Punctuated shortening and subsidence in the Altiplano Plateau of southern Peru: Implications for early Andean mountain building

Tectonic reconstructions of the Andes commonly invoke multiple phases of shortening, uplift, and sedimentation. The conceptual framework and delineation of Andean orogenic phases originated largely from structural and stratigraphic relationships in the Peruvian Andes (Steinmann, 1929; Newell, 1949; Audebaud et al., 1973; Noble et al., 1974; Mégard, 1978; Ellison et al., 1989; Benavides-Cáceres, 1999). Although the proposed pulses of regional deformation from Cretaceous to present have been linked to rapid Nazca–South America convergence (Pardo-Casas and Molnar, 1987; Jaillard and Soler, 1996), records of enhanced Neogene deformation and uplift paradoxically correlate with reduced rates of convergence (Somoza and Ghidella, 2012). Moreover, improved age control and structural and stratigraphic constraints have revealed significant temporal and spatial variations, obscuring the timing and extent of deformation pulses (Clark et al., 1990; Noblet et al., 1996; DeCelles and Horton, 2003). Nevertheless, the concept of punctuated orogenesis has continued to govern many reconstructions of Andean shortening and basin evolution.

Although Cenozoic nonmarine deposits in the Peruvian Andes have been assigned different, often conflicting, basin names (Marocco et al., 1995; Carlotto, 2013), these belts may represent small independent basins or isolated remnants of once-contiguous basins (Jordan and Alonso, 1987; Horton, 2005). The Cenozoic successions range in thickness (<0.2 km to >5 km) and commonly contain internal unconformities, with some studies emphasizing regional extrapolation of local unconformities (e.g., Audebaud et al., 1976; Mégard et al., 1984; Noble et al., 1990; Wise et al., 2008; Pfiffner and Gonzalez, 2013). Whereas younger successions have better chronologic records from dated volcanic rocks (e.g., Rousse et al., 2005; Giovanni et al., 2010; Horton, 2012), most pre–middle Miocene deposits have incomplete age control with limited fossils and few volcanic horizons.

Debate persists over the geodynamic settings for Andean basin evolution. Although shortening dominated the Cenozoic history, some Peruvian basins have been attributed to extension or strike-slip deformation, including interpretations of large-displacement transcurrent faulting (Sempere et al., 2004; Roperch et al., 2011). In addition, there exists a potential for pulses of rapid uplift or subsidence driven by phases of inboard magmatism, flat-slab subduction, and foundering or delamination of lower lithosphere (e.g., Garzione et al., 2006, 2008; DeCelles et al., 2015). Such processes are particularly relevant for the northern Altiplano Plateau and its temporally and spatially overlapping records of deformation, arc magmatism, surface uplift, and basin accumulation (Schildgen et al., 2007; Mamani et al., 2010; Carlotto, 2013; Saylor and Horton, 2014). Nevertheless, a persistent problem in southern Peru involves the difficulty in identifying the major faults responsible for crustal thickening, potentially due to overprinting by Neogene arc magmatism and hinterland sedimentation.

In this paper, we seek to define the depositional, chronological, and provenance history of a >4-km-thick Cenozoic succession in the northern Altiplano, between the Western Cordillera magmatic arc and Eastern Cordillera fold-thrust belt. The study area and surrounding regions of southern Peru show a dynamic history of subduction, magmatism, and pre-Andean deformation, which helped to focus Andean shortening and basin subsidence. Finally, our study assesses whether Cenozoic subsidence was genetically linked to upper-crustal deformation, magmatism, or lithospheric dynamics, and the relative importance of these processes in driving punctuated or sustained mountain building.

The Andes are the type example of a contractional orogen along an ocean-continent convergent margin (Dewey and Bird, 1970; James, 1971). The subducting Nazca slab beneath South America varies from a moderately (∼30°) east-dipping segment in the central Andes to zones of flat-slab subduction to the north (5–14°S) and south (27–33.5°S; Cahill and Isacks, 1992; Ramos, 1999). A similarly dynamic slab configuration apparently characterized the Cenozoic record, with phases of shallow and steep subduction dictating advance and retreat of arc magmatism (Ramos and Folguera, 2009; Mamani et al., 2010).

Topographic and structural elements of southern Peru (Fig. 1) display the NW-SE trend defining the northern end of the Bolivian orocline (Rousse et al., 2005; Roperch et al., 2006, 2011) and central Andean Plateau (defined by >3 km elevation at 13–27°S; Isacks, 1988; Allmendinger et al., 1997). Tectonomorphic provinces in southern Peru (13–17°S) include a coastal plain, forearc slope, Western Cordillera, Altiplano Plateau, Eastern Cordillera, Subandean zone, and Amazon foreland basin.

