This study examines the role of erosion on exhumation and deformation history of accretionary orogens through experimental and geological methods. Fault propagation, kinematics, and exhumation rates in an eroded thrust wedge are analyzed through scaled sandbox simulations by testing different parameters (friction, rheologic layering, and sediment input/output ratio). A model Coulomb wedge is submitted to erosion under flux steady state conditions; that is, the volume of eroded material remains equal to the volume of newly accreted material, maintaining a constant surface slope during shortening. Results show that the way of exhumation depends on the internal dynamics of thrust wedges and, conversely, on how this dynamics is modified by erosion. In the eroded thrust wedges the diversity of exhumation patterns is controlled by the mode of fault propagation, depending on the basal friction (high or low). The vertical component of exhumation is generally higher for the wedges with high basal friction than for low-friction wedges. The uplift of material occurs along a cluster of subvertical thrusts in the middle part of the eroded thrust wedge with low basal friction. The material is exhumed along a series of inclined (20°–50°) thrusts in the rear of the high-friction wedge. The zone of maximum exhumation is generally localized in the middle part of the thrust wedge and migrates toward the backstop with continued shortening. The vertical exhumation rate increases with time, and the material accreted later is rapidly transferred to the main exhumation zone compared to the material accreted during the early stages. The exhumation occurs at rates 2 times faster in the wedges with half the thickness than in full thickness wedges. The total quantity of eroded material at the end of convergence constitutes 36–50% of initial model area for the thrust wedges with low basal friction and 20–40% for the high basal friction wedges. The extent of basal underplating increases with total shortening. The area of basal underplated material constitutes up to 30 and 40% of the eroded thrust wedge area for the models with low and high basal friction, respectively. Presence of décollements in the accreted series allows underplating of thrust units developing an anticlinal stack, whose growth and location is favored by erosion. Our results are compared to different present-day active and ancient convergent orogens (Himalaya, Taiwan, Southern Alps of New Zealand, and Cascadia orogen), allowing the different specific exhumation processes to be characterized.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
 It has long been understood that erosion has a significant impact on the tectonic evolution and structure of compressional belts. Burbank  demonstrated that spatial variations in erosion rates mimic variations in bedrock cooling rates at timescale of 106 yr. in areas under dynamic equilibrium. Recent studies provide a clear indication that surface processes, such as river incision, bedrock landsliding and glacial erosion, are able to produce erosion rates exceeding 5 km/Ma [Dahlen and Suppe, 1988; Gardner and Jones, 1993; Burbank et al., 1996; Hallet et al., 1996; Hovius et al., 1997; Norris and Cooper, 1997; Lavé and Avouac, 2000; Hovius, 2000; Dadson et al., 2003] that are sufficient to account for many of the rapid rates of unloading or cooling that are revealed by geobarometric or thermochronological studies. Exhumation rates from upper and midcrustal rocks in collisional mountain belts are commonly in the range 1–10 km/Ma like in Nanga Parbat, Pakistan, Nepalese Himalaya, Southern Alps of New Zealand, Olympic Range, Washington, Taiwan, western and central Alps [Tippett and Kamp, 1995; Harrison et al., 1997; Brandon et al., 1998; Zeitler et al., 2001; Bousquet, 1998; Bousquet et al., 2002; Dadson et al., 2003]. Thus it appears that combined tectonic and erosional processes are able to produce sustained exhumation rates in excess of 5 mm/yr. Interactions and relative importance of both tectonic and erosional processes are less well established. Regional field studies within many orogenic wedges (Southern Alps of New Zealand, European Alps, Andes, Taiwan, Tien Shan and the Himalayas) indicate that erosion in turn influences the distribution and propagation of deformation. Thus the way of exhumation depends on the internal dynamics of thrust wedges and conversely, on how this dynamics is modified by erosion.
 The impact of erosion on the dynamics of thrust wedges was studied in a number of papers using both analogue and numerical modeling techniques. Formation and evolution of discrete fault paths was studied in thrust wedges by numerical modeling using the hypothesis of minimum work [Hardy et al., 1998; Masek and Duncan, 1998], establishing that erosion as a parameter affecting topography can determine structural geometry (fault paths) of mountain belts. Complementary results were obtained by Beaumont et al. , who found that the highest rate of rock uplift correlate with zones of intense precipitation and erosion. Other models have demonstrated that syntectonic erosion affects the critical state of the orogenic wedge and that the dynamics between tectonics and erosion contain important feedback mechanisms such that orogenic systems tend toward a steady state [Jamieson and Beaumont, 1989; Avouac and Burov, 1996; Willett, 1999; Willett and Brandon, 2002; Hilley and Strecker, 2004]. Some of these models have been directly applied to specific orogens: Pfiffner et al.  discuss the effects of erosion on the evolution of the internal structure of the Swiss Alps, thermal and kinematic modeling of thermochronological ages are used by Batt et al.  to investigate the development of Olympic Mountains segment of Cascadia accretionary wedge and demonstrate a flux balance across the margin. Complementary studies have used analogue modeling techniques to study the role of erosion or sedimentation at different scales during the tectonic evolution of orogens. Some studies consider thrust development at regional scale, suggesting that synkinematic erosion-sedimentation influence fault propagation and geometry, favoring out-of-sequence thrusting and back thrusting [Cobbold et al., 1993; Larroque et al., 1995; Merle and Abidi, 1995; Storti and McClay, 1995; Mugnier et al., 1997; Leturmy et al., 2000; Casas et al., 2001; Barrier et al., 2002; Marques and Cobbold, 2002; Persson and Sokoutis, 2002; Del Castello et al., 2004], other studies deal with processes at the scale of the orogen [Koons, 1990; Davy and Cobbold, 1991; Persson and Sokoutis, 2002]. Combined analogue-numerical modeling was applied to better understand the interplay between surface (river) transport running parallel to the evolving mountain belt and deformation in compressional belts [Persson et al., 2004]. More recently, using scaled sandbox simulations, S. Hoth et al. (Influence of erosion on the kinematics of bivergent orogens: Results from scaled sandbox simulations, submitted to Journal of Geophysical Research, 2004) discussed the influence of erosion on the kinematics of bivergent orogens, showing that deformation responds immediately to erosion.
 However, numerous questions remain concerning the effects of erosion both on internal kinematics and exhumation paths in thrust wedge. In this paper, using sandbox experiments, we study the effects of erosion on structure, evolution, spatial pattern and exhumation rates in thrust wedges. A quantitative analysis of material transfer kinematics across model thrust wedges allows us a determination of the different modes of exhumation in response to erosion. We studied thrust wedges eroded under flux steady state conditions as defined by Willett and Brandon : the volume of eroded material remains equal to the volume of newly accreted material. A model Coulomb wedge is submitted to uniform surface slope erosion, i.e., maintaining a constant slope of erosion surface during shortening. The different patterns of fault propagation and exhumation style by erosion are described with respect to various controlling parameters such as basal friction, ratio of sediment input to output, action of a subduction window, sediment rheology, taper angle increase. The model results are used to better understand exhumation processes (see definitions by England and Molnar  and Ring et al. ) occurring in different examples of natural convergent orogenic wedges.
