Shortening and structural architecture of the Andean fold-thrust belt of southern Bolivia (21°S): Implications for kinematic development and crustal thickening of the central Andes

The ∼7000-km-long Andean orogen of South America is the type example of an active continental margin orogenic system, and the central Andes have emerged as a focal point in debates over multiple orogenic processes (e.g., Oncken et al., 2006; DeCelles et al., 2015). In cordilleran-style orogenic belts, primary features observed in the overriding continental plate (e.g., deformation, topographic growth, magmatism, synorogenic sedimentation) are attributed to a diverse array of geodynamic processes, including the dynamics of plate convergence and properties of the subducting oceanic slab (e.g., Pardo-Casas and Molnar, 1987; Coney and Evenchick, 1994; Gephart, 1994; Lallemand et al., 2005), processes operating within the mantle lithosphere such as thermal weakening, mantle flux, and delamination of dense lithospheric roots (e.g., Isacks, 1988; Allmendinger et al., 1997; Lamb and Hoke, 1997; Ducea, 2001; DeCelles et al., 2009), or erosional modification of the orogenic wedge due to climatic variations (e.g., Horton, 1999; Willett, 1999; Montgomery et al., 2001; Sobel et al., 2003; McQuarrie et al., 2008a). Commonly the record of such processes is only preserved in the overriding plate. Geologic mapping and balanced cross sections provide an important framework within which to analyze the mechanisms of orogenesis because they provide plausible constraints on the geometry, kinematics, and magnitude of shortening, which are fundamental parameters for quantifying deformation rates and crustal thickness budgets (e.g., Sheffels, 1990; Kley, 1996; Kley and Monaldi, 1998; McQuarrie, 2002; Gotberg et al., 2010; Eichelberger et al., 2015).

In southern Bolivia (21°S; Fig. 1), efforts to understand processes operating within the central Andes are hindered by variable and uncertain published shortening magnitudes; this precludes assessment of how deformation in the thrust belt may be linked to differing geodynamic models. At this latitude, the retroarc thrust belt contains one of the best exposed stratigraphic sections in the central Andes (Sempere, 1995; Horton et al., 2001), is one of the few locations where hinterland intermontane basins preserve a record of Eocene to Miocene deformation (Horton, 1998), and is the site of published geophysical data sets that extensively characterize lithospheric-scale structure (e.g., Götze et al., 1994; Wigger et al., 1994; Schmitz and Kley, 1997; Beck and Zandt, 2002), making it an ideal location to illustrate deformation geometry and quantify crustal shortening. However, a regional cross section through the entire retroarc at 21°S is currently lacking. Current kinematic interpretations and shortening estimates rely on a composite of published cross sections from individual tectonomorphic zones, resulting in inconsistent shortening magnitudes (199–320 km) (Dunn et al., 1995; Kley, 1996; Baby et al., 1997; Müller et al., 2002; Elger et al., 2005; Oncken et al., 2006), and differing interpretations of how basement deformation is linked to deformation in the overlying sedimentary cover sequence (e.g., Baby et al., 1997; Müller et al., 2002; McQuarrie, 2002).

Recent paleoaltimetry data sets at ∼21°S suggest rapid surface uplift (∼2–3 km) of the Altiplano and Eastern Cordillera in the middle-late Miocene has been attributed to rapid removal of dense, eclogitic lower crust and mantle lithosphere (Sobolev and Babeyko, 2005; Garzione et al., 2006, 2014; Ghosh et al., 2006; Molnar and Garzione, 2007; Hoke and Garzione, 2008; Cadena et al., 2015; Wang et al., 2015). In addition, delamination-related magmatism and the felsic composition of the 60–70-km-thick crust below the southern Altiplano suggest that loss of lower crust has occurred in the central Andes (Kay and Kay, 1993; Beck and Zandt, 2002). However, a longstanding debate regarding the central Andes has focused on whether shortening can fully account for the observed 60–70-km-thick crust in the Altiplano, because many early shortening estimates could only account for 70%–80% of the observed crustal area (Sheffels, 1990; Kley, 1996; Allmendinger et al., 1997; Baby et al., 1997; Kley and Monaldi, 1998; McQuarrie, 2002; Gotberg et al., 2010). Recent orogen-scale cross sections, map-view and paleomagnetic rotation reconstructions, and foreland basin studies interpret that crustal shortening and accompanying underthrusting of foreland lithosphere are sufficient to account for the crustal area of the Andes between 15° and 22°S (Kley, 1999; McQuarrie and DeCelles, 2001; McQuarrie, 2002; DeCelles and Horton, 2003; McQuarrie et al., 2005; Hindle et al., 2005; Arriagada et al., 2008; Eichelberger and McQuarrie, 2015; Eichelberger et al., 2015). However, the wide range of existing shortening estimates still falls short of accounting for crustal area at 21°S (Kley and Monaldi, 1998). Given the current debates over the links between shortening, delamination, and surface uplift, the uncertain shortening magnitude at this latitude calls for reassessment of the crustal thickness budget. Such a study would provide a benchmark for understanding if there was sufficient excess material available for removal that would significantly affect surface elevations. In addition, a structural analysis across the entire backarc in southern Bolivia would provide key context for validating other geodynamic models in the central Andes, such as oroclinal development (e.g., Eichelberger and McQuarrie, 2015) or testing proposed links between several primary cordilleran processes (e.g., DeCelles et al., 2009).

To address these problems we incorporate new geologic mapping with published geophysical data to construct a regional cross section that encompasses the entire Bolivian retroarc at 21°S. This allows us to calculate a comprehensive estimate of crustal shortening, present a new kinematic model for thrust belt development, assess upper and lower bounds for the crustal thickness budget, and calculate constraints on the average crustal thickness during thrust belt construction.

Late Cretaceous–Cenozoic construction of the Andean orogen has been marked by progressive crustal thickening, magmatic arc development, and eastward advance of a retroarc fold-thrust belt and foreland basin system within the South American continental plate (Coney and Evenchick, 1994; Sempere et al., 1997; Horton and DeCelles, 1997; DeCelles and Horton, 2003; McQuarrie et al., 2005). The central Andes (∼16°–24 °S) are the widest part of the orogen, and have accommodated the greatest amount of crustal shortening (e.g., McQuarrie, 2002).

The central Andean retroarc has been divided into five tectonomorphic zones that are defined by major steps in mean topographic and structural elevation (Fig. 1; Kley, 1996; McQuarrie, 2002). From west to east, they include the hinterland plateau, the Altiplano, a broad fold-thrust belt that has been divided into the Eastern Cordillera (EC), Interandean zone (IAZ), and Subandean zone (SAZ), and the modern foreland basin, the Chaco plain. The Altiplano is a high-elevation, low-relief, internally drained basin bordered on the west by the Western Cordillera (the modern volcanic arc) and on the east by the EC (Isacks, 1988; Allmendinger et al., 1997; McQuarrie, 2002). The EC is the structurally highest part of the thrust belt, elevated relative to the adjacent Altiplano and IAZ. The EC is bivergent, characterized by west-vergent backthrusts in the west and east-vergent forethrusts in the east (McQuarrie and DeCelles, 2001; McQuarrie, 2002; Müller et al., 2002). Major changes in structural elevation between the EC and IAZ, and IAZ and SAZ, are interpreted as the result of the emplacement and passage of concealed basement megathrust sheets over crustal-scale footwall ramps (e.g., Kley et al., 1996; McQuarrie, 2002; Müller et al., 2002). The IAZ is a narrow transitional zone of intermediate structural and topographic levels between the EC and SAZ (Kley, 1996), and the SAZ encompasses the foothills of the eastern Andes (Dunn et al., 1995; Uba et al., 2009). Both the IAZ and SAZ are characterized as primarily east-vergent thrust systems detached from a basal décollement in Silurian shale (Baby et al., 1992; Dunn et al., 1995; Moretti et al., 1996; Kley, 1996). The modern foreland basin system is located in the Chaco plain, east of the Andean topographic front (Horton and DeCelles, 1997).