The Western Cordillera (Fig. 1) is a Cretaceous–Neogene magmatic arc, with >4.5–6 km volcanic peaks and extensive ignimbrite sheets, which overprints a Cretaceous basin system (West Peruvian Trough) and Precambrian basement (Arequipa massif; Wilson, 1963; Cobbing, 1978; Cobbing et al., 1981; Atherton et al., 1983; Mégard, 1987). The 100-km-wide Altiplano Plateau represents an internally drained hinterland basin (Horton, 2012) at ∼4 km elevation. Shortened crust of the central Andean Plateau attains thicknesses of 60–70 km beneath the Altiplano and elevated Western and Eastern Cordillera margins (Kono et al., 1989; James and Sacks, 1999; Beck and Zandt, 2002). Cratonward of the Altiplano, there lies the retroarc fold-thrust belt and foreland basin of the Eastern Cordillera, Subandean zone, and Amazon plain. The rugged Eastern Cordillera, a bivergent fold-thrust system involving thick Paleozoic strata and a discontinuous, unconformable cover of Cretaceous (East Peruvian Trough) and Cenozoic clastic fill, is composed of a hinterland-directed backthrust zone bordering the Altiplano and a forethrust zone bordering the Subandean zone (Newell, 1949; Sempere et al., 1990; Gubbels et al., 1993; McQuarrie and DeCelles, 2001; Gotberg et al., 2010). The mountain front is defined by the active, thin-skinned Subandean fold-thrust system (Suárez et al., 1983; Baby et al., 1997), with intervening wedge-top basins that grade eastward into the Amazon (Madre de Dios and Beni) foreland basin (Horton and DeCelles, 1997; Roddaz et al., 2005).

The regional configuration is instrumental in assessing sediment provenance and the compositional and U-Pb age signatures of potential source regions. The Peruvian Andes are largely constructed upon western basement of the Proterozoic Arequipa terrane (Loewy et al., 2004; Ramos, 2008) and eastern basement of probable lower Paleozoic (ca. 450–500 Ma) metamorphic rocks of the Marañon complex (Chew et al., 2007) and underthrust Proterozoic crystalline rocks of the Brazilian Shield (Mégard, 1987). Mesoproterozoic amalgamation of the Arequipa terrane, broader Amazonian craton, Laurentia, and other blocks formed the Rodinia supercontinent during the Grenville-Sunsás orogeny (Dalziel, 1997; Loewy et al., 2004). Recycling of Sunsás and Arequipa basement has distributed 900–1200 Ma age signatures in clastic sediment throughout South America (e.g., Horton et al., 2010; Bahlburg et al., 2011). Neoproterozoic ages in the Andes and proximal foreland document late-stage breakup of Rodinia from ca. 650 Ma to 530 Ma. Although these signatures are best expressed in the Pampean orogen of Argentina and Putumayo orogen of Colombia (Ramos, 2008; Ramos et al., 2010; Cardona et al., 2010; Ibanez-Mejia et al., 2011), comparable Neoproterozoic ages are reported in Peru and Bolivia (Lehmann, 1978; Dalmayrac et al., 1980a; Troëng et al., 1994; Tosdal, 1996; Aceñolaza et al., 2002; Chew et al., 2007).

Phanerozoic deformation and magmatism conditioned large parts of the Andean belt. Early Paleozoic shortening (Famatinian and Ocloyic orogenesis) affected much of western South America, followed by subduction-related arc magmatism, late Paleozoic shortening (Chanic, Hercynian, and Gondwanide deformation), and deposition of volcaniclastic sediments in southern Peru (Laubacher, 1978; Dalmayrac et al., 1980b; Martinez, 1980; Carlier et al., 1982; McGroder et al., 2015). Late Permian–Triassic backarc extension resulted in accumulation of the up to 3-km-thick Mitu Group (Mégard, 1978; Kontak et al., 1985; Sempere et al., 2002). Deposition of craton- and subordinate arc-derived sediment (in the East Peruvian Trough and West Peruvian Trough) characterized the Cretaceous until a ca. 90 Ma renewal and cratonward advance of magmatism began to significantly affect western Peru (Mamani et al., 2010; Demouy et al., 2012; Scherrenberg et al., 2012). During the Cenozoic, arc magmatism variably affected forearc regions, the Western Cordillera, and, for periods of flat-slab subduction, the Altiplano and Eastern Cordillera (Mamani et al., 2010; Decou et al., 2011). Coeval retroarc shortening in the Eastern Cordillera and Subandean zone (Gotberg et al., 2010; Espurt et al., 2011) partitioned and isolated the Altiplano basin within the central Andean hinterland, which has since experienced sediment influx from both plateau margins (Rousse et al., 2005; Horton, 2012).