2. Experimental Method
 Scaled 2-D sandbox simulations are used to analyze deformation and exhumation processes occurring in thrust wedges suffering erosion. They have been chosen because they enable us to simulate large magnitudes of convergence. We describe 13 experiments chosen among the 18 realized for the study: 2 reference experiments without erosion and 11 experiments with erosion. Each experiment was performed under normal gravity using an apparatus (Figure 1) which simulates the basic geometry and the main mechanisms of a subduction zone. The classical sandbox device, close to the one used by Malavieille , is constituted by a horizontal basal plate bounded by two lateral glass walls. To reduce the amount of sidewall friction, a lubrication of glass walls was done before sand deposition. A 10 cm wide Mylar sheet lying on the horizontal basal plate, is pulled beneath a vertical rigid backstop allowing more than 150 cm of convergence for each experiment. Use of plastic sheets with a different degree of roughness allows us to simulate the effects of different frictions at the base of the layered incoming sand. A polished Mylar film produces low basal friction (LF) and a rough Mylar sheet surface simulates a high basal friction (HF). The rigid backstop mimics the upper plate against which the thrust wedge develops. A space can be opened at its base allowing gradual removal of a chosen amount of model material, thus simulating a subduction window. It was set to a height of 1 cm such that part of the section is “subducted” forcing a décollement to develop at the height of the window, so that the subducted section is significantly thinner than the incoming section.
 On the Mylar sheet, successive colored sand layers simulated the oceanic sediments. The sand input or initial thickness of sand layers (h) is 3.1 cm for the LF models and 3.6 cm for the HF models. Their accumulation builds an initial accretionary wedge which is subsequently eroded. Before shortening, a protowedge is built against the buttress to rapidly obtain a critically accreting thrust wedge. The height of the protowedge is 5.4 cm for LF and 10.4 cm for HF models, its slope angle is 15°.
 The model materials have frictional properties satisfying the Coulomb theory [Dahlen et al., 1984; Dahlen, 1984] and thus they are fair analogues of upper crustal rocks (see discussions by Lohrman et al. [2003, and references therein]). The aeolian sand used is rounded with a grain size of less than 300 μm and a density of 1690 kg/m3. The internal coefficient of friction is 0.57 and the cohesion Co = 20 Pa. A basal friction corresponding to these parameters is around 12 degrees for the LF model and around 24 degrees for the HF model. Décollement levels are created by introducing in the model thin (1–2 mm) layers of glass microbeads. They are a Coulomb material and their density and size are almost the same as those of dry sand, however due to their close to perfect roundness their coefficient of internal friction is about 23% smaller (0.44), with cohesion almost negligible. Resistant layers in the models are made of dry silica powder, which has a significantly higher cohesion (150 Pa) than sprinkled sand (20 Pa) to simulate stronger material. Cohesion and size are scaled with a factor of 105: 1 cm in our experiment is roughly equivalent to 1 km in nature (scaling and characterization of model materials are also discussed by Lallemand et al. ; Kukowski et al. , Gutscher et al. [1996, 1998b], and Kukowski et al. ). Two cameras and one video camera record all intermediate stages of the experiment (each 1 cm of shortening) allowing structural interpretations and quantitative analysis to be made.
 For each experiment, an inclined line, marking the projection of the chosen erosion surface, is plotted on the glass wall after 20 cm of shortening (Figure 2). The profile runs along the initially formed model thrust wedge and cuts the x axis at its toe. The imposed slope angle of the erosion surface is determined by the critical accreting taper angle measured from the experiments without erosion: 4° for LF and 8° for HF models (see below). Location of the erosion surface remains constant through shortening. Accretion of new material by frontal thrusting and/or basal underplating leads to vertical growing of the model wedge above the erosion surface. All material rising above the erosion surface is scraped off and removed by a vacuum cleaner. Erosion is allowed elsewhere above the erosion line, even near the rear or near the front of the wedge. The uplift of wedge topography occurred unequally across its strike, being the most important in the area of active faulting. Erosion is made gradually each 2 cm of shortening. Between erosion steps, the wedge is allowed to try to obtain its own critical taper. The natural equivalent of erosion during which a constant slope is maintained corresponds to a basis of erosion for the orogen under topographic steady state conditions, i.e., the elevation of Earth's surface does not change with time [Willett and Brandon, 2002]. Next, we describe and compare the results obtained on some significant experiments (chosen among the 18 realized for the study) varying important parameters, basal friction (LF and HF), ratio of sediment input (1/2 of thickness h), action of a subduction window, sediment rheology (impact of décollement levels or cohesive levels), taper angle increase.
3. Results of Modeling
3.1. Wedge Geometry, Mode of Fault Propagation, and Exhumation
 Two experiments without erosion were run, to be used as reference models when comparing the impact of various parameters tested in experiments with erosion. Main characters are the following. Model thrust wedges grow mostly by frontal accretion due to forward propagation of the thrust sequence [e.g., Davis et al., 1983; Malavieille, 1984; Malavieille et al., 1991; Mulugeta, 1988; Storti and McClay, 1995; Gutscher et al., 1998a, 1998b; Cobbold et al., 2001; Persson and Sokoutis, 2002; Persson et al., 2004; Smit et al., 2003]. The low basal friction wedge (LF0) grows by successive accretion of small thrust units bounded by pared antithetic faults (a forward propagating thrust coupled with a series of small back thrusts) at the wedge front (Figure 3a). The high basal friction wedge (HF0) grows primarily through underthrusting of long imbricated thrust slices (Figure 4a). Continuous accretion (the sand input is nearly equal, see legend of Figure 1) leads to frontal growth both of the LF and HF wedges, the former being 1.5 times longer than the latter after 100 cm of shortening. A zone of back thrusting controls the growth at the rear of both wedges involving the frontal part of the protowedge. The total vertical thickening relative to initial model setting reaches 274% for LF thrust wedge and 340% for the HF thrust wedge at about 100 cm of convergence. The slope angles of the thrust wedges obtained without erosion are measured after about 100 cm of convergence. They are respectively 4° for low basal friction and 8° for high basal friction. These angles, that reflect the imposed frictional strength, are maintained during shortening respectively in the next (high and low basal friction) experiments with erosion and assumed as critical angles for accreting wedges growth. Critical Accreting Taper Angle = CATA. Below, the experimental results are presented for eroded thrust wedges with respect to various controlling parameters.