Orogenesis of the central Andes initiated and was likely focused along the western margin of the South American continent from the Late Cretaceous to mid-Paleogene (ca. 70–45 Ma) (e.g., Horton and DeCelles, 1997; Sempere et al., 1997; McQuarrie et al., 2005). The thrust front made a major eastward advance (∼400 km) to the EC by ca. 40 Ma (Horton, 2005; Elger et al., 2005; Ege et al., 2007), and has continued to propagate eastward to its present position in the Chaco plain (Horton and DeCelles, 1997; Brooks et al., 2011).

This study presents new geologic mapping at scales of 1:50,000 and 1:24,000, and is displayed in Plate 1 (scale 1:200,000). Mapping was conducted on 1:50,000 Instituto Geografico Militar Bolivia topographic base maps (retrieved from http://www.igmbolivia.gob.bo/), and 1:24,000 Bing Maps aerial imagery overlain with a 20 m topographic contour interval. Mapping was focused between 21°10′S and 22°10′S along an ∼300 km east-west transect from the Chaco plain to the eastern Altiplano. Map data were collected principally on the roads from Villamontes to Tarija, Tarija to Tupiza, and Tupiza to San Vincente, with several ancillary transects near Tarija and Tupiza (see Plate 1 for locations of transect routes). In the Chaco plain, seismic reflection profiles tied to well logs from Uba et al. (2009) were extrapolated to the surface. In addition, the new mapping was integrated with published geologic maps (Servicio Geologico de Bolivia, 1992; Horton, 1998; Müller et al., 2002; Servicio Nacional de Geologica y Tecnico de Minas, 2009, 2010). ArcGIS software (https://www.arcgis.com/) was used for compilation of published maps and digitization of field sheets. Stratigraphic and structural contacts were extended along strike using a combination of topographic maps, aerial imagery, and published geologic maps.

A balanced cross section, oriented perpendicular to the regional strike of thrust faults and folds, was constructed based on field data projected from east-west mapping traverses. Deformed and restored sections were hand-drafted using line-length balancing methods (e.g., Dahlstrom, 1969; Elliott, 1983), and are both displayed at 1:200,000 scale (Plate 2). Across-strike stratigraphic variation and near-surface geometry are obtained from apparent dips calculated from our mapping, and the manner in which the space above the basal décollement of the thrust belt is filled was supplemented by available published stratigraphic and geophysical constraints (e.g., Woodward et al., 1989; McQuarrie et al., 2008b). The final line-length balanced section was compared to an area-balanced cross section to ensure that shortening estimates were reliable, and to quantify total shortening uncertainties (e.g., Judge and Allmendinger, 2011; Allmendinger and Judge, 2013). Detailed justifications for individual drafting decisions on the cross section are annotated in Plate 2. More detailed discussions of the methods of cross section construction and the assessment of the uncertainty are found in the Supplementary Material¹.

The stratigraphic section observed along the mapped transect consists of an ∼10–15-km-thick package of predominantly siliciclastic sedimentary rocks ranging in age from Cambrian to Quaternary (Sempere, 1995; Tankard et al., 1995; Welsink et al., 1995). However, significant along- and across-strike variations in the stratigraphic thicknesses of Paleozoic–Mesozoic rocks (Plate 2), as well as regional erosional unconformities, are observed across the thrust belt, and are interpreted as a result of pre-Andean tectonism (Sempere, 1995; Tankard et al., 1995). Because formation names vary regionally, stratigraphic units are described here based on formation age. A detailed stratigraphic correlation chart with individual unit names can be found in Plate 1 and Figure 2.

In southern Bolivia, the Chaco plain, SAZ, and IAZ deform an ∼10–12-km-thick section of Silurian to Cenozoic clastic sedimentary rocks (Fig. 2; Baby et al., 1992; Dunn et al., 1995; Sempere, 1995). Gravimetric modeling indicates that Silurian rocks overlie crystalline basement, rather than the thick sedimentary section of Ordovician and Cambrian rocks observed to the west in the EC (Fig. 2; Plate 2; Kley et al., 1996; Schmitz and Kley, 1997). Across-strike thickness trends of Silurian rocks at 21°S are not precisely known because a complete section is only exposed in one location in the IAZ (Plate 1), and the base of the Silurian section is poorly imaged under the Chaco plain (Baby et al., 1992; Dunn et al., 1995). However, Silurian rocks are exposed across the SAZ, IAZ, and EC in central and northern Bolivia, and gently thicken westward (McQuarrie, 2002; Eichelberger et al., 2013). Silurian rocks at 21°S are therefore assumed to thicken westward from ∼1.2 km in the Chaco plain (Baby et al., 1992) to ∼1.6 km in the IAZ (Plate 2). Due to extensive surface exposures across the SAZ and IAZ (Plate 1), and well control and seismic data in the Chaco plain (e.g., Baby et al., 1992; Dunn et al., 1995; Uba et al., 2006, 2009), the thicknesses of Devonian through Cenozoic rocks are well constrained. The Devonian section maintains a thickness of ∼2.8 km across the Chaco plain and eastern SAZ, but thins westward to ∼1.5 km at the IAZ-SAZ boundary. The Carboniferous section thins westward, from ∼2.2–2.4 km in the Chaco plain and eastern SAZ (Baby et al., 1992; Dunn et al., 1995), to ∼1.5 km in the central SAZ, to ∼0.5 km at the IAZ-EC boundary (Plate 2). Permian rocks are ∼0.6 km thick across the Chaco plain, SAZ, and IAZ.

Triassic red beds and Jurassic basalts (Sempere, 1995; Bertrand et al., 2014) vary between 200 and 500 m thick, and are restricted to the western SAZ. Overlying Jurassic and Cretaceous rocks thicken from tens of meters in the Chaco plain (Uba et al., 2009) and eastern SAZ to ∼1.5 km in the central and western SAZ (Plate 2). As much as ∼4.3 km of Cenozoic synorogenic sedimentary rocks are preserved along the transect (Plate 2), and record the eastward advance of the Andean foreland basin across the SAZ and Chaco plain from ca. 27 Ma to the present (Horton and DeCelles, 1997; Echavarria et al., 2003; Uba et al., 2009; Hulka and Heubeck, 2010). No Cenozoic rocks are preserved in the IAZ, but Cenozoic apatite fission track (AFT) ages obtained there suggest that at least ∼3.0 km of Mesozoic–Cenozoic overburden was present prior to the onset of Andean deformation ca. 25–12 Ma, but has since been eroded (Ege et al., 2007).

The stratigraphy of the EC is characterized by a >10-km-thick section of upper Cambrian to Ordovician, low-grade (epizone-anchizone) metasedimentary rocks (Jacobshagen et al., 2002), which are unconformably overlain by narrow, north-south–trending outcrop belts of unmetamorphosed Cretaceous and Cenozoic rocks, which in turn are beveled by the Miocene–Pliocene San Juan del Oro erosional surface (Plate 1; Fig. 2) (Gubbels et al., 1993; Kennan et al., 1995, 1997; Horton, 1998, 2005; Müller et al., 2002; DeCelles and Horton, 2003). Early Paleozoic tectonism (Egenhoff, 2007) and a Late Devonian to Carboniferous orogenic event (Sempere, 1995; Jacobshagen et al., 2002) resulted in significant variation in Paleozoic stratigraphic thicknesses across the EC-IAZ boundary (Plate 2; Kley, 1996). In particular, the ∼4.3-km-thick section of Silurian–Permian rocks at the IAZ-EC boundary is not observed in the hinge zone of the Camargo syncline ∼50 km to the west, where Mesozoic–Cenozoic rocks unconformably overlie Ordovician rocks (Plate 1). Cambrian–Early Ordovician rocks thin to the west and Middle and Late Ordovician rocks thin to the east (Egenhoff, 2007). Early–Middle Ordovician rocks are not exposed west of Tupiza (Plate 1), but a velocity discontinuity at ∼8–10 km interpreted as the basement-cover contact (Wigger et al., 1994; Schmitz and Kley, 1997) coupled with the known thicknesses of overlying Late Ordovician rocks yields a thickness estimate of ∼5 km for Early–Middle Ordovician rocks in the subsurface of the western EC (Plate 2, annotation 17).