Although Precambrian basement of the Arequipa terrane is exposed locally in the Western Cordillera, the Ordovician Calapuja Formation, Silurian–Devonian Chagrapi Formation, and Permian–Triassic Mitu Group make up the western front of the Eastern Cordillera (Fig. 2). Ordovician–Devonian strata are composed of dark shale with intercalated siltstone, sandstone, and their metasedimentary equivalents (Newell, 1949; Callot et al., 2008). The Permian–Triassic Mitu Group contains diverse nonmarine clastic facies, including red to purple sandstone, shale, and conglomerate, as well as volcanic flows and intrusive igneous bodies (Jenks, 1951; Mégard, 1978; Sempere et al., 2002; Callot et al., 2008). Although shallow-marine clastic facies along with local carbonate and evaporite are represented in upper Mitu levels, an extensive succession of Jurassic marine carbonate is absent in the northern Altiplano.

The Cretaceous section can be divided into a lower clastic, nonmarine interval and an upper marine interval of mixed clastic and carbonate rock (Wilson, 1963). The Lower Cretaceous succession (principally Lagunillas Group) recorded deposition of sandstone, conglomerate, and red mudstone, and small-scale transgressive marine carbonate (Wilson, 1963; Portugal, 1974; Callot et al., 2008). A major mid-Cretaceous transgression is marked by the marine Muni and Arcurquina Formations, denoted by red mudstone, sandstone, and organic-rich micritic limestone (de la Cruz, 1995; Callot et al., 2008). Late Cretaceous felsic to intermediate intrusive bodies, present locally (de la Cruz, 1995), are coeval with arc magmatic rocks farther west (Mamani et al., 2010; Demouy et al., 2012).

The focus of this study, the Paleogene San Jerónimo Group, is continuous with northern exposures near Cusco and southern exposures of the Puno Group near Lake Titicaca (Fig. 1; Newell, 1949; Klinck et al., 1986; Carlotto, 1998, 2013; Carlotto et al., 2005). In the northern Altiplano (Fig. 2), the >3–5-km-thick San Jerónimo succession of upper Eocene–Oligocene nonmarine fill unconformably overlies Upper Cretaceous carbonate and mudrock and consists of gray and reddish-brown arkosic sandstone, conglomerate, shale, and rare lacustrine carbonate (Noblet et al., 1987; de la Cruz, 1995; INGEMMET, 1995a, 1995b; Cerpa and Meza, 2001; Latorre and Orós, 2000; Latorre et al., 2004; Carlotto et al., 2005). New detrital zircon U-Pb ages for this succession provide a minimum depositional range of 37–28 Ma.

The Altiplano study region lies between the NW-trending Western Cordillera and Eastern Cordillera (Figs. 1 and 2). The Paleogene San Jerónimo Group is exposed near the villages of Macari (14.76°S, 79.96°W) and Llalli (14.95°S, 70.88°W) at 3.9–4.8 km elevation between oppositely dipping faults (Fig. 2; Latorre et al., 2004). In the east, the major NE-dipping Ayaviri thrust fault defines the Altiplano–Eastern Cordillera boundary and places east-dipping Ordovician–Silurian and Cretaceous rocks on Cenozoic basin fill. In the west, the SW-dipping Pucapuca-Sorapata and Lagunillas fault systems (de la Cruz, 1995), the latter approximating the Altiplano–Western Cordillera boundary, bring Cretaceous and Permian–Triassic strata to the surface. These fault systems may have formed during Late Permian–Triassic or possible Early Cretaceous extension, with later reactivation during Cenozoic compression or transpression (Carlotto, 1998; Carlotto et al., 2005).