3.1.1. Interplay of Basal Traction and Body Forces in a Neutrally Eroding Thrust Wedge
 The geometry of eroded thrust wedge results from a combination of underplating and erosion at the surface. The eroded thrust wedge seeks to reproduce the CATA corresponding to a steady state topographic condition. The fluctuations of the frontal thrust location are limited in the eroded thrust wedge, and the geometry of the wedge remains relatively stable through time of convergence (Figure 2). The length of the eroded thrust wedge relative to the length of the thrust wedge without erosion at about 100 cm of convergence is about 55% for the LF wedge and 75% for the HF wedge. The total vertical thickening relative to the initial model setting reaches 219% for LF thrust wedge and 200% for the HF wedge (Table 1).
Table 1. Measured and Calculated Data on Experimental Thrust Wedges With High (HF) and Low (LF) Basal Frictiona
Sinit, initial model area; Sinp, area of inputted material; Soutp, area of outputted material; Serod, area of eroded material; Sfinal, thrust wedge area at the end of convergence; Eeros, extent of erosion; Sund, area of underplated basal layer in eroded thrust wedge at the end of convergence; Si, area of basal layer material in initial model within the length equal to the final thrust wedge length; Eund, extent of basal underplating; Rund, ratio of basal underplating area (Sund) with respect to final area; Lconv, total convergence; Reros, rate of erosion; Lc, relative length compression; Ler, length of final eroded thrust wedge; Lner, length of thrust wedge without erosion at about 100 cm of convergence; Tv, total vertical thickening; H, thickness of final eroded thrust wedge in front of initial protowedge; h, thickness of initial model in front of protowedge. Sfinal = Sinit + Sinput − Serod − Soutp, Serod = Sinit + Sinput − Sfinal− Soutp, Eeros = Serod/(Sinit + Sinput) * 100%. Eund = Sund/Si * 100%, Rund = Sund/Sfinal * 100%, Reros = Serod/Lconv, Tv = H/h * 100%, and Lc = Ler/Lner * 100%.
 In contrast to constant geometry, the internal structure of an eroded thrust wedge evolves during shortening, newly accreted material is transferred through the wedge and exhumed at the surface. The mode of underplating and exhumation rates of accreted material differ markedly between the eroded thrust wedges with high and low basal friction.
 The exhumation in the eroded critical wedge with low basal friction (LF4) occurs along a series of high-angle to vertical thrusts in the middle part of the thrust wedge (Figures 2a and 3b). The gently dipping frontal thrusts (30°) become steeper (50°) to near vertical (90°) with continued shortening and migrate through the wedge toward the backstop. The forward thrusts becoming steeper go progressively inactive and most of the thickening at the rear of the wedge is achieved by the play of out-of-sequence back thrusts, favoring thrusts steepening. The backward thrust migration results from a combination of continuous frontal accretion of new material and erosion at the surface of the thrust wedge. Later, thrust faults are cut by a series of coupled back thrusts, become inactive and rotate out of a preferential slip orientation being passively displaced toward the rear part of thrust wedge. When a new thrust is formed in front of the eroded wedge to reach the conditions of critical taper, a series of coupled back thrusts is “moved” to the front of the wedge and the migration cycle is repeated. Through convergence, the frontal part of protowedge in the LF model is affected by one or two major back thrusts in the rear of the thrust wedge that is emplaced onto the protowedge.
 The surface uplift in the LF model occurs mostly in the thrust wedge between the major back thrust and frontal thrust. The frontal part of the protowedge affected by the major back thrust is eroded while its main body remains undeformed. The maximum of material uplift and erosion occurs at the location of active faults, a frontal thrust and a series of coupled back thrusts. The area of exhumation becomes progressively wider and narrower following the migration of the series of back thrusts toward the backstop and back to the front of the thrust wedge. The locus of maximum erosion follows the migration of the denudation area as determined by the internal structure of the thrust wedge.
 Continuous shortening and erosion of the LF thrust wedge lead to the final localization of the area of high-angle thrusts in the middle part of the wedge, where the most rapid exhumation of the underplated material occurs. The area of maximum exhumation is found at the front of the protowedge with respect to its initial setting.
 The exhumation in the eroded critical wedge with high basal friction (HF8) occurs along a group of differently inclined (20° to 50°) thrusts at the rear of the thrust wedge (Figures 2b and 4b). The newly accreted material forms long thrust slices that underplate the thrust wedge. The gently dipping (20°) frontal thrust faults of the eroded wedge become steeper (50°) with increasing convergence due to continuous accretion and basal underplating of new material. Through continuous convergence and erosion, the thrust wedge is internally compressed and thrust faults migrate toward the backstop becoming progressively inactive. The backward thrust migration occurs along a major back thrust that develops at the rear of the eroded thrust wedge and affects the frontal part of the protowedge. The protowedge for the HF model is deformed and rotated backward due to underplating.
 The material uplift in the HF model is observed both in the protowedge and the thrust wedge areas because of basal underplating. The zone of maximum uplift is located above active faults: a frontal thrust, out-of-sequence thrusts and a major back thrust. The eroded area is concentrated mostly in the thrust wedge, but the bent frontal part of the protowedge is eroded too. The most rapid denudation occurs in the area of active faulting.
 The maximum of accumulation and exhumation of underplated material is localized under the frontal part of the protowedge. The basal underplating lead to the whole protowedge area is also submitted to uplift with about 110% relative to its initial setting (Figure 4b).
3.1.2. Sediment Rheology (Cohesive Layer)
 The presence of the cohesive layer (silica powder) in the layered sequence of the initial model does not change significantly the mode of fault propagation and the exhumation of material in both low (LF4CL) and high (HF8CL) basal friction eroded wedge. The internal geometry of the critical wedges with and without cohesive layer (CL) remains similar (Figures 3c and 4c). However, the total vertical thickening relative to initial model for the CL wedges is slightly higher (230%) than the same parameter for the critical wedges without CL (219% for LF and 200% for HF) (Table 1).
3.1.3. Taper Angle Increase
 In the two following models, the new taper is imposed by a higher-angle erosion plane relative to initial critical taper. This setting likely represents an active orogen, where tectonic processes produce vertical surface uplift at rates higher than erosion rates. The new imposed erosion plane angle is 6.5° for thrust wedge with low basal friction (LF6) and 12° for thrust wedge with high basal friction (HF12). The new geometry of eroded thrust wedge differs for both the LF and HF models. Low and high basal friction wedges with a higher-angle erosion plane are about 12–13% shorter than the critical wedges of respective basal friction (Figures 3d and 4d). Their length consists of respectively 48% and 63% relative to the length of non eroded wedges for LF and HF model. The total vertical thickening relative to initial model at the frontal part of the protowedge area remains the same for critical and overcritical thrust wedges of low and high basal friction, respectively (Table 1). The high-angle LF wedge (LF6) is characterized by similar mode of fault propagation if compared to the critical one. However, the area of maximum exhumation is localized closer to frontal part of the high-angle wedge and not in the middle part as observed for the critical wedge. The exhumation is controlled by the combined displacements along both a major back thrust that is stabilized in the middle part of the thrust wedge and a series of frontal thrusts which rotate, become steeper and less active with increasing shortening. The underplated basal material is exhumed to form a dome-like structure in the frontal part of the wedge (Figure 3d). Because of these conditions, the zone of maximum exhumation moves to the front because more material is removed close to the toe of the model.
