasing the MT, the basement deforms in a more ductile manner and deformation is less localized near the indenter. As a result, due to the lower degree of deformation of the hinterland basement in hotter experiments, the cover deformation is similar in the salt and partial-salt models (Figures 5f, 5g, 6c, and 6d). Boundary effects near the indenter also contribute to the local model deformation, but as model results show (Figures 5e–5g and 6b–6d), the influence of the salt detachment near the indenter is the more dominant effect and completely changes the mechanism of deformation especially in cold models.
 In the following section, we present models with no-salt detachment and discuss the role of coupling between cover and basement deformation in the presence of pre-existing weak zones (faults) in the basement.
3.3 Pre-existing Basement Faults and Moho Temperature
 In order to study the influence of basement faults during thick-skinned deformation, we ran three partial-salt experiments (with three different MTs) containing three basement faults located in the same position as the initial basement steps inherited from continental rifting (Figures 3 and 6e–6h). After 50 km shortening, the partial-salt models with basement faults are strongly sensitive to the temperature in the basement and deform differently from the models without any basement faults (cf. Figures 6b–6d and 6f–6h). In the MT = 400°C experiment, the frontal part deforms similarly to the partial-salt model without any basement faults because there is little ductile flow of lower crust (Figures 6b and 6f). In this model, deformation mainly occurs in the cover sediments and basement is mainly involved near the indenter where the first pre-existing basement fault shows the most reactivation (Figure 6f). By increasing the Moho temperature (i.e., MT = 500°C and MT = 600°C), the basement faults far from the indenter are reactivated and take up significant displacement (Figures 6g and 6h). In these “hotter” models, the basement is folded and pre-existing faults localize the large amount of deformation in the basement owing to the ductile lower crust that transfers the deformation forward toward the foreland. In these hotter models, larger faulted blocks form in the cover relative to the cold models where small pop-up structures are developed (Figures 6f–6h). This implies that in colder salt-based fold and thrust belts, the displacement related to reactivation of pre-existing basement faults is greater near the indenter. Increasing the MT increases ductile flow in the basement, and consequently, the distal basement faults in the middle and frontal parts of fold and thrust belt are also reactivated.
 In order to illustrate systematically how pre-existing basement faults are reactivated sequentially from the hinterland toward the foreland, and how deformation is taken up by displacements along these faults, we changed the spacing of the basement faults to an arbitrary equal distance of 30 km and ran three more salt models (Figures 7a–7d). The MT was varied from 400°C to 500°C and 600°C, and the models were shortened up to 50 km from one side. We clearly observe that by increasing the MT, the extent of basement deformation and basement fault reactivation are increased while the amount of cover deformation is decreased in the hinterland (Figures 7b–7d). In the cold model (MT = 400°C), the basement deformation is mostly localized near the indenter and the amount of displacement along the basement faults decreases toward the foreland (Figure 7b). Imbrication of the basement blocks and gravity gliding of the cover sediments above the salt layer in the cold experiment introduces significant extension in the cover (see section 3.6). In contrast, in models with hotter MTs (500°C and 600°C), the cover sediments above the basement faults are less deformed. This happens because the basement blocks can easily rotate and displace along the pre-existing basement faults and take up more deformation than the cover sediments. Basement blocks rotate more in models with hotter MTs due to the increased ductility of the lower crust (Figures 7a–7d). Consequently, the salt detachment is segmented into triangular salt zones and salt flows toward the hinterland.
 The rotation of basement blocks in the hotter models steepens the dip of the pre-existing basement faults. In these hotter models, all pre-existing faults are reactivated as blind faults that do not cut through the cover, which is decoupled by the weak salt detachment.