The Ayaviri thrust displacement is constrained by footwall growth strata in the uppermost San Jerónimo Group (Fig. 3), as defined by an ∼90° dip variance (from subvertical to subhorizontal), systematic thinning of bedding packages toward the structure, and internal angular unconformities within the succession. Fault timing is revealed by U-Pb ages of syndepositional zircon grains, which show young age populations ranging from 27.9 ± 0.4 Ma to 26.3 ± 1.6 Ma (Perez and Horton, 2014). Subhorizontal (<5–10°) beds in uppermost stratigraphic levels postdate nearly all of the ∼90° of tilting accomplished during fault-related folding. Therefore, we interpret a main phase of fault displacement underway by ca. 29 Ma, with cessation of growth strata development by ca. 26 Ma.

The San Jerónimo Group constitutes the lower half of the 5–10 km accumulation of Cenozoic nonmarine clastic fill in the northern Altiplano. The exposed 3–4-km-thick succession was analyzed at two principal sites, near Macari and Llalli (Fig. 2), with provenance constrained by paleocurrent, compositional, and geochronological analyses. A new depositional chronology is defined by zircon U-Pb ages for sandstones and rare reworked tuffs, along with magnetic polarity stratigraphy of continuous exposures from the upper 1600 m of the succession.

The main sedimentologic features are depicted in a generalized stratigraphic section for the northern (Macari) locality (Fig. 4) and a detailed section for the southern (Llalli) locality (Fig. DR1¹). On the basis of individual lithofacies and broader facies associations defined here (Tables DR1 and DR2 [see ^{footnote 1}]), exposures in the study area can be divided into a lower sandy unit and an upper conglomeratic unit (corresponding to San Jerónimo Members C, D, and E of Latorre et al., 2004).

The succession shows similar facies, facies assemblages, and stratigraphic trends at both localities (Figs. 5–7). Nine major lithofacies are identified (Table DR1 [see ^{footnote 1}]) according to standard codes and procedures (e.g., Miall, 1996; Uba et al., 2005). Laminated siltstone (Fl), often capping coarser sandstone or arranged in upward-fining packages (Figs. 5 and 6), is common in the lower succession and interpreted as overbank or waning flood deposits. Medium to very coarse sandstones are divided into massive (Sm), horizontally stratified (Sh), trough cross-stratified (St), and ripple cross-stratified (Sr) sandstone lithofacies (Figs. 5 and 6). Whereas Sm is indicative of rapid deposition by sheet or sediment gravity flow, Sh was deposited by planar bed flow. St and Sr represent migrating dunes and ripples, respectively, and are typical in channels or laterally extensive sandy units displaying basal scour and amalgamated poorly developed channels. Gravelly lithofacies (Figs. 6 and 7) include: clast-supported, weakly trough cross-stratified granule-pebble conglomerate (Gt) representative of transverse bars in broad channels; horizontally stratified conglomerate (Gh) deposited by traction bed load with local weak imbrication; and poorly sorted, disorganized clast-supported (Gcm) and matrix-supported (Gmm) pebble-cobble conglomerates interpreted as hyperconcentrated and debris flows, respectively.

Five fluvial to alluvial-fan facies associations are defined (Table DR2 [see ^{footnote 1}]). Facies association F1 contains laminated siltstone (Fl) and thinly interbedded Sh, Sr, and Sm sandstone (Fig. 5A) indicative of a fluvial overbank environment. Association S1 (Figs. 5B and 5C), composed of thick Sh and St sandstone with limited basal scour and localized Sr and Sm, is indicative of unconfined sheet flow in a fluvial setting. Association S2, dominated by St with local Sr and Sh, is found in erosive channels of a sandy fluvial environment (Figs. 5D and 6A). Association G1, containing Gt with interbedded St and Sh in broad channels (Figs. 6B, 6C, 7A, and 7C), is representative of large, shallow channel fill on a fluvial braid plain or alluvial fan. The G2 association, composed of massive Gcm and Gmm lithofacies (Figs. 7B and 7D), is interpreted as hyperconcentrated or debris flows on an alluvial fan.

From these facies associations and interpretations of depositional systems, the succession was subdivided into two units. A lower sandy unit is dominated by S1 sheets with interbedded F1 overbank deposits and local S2 channel complexes in upper levels. An upper conglomeratic unit is dominated by channelized S2 and G1 deposits, with scattered S1 sheet-flow deposits and an interval of G2 sediment gravity flow deposits.