 The modes of fault propagation and exhumation in the high-angle HF eroded wedge (HF12) are similar to the one of the HF critical taper. Exhumation occurs along a series of differently inclined thrusts in the rear of thrust wedge, under the frontal part of the protowedge area (Figure 4d). However, the extent of underplating of basal material under the protowedge is essentially higher in the high-angle thrust wedge resulting in a more significant vertical material flow in the protowedge domain, which reaches 121% relative to the initial model and 110% relative to the critical thrust wedge (HF4).
3.1.4. Sediment Input (Half Thickness Wedges)
 Two experiments with input of a sand layer of 1/2 thickness are studied for low (LF4_1/2) and high (HF8_1/2) basal friction. Thickness of the sand input is important because the yield stress is pressure-dependent (and therefore dependent on thickness). They are 1.6 for LF and 1.8 cm for HF models instead of 3.1 and 3.6 cm, respectively. The height of the protowedge is 5.4 (LF model) and 5.2 cm (HF model), maintaining the slope angle of the protowedge (15°) equal to the one chosen in all others models. The total shortening is more important for the models with half the thickness than for other experiments (Figures 3 and 4). The final geometry of the wedges is characterized by a greater degree of total vertical thickening relative to the initial model: 275% for LF wedge and 322% for HF wedge. These values make them 1.3 and 1.6 times higher when compared to the eroded critical wedges of LF and HF, respectively (Table 1).
 Decreasing the model thickness by a factor of two does not significantly change the modes of fault propagation in the eroded thrust wedges (Figures 3e and 4e). The domain of maximum exhumation takes place in the rear part of the eroded thrust wedge, being controlled by a major back thrust. This is likely resulting from the large total shortening achieved in these experiments. It is not observed in the protowedge area for the LF model with half the thickness. The domain of maximum exhumation for the HF wedge takes place in both the protowedge and the neothrust wedge. The upward material flow in the protowedge area results from important basal underplating in the HF wedge and thickens 143% relative to initial model and 130% with respect to the critical thrust wedge (HF4).
3.1.5. Presence of a Subduction Window
 The deformation of eroded thrust wedges subjected to a deep “tectonic erosion” is studied for low (LF6SUB) and high (HF14SUB) basal friction models. The material accreted in these models is removed both by surface erosion and by transfer to a “subduction window”, thus the steady state is not maintained in these models. The subduction window induces the development of a major long-lived out-of-sequence thrust fault that bounds the base of protowedge (Figures 3f and 4g) and controls transfer of material across the wedge toward the window. Subduction of 1/3 of accreted material along this thrust leads to change the geometry of the developed thrust wedge. The wedges are shorter and represent 77% length of wedges without subduction window for the LF model and 93% for the HF model. The total vertical thickening relative to initial model represents 180% for the LF wedge and 225% for the HF wedge (Table 1). Fault propagation within the thrust wedge is similar to that observed in wedges without a subduction window. The exhumation occurs along subvertical and vertical thrust faults for the LF6SUB wedge and along a series of inclined (20–50° dipping) thrust faults in the HF14SUB wedge. The zone of maximum exhumation migrates toward the backstop both for LF and HF wedges, following the transfer of material toward the subduction window. Exhumation in the LF6SUB model is observed both in the thrust wedge and in the protowedge areas. It results from the uplift of material along a major back thrust developed in the protowedge. As a result, both the neothrust wedge and protowedge areas underwent denudation. In the HF14SUB wedge, denudation occurred mostly in the neothrust wedge area and in the frontal part of bending protowedge.
3.1.6. Presence of a Décollement Layer
 The eroded thrust wedge (HF6DL) is characterized by high basal friction and the presence of a low-friction décollement level (thin layer of glass microbeads). A 6° slope angle has been chosen for the imposed erosion profile to represent an overcritical taper with low basal friction setting.
 The décollement acts from the beginning of shortening to control the thickening of the wedge. The material located at depth below the weak layer is underplated under the frontal part of the protowedge (Animation 1) as occurring in other HF model wedges, while above the décollement the neothrust wedge is thickened mainly by frontal accretion (typical mechanism for LF wedge).
 The exhumation of the basal layers from below the detachment level is localized under the frontal part of the protowedge inducing the formation of a dome-shaped structure (Figure 4f). The exhumation of the basal material occurs along a series of inclined thrusts (20°–50°) that become steeper up to 60°–70° with continued shortening. The main back thrust growth at the rear of the thrust wedge affecting the protowedge at the final stages of shortening. It controls the farther upward transfer of the basal layers material within the area of maximum exhumation.
 The upper thrust wedge creeps along the décollement leading to forward propagation of the wedge by frontal thrusting. The exhumation in the upper thrust wedge occurs through thrust faults that become progressively steeper, to near vertical due to continued shortening. The upper wedge is compressed, and the former frontal thrust faults migrate toward the backstop as a result of combined frontal accretion and surface erosion. The coupling of the thrust faults in the upper wedge at the final stages of shortening leads to the formation of a synformal “klippe” completely detached from the basal layers of the model (Figures 4f and 10e–10g).
 The domain of maximum vertical uplift in the thrust wedge HF6DL develops between the main back thrust and the frontal thrust. The area of the protowedge is undeformed.
 The length of the wedge with décollement level at the end of shortening is 12% shorter than the HF thrust wedge without erosion and 27% shorter than LF wedge without erosion. The vertical thickening of the wedge reaches 222% relative to the initial model setting (Table 1).
3.2. Trajectories (Particle Transfer and Exhumation Paths)
 For all the experiments, markers of colored sand (particles) are included in the sand cake. They are located at 1 cm high from the model bottom and distributed with an interval of 5 cm along the initial model (Figure 2). The chosen particle paths and their final position are shown after total shortening be achieved (Figures 3 and 4). The particles were selected to characterize the transfer of material through the wedge at the early, middle and late intervals of shortening, similarly for all the models.
 Particle transfer paths are observed in non eroded thrust wedges at the base of the models, for both low and high basal frictions (LF0, HF0). The particle paths within these thrust wedges are inclined toward the backstop and do not vertically exceed the relief of the initial model (Figures 3a and 4a). The angle of the particle transfer path is gentle (10°) in the middle part of the LF wedge, becoming steeper (40°) near the major back thrust. The transfer paths at the rear of the HF wedge are inclined at 30°, subparallel to the back thrust.