The Mesozoic–Cenozoic rocks of the EC are differentiated into separate western (Tupiza region; see detailed discussion in Supplementary Material, and Figs. SM1, SM2 [^{footnote 1}]) and eastern (Camargo syncline; Plate 1) outcrop belts based on contrasting sediment dispersal patterns, depositional setting, and lithology (Sempere, 1994; Sempere et al., 1997; Horton, 1998, 2005; DeCelles and Horton, 2003). Nearly ∼700 m of distinctive upper Cretaceous to Paleocene restricted marine rocks with pervasive paleosol development are preserved in the Camargo syncline, and directly overlie Ordovician rocks (Plate 1). These strata (El Molino Formation, Fig. 2) are significant as they are correlative throughout the central Andes, and reflect regionally extensive low topography prior to Andean deformation (Sempere, 1994; Welsink et al., 1995; Sempere et al., 1997; Horton et al., 2001). Overlying Cenozoic rocks preserve an ∼2.5-km-thick Paleogene foredeep section (DeCelles and Horton, 2003; Horton, 2005).

In the western Cenozoic outcrop belt, late Oligocene AFT ages suggest erosion of ≥3 km of Mesozoic–Cenozoic rocks, followed by Oligocene to late Miocene intermontane basin development (Horton, 1998, 2005; Müller et al., 2002; Ege et al., 2007). Near Tupiza (Fig. 1), an ∼750-m-thick section of Cretaceous sandstone, mudstone, siltstone, and interlayered basalt is preserved in a narrow north-south–trending syncline (Sempere et al., 2002), but the distinctive upper Cretaceous to Paleocene El Molino Formation is only present as recycled clasts within Cenozoic synorogenic sedimentary rocks (Horton, 1998). After erosion of the Paleogene foredeep section, between ∼0.6 and 2.3 km of sedimentary and volcanic rocks (Fig. 2) were deposited across 5 separate basins (Nazareno, Eastern Tupiza, Central Tupiza, Western Tupiza, and Estarca basins; Plate 1; Fig. SM2) during the Oligocene to late Miocene (Hérail et al., 1996; Horton, 1998; Müller et al., 2002). The San Juan del Oro erosional surface is a broad erosional pediment developed between ca. 12 and 3 Ma that beveled deformed Paleozoic to Cenozoic rocks and signified the end of major deformation in the EC (Gubbels et al., 1993; Kennan et al., 1995, 1997).

In the easternmost Altiplano, nearly 4 km of Cenozoic synorogenic sedimentary and volcanic rocks overlie the upper Cretaceous–Paleocene El Molino Formation (Fig. 2; Welsink et al., 1995; Elger et al., 2005; Servicio Nacional de Geologica y Tecnico de Minas, 2010).

In the following we describe, from east to west, the map patterns, structural style, and deformation geometry of each tectonomorphic zone of the Andean fold-thrust belt, as determined from our new mapping (Plate 1) and balanced cross section (Plate 2).

The western Chaco plain comprises the wedge-top portion of the modern Andean foreland basin system and encompasses the region between the main topographic front to the frontal-most thrust ∼60 km to the east (Tankard et al., 1995; Moretti et al., 1996; Horton and DeCelles, 1997; Uba et al., 2006, 2009). Surface exposures consist of Neogene synorogenic sedimentary rocks and Quaternary sediment (Plate 1). One prominent feature is a north-trending exposure of gently folded Neogene sedimentary rocks in the hanging wall of the surface-breaching Mandeyapecua thrust, which marks the deformation front of the thrust belt. It is debated how much active shortening is distributed on the Mandeyapecua thrust (e.g., Mugnier et al., 2006; Brooks et al., 2011; Yagupsky et al., 2014), but the trace of the gentle topographic ridge associated with the fault is continuous along strike from ∼18° to 22°S and shows some evidence of Quaternary rupture (Costa et al., 2006; Brooks et al., 2011).

In the Chaco plain, the basal décollement is located within shales near the base of the Silurian section (Baby et al., 1992; Dunn et al., 1995). Based on seismic reflection data ∼65 km to the south (Baby et al., 1992), we assign a 10 km depth to décollement at the eastern limit of the cross section (see Plate 2, annotation 3, and discussion of uncertainty in Supplementary Material [^{footnote 1}]). Seismic reflection data tied to well logs also constrain a 1°–2° west dip for the basal décollement (Baby et al., 1992; Dunn et al., 1995; Moretti et al., 1996; Uba et al., 2006, 2009).

The subsurface structure of the Chaco plain is characterized by gentle (5°–15° limb dips) fault-propagation folds above east-vergent, low-offset (∼1–4.5 km) thrust faults that ramp up from the basal décollement (Plate 2). However, the Mandeyapecua thrust is an emergent thrust with a hanging-wall syncline-anticline geometry produced from fault-bend folding above a multibend thrust fault. In the footwall of the Mandeyapecua thrust, Devonian–Silurian rocks are folded and overlying Carboniferous–Cenozoic rocks are elevated above their regional level (e.g., Uba et al., 2009), suggesting the presence of a concealed thrust that tips out within the Carboniferous section ∼10 km east of the emergent thrust front (Plate 2). Sedimentological and geochronologic data record arrival and unsteady advance of the deformation front in the Chaco plain by ca. 5.9 Ma, followed by westward retreat of deformation to the SAZ between ca. 4.0 and 2.1 Ma, and refocusing of the deformation front at the Mandeyapecua thrust since ca. 1.5 Ma (Uba et al., 2009).

The SAZ is defined by a Valley-and-Ridge–style topography characterized by anticlinal ranges separated by synclinal valleys, which are continuous for hundreds of kilometers along strike (Fig. 1; Plate 1). Folds and thrust faults are east vergent, with an overall north-northeast–trending (∼005°–020°) structural grain (Plate 1). Folds generally exhibit broad wavelengths (5–10 km), with Carboniferous to Cretaceous rocks exposed in the cores of anticlines and extensive exposures of Neogene synorogenic sedimentary rocks in the hinges of intervening synclines.

Teleseismic data image a subhorizontal velocity discontinuity at ∼10–12 km depth across the SAZ, which is interpreted as the contact between Silurian sedimentary rocks and denser basement (Wigger et al., 1994). This is consistent with an interpreted 1°–2° west dip for the basal décollement projected westward from the Chaco plain under the SAZ (Plate 2, annotation 1; Dunn et al., 1995). The anticlinal ranges of the eastern SAZ are narrow in width and structurally simple, with smaller offset (∼5–7 km) thrust faults branching off the basal décollement, breaching the cores of several anticlines as a result of progressive growth of fault-propagation folds. In the western SAZ, large-scale fault-bend folding of thrust sheets occurs as larger offset (∼8–20 km) thrust faults cut upsection from the Silurian décollement to secondary detachment levels in Devonian shale and Triassic evaporites (Plate 2, annotations 7–8; Dunn et al., 1995).

In map view, the SAZ-IAZ boundary is demarcated by a westward transition to thin, closely spaced (∼2–3 km) thrust sheets that carry upper Devonian to Carboniferous rocks, indicating an increase in structural level. In addition, the widely spaced Valley-and-Ridge topography of the SAZ gives way to more closely spaced, sharp-crested ranges separated by deeply incised valleys, with a gradual westward increase in elevation (Fig. 1). The basal décollement under the western SAZ is inferred to gradually steepen toward the mountain belt interior due to loading and flexure of the underthrusted Brazilian shield (Lyon-Caen et al., 1985). The dip of the décollement increases to 4° (e.g., McQuarrie, 2002) at approximately the SAZ-IAZ boundary, where the elevated structural level implies increased loading on the underlying Brazilian shield (Plate 2, annotation 5). A dramatic increase in the sedimentation rates of synorogenic rocks in the western SAZ beginning ca. 12–8 Ma is interpreted as the result of enhanced crustal flexure induced by the approaching deformation front, and is inferred to be the time frame for the onset of SAZ deformation (Echavarria et al., 2003; Uba et al., 2009).