3.4 Pre-existing Basement Faults and Salt Detachment
 To emphasize the effect of salt on deformation decoupling, we ran three experiments without a salt detachment but containing five equally spaced basement faults (Figures 7e–7h). The MT was varied as before, and the models were shortened by 50 km from one side. In these experiments and for all MTs, the pre-existing basement faults were all reactivated during shortening and they cut through the cover units to emerge to the surface (Figures 7f–7h). Pop-up structures were localized in the hanging walls of pre-existing basement faults and developed more in colder models (i.e., MT = 400°C and 500°C; Figures 7e–7g). In the hottest model (MT = 600°C), the basement blocks display the greatest amount of clockwise rotation (about 20° clockwise relative to about 15° in salt-present models measured from dip change in basement faults) and pop-up structures did not develop. The deformation mechanism is significantly different in these no-salt models compared to models with salt (Figures 7a–7h). Deformation is mainly localized across half of the model in the no-salt models, whereas in models with salt deformation is distributed over a wider area and cover sediments are deformed far into the foreland. Due to coupling of the basement and cover in the hotter no-salt models, the cover units rotate as a consequence of basement block rotation (Figures 7g and 7h). In contrast, the cover units in hotter salt-present models dip slightly toward the foreland.
3.5 Deformation Localization and Distribution
 We illustrate strain rate localization and distribution in different stages of model shortening by including one more partial-salt model with five equally spaced (30 km) pre-existing basement faults and MT = 400°C (Figure 8). After 2 km shortening, weak zones, which coincide with pre-existing faults and the viscous basal detachment, record relatively higher strain rates than other deformation zones in the cover and the basement (Figure 8b). Resistance along the no-slip boundary in the bottom of the model also accommodates higher strain rates. At the 2 km shortening stage, we observe a high strain rate zone associated with the pre-existing basement fault located closest to the indenter where salt is missing, which extends all the way from the basement to the cover. This basement fault has therefore localized a relatively high strain rate and propagated upward into sediments in the no-salt zone. The second and the third pre-existing faults closest to the indenter also localize high strain rates. At this early stage, deformation in the cover sediments extends to about 240 km away from the indenter whereas basement deformation is confined to about 100 km from the indenter. This shows that deformation is transferred quickly forward to the distal end of salt layer and indicates that deformation does not take place simultaneously in the cover and basement. In the distal part of the system, the involvement of the basement is preceded by a phase during which only the cover is deformed (thin-skinned phase).
 By further shortening to 8.1 and 16.2 km, pop-up structures develop in the cover units above the salt detachment. The steps in the basement (Figures 3b, 8b, and 8c) and the distal end of the viscous layer initiate relatively higher strain rate zones resulting in shear zones and pop-up structures in the cover at early stages of shortening. The relatively hotter lower crust below 30 km also records accommodation and transfer of the high strain rates zones in the first half of the model. The strain rate plots also illustrate that the higher strain rate zones in the lower, ductile part of the experiments (below 30 km depth) are linked by higher strain rate zones that are coincident with pre-existing faults in the upper crust (Figure 8).
 We observe that after 24.3 km of shortening (Figure 7e), the zone near the indenter does not record high strain rates, indicating that the first basement fault becomes inactive and the whole block near the indenter is pushed forward as an almost rigid block. Moving away from the indenter, basement faults are reactivated sequentially with progressive shortening and as horizontal stress is transferred forward. With further shortening, pre-existing weak zones in the basement are preferred zones of localized shear strain accumulation and only one major back thrust is formed in the basement at a late stage (Figures 8g and 8h, 50 km).
 After 24.3 km shortening (Figures 8e–8h), higher strain rates in the cover are partly associated with extension in the cover due to the gravity gliding above the salt detachment above and after the second basement fault. In next section, we explain in detail how gravity gliding increases the amount of cover shortening relative to the basement in the salt-based models.
3.6 Shortening Velocities
 All experiments were shortened at a constant horizontal velocity of 8 mm/yr. In general, the local shortening velocity within the models decreases from the hinterland toward the foreland but also increases in cover sediments in the salt-present models. To illustrate these velocity variations, we plot in Figure 9 the horizontal velocities of Eulerian grid points for three different models: no-salt, partial-salt, and salt models, each having five pre-existing basement faults and MT = 400°C at 36 km depth. In the no-salt model, horizontal velocities decrease gradually from the hinterland toward the foreland, and cover and basement rocks shorten simultaneously at almost the same rate. As stated before, in the no-salt models, the basement and cover are coupled, and shortening is taken up by formation of structures in the cover and basement. Shortening velocities gradually decrease in partial-salt and salt models from the indenter toward the frontal part of the model, but because of the imbrication of basement blocks and uplift of the cover units, the sediments above the salt glide, due to gravity, toward the foreland, resulting in higher velocities. Gravity gliding is defined as downslope movement of a rock mass above a weak detachment surface or zone [Schultz-Ela, 2001]. The gravity gliding in our models locally increases horizontal velocity so that horizontal shortening rates can be higher in the cover than in the basement (Figure 9). In these models, horizontal velocities reach up to 16 mm/yr, almost double the indenter velocity, and in salt and partial-salt models, their distribution is heterogeneous (Figure 9). In the salt and partial-salt models, the cover units deform faster, and they are decoupled from the basement.