 The material transfer in eroded thrust wedges follows paths inclined toward the backstop (Figures 3b–3f and 4b–4g). Some material particles are transferred within the wedge to zones of stagnation or to the subduction window, and will never be exhumed. Other particles are transferred at different vertical rates up to the surface in the zone of maximum exhumation. The paths of particle transfer to stagnation zones are inclined at 40° while the paths allowing exhumation are inclined at 50°–60° in the HF eroded wedges. The same types of paths for the LF eroded wedges are inclined at 30°–40° and 20°–30°, respectively. Particle transfer to the subduction window is accomplished along curved paths.
3.2.1. LF Eroded Thrust Wedges
188.8.131.52. Wedge With Critical Taper (LF4)
 Particles trajectories in zones of maximum exhumation are similar to one another (Figure 3b), but they describe upward motions at different vertical rates. Particle 2 reaches the surface at half of the total convergence, while particle 3 only reaches the surface at the end of the shortening (Figure 5b). Particle 1 remains in a zone of stagnation near the major back thrust and is not exhumed when shortening has ceased (Figures 3b and 5b).
184.108.40.206. Wedge With Cohesive Layer (LF4CL)
 The transfer paths of particles in zones of maximum exhumation are uniform (Figure 3c). Particles 2 and 3, accreted at different times, reach the surface simultaneously at the end of convergence (Figure 5c). This means that the later accreted particle 3 is exhumed through the wedge at a higher vertical rate than particle 2. Particle 1 attained 2/3 of the wedge thickness at the end of convergence and is not exhumed, but remains in a stagnation zone close to the major back thrust (Figures 3c and 5c).
220.127.116.11. Wedge With Higher Erosion Plane Dip Angle (LF6)
 The paths of particle 2 and 3 characterize material motion in zone of maximum exhumation with different vertical rate (Figure 3d). Particle 3 is accreted later than particle 2, but it is exhumed more rapidly at 2/3 of total shortening (Figure 5d). This results from different mode of the material transfer. Particle 2 is transferred mostly along the back thrust contouring the zone of maximum exhumation where material transfer is slow. Particle 3 is transferred more rapidly with the frontal thrust that steepens with increasing shortening (Figure 5d). Particle 1 remains in a stagnation zone in the rear of the thrust wedge.
18.104.22.168. Half Thickness (Sand Input = 2 cm) of the Model (LF4_1/2)
 Particles 1 and 2 are transferred across the wedge along similar paths (Figure 3e). The particles reach simultaneously the surface in zones of maximum exhumation at the rear of the wedge at half the total convergence (Figure 5e). The rate of vertical motion of later accreted particle 2 is higher relative to particle 1. The exhumation of particles in this thrust wedge is two times faster with respect to the wedge with double thickness (LF4) (Figures 5b and 5e).
22.214.171.124. Wedge With Subduction Window (LF6SUB)
 The particle paths illustrate different modes of material transfer through the wedge (Figure 3f). The formerly accreted particle 1 is transferred directly to the subduction window along a flat (3°) trajectory. This particle never reaches the surface of the wedge. The later accreted particle 2 is transferred along a trajectory that is inclined at 20°, then becomes flat (2°), following the backward movement of material toward the subduction window. At the end of shortening, this particle is accreted at the base of the protowedge above the major out of sequence thrust and it is also never exhumed (Figure 5f). Particle 3, accreted at later stages of shortening, is transferred along steeper (25°) trajectories and is exhumed in the central part of the thrust wedge.
3.2.2. HF Eroded Thrust Wedges
126.96.36.199. Wedge With Critical Taper (HF8) and With Cohesive Layer (HF8CL)
 Particles 1, 2, and 3 are accreted successively to different portions of the wedge at different time and they reach different zones at the end of convergence (Figures 5h and 5i). Particle 1 reaches the stagnation zone below the major rear back thrust (Figures 4b and 4c). Particle 2 is uplifted near the surface, while the particle 3 is exhumed in the zone of maximum exhumation. The vertical rate of exhumation is higher for particle 3 accreted later to the wedge.
188.8.131.52. Wedge With Higher Erosion Plane Dip Angle (HF12)
 The highest rate of vertical exhumation is observed for particle 2, which is exhumed at half of total convergence (Figure 5j). Particle 3 is exhumed at the end of convergence. Particle 1 is transferred toward the stagnation zone below the major back thrust (Figure 4d) and uplifted only up to half of the wedge thickness at the end of convergence (Figures 5j).
184.108.40.206. Half Thickness (Sand Input = 2 cm) of the Model (HF8_1/2)
 Particles 1 and 2 are transferred across the wedge following similar exhumation paths (Figure 4e). However, particle 1 is exhumed at 45% of convergence while particle 2 is uplifted to the same place at 64% of total convergence (Figure 5k). The vertical exhumation rate is two times higher for particle 1 in this model relative to particles of the wedge with double thickness (HF8) (Figures 5h and 5k).
220.127.116.11. Wedge With Décollement Level (HF6DL)
 The trajectories of the particles 1–3 characterize the transfer of basal material from a zone below the décollement toward the surface in the main exhumation zone (Figure 4f). Particles 2 and 3 are exhumed simultaneously at 83% of convergence, even though particle 3 was accreted later than particle 2 (Figure 5l). Particle 1 is exhumed at the end of convergence being the earliest accreted particle relative to 2 and 3 particles. The vertical exhumation rate for the basal material increased with shortening. The rate of exhumation of basal material in the dome-like structure of this wedge (Animation 1) is one of the highest with respect to all other high-friction wedges (Figure 5h and 5l).
18.104.22.168. Wedge With Subduction Window (HF14SUB)
 Particle 1 is transferred along a flat trajectory toward the subduction window at the end of convergence (Figure 4g). Particle 2 is uplifted to 2/3 of thickness of the wedge at the same time and then remains in a stagnation zone. Particle 3 is exhumed in the middle part of the wedge at the end of convergence, demonstrating the higher vertical rate of material uplift (Figure 5m).
3.3. Extent of Erosion and Basal Underplating
 To estimate the extent of erosion in experimental thrust wedges we identified the following areas (Figure 6 and Table 1): (1) Sinit, initial wedge area; (2) Serod, total eroded area (removed from the surface); (3) Sinp, total area added as “input” at the front (or accreted at the base) of the wedge; and (4) Soutp, total area removed as “output” at the subduction window. The Sfinal thrust wedge area at the end of convergence is measured from digitized model wedge. Sinit, Sinp, and Soutp are calculated from model wedge and experiment parameters.