In the IAZ, the structural grain has a dominant north-northeast (∼005°–020°) trend, and faults and fold axes are more closely spaced (1–3 km). The western half of the IAZ is deformed into an imbricate stack of dominantly west-vergent thrust faults, whereas the eastern IAZ is deformed into an imbricate stack of dominantly east-vergent thrust faults (Plate 2). The stratigraphic exposure level changes from the SAZ to the IAZ, and is dominated by Devonian and Silurian rocks (Plate 1). The deepest stratigraphic levels are exhumed in the central IAZ, where the change in vergence direction produced a structurally complex synformal klippe (Plate 2).

Across the IAZ, the basal levels of tightly folded synclines containing Carboniferous rocks define a regional structural level that suggests a horizontal décollement at a relatively shallow depth (∼5–6 km) (Plate 2; Kley et al., 1996). Upper Silurian rocks are the deepest exposed rocks in the IAZ (Plate 1), and the décollement under the IAZ is therefore interpreted to be located within the same Silurian stratigraphic horizon as the basal SAZ décollement. On the basis of the known stratigraphic thickness of the Paleozoic section from surface data (Fig. 2), the décollement in the IAZ is elevated ∼11 km relative to the basal décollement in the western SAZ (Plate 2). The increase in structural elevation at the IAZ-SAZ boundary is also coincident with the eastern edge of a residual Bouguer gravity anomaly and a lateral change in the velocity structure of the upper crust (Plate 2, annotation 12; Wigger et al., 1994; Kley et al., 1996; Schmitz and Kley, 1997). Such a regional change in exposure level requires the presence of a major footwall or hanging-wall ramp, and the filling of space below the elevated Silurian décollement in the IAZ can be accomplished in two end-member scenarios: (1) structural repetition of Silurian–Neogene sedimentary rocks, thereby doubling the structural thickness of the section, or (2) by emplacement of a thick thrust sheet of pre-Silurian basement rocks (Kley et al., 1996; Dunn et al., 1995; McQuarrie, 2002). Gravimetric modeling and velocity models from teleseismic data favor the latter, and the gravity data suggest density values typical of true crystalline basement and not the Ordovician–Cambrian metasedimentary rocks observed in the eastern EC (Wigger et al., 1994; Kley et al., 1996; Schmitz and Kley, 1997). Although crystalline basement is not observed along the mapped transect, the crystalline composition of the concealed basement thrust sheet is justified by (1) the history of early Paleozoic extension and basin formation along the axis of the eastern EC that resulted in eastward tapering and erosion of the Cambrian–Ordovician section and likely deposition of Silurian rocks disconformably on crystalline basement in the IAZ and possibly the SAZ (Kley et al., 1996; Egenhoff, 2007); and (2) Precambrian metagranodiorite exposed at the EC-IAZ boundary at 22°S (Servicio Geologico de Bolivia, 1992; Kley, 1996), 75 km south of our transect, that is involved in thrusting in the eastern EC in northern Argentina (Mingramm et al., 1979; Allmendinger et al., 1983; Mon and Hongn, 1991).

On our cross section, the basement thrust sheet interpreted below the IAZ is drafted as a single intact sheet (e.g., McQuarrie, 2002) that feeds slip forward into the basal Silurian décollement under the SAZ. This geometry requires a corresponding footwall ramp through basement rocks, the location of which is dictated by the total shortening accommodated in the SAZ and Chaco plain. The cumulative slip documented in the SAZ and Chaco plain (Plate 2) restores the eastern edge of the basement thrust sheet to the subsurface projection of the eastern limb of the Camargo syncline, a regional-scale fold in the eastern part of the EC (Plate 2). The 35°–40° west-dipping rocks in the eastern limb of the Camargo syncline are therefore interpreted to reflect the subsurface geometry of a basement footwall ramp. This interpretation is validated by seismic reflection data at ∼75 km along strike to the south that reveal an interpreted 35°–40° west-dipping ramp at the same across-strike location and depth below the Camargo syncline as our basement footwall ramp (Plate 2, annotation 12; Allmendinger and Zapata, 2000). The thickness of the interpreted basement thrust sheet is equal to the increase in structural elevation across the SAZ-IAZ transition (∼11 km), and the geometry of the leading edge corresponds to the footwall cutoff angle imaged below the Camargo syncline (∼35°–40°) (Plate 2).

Our IAZ geometry differs from published cross sections along this transect, which have characterized the subsurface geometry of the cover sequence as dominated by east-vergent thrust faults and folds, and have interpreted the geometry of the leading edge of the basement thrust sheet as an imbricate fan rather than a single intact sheet (Kley, 1996; Kley et al., 1996; Müller et al., 2002; Elger et al., 2005; Oncken et al., 2006). Both interpreted basement geometries are plausible, as the surface geometry data do not favor one model over the other, and no basement rocks are exposed at the IAZ-SAZ boundary. However, the imbricated basement sheet model implies a different footwall ramp position (further east) than the single intact thrust sheet model, and requires a footwall geometry characterized by a series of ramps that progressively step up through ∼11 km of basement rock over an east-west distance of ∼75 km (e.g., Kley, 1996), rather than a flat décollement underlying the basement thrust sheet. In addition, our interpreted subsurface geometry is consistent with the height, dip, and across-strike location of the footwall ramp imaged in seismic data ∼75 km to the south (e.g., Allmendinger and Zapata, 2000) and the subhorizontal velocity discontinuity that extends from the Chaco plain to the top of this footwall ramp (Wigger et al., 1994; Schmitz and Kley, 1997). Furthermore, our interpreted intact basement sheet geometry is mechanically favorable because the weak basal Silurian shale, which is interpreted as a regional décollement level (e.g., Baby et al., 1992; Dunn et al., 1995), is continuously exploited as slip transitions from the IAZ to the SAZ, instead of breaking new, steeply dipping faults through stronger crystalline basement. This geometry also allows us to more clearly define the IAZ-SAZ boundary as the eastern tip of the basement thrust sheet, and the IAZ-EC boundary at the westernmost backthrust just north of the town of Tarija (Plates 1 and 2).

The timing of Andean deformation within the IAZ is broadly bracketed between ca. 25 and 10 Ma (Gubbels et al., 1993; Horton, 2005; Ege et al., 2007), but the sequence and rate at which the deformation progressed are unknown. Therefore, deformation in the IAZ is assumed to have progressed in sequence from west to east.

The EC is bound on the east by the Sama-Yunchara anticlinorium, and on the west by the west-vergent San Vincente thrust (Plates 1 and 2). The EC is structurally elevated with respect to the Altiplano and IAZ, and the change in exposure level from the IAZ (mainly Devonian–Silurian rocks) to the EC (lower Ordovician–Cambrian rocks) requires that the primary décollement horizon is located within a deeper stratigraphic level than in the IAZ and SAZ. Basement-involved deformation is required to explain the structural elevation of the EC because the basal décollement of the SAZ-IAZ projects below the deepest Paleozoic sedimentary rocks in the EC and Altiplano (Plate 2; e.g., McQuarrie, 2002; McQuarrie et al., 2008b). Furthermore, seismic refraction and gravity data across the EC preclude imbrication or doubling of the Paleozoic sedimentary section to fill this space (Götze et al., 1994; Wigger et al., 1994; Kley et al., 1996).

The interpreted basement geometry in the EC is important because it influences how deformation at depth balances shortening observed at the surface, and the kinematic style through all levels of the thrust belt (e.g., cf. Plate 2 in this paper and fig. 12 of Müller et al., 2002). Two end-member geometries for basement deformation of the EC have emerged: (1) an east-vergent stack of long, thick basement thrust sheets that feed slip into a principal décollement located at the basement-cover interface via a roof thrust system (e.g., McQuarrie, 2002; McQuarrie et al., 2005, 2008b; Eichelberger et al., 2013), or (2) basement rocks that are deformed through duplexing, or by multiple, deep-seated east- and west-vergent basement faults that both feed slip into and cut across the sedimentary cover interface (e.g., Baby et al., 1997; Müller et al., 2002; Elger et al., 2005; Oncken et al., 2006).