4 Discussion and Conclusions
4.1 Implications for the Zagros
 The geometrical, mechanical, and thermal parameters (i.e., cover thickness, Moho depth, salt distribution, Moho temperature, shortening rate, and total shortening) are different throughout the crust underling the Zagros, which must change the geometry, kinematics, and dynamics of deformational structures across and along the belt (Figure 1) [e.g., Sherkati and Letouzey, 2004, Jahani et al., 2009; Mouthereau et al., 2012]. Therefore, we avoid selecting any specific model to represent deformation in the Zagros. Rather, we discuss possible applications of our model to the Zagros fold and thrust belt in order to highlight processes that have influenced the evolution of this active orogen.
 From our simplified modeling results (e.g., Figures 6-8), we observe both imbrication of the basement that is decoupled from the cover [Molinaro et al., 2005; Mouthereau et al., 2006, 2007; Sherkati et al., 2006] and large displacements on pre-existing basement faults that cut through the cover, as proposed by Blanc et al.  and Alavi . Our modeling results find that salt distribution and geothermal gradient are key factors for controlling the crustal-scale deformation of the Zagros. The presence of a relatively thick (1–2 km) salt layer at the basement-cover interface together with a hot geotherm (higher Moho temperature) can prevent reactivated basement faults from propagating into the cover units. Such hidden (“blind”) faults have been discussed, for example, by Berberian  and Bahroudi and Talbot , and they are very important for earthquake studies of the Zagros [e.g., Nissen et al., 2011]. Moreover, future studies of these hidden basement faults might aid in improving balanced cross sections and determining the total shortening of the cover and basement across the Zagros more reliably. However, recent research by Mouthereau et al.  explored differential topographic uplift due to displacements on some of these basement faults in the Fars region (Figure 1). This means that although basement faults can be hidden beneath cover units, their reactivation can be observed in surface topography data.
 The wider extent of cover deformation in salt-based models is in agreement with previous research [Davis and Engelder, 1985; Cotton and Koyi, 2000; Bahroudi and Koyi, 2003; Nilforoushan and Koyi, 2007; Nilfouroushan et al., 2012] and indicates that salt distribution has a major influence on the structural and final wedge geometry of fold and thrust belts like the Zagros. Different distributions of salt beneath the cover sediments in the northwest and southeast regions of the Zagros can partly explain faster GPS-based shortening rates observed in the southeast compared to the northwest [Hessami et al., 2006; Nilforoushan and Koyi, 2007; Walpersdorf et al., 2006].
 Our modeling results also find that the rate of shortening of the cover and the basement can vary considerably. In cold models (MT = 400°C), imbrication of basement blocks occurs in the hinterland near the indenter, which can cause uplift of the cover units and consequently their gravity gliding above a relatively thick salt layer. In the Zagros, with a salt detachment thickness of 1–2 km, our modeling results support Molinaro et al.’s  proposal that more shortening has occurred in the cover than the basement if a cool geotherm is assumed (Figure 9). Therefore, different amounts of shortening and styles of deformation of cover and basement rocks in salt-based fold and thrust belts like the Zagros can be expected [e.g., Molinaro et al., 2005]. Molinaro et al. , however, suggested that multiple phases of deformation occurred in the Zagros in which thin-skinned cover deformation started first, subsequently followed by thick-skinned deformation expressed as out-of sequence faulting in the cover and reactivation of basement faults. Similarly, our model results also indicate that in the distal part of the system, the involvement of the basement is preceded by a phase during which only the cover is deformed (thin-skinned phase). This supports the kinematic scenario proposed by Molinaro et al.  or Sherkati et al.  and contrasts Mouthereau et al. , who suggested that the basement deformation is activated early, even at the deformation front. Further investigation using thermomechanical models can potentially resolve the issue of multiphase shortening in the Zagros.