 The area of eroded material is calculated as follows:
and the extent of erosion is estimated
The total eroded material at the end of convergence represents 36–50% of the sum of initial model area and area of inputted material for the LF thrust wedges and 20–40% for the HF thrust wedges. A 4 mm thick basal layer outlined by a specific color is located at the base of each model. The material of basal layer is underplated in the eroded thrust wedges to differing degrees. The area of basal underplating is defined as the area composed of basal layer material (detected due to its specific color) within eroded wedge. It is measured at the end of convergence on model wedges from different experiments. It represents 10–30% of eroded thrust wedge area for the LF models and 10–40% for HF models (Figures 6 and 7 and Table 1). The ratio (Rund) of basal underplating area (Sund) with respect to final area (Sfinal) is calculated as follows:
The extent of basal underplating (Eund) is calculated as a ratio of the area of underplated basal material (Sund) in eroded thrust wedge at the end of convergence with respect to the area of basal layer in initial model within the length equal to the final thrust wedge length (Si) (Figure 6):
The extent of basal underplating (Table 1) increases with total shortening in eroded thrust wedges (Figure 8a). The maximum extent of underplating is observed for the thrust wedges with a half of thickness (sand input = 1.6 and 1.8 cm for LF and HF models, respectively) that are characterized by highest total shortening. The extent of basal layer underplating in these wedges represents 334% for the LF model and 448% for the HF model with respect to the area of basal layer in initial model (Table 1). The extent of erosion likely has a linear correlation with the extent of basal underplating (Figure 8b). These empiric correlations obtained for model thrust wedges confirm that erosion directly influences exhumation.
 The geometry of the eroded thrust wedge remains stable throughout convergence. We used the following size parameters for the eroded thrust wedges: Lc, relative length compression, Tv total vertical thickening, and Sfinal – thrust wedge area at the end of shortening (Table 1). The values of these parameters are determined by the erosion surface. The erosion surface is not applied voluntarily, but it is imposed to the wedge after 20 cm of shortening running without erosion. The geometry of the initial wedge is determined by the internal wedge kinematics and fault propagation. Thus the chosen parameters are model consequences.
 The eroded thrust wedge represents 1/2 or 3/4 of the length of thrust wedge without erosion after about 100 cm of shortening for low and high basal friction models, respectively (Figure 6 and Table 1, Lc, %). The total vertical thickening (Tv, %) relative to the initial setting of model is doubled both for LF and HF eroded thrust wedges (Table 1). The same parameter for LF and HF non eroded thrust wedge at 100 cm of convergence is 2.7 and 3.4 times, respectively (Figure 6 and Table 1). The vertical surface uplift of the protowedge area occurs for the HF eroded thrust wedges due to basal underplating. This process is not observed for the LF thrust wedges where the vertical growth occurs mostly in the thrust wedge area.
 The different patterns of fault propagation in LF and HF eroded thrust wedges likely controls the diversity of exhumation modes in eroded thrust wedges. The uplift of material is achieved along a series of vertical thrusts in the middle part of the LF eroded thrust wedge (Figure 2a). The material is exhumed along a series of inclined (20°–50°) thrusts in the rear of the HF eroded thrust wedge (Figure 2b). The zone of maximum exhumation in eroded thrust wedge (i.e., the surface projection of the domain where maximum total underplating and uplift of basal material occurs), is located in front of protowedge for the LF initial model and under the frontal part of the protowedge for the HF model. The zone of maximum exhumation migrates toward the backstop as a result of internal shortening of eroded thrust wedge. At the end of convergence, exhumation occurs mostly due to the major back thrust developing at the rear of eroded thrust wedge.
 Generally, the vertical component of exhumation is higher for the wedges with high basal friction than for the wedges with low basal friction (Figure 9). During evolution of a single wedge, the vertical exhumation rate is generally higher for the particles accreted later, which are transferred to the main exhumation zone more rapidly along a shorter path than the particles accreted earlier. In some cases, the transfer paths are perturbed, and the early accreted particles are exhumed through the wedge more rapidly (and along steeper paths) than the later accreted particles (LF4, HF12 and HF8_1/2). Transfer path perturbation can occur due to either local activation of back thrust (LF4) or basal underplating leading to material uplift in former accreted units (HF12, HF8_1/2). Exhumation occurs twice as rapidly in the wedges with a half of thickness when compared to the full thickness wedges.
 The particles that are exhumed at nearby location may arrive simultaneously at the surface, or at different stages of the experiment. The early exhumed particles may be found closer to front of the wedges due to their steeper (and shorter) exhumation paths (LF4, LF6, HF12, HF6DL). The early exhumed particles may also be found closer to the rear side of wedge because of their steeper transfer paths through the wedge (HF8_1/2).
5. Natural Examples
 Since this study is based on analogue models with Coulomb behavior, our comparisons with natural orogenic wedges consider parts of low-temperature orogens where brittle behavior prevails.
5.1. Himalaya Orogen Wedge
 Several geographical domains can be distinguished from north to south across the Himalayan orogen: Tibet, High Himalaya, Lesser Himalaya, sub-Himalaya, and the Gangetic plain. The tectonic structure of Nepal Himalaya at the longitude of Katmandu is shown on the interpretative cross section (Figure 10a). The Main Himalaya Thrust (MHT) reaches the surface at the front of the Siwalik hills, where it coincides with the Main Frontal Thrust (MFT). The Main Boundary Thrust (MBT) separates Lesser Himalayan metasediments from the molasse deposits of the sub-Himalaya (the Siwalik hills). The Main Central Thrust (MCT) places the high-grade metamorphic rocks of the High Himalayan crystalline units over the Lesser Himalayan metasediments [Avouac, 2003]. Friction on the flat portion of the MFT-MHT is calculated to be low of the order of 0.1 or 0.3 at most, if a lithostatic pore pressure is assumed [Avouac, 2003]. The observation of microseismic activity along the fault [Avouac, 2003] provides additional constraint on the friction along MFT-MHT fault zone. As our study mainly concerns exhumation processes occurring in the southern frontal part of the Himalayan wedge, our results are complementary to the ones obtained from numerical modeling by Beaumont et al. [2001, 2004], which emphasize the role of channel flow on exhumation and normal faulting development in the High Himalaya.
 The Himalayan wedge has resulted from a combination of brittle-ductile underplating and erosion at the surface [Avouac, 2003]. The structural evolution of the Himalayan orogen includes the following principal events (Figures 10b–10d): emplacement of the MCT crystalline thrust sheet (45–16 Ma), emplacement of the Lesser Himalaya thrust sheet and development of the duplex (15–6 Ma), major phase of exhumation of the Lesser Himalaya duplex (5–0 Ma). Deformation and anatexis in the MCT zone occurred at about 22 Myr [Copeland et al., 1991; Hodges et al., 1996], an age consistent with slightly younger cooling ages in the rocks of the hanging wall, attributed to the emplacement of the MCT [Hubbard, 1989; Copeland et al., 1991; Hubbard et al., 1991; Hodges et al., 1996]. The onset of erosion of the High Himalayan crystalline rocks during the Early Miocene is supported by isotopic provenance data from the Dumri formation [Robinson et al., 2001]. Out-of-sequence reactivation of the MCT may have occurred by late Miocene to Early Pliocene times as indicated from monazite ages of rocks exhumed from midcrustal depths [Harrison et al., 1997; Catlos et al., 2001].