The lower sandy unit (Fig. 4; Fig. DR1 [see ^{footnote 1}]) is composed of facies association S1, containing extensive sheets of Sh and St, with 0.05–0.5 m beds amalgamated into packages 1–15 m thick. Upward gradation from Sh to St is common. Association F1, containing thinly bedded fine sandstone and siltstone (Fl, Sh, Sr) is uncommon but present in upward-fining sequences of the northern (Macari) section. The unit coarsens upward, with channelization and gravel clasts increasingly common up section. S2 and G1 channel-fill deposits include lenses of St, Sr, Sh, and Gt. Conglomerate beds contain clasts up to 6 cm and display ∼10 cm of basal relief. Successive channel deposits are commonly amalgamated, with limited (<20 cm) basal scour, poorly defined upper channel margins, and thicknesses up to 2 m.

The lower sandy unit is attributed to sheetflood and distal braided deposition in a flashy, ephemeral fluvial system. Gradation from Sh to St within S1 sheet-flow deposits is consistent with deposition during peak discharge and intervening lower-flow periods (Bromley, 1991). Amalgamated S2 and G1 deposits are attributed to in-channel deposition with routine overtopping of lateral channel margins during flood events (Miall, 1996). Upward-fining F1 packages of siltstone and sandstone (Fl, Sh, Sr) represent overbank deposition from crevasse splays and suspension fallout. Alternation between sheet-flow (S1) and shallow channel deposition (S2, G1), with accompanying overbank processes (F1), is suggestive of a poorly organized braided fluvial system with limited scouring in a highly aggradational system (e.g., Hampton and Horton, 2007). The up-section coarsening and increasing proportion of channel facies indicate a gradual shift to more proximal settings.

The upper conglomeratic unit (Fig. 4; Fig. DR1 [see ^{footnote 1}]) contains thick amalgamated channel complexes of facies associations G1 and S2, with subordinate G2 mass-flow deposits in lower levels. G1 deposits of Gt and Gh are laterally continuous (>10–100 m), with lenticular channel morphology and thicknesses of 1–5 m, locally up to 12 m. Intercalated St and Sh beds are 0.1–2 m thick. Intervals between G1 channel-fill conglomerates are mostly composed of association S2, with coarse to very coarse St beds arranged into packages up to 25 m thick, and thin isolated Gt and Gh channels (up to 2 m thick) and pebble stringers. Covered intervals of considerable lateral extent (>200 m) are likely indicative of finer S1 and F1 deposits. Massive G2 conglomeratic intervals are restricted to the lower half of the unit, where amalgamated stratified (Gh) and structureless (Gcm) granule to cobble conglomerates with indistinct bedding contacts are stacked vertically and occasionally separated by isolated 0.1–0.5 m Sh and St lenses.

The upper conglomeratic unit represents a basinward facies shift and more proximal braided fluvial or alluvial-fan deposition. The unit is dominated by the G1 and S2 facies associations, with minor covered intervals indicative of sandy crevasse splay (S1) and silty overbank (F1) deposition in transient floodplains. The extent of G1 conglomerate complexes indicates shallow interconnected channels, which migrated laterally over large areas (e.g., Kelly and Olsen, 1993). Intervals of poorly to well-developed Gt are representative of transverse bars, while Gh likely represents longitudinal bar deposits emplaced during peak discharge (Miall, 1996). St lenses were emplaced as three-dimensional migrating dunes along the channel floor during periods of diminished flow. The massive, highly amalgamated G2 conglomerates, present in lower levels of the unit, represent deposition of hyperconcentrated flows (Gmm, Gcm), sheet flows (Gh), and rare debris flows (Gmm) in shallow, poorly defined channels of an alluvial fan (Nemec and Steel, 1984). Above these fan deposits, the upper levels record a shift back to a braided fluvial system with large, shallow, bed-load channels, potentially within a fluvial megafan (e.g., Horton and DeCelles, 2001; Hartley et al., 2010).

We measured 441 paleocurrent orientations in trough cross-stratified sandstones and imbricated conglomerates at 22 sites and then restored them for bedding tilt. Measurements of imbricated clast orientations (n > 10) were linearly averaged for each site. For trough cross-stratified sandstones, measurements of right (n > 10) and left (n > 10) limb orientations allowed us to determine the mean trough-axis orientation for each site (e.g., DeCelles et al., 1983; Dasgupta, 2002).

Paleocurrent results show a dominant flow direction to the north and northeast (Fig. 4). Given the NW-trending structural fabric (Fig. 1), the main sediment source was likely in the Western Cordillera rather than the Eastern Cordillera. Although paleocurrents for the northern (Macari) section are relatively uniform, the southern (Llalli) section shows a potential up-section shift from principally NE-directed flow for the lower sandy unit to N-directed flow for the upper conglomeratic unit (Fig. 4; Fig. DR1 [see ^{footnote 1}]), possibly reflecting a transition to more axial flow.