4.2 Application to Other Fold and Thrust Belts (The Jura Mountains)
 Although we selected our model parameters for the Zagros, the models presented here also have implications for other mountain belts that are tectonically similar to the Zagros fold and thrust belt. For example, the Jura Mountains are a salt-based fold and thrust belt formed over a younger and hotter basement [Sommaruga, 1997, 1999; Mosar, 1999]. The Jura Mountains and the Swiss molasse basin represent the youngest deformation zone of the northwestern Alps [Sommaruga, 1999]. Here the Mesozoic and Cenozoic cover units were deformed above a weak basal detachment comprising Triassic evaporites [Smit et al., 2003]. The thickness of evaporates reaches 1 km, decreasing toward the frontal part of the orogenic wedge [Sommaruga, 1999]. The temperature at the brittle-ductile transition zone (BDT) is estimated to be around 450°C, basement faults extend to 15–20 km depth, and the Moho depth is around 25 km [Mosar, 1999]. Deformation in the Jura Mountains is distributed in several contrasting domains: long wavelength, low amplitude folding in the Molasse basin; thrusting and box folding in the High Jura; a mostly undeformed Jura Plateau; and imbrication in the frontal Faisceau zone [see Smit et al., 2003, Figure 17a; Sommaruga, 1999, Figure 3]. The mostly undeformed Jura Plateau is comparable to the results of our experiments. For example, experiments 5b and 5c, which have a similar thermal signature to the Jura mountains, are characterized by a less deformed wide plateau in the cover sediments in hinterland. In these models, cover deformation is mainly observed in the foreland, near the salt pinch out. Compared to our experiments, salt detachment and hot basement in the Jura fold and thrust belt probably contributes to formation of a less deformed plateau.
 A series of 2-D thermomechanical numerical experiments that focus on the Zagros fold and thrust belt, evaluated the possible interaction between pre-existing faults in Precambrian crystalline basement and its sedimentary cover containing a 1–2 km thick intervening layer of weak Hormuz salt. The results find that the degree to which pre-existing basement faults are reactivated is correlated to temperature-dependent ductile flow of the lower crust. A cooler lower crust prevents the transfer of deformation in the basement toward the foreland and only reactivates the basement faults in the hinterland. In relatively warmer models, the lower crust deforms by ductile flow, allowing the basement blocks to rotate and segment the salt layer. In general, salt-based experiments with and without pre-existing basement faults suggest that a cold rheology model simulates better the present structure of the Zagros, in which many detachment folds and thrust faults are observed in the cover. In hotter models with and without pre-existing basement faults, the cover is much less deformed owing to lower crustal ductile flow and most of the deformation occurs in the basement by folding, thrusting, or displacements along pre-existing faults. However, it is worth noting that other factors, including the geometry of pre-existing faults and the magnitude of the imposed strain rates, are likely also important for studies of hidden basement faults beneath a salt-based fold and thrust belt, and should be included in future investigations.
 The presence of a salt detachment layer near the indenter favors the uplift of basement blocks, resulting in a large amount of cover extension due to gravity gliding, which in turn drives shortening in the foreland. Our results indicate that the amount and style of tectonic deformation in the cover and basement and the degree of decoupling between them are strongly governed by the presence and distribution of the salt detachment in the Zagros.
 The thermal and mechanical parameters and the crustal configuration we employed for our numerical modeling were selected to study systematically the thick-skinned deformation of an idealized salt-based fold and thrust belt like the Zagros. We did not attempt to make our experiments simulate fully the complex tectonic evolution of the Zagros itself. However, the results provide insights into “Zagros-like” thick-skinned deformation and are a step further to understanding the interaction of cover and basement rocks by including a salt detachment, temperature effects, and pre-existing basement faults.
 Research Council of Sweden (VR) funds FN and HK. GMT free software was used for making Figure 1. RP and ARC were funded by Natural Sciences and Engineering Research Council of Canada (NSERC) Discovery grants. We are grateful for the detailed and constructive reviews of Jürgen Adam and Dominique Frizon de Lamotte. We also acknowledge Onno Oncken and Paola Vannucchi for handling our manuscript.