 The structural evolution of the Himalayan orogen may be compared with formation of experimental eroded thrust wedge with detachment level (HF6DL) (Figures 10e–10g). (1) The emplacement of the first thrust sheets in the model wedge occurred along the thrust that follows the main décollement level at the base of the protowedge (Figure 10g). This stage likely corresponds to the MCT crystalline sheet emplacement (Figure 10d). (2) During the following steps of shortening, the basal material of the sand model is thrust under the frontal part of protowedge and exhumed like a dome in the rear side of developed thrust wedge (Figure 10f). This domal structure likely corresponds to the exhumed Lesser Himalaya duplex (Figure 10c). A series of thrust sheets that are completely detached from the base slides above the growing domal structure moving to the front of the model wedge. The main décollement level is reorganized into a frontal thrust, its flattened portion under the series of thrust sheets and two faults that control exhumation of basal material in the domal structure (Figure 10f). (3) A new series of thrust sheets forms in front of the model wedge with continuous shortening (Figure 10e). The former series of detached thrust sheets is compressed into a synform. The main active faults become a new frontal thrust and a major back thrust at the rear of the thrust wedge. The forward propagation of frontal thrusts in the model wedge likely corresponds to propagation of deformation from MBT to MFT at about 5 Ma in the Himalaya orogen (Figure 10b). Similar processes could be applied to the NW Indian Himalaya, where underplating induces the growth of antiformal domes (e.g., Zanskar crystalline dome) and where late stage back thrusting develops, favoring late exhumation [Steck, 2003].
 Another comparison point arises from the observation of our models which suffered erosion during large deformation. Estimates of total shortening in the Himalayan wedge are obviously largely underestimated. Not only due to deformation and subsequent volume loss in the internal zones, but mainly because a substantial length of the imbricated thrust units in the foreland domains have been removed by erosion, making it impossible to measure shortening from balanced cross sections. This may explain the important discrepancies obtained when using different methods of calculation for the estimation of total shortening accounted for by Himalayan deformation since the onset of India Asia collision. Indeed, estimates from plate kinematics reconstructions (although widely discussed, are roughly 1500 to 2500 km), are significantly greater than values obtained from other geologic approaches (for example, see discussion by Matte et al. ). During the early stages of collision, about 250 km of the shortening was absorbed by the subduction of the Indian continental crust to depths as great as 250 km [e.g., Boutelier and Chemenda, 2003; Guillot et al., 2003], which is probably a maximum for this process. Another part was widely distributed in the deformation of the Asian lithosphere far from the Indian indenter. The last part occurred during the growth of the Himalayan thrust wedge, involving a large progressive shortening (possibly more than 1000 km) by thrust imbrication, that was probably the greatest part, but is more difficult to quantify. The missing volume of continental crust was removed by continuous erosion and deposited as sediments, far from the belt, in the Indus fan, Bengale fan and the Burma accretionary prism.
5.2. Taiwan Orogenic Wedge
 The young Taiwan thrust wedge has been studied for many years, mainly on land but also offshore [Malavieille et al., 2002, and references therein]. The relative motion between the Philippine Sea plate and the Eurasian plate has resulted in the progressive subduction of the Chinese continental margin and the development of the Taiwan thrust wedge. The sedimentary cover of the Chinese continental margin was accreted in the northwestern part of the island, and it is still accreting in the southwestern part, against a backstop formed by the pre-Tertiary rocks of the Central Range (Figure 11a) [e.g., Lu and Malavieille, 1994]. Shallow marine sequences of the passive continental margin and foreland sequences constitute the deformed units of the Coastal Plain, Western foothills and Hsueshan Range. During convergence, these were progressively accreted to the collision prism along a series of east dipping thrusts. The Central Range includes the Eocene and Miocene exhumed metamorphic rocks of the Backbone Range. The Western Foothills correspond to a fold-and-thrust belt affecting Oligocene and Miocene strata overlain by a 4-km-thick sequence of Pliocene-Quaternary molasse. About 8 cm/yr of plate convergence is taken up across the whole orogen, inducing intense shortening, thickening of the crust and exhumation of metamorphic rocks by combined erosion and underplating.
 The zircon fission track ages of surface rocks from Taiwan orogen are consistent with west to east accretion and shortening [Liu et al., 2001]. The erosion within the orogen is sufficient to exhume reset thermochronometers. The patterns on surface reset zircon fission track ages reflect synorogenic temperatures and exhumation rates (Figure 11b). The reset ages zones are located adjacent to the eastern mountain front and widen to the north, with increasing time of the arc-continent collision process (Figure 11c).
 The main structural features of Taiwan orogen may be compared with the eroded thrust wedge HF8_1/2. Both the Taiwan orogen and the model wedge are formed by a series of thrusts that become steeper from the frontal parts of these wedges to their rear areas (Figures 4e and 11a). The main exhumation zone is located at the rear of both thrust wedges (Figures 4e and 11c). The early exhumed particle in the model wedge is located closer to the major back thrust and the later exhumed particle, closer to the frontal part (Figures 4e and 5k). This feature is similar for the Taiwan orogen where the total reset zircon zone (low greenschist facies) is found closer to the retrodeformation front of the orogen in the east, while the partial reset zircon zone (prehnite-pumpellyite facies) is observed more to the west (Figure 11b) [Liu et al., 2001].
5.3. Cascadia Orogenic Wedge
 The Cascadia forearc high separates a relatively continuous forearc depression in the east from the accretionary wedge offshore to the west. The Cascadia accretionary wedge underlies most of the offshore continental margin. Rocks of the wedge are locally uplifted and exposed in the Olympic Mountains in northwest Washington [Tabor and Cady, 1978], where it is known as the Olympic subduction complex [Brandon and Vance, 1992]. This area represents the most deeply exhumed segment of the coastal mountain range extending along the Cascadia margin. Exhumation of this part of the Cascadia forearc high is assumed to have been dominated by erosion and not by extensional faulting as shown by apatite fission track data [Brandon et al., 1998].
 The Cascadia accretionary wedge has grown by frontal accretion and underplating in front of the Coast Range terrane (Figure 12a) [Clowes et al., 1987; Brandon et al., 1998]. Accretion of sediments from Cascadia basin was driven by northeast subduction of the Juan de Fuca plate. Only minor internal deformation is recorded in the deposits of the modern shelf basins, Olympic and Willapa-Grays Harbor basins [Adams, 1984]. These basins are inferred to move slowly to the northeast relative the North American plate due to horizontal shortening within the Olympic Mountains uplift [Brandon et al., 1998]. At the surface, horizontal shortening appears to be mainly localized in the vicinity of the uplift. The continuous movement of material into the Olympic Mountains uplift ensures that topography and rapid erosion of the mountains are sustained. The material flow through the accretionary wedge is suggested to account for the fanning of cleavage across the uplift [Brandon et al., 1998].