The geophysically imaged crustal structure across the EC at 21°S is characterized by subhorizontally oriented velocity domains, with an interpreted basement-cover interface located at a relatively shallow depth of ∼7–10 km below sea level (Plate 2, annotation 17; Wigger et al., 1994; Schmitz and Kley, 1997). In addition, a subhorizontal band of strong mid-crustal reflections seismically imaged at ∼7–13 km below sea level under the Camargo syncline at 22°S is interpreted as the basal EC décollement (Plate 2, annotation 12; Allmendinger and Zapata, 2000). The subsurface projection of the basement-cover interface below the Camargo syncline matches with the depth and geometry of horizontal mid-crustal reflections imaged ∼75 km to the south (Allmendinger and Zapata, 2000) as well as the ∼8–10 km depth to the velocity discontinuity across the central and western EC at ∼21°S (Wigger et al., 1994). We therefore interpret the main décollement horizon for upper crustal shortening to be relatively flat, and located at the basement-cover interface (e.g., McQuarrie, 2002; McQuarrie et al., 2008b; Eichelberger et al., 2013). At deeper levels, teleseismic data show subhorizontal low-velocity zones at ∼15–20 km depth across the western EC, and at ∼30 km depth across the central EC at ∼20°S (Beck and Zandt, 2002). The depth to these mid-crustal low-velocity zones is consistent with the depth to décollement for a stack of two ∼10–12-km-thick basement thrust sheets (Plate 2; McQuarrie et al., 2005). Interpretation of multiple east- and west-vergent basement faults or duplexing of basement would impart a complex structural topography that is not observed in the EC, and is not demonstrably kinematically viable (McQuarrie, 2002; McQuarrie et al., 2008b). Also, flexural modeling suggests that deep-seated basement faults in the EC would have produced excessive subsidence and much thicker synorogenic sedimentary thicknesses than are observed in the Chaco plain (Prezzi et al., 2009).

The stacked basement thrust sheet model, with an upper crustal décollement located at the basement-cover interface, is therefore favored in this reconstruction because it is most compatible with the observed features discussed here. It is also the simplest kinematic approach to balancing basement and upper crustal shortening, with a geometry that accounts for the structural steps across the tectonomorphic zones (McQuarrie et al., 2008b). Such a basement geometry has long been argued for the central Andes (McQuarrie and DeCelles, 2001; McQuarrie, 2002; McQuarrie et al., 2005, 2008b; Horton, 2005; Eichelberger et al., 2013), and similar basement megathrust sheets are documented in other orogens (e.g., Mitra, 1978; Boyer and Elliott, 1982; Hatcher and Hooper, 1992; DeCelles and Mitra, 1995). Furthermore, numerical models show that orogens loaded by moderate to thick syntectonic sedimentary deposits can develop long, thick, crustal-scale basement thrust sheets (Erdos et al., 2015). However, the present-day average crustal velocities observed for the Altiplano and the presence of localized partial melts at ∼14–20 km depths at the Altiplano-EC transition (Beck and Zandt, 2002) are unlikely to sustain the development of the discrete basement faults that we interpret at the western edge of our cross section. Instead, we acknowledge that deformation at these depths may be more diffuse, perhaps occurring along shear zones of as much as kilometer-scale thickness.

Based on the dominant vergence direction of structures, the EC is further divided into two zones (e.g., McQuarrie and DeCelles, 2001; McQuarrie, 2002), discussed in the following.

The eastern forethrust zone encompasses the region from the IAZ-EC boundary to the Nazareno basin (Plate 1), and is characterized by two regional-scale structures: the Sama-Yunchara anticlinorium and the Camargo syncline (Plate 1). The ∼55-km-wide, north-plunging Sama-Yunchara anticlinorium (Kley, 1996) is a structural high observed continuously along the eastern margin of the EC from northern Argentina to northern Bolivia (Allmendinger and Zapata, 2000; McQuarrie, 2002). The eastern limb consists of an ∼10°–20° east-dipping homocline of an ∼12-km-thick, structurally intact section of Ordovician to Devonian rocks (Plates 1 and 2). The western limb dips ∼5°–40° west toward the hinge zone of the Camargo syncline, where lower Ordovician rocks are unconformably overlain by Cretaceous rocks (Plate 1). North-trending, ∼3–7-km-wavelength fault-propagation folds are superimposed on the broader anticlinorium, and grew above low-offset (∼1–3 km), generally east-vergent thrust faults (Plate 2).

The western limb of the anticlinorium is interpreted as a ramp anticline above a deep-seated basement footwall ramp, and the eastern limb corresponds to a hanging-wall ramp that must cut upsection to the east from basement through the Cambrian–Ordovician section (Kley, 1996; Kley et al., 1996). Deformation in the EC and IAZ is linked to progressive eastward emplacement of the same basement thrust sheet, which is stacked above the basement thrust sheet that feeds slip into the SAZ. However, we interpret the structurally higher basement thrust sheet as a tectonic wedge that distributes a component of its slip into the sedimentary cover sequence in the EC via a west-vergent roof thrust system at the basement-cover interface (e.g., McQuarrie, 2002; Horton, 2005) and the remainder of its slip forward into the sedimentary cover of the IAZ (Plate 2). This model allows the simplest basement geometry interpretation because it requires only one major basement footwall ramp for EC-IAZ deformation (Plate 2).

The Camargo syncline is traceable for ∼780 km from northern Argentina to central Bolivia. On our transect, the syncline exhibits a gently dipping eastern limb and a steeply dipping to overturned western limb. The western limb is bound by an east-vergent thrust fault that places Ordovician rocks over Mesozoic–Paleocene rocks. The difference in metamorphic grade of Ordovician rocks on either side of the syncline suggests pre-Cretaceous vertical offset (Müller et al., 2002; Jacobshagen et al., 2002). Where the depositional contact is exposed, Cretaceous rocks in the western limb overlie Ordovician rocks that are ∼3 km stratigraphically deeper than on the eastern limb (Servicio Geologico de Bolivia, 1992; Plate 1). However, the overlying Paleogene rocks are folded into a tight anticline inset within the larger syncline (Plate 2, annotation 14). Therefore, we interpret a west-dipping, pre-Andean thrust fault below the Camargo syncline that was reactivated during Andean deformation.

Between the Camargo syncline and the Nazareno basin (Plate 1), lower Ordovician rocks are deformed by tight folds (∼1–2 km wavelength) and closely spaced (∼1–2 km) thrust faults with stratigraphic separations typically ≤2 km (Plate 2). This interpretation is in contrast with that of Müller et al. (2002), who interpreted the region between the Camargo syncline and Nazareno basin as a west-vergent imbricate fan.

The western backthrust belt of the EC has long been recognized as a continuous along-strike feature throughout much of the central Andes (Kley et al., 1997; Baby et al., 1997; McQuarrie and DeCelles, 2001; Müller et al., 2002). However, differing structural interpretations in the Tupiza region (Horton, 2000; Sempere, 2000) have hindered efforts to obtain reliable shortening estimates through a significant portion of the western EC along our transect (Müller et al., 2002). Through reexamination of key structural and stratigraphic relationships (Figs. SM1–SM5; see ^{footnote 1}), and construction of a sequentially restored section (Fig. SM6) that matches the synorogenic depositional history of the Cenozoic sedimentary basins near Tupiza (e.g., Hérail et al., 1996; Horton, 1998; Sempere, 2000; Müller et al., 2002), we show that thrust faults and folds are dominantly west vergent from the western edge of the Nazareno basin to the easternmost Altiplano (Plates 1 and 2). This interpretation facilitates a balanced solution through the Tupiza region that kinematically links deformation between the eastern and western halves of the EC, and allows us to quantify the magnitude of out-of-sequence shortening that was previously unconstrained. A detailed discussion of the sequential restoration through the Tupiza region is included in Supplementary Material sections 2 and 3 (see ^{footnote 1}).

The eastern half of the backthrust belt (Nazareno basin to Estarca basin) is characterized by five Cenozoic basins bound on at least one of their margins by thin, west-vergent thrust sheets (∼1–2 km) carrying middle-upper Ordovician rocks. The structural grain is north-trending (350°–010°), faults are closely spaced (∼1–4 km) with maximum displacements of ∼2–4 km, and folds are short wavelength (<2 km). The western half of the backthrust belt was constructed as an imbricate thrust system that soles into a higher, secondary décollement at the base of the middle Ordovician section (Plate 2). Thrusts carrying thick, east-dipping sheets of upper Ordovician rocks are widely spaced (∼5–12 km), with maximum displacements of ∼5–10 km. West of the Estarca basin, thrust faults and folds trend 345°–350°. This oblique structural grain is only observed in Ordovician rocks and may be attributed to a pre-Andean structural fabric that was reactivated during Andean deformation (Müller et al., 2002).

We include the easternmost Altiplano in our description of the EC, as these two zones are kinematically linked (Elger et al., 2005). In the footwall of the San Vincente thrust, well logs and seismic data reveal ∼2–3 km of Cenozoic synorogenic rocks above ∼11 km of Paleozoic–Mesozoic sedimentary rocks (Welsink et al., 1995; Elger et al., 2005). This constrains the depth to basement in the eastern Altiplano at ∼13 km below sea level and reveals ∼6 km of vertical offset of the basement-cover interface across the Altiplano-EC boundary (Plate 2, annotation 22). This structural step has been attributed to a west-dipping monocline of basement rock (Müller et al., 2002; Elger et al., 2005). However, because the EC is modeled here as a roof thrust system above a basement thrust sheet, we interpret this structural step as the basement footwall ramp that corresponds to the hanging-wall ramp described at the IAZ-EC boundary (e.g., McQuarrie and DeCelles, 2001; McQuarrie, 2002). West of the San Vincente thrust (Plate 1 and Plate 2, annotation 23), mapping (Servicio Nacional de Geologica y Tecnico de Minas, 2010) and seismic reflection data (Elger et al., 2005) define a broad wavelength (>20 km) anticlinorium of Paleozoic rocks with secondary, shorter wavelength (∼10 km) folds in overlying Cenozoic synorogenic rocks. A third, westernmost basement thrust sheet (Plates 1 and 2, western basement sheet) is interpreted to structurally elevate Paleozoic rocks by passing over a footwall ramp and feeding slip into the basement-cover interface below the EC and easternmost Altiplano (e.g., McQuarrie, 2002, Fig. 3 therein).

Restoring the EC to the upper Cretaceous horizontal datum reveals the presence of the probable Jurassic Tupiza rift basin (e.g., Sempere, 1994; Sempere et al., 2002), and a pre-Mesozoic (likely late Paleozoic) west-dipping monocline (e.g., Müller et al., 2002) (Plate 2, annotation 22). Cenozoic AFT ages support a locally thick Mesozoic section near Tupiza (Ege et al., 2007), but we show that extension and later inversion of normal faults was likely low magnitude (Plate 2, annotation 19; Fig. SM6 [see ^{footnote 1}]). A pre-Cretaceous (likely Carboniferous) crustal shortening event has also been interpreted in the EC; however, this event produced subvertical penetrative cleavage and small-scale folding in Ordovician rocks rather than a system of discrete, large-offset faults (Kley et al., 1997; Jacobshagen et al., 2002; Müller et al., 2002). Except where pre-Andean offsets are noted, all shortening in our estimate is related to Andean deformation.

AFT ages are interpreted to mark the onset of deformation in the central EC ca. 40–36 Ma, which spread toward the Altiplano and IAZ from ca. 33 to 25 Ma (Ege et al., 2007). Later phases of deformation in the EC are typically not well resolved (Ege et al., 2007), but the sequential restoration through the western Cenozoic outcrop belt (Fig. SM6 [see ^{footnote 1}]) reveals 20 km of west-vergent, out-of-sequence deformation between ca. 29 and 10 Ma, contemporary with eastward advance of the deformation front through the IAZ (Horton, 2005; Ege et al., 2007).

Crustal shortening is calculated from the measured difference between the line length-balanced deformed and undeformed cross sections in Plate 2 (results summarized in Table 1). Uncertainty in the shortening estimate is assessed by producing an area-balanced solution, which permits formal propagation of errors on input parameters such as stratigraphic thickness, depth to décollement, and eroded hanging-wall cutoffs (e.g., Judge and Allmendinger, 2011; Allmendinger and Judge, 2013) (detailed discussion in Supplementary Material [see ^{footnote 1}]). The line-length balanced and area-balanced solutions for total shortening produce similar results, both for the thrust belt as a whole and for individual tectonomorphic zones (Table 1; Tables SM1 and SM2 [see ^{footnote 1}]). Error magnitudes are reported assuming a Gaussian distribution of uncertainty (Judge and Allmendinger, 2011).

Shortening estimates for each tectonomorphic zone are shown in Table 1. The EC accommodated 120 km of shortening (37%), and the IAZ accommodated 70 km (70%). Individual estimates of uncertainty for the EC and IAZ were not possible due to the limitations of the area balance software for the complex deformation geometry at the IAZ-EC boundary (e.g., Judge and Allmendinger, 2011; Supplementary Material [see ^{footnote 1}]). However, we combined the two zones because they are kinematically linked, resulting in a shortening estimate of 190 ± 46 km (43% ± 6.9%). The SAZ accommodated 82 ± 21 km of shortening (36% ± 6%). The total shortening estimate from the EC, IAZ, and SAZ is 272 ± 67 km (42% ± 7.8%). The total retroarc shortening was determined by adding our estimate to the 65 km of shortening estimated for the Altiplano at this latitude (Elger et al., 2005), for a total of 337 ± 69 km of shortening (36% ± 7%).

Previous total shortening estimates at 21°S are based on a composite of cross sections through individual tectonomorphic zones (Table 2), and vary depending on the combination of structural studies used, from a minimum of 199 km (Dunn et al., 1995; Kley, 1996; Müller et al., 2002; Elger et al., 2005) to a maximum of 319 km (Kley, 1996; Baby et al., 1997; Müller et al., 2002). The maximum composite shortening estimate is within our uncertainty range, but our range is above the 199 km minimum. Although our estimate of SAZ shortening is ∼20 km less than the previous estimates at this latitude (Tables 1 and 2), it is comparable to other SAZ shortening estimates between 15° and 23 °S (60–86 km; Table 3). The overall increase in total shortening arises from our deformation geometries in the IAZ and EC that differ from published cross sections, and by accounting for the unbalanced Tupiza segment in the EC.

Given the significant errors that we estimate for individual tectonomorphic zones, it is important to place these uncertainties in context (e.g., Eichelberger et al., 2013). The area-balance solution is not constrained by any particular kinematic model, and provides a shortening estimate that encompasses all line-length balance solutions for a given area (Judge and Allmendinger, 2011). In contrast, the line-length balanced cross section (Plate 2) considers geometric constraints from geophysical and map observations, and is balanced by slip on the underlying basement thrust sheets (e.g., viability; Woodward et al., 1989). As a consequence, increasing or decreasing the magnitude of shortening will result in a change in footwall and hanging-wall cutoff positions. Given the known thicknesses of the Paleozoic–Mesozoic section from geologic mapping, the location of the Camargo syncline, and the geophysical constraints on the positions of the footwall and hanging-wall cutoffs of the SAZ basement thrust sheet (Plate 2, annotations 6a and 12), errors as large as ±20 km in the SAZ are difficult to reconcile. However, given the lack of constraints on the basement geometry below the EC (Plate 2, annotations 6a and 22), it is possible to allow for a ±46 km (or even ±69 km) change in the position of the EC-IAZ basement sheet footwall ramp, or a different basement geometry altogether (e.g., Baby et al., 1997; Müller et al., 2002). We state this with the caveat that any change to the basement geometry must still honor the consistent structural level observed across the EC and the changes in structural elevation at the EC-IAZ and EC-Altiplano boundaries.

At a broader scale, the shortening estimate must also be compatible with the regional kinematics required for the development of the Andean oroclinal bend. Our central estimate of total shortening is comparable with the patterns of retroarc shortening estimated from balanced cross sections throughout the central Andes (Fig. 3). The highest shortening magnitudes occur near the central core of the orocline (17°–21°S), but decease substantially along strike to the north and south. Given this along-strike shortening gradient, our shortening estimate range (268–406 km) should produce clockwise rotations of ∼5°–15° in the southern limb of the orocline (e.g., Kley, 1999). This is consistent with ∼6°–14° limb rotations predicted from simplified reconstructions of the orocline and observed paleomagnetic rotations (Eichelberger et al., 2013; Eichelberger and McQuarrie, 2015). Although the upper end of our estimate is rather large, it is still compatible with (but at the limit of) the crustal shortening predicted by regional models of oroclinal bending (Fig. 3; Arriagada et al., 2008). However, more robust three-dimensional (3D) kinematic models that consider the complex map-view kinematics that occur during development of curvature in the core of the central Andean orocline predict ∼320–350 km of shortening in the southern limb (Fig. 3; Eichelberger and McQuarrie, 2015), on par with our shortening estimate of 337 km.