 The Cascadia wedge shows a nested pattern of reset age zones (Figure 12b) [Batt et al., 2001]. The reset ages are limited to the core of the range. With time, the reset age zones are expected to expand to the northeast margin of the orogen [Willett and Brandon, 2002]. Exhumation of deep levels of the Cascadia accretionary wedge is consistent with the observed topographic precipitation characterized by dominant wind in the direction of subduction motion [Willett, 1999].
 We observe a wide domal pattern of exhumed material in the experiments with low basal friction. The most pronounced exhumation of basal underplated material in a dome-like structure occurs in the model wedge with overcritical taper (LF6) (Figure 3d). When we changed the angle of the erosion surface to a maximum critical value, the domal structure was gradually displaced to the rear of the thrust wedge stimulating exhumation of the material particles previously accreted in the stagnation zone (Figure 13). To explain the anomalous surface uplift and deep exhumation in the Olympic Mountains, the presence of a 10-km-high arch in the underlying Juan de Fuca plate beneath the orogen was proposed by Brandon and Calderwood . Interestingly, in the wedge LF6, the dome-like exhumation is observed due only to overcritical erosion and produced specific thrust fault propagation. Neither extensional tectonics nor an arch in the downgoing plate are needed in this case.
 The zone of maximum exhumation is found in the middle parts of the LF model wedges (Figures 3b–3d), and this zone moves to the major back thrust with continued shortening as observed in the wedge with a maximum shortening (LF4_1/2) (Figure 3e). The exhumed material is transferred by steep, to near vertical thrusts, or by a combination of these thrusts and back thrusts (LF6). In contrast, the inclined (17°–20°) out of sequence thrusts, surround the main exhumation zone in the model wedge with a subduction window (Figure 3f).
5.4. New Zealand Orogenic Wedge
 The Southern Alps of New Zealand represent an example of young collisional orogen with high erosion rates [Beaumont et al., 1996; Batt et al., 2000]. The Southern Alps lie east of the Alpine Fault, which marks the boundary between the Australian and Pacific Plates (Figure 14). The Alpine Fault is primarily a transform fault that connects the west dipping Hikurangi subduction Zone in the north to the east dipping Puysegur subduction zone to the south. Prior to about 10 Ma, the Alpine fault accommodated pure strike-slip motion, but following the change in the relative plate motions at about 5 (9.8) Ma, the plate boundary became obliquely convergent. Currently, the Alpine fault must accommodate approximately 11 mm/yr of convergence in addition to about 38 mm/yr of transcurrent motion [Norris et al., 1990]. The consequence of this relative convergence is crustal shortening, thickening and the formation of the Southern Alps.
 The Southern Alps orogen has been interpreted as a result of middle to lower crustal detachment of the Pacific plate with down to the west subduction [Wellman, 1979; Norris et al., 1990; Koons, 1989; Beaumont et al., 1996]. The accretion and erosion are likely responsible for an east to west kinematic transport. The highest-grade rocks are exhumed and exposed in a zone adjacent to the main thrust (Alpine fault). The surface uplift in the Southern Alps is estimated to be as high as 10 mm/yr and the highest surface uplift and exhumation rates are observed along the western mountain front a few kilometers east of the Alpine fault [Walcott, 1984; Tippett and Kamp, 1993, 1995; Batt et al., 2000].
 The zone of maximum exhumation is located adjacent to the retro-step-up shear zone as a consequence of the basic asymmetry in the numeric model of Beaumont et al., 1996. The Southern Alps exhibits the climate and exhumation asymmetry characteristic of wind in the direction opposite to motion of the subducting plate [Willett, 1999].
 On the basis of numerical models, the asymmetric exhumation pattern with a sharp exhumation front on the retroside of eroded thrust wedge is interpreted to be a function of the subduction polarity [Beaumont et al., 1996; Willett, 1999]. In our experiments, the zone of maximum exhumation is located in close proximity to the major backstop in the model wedges with maximum of shortening (LF4_1/2, HF8_1/2) (Figures 3e and 4e).
 Analysis of model thrust wedges show that the way of exhumation depends on the internal dynamics of thrust wedges and conversely, on how this dynamics is modified by erosion [e.g., Horton, 1999]. A different style of fault propagation is observed in the eroded thrust wedges with low and high basal friction, respectively. The diversity of exhumation mode is mainly controlled by the geometry of imbricate faults and the mode of fault propagation. The material is exhumed along a series of vertical and subvertical thrusts in the middle part of the thrust wedge with low basal friction. It is exhumed along a network of inclined (20°–50°) thrusts in the rear of the wedge with high basal friction, respectively. As the entire wedge undergoes internal deformation, erosion controls the lifetime of individual faults, the degree of rotation of the faults and the location of the zone of maximum exhumation. This zone is generally located in the central part of the eroded thrust wedge and migrates toward the backstop with continued shortening. The vertical exhumation rate increases with time, and the material accreted later is rapidly transferred to the main exhumation zone, compared to the material accreted in the early stages. The exhumation rate is twice as fast in the model wedges with half the initial thickness with respect to the full thickness wedges. The vertical component of exhumation is generally higher for wedges with high basal friction than for wedges with low basal friction. The total eroded material at the end of convergence constitutes 36–50% of the sum of initial model area and area of inputted material for the thrust wedges with low basal friction and 20–40% for the thrust wedges with high basal friction. The extent of basal underplating increases with total shortening. The area of basal underplated material constitutes up to 30% and to 40% of the eroded thrust wedge area for the models with low and high basal friction, respectively.
 Comparison of experiments with different present-day active convergent orogens (Himalaya, Taiwan, Southern Alps of New Zealand, Cascadia orogen) emphasizes the role played by erosion and allows to better understand, how it influences the distribution and propagation of deformation which in turn controls the evolution of thrust wedges. Observation of analogue thrust wedges which suffered erosion during large deformation, suggests that estimates of global shortening in orogenic wedges may be largely underestimated, not only due to deformation and subsequent volume loss in the internal zones, but also when measuring shortening from balanced cross sections in the foreland domains where a substantial length of the imbricated thrust units is removed by erosion.
 We thank the reviewers, and the Editor, Peter van Keken, for their constructive comments. We particularly acknowledge Stephane Dominguez for help and constructive comments during experimental work, Marc-André Gutscher for his scientific suggestions to improve the manuscript and help with the English and Christian Romano for technical support. Elena Konstantinovskaia benefited of grants from the French MENRT to fund the Associate Professor position in Montpellier (2000–2003) and of grants from Russian Science Support Foundation (2004).