For these reasons, we consider our central estimate to be the most reliable estimate, because it is derived from a kinematically consistent reconstruction, and also closely matches the crustal shortening predicted from the 3D kinematic model of the central Andean orocline. However, the minimum and maximum shortening estimates cannot be totally discounted, because looser constraints on the EC and Altiplano basement geometry allow for the freedom to modify the crustal ramp locations without violating regional kinematic constraints.

Our cross section provides a new geometric and kinematic model for the development of the retroarc thrust belt at 21°S that is compatible with published cooling ages (Ege et al., 2007) and the foreland basin migration history (Gubbels et al., 1993; Horton, 1998, 2005; Echavarria et al., 2003; Uba et al., 2009) in southern Bolivia (Fig. 4), and is consistent with recent models of the Andean thrust belt in central and northern Bolivia (McQuarrie, 2002; McQuarrie et al., 2005, 2008b; Horton, 2005; Eichelberger et al., 2013). The majority of deformation in the EC and IAZ is linked to earlier emplacement of the upper basement thrust sheet (Figs. 4B–4D), with a separate, westernmost basement thrust sheet that fed a minor amount of slip (∼30 km) into the EC and easternmost Altiplano (Fig. 4C). Deformation in the SAZ is linked to emplacement of the younger, lowermost basement thrust sheet (Fig. 4E).

Although climatic effects such as orographic precipitation may exert a control on the distribution of cooling ages, especially in the northern Bolivian Andes (Barnes et al., 2012), the timing of deformation and initial rapid exhumation are synchronous in the central Andes (Barnes and Ehlers, 2009). In particular, AFT ages demarcating initial cooling are related to uplift and exhumation accompanying eastward propagation and passage of basement thrust sheets over large-scale footwall ramps (Barnes et al., 2008; McQuarrie et al., 2008b). In our cross section, the footwall ramp of the upper basement thrust sheet is restored to a position directly below the central EC (Fig. 4A), where ca. 40–36 Ma AFT ages (Ege et al., 2007) are interpreted to mark the initiation of deformation and uplift as the basement sheet began propagating eastward (Fig. 4B). AFT ages of ca. 33–25 Ma define the westward and eastward expansion of deformation into the backthrust and forethrust belts of the EC (Ege et al., 2007), contemporary with deposition and eastward migration of foredeep deposits (Fig. 4C; DeCelles and Horton, 2003; Horton, 2005). This bivergent expansion of exhumation is explained by insertion of the upper basement sheet as a tectonic wedge, because vertical uplift should occur both above the active basement footwall ramp and above the advancing tip of the basement sheet as it was wedged eastward (Fig. 4C). As deformation was distributed into the sedimentary cover in both the EC backthrust and forethrust belts via a west-directed roof thrust system, the upper basement thrust sheet was progressively wedged eastward, which drove flexural subsidence and eastward migration of the foredeep (Fig. 5C). The transition from foredeep to intermontane sedimentation in the EC after ca. 25 Ma, combined with a ca. 18–15 Ma AFT age in the IAZ (Ege et al., 2007), indicates diminished shortening within the EC and eastward migration of the deformation front into the IAZ. The deformation front migrated eastward across the IAZ from ca. 25 to 10 Ma by imbrication of the sedimentary cover in front of the advancing upper basement sheet (Fig. 4D). The kinematics of basement deformation shifted from tectonic wedging to translation of basement slip forward into the décollement below the IAZ (Figs. 4C–4D), possibly as a consequence of thickness changes of Paleozoic rocks across the IAZ-EC boundary, or as weaker Silurian shale at the basement-cover interface was encountered at the western edge of the IAZ. Backthrusting in the western half of the IAZ may have acted as a taper building mechanism in response to a subcritical wedge state, which could have been induced by erosional removal of material from the wedge (Davis et al., 1983; Willett, 1999; Malavieille, 2010), negative isostatic buoyancy from a dense lithospheric root (Kay and Kay, 1993; DeCelles et al., 2009, 2015; Wells et al., 2012), or a change in basin taper due to sedimentary thickness changes between the EC and IAZ (e.g., Boyer, 1995).

Although the majority of shortening between ca. 25 and 10 Ma took place in the IAZ, the west-directed roof thrust system below the EC remained active, at least intermittently, because we estimate >20 km of west-directed, out-of-sequence shortening in the Tupiza region between ca. 25 and 10 Ma (Fig. 4D). The development of the ca. 10 Ma San Juan del Oro erosion surface marks the cessation of deformation in the EC and eastward migration of the deformation front into the SAZ, suggesting that emplacement of the upper basement sheet was complete by this time (Gubbels et al., 1993; Horton, 1998, 2005). A dramatic increase in sedimentation rates of synorogenic rocks in the eastern SAZ beginning ca. 12.4–8.5 Ma, interpreted as the result of enhanced crustal flexure induced by initiation of the lower basement sheet, is inferred as the timing of onset for SAZ deformation (Echavarria et al., 2003; Uba et al., 2009). The deformation front migrated eastward across the SAZ as the sedimentary cover sequence was imbricated in front of the advancing lower basement sheet (Fig. 4E). During this time, the advancing basement thrust sheet drove the flexural foredeep eastward to its current position (Horton and DeCelles, 1997; Prezzi et al., 2009; Uba et al., 2009) while passively translating rocks in the EC and IAZ eastward.

Although magmatic addition may be an important mechanism for crustal thickening in Cordilleran orogens (e.g., Allmendinger et al., 1997), numerous studies have demonstrated that horizontal shortening is the most significant contributor to crustal thickening and resultant surface uplift in the central Andes (Isacks, 1988; Sheffels, 1990; Schmitz, 1994; Allmendinger et al., 1997; Baby et al., 1997; Kley and Monaldi, 1998; Kley et al., 1999; McQuarrie and DeCelles, 2001; McQuarrie, 2002; McQuarrie et al., 2005). The crust of a retroarc fold-thrust belt is tectonically thickened as foreland lithosphere is subducted toward the hinterland by an amount equal to the magnitude of upper crustal shortening (Fig. 5A) (DeCelles and DeCelles, 2001). In this basic model, the crustal budget, or the expected cross-sectional area of the crust, is calculated by multiplying the undeformed original length of the thrust belt by an initial preorogenic crustal thickness (Fig. 5A; Kley and Monaldi, 1998; McQuarrie, 2002).

A major uncertainty in the crustal budget calculation is the initial thickness of South American continental crust prior to Cenozoic shortening, which may have varied across strike (McQuarrie, 2002). The undeformed foreland lithosphere at ∼21°–20°S has an observed thickness of 35–40 km from teleseismic data (Wigger et al., 1994; Beck et al., 1996; Beck and Zandt, 2002; Baumont et al., 2002). Although Late Triassic to Middle Jurassic (Sempere et al., 2002) and Late Cretaceous (Salfity and Marquillas, 1994) rifting affected southern Bolivia (McGroder et al., 2015), lower crustal granulite xenoliths from the Salta Rift of northern Argentina (∼23°–26°S) suggest that crustal thinning was minimal, and that the crust was at least ∼35 km thick ca. 90 Ma (Lucassen et al., 1999). In addition, geochemical proxies for crustal thickness (Sr/Y and La/Yb ratios) suggest that the western edge of the South American crust was ∼35–43 km thick at the onset of Andean shortening (ca. 80–70 Ma; Profeta et al., 2015). Expected crustal thicknesses are compared to the modern crustal area (Fig. 5B), which includes the geophysically determined crustal area of the orogen at this latitude (∼32,563 km²; e.g., Beck et al., 1996; Beck and Zandt, 2002), as well as the area of rock projected above the erosion surface (1023 km²). Along-strike estimates of the crustal thickness budget for the central Andes are displayed in Figure 3.

For a 35-km-thick initial crust, the expected crustal area for the minimum, central, and maximum shortening estimates for the entire retroarc result in a deficit of 3486 km² (90%), a deficit of 1071 km² (97%), and an excess of 1334 km² (104%) compared to the observed crustal area at 21°S (Table 4). For a 40-km-thick initial crust, our range of shortening estimates predicts 2%–18% of excess crustal area (Table 4). Therefore, our shortening estimate range is in agreement with other studies along strike that show that upper crustal shortening can account for, and in some cases exceed, observed modern crustal thicknesses in the central Andes (McQuarrie and DeCelles, 2001; McQuarrie, 2002; DeCelles and Horton, 2003; Hindle et al., 2005; McQuarrie et al., 2005; Eichelberger et al., 2015).

Excess crustal area implies a loss or redistribution of material, either by lower crustal flow or delamination of dense lower crust (Beck and Zandt, 2002). Accordingly, because the thick crust below the central Andes of southern Bolivia is predominantly felsic, the mafic component of the lower crust is interpreted to be missing (Beck and Zandt, 2002). Delamination of dense eclogitic lower crust and mantle lithosphere has been called upon as a mechanism to explain the absence of a mafic lower crust, as well as a cause for rapid Miocene surface uplift events in the EC and Altiplano (Garzione et al., 2006, 2008, 2014; Molnar and Garzione, 2007). Petrologic and numerical models indicate that crustal thicknesses of ≥50 km are required to generate eclogitic phases from mafic lower crust (Kay and Kay, 1993), and that thicknesses of ≥45–50 km must be achieved to drive lower crustal flow (Husson and Sempere, 2003). Furthermore, numerical models emphasize that a sufficient thickness (∼10 km) of eclogitic lower crust must be lost along with the mantle lithosphere in order to significantly affect surface elevation (Garzione et al., 2006; Hoke and Garzione, 2008; Krystopowicz and Currie, 2013; Wang et al., 2015). Therefore, in order to provide insight into the potential geodynamic processes that operated in the Andean retroarc at the latitude of our study area, it is necessary to highlight where crustal thickening was distributed in the retroarc region (e.g., Eichelberger et al., 2015), and when critical thicknesses may have been achieved.

The mode of deformation that accommodates lithospheric thickening (i.e., pure shear versus simple shear), which will affect the distribution of crustal thickening (e.g., Allmendinger and Gubbels, 1996), likely differs across the retroarc in both space and time. The crust beneath the Altiplano and EC in southern Bolivia is weak and isostatically compensated, which favors pure shear deformation (Beck et al., 1996; Beck and Zandt, 2002). In contrast, the IAZ and SAZ are flexurally supported by the rigid Brazilian lithosphere, and therefore likely deform by simple shear (Lyon-Caen et al., 1985; Dorbath et al., 1993; Beck and Zandt, 2002; Eichelberger et al., 2015). We modify our crustal budget calculation (Fig. 5A) to account for differing modes of deformation, similar to the models presented in Isacks (1988, Fig. 4 therein) and Eichelberger et al. (2015, Fig. 5 therein). We assume that horizontal upper crustal shortening in the Altiplano and EC was accompanied by vertical thickening of the underlying lower crust via pure shear. Shortening of the SAZ-IAZ occurs by upper crustal imbrication focused along the eastern margin of the thrust belt, and the rigid lithosphere compresses and thickens the ductile lower crust beneath the Altiplano-EC. Because the locus of thickening occurs below the EC-Altiplano, crustal thicknesses are approximated by dividing the area added during each major shortening increment (e.g., Fig. 5; 40–27 Ma, 25–10 Ma, 10 Ma to present) by the deformed length of the EC-Altiplano. The nominal thickening of the IAZ and SAZ by imbrication of basement thrust sheets is accounted for and not included in the thickening below the EC-Altiplano. Because it is difficult to ascribe the precise location and timing of our shortening errors (±69 km), we assign the shortening uncertainties to the EC and Altiplano, and divide them between the periods ca. 40–27 Ma and ca. 25–10 Ma. Although this model is simplistic, it sets a basic benchmark calculation for the distribution of crustal thickness through time.

Given the general timing constraints (Fig. 4), our range of shortening estimates, and an initial crustal thickness of 35 km, the EC and Altiplano are predicted to have achieved thicknesses of 43 ± 3 km by ca. 27 Ma, 56 ± 5.5 km by 10 Ma, and 64 ± 5 km by the present. Critical thicknesses (>45–50 km) required for phase changes of mafic material to dense eclogite, or for initiation of lower crustal flow, were achieved by 10 Ma. However, because the final estimated crustal thicknesses closely match the modern average thickness (∼65 km) observed across the EC and Altiplano (Fig. 5), it is unlikely that significant thicknesses of lower crust were removed.

For an initial crustal thickness of 40 km, the EC and Altiplano are predicted to have achieved thicknesses of 49.5 ± 3.5 km by ca. 27 Ma, 64.5 ± 5.5 km by 10 Ma, and 74 ± 5.5 km by the present. Critical thicknesses in the EC and Altiplano were achieved by ca. 27 Ma, and predicted crustal thicknesses match modern thicknesses (∼65 km) by ca. 10 Ma. Predicted thicknesses of 74 ± 5.5 km by the present exceed the observed thicknesses, suggesting that loss of an ∼2.5–14.5-km-thick section of lower crust across the EC and Altiplano is possible, which is large enough to have a significant effect on hinterland surface elevation.

The SAZ, defined by 10–20-km-wavelength, 4–6-km-amplitude fault-bend folds above a 10–12-km-deep regional décollement in Silurian rocks, accommodated 82 km (36%) of east-west shortening. The IAZ, a bivergent zone of 2–4-km-thick thrust sheets of mainly Silurian–Devonian rocks that are structurally elevated ∼10 km relative to adjacent SAZ, accommodated 70 km (70%) of shortening. The EC, a bivergent zone of 2–10-km-thick thrust sheets of Cambrian–Ordovician rocks structurally elevated ∼6 km relative to both the IAZ and Altiplano, accommodated 120 km (37%) of shortening. Combined with a published estimate of 65 km of shortening for the Altiplano, total shortening of the Andean retroarc thrust belt at 21°S is 337 ± 69 km (36% ± 7%).

Thrust belt development was controlled by emplacement of two long thick basement thrust sheets. The upper basement sheet was tectonically wedged eastward, and fed slip into the sedimentary cover of the overlying EC via a west-vergent roof thrust system from ca. 40 to 25 Ma. The sedimentary cover in the IAZ was deformed in front of the upper basement sheet, which fed slip into a décollement within Silurian shale, from ca. 25 to 10 Ma. Approximately 25 km of out-of-sequence deformation in the western EC, between ca. 29 and 10 Ma, was contemporaneous with eastward migration of the deformation front through the IAZ. The SAZ was deformed in front of a lower basement sheet, which also fed slip into a décollement within Silurian shale, beginning ca. 12–8 Ma. This model is consistent with the patterns of published cooling ages and the record of deposition and eastward migration of the central Andean foreland basin system.

Crustal thickness budgets calculated from our shortening estimates can account for 90%–104% of the crustal area of the retroarc observed today, assuming an initial crustal thickness of 35 km, but may have produced 2%–18% excess area for initial crustal thickness of 40 km, indicating the possibility of crustal flow or delamination at 21°S.

The locus of crustal thickening was likely beneath the EC and Altiplano. For an initial crustal thickness of 35 km, the predicted crustal thicknesses for the EC and Altiplano match present-day observed thicknesses, but were not achieved until the present. For a 40-km-thick initial crust, the modern observed thicknesses of the EC and Altiplano may have been achieved by ca. 10 Ma, and exceeded modern thicknesses by ∼2.5–14.5 km by the present, suggesting crustal losses significant enough to have affected the surface elevation of the hinterland.

This work was funded by National Science Foundation grant EAR-1250510 to Long and Horton. We thank Ramiro Matos for assistance with logistics, and Javier Matos for driving, translating, logistics, and assistance in the field. Constructive reviews from Richard Allmendinger and Nathan Eichelberger significantly improved this manuscript.