Clast counts were conducted at 10 sites using 30 × 30 cm grids on outcrops of pebble conglomerate. Clasts were identified according to rock type, mineralogy, and grain size and then classified into four groups: volcanic, plutonic, quartzite/clastic sedimentary, and limestone. Igneous rocks suggest sediment derived from the western magmatic arc, <50 km WSW of the study area (Figs. 1 and 2; Mamani et al., 2010). In contrast, quartzite/clastic sedimentary and limestone clasts mostly originated from Cretaceous and possibly Paleozoic strata, which have multiple potential source regions around the study area.

Conglomerate clast compositions (Fig. 4; Fig. DR1 [see ^{footnote 1}]) for the uppermost lower sandy unit and base of the upper conglomeratic unit show >95% volcanic clasts, with limited limestone, clastic (including quartzite), and plutonic clasts. Middle to upper levels of the upper conglomeratic unit show an up-section decrease in volcanic clasts and increase in clastic/quartzite and limestone categories. The uppermost count includes significant plutonic clasts. The volcanic composition of lower levels is consistent with influx from the western magmatic arc, an interpretation supported by paleocurrent data. Increased proportions of sedimentary and plutonic rocks in the upper unit indicate the introduction of Cretaceous and older rocks in the drainage systems feeding the basin, which paleocurrent data suggest were situated WSW of the study area.

Petrographic point-count analyses were conducted on 19 medium-grained sandstones. Thin sections were constructed for each sample, with staining for K-feldspar and plagioclase to facilitate grain identification and injection of blue dye epoxy to fill pore space and aid in recognition of dissolved grains. Point counts were performed using the Gazzi-Dickinson classification scheme (Gazzi, 1966; Dickinson, 1970, 1985; Dickinson and Suczek, 1979), with identification of 350–400 detrital grains per thin section. Grain categories are outlined in Table 1, and results are plotted on ternary diagrams (Fig. 8) following various classification schemes (Folk, 1980; Dickinson and Suczek, 1979; Dickinson et al., 1983; Dickinson, 1985).

Analyzed sandstones have limited quartz (Q) but significant feldspar (F) and lithic fragments (L) (Fig. 8). Nearly all are lithic arkose (Fig. 8A) and conform to a magmatic arc provenance (Figs. 8B–8D). The dominant constituents are plagioclase feldspar (P) and lithic volcanic fragments (Lv), many containing plagioclase phenocrysts. Sandstones of the southern (Llalli) transect are richer in quartz, with a mean composition of Q₃₃F₃₈L₃₀ and no discernible up-section shifts. In contrast, samples from the northern (Macari) section have a mean composition of Q₁₈F₄₄L₃₈, and the upper conglomeratic unit can be subdivided into an arkosic lower interval (Q₄F₅₆L₄₀) and lithic arkosic upper interval (Q₁₀F₇₀L₂₀). The more arkosic signature is expressed by enhanced feldspar (F), particularly plagioclase feldspar (P), and reduced lithic fragments (L), particularly volcanic fragments (Lv).

The dominant magmatic arc provenance signature is consistent with the basin location adjacent to the Paleogene arc of the Western Cordillera and paleocurrent data indicating NE-directed sediment dispersal. During late Eocene–Oligocene sedimentation, the Andahuaylas-Anta (45–30 Ma) magmatic arc was within 50 km of the study area (Mamani et al., 2010). The overwhelming arc signal in the detrital record eclipses other petrographic changes that might otherwise be present.

Although the transition from the lower sandy unit to upper conglomeratic unit and corresponding increase in feldspar may suggest an increase in basement input, there is no exposed basement nearby. Therefore, these shifts in facies and grain composition more likely reflect (1) exhumation of newly exposed intrusive rocks of older Cretaceous–Paleogene arcs or (2) introduction of more-proximal sources of syndepositional volcanic units with shorter transport distances and limited weathering. Up section, a final shift back toward more lithic fragments, including aphanitic volcanic clasts, could be correlated with (1) increased local volcanism associated with the closer proximity of the Tacaza (30–24 Ma) magmatic arc (Mamani et al., 2010) and/or (2) erosional removal of more-proximal sources.

In total, 11 samples were collected in the northern (Macari) transect for U-Pb geochronology. Nine medium-grained sandstone samples (0–3106 m level) and two reworked tuff samples (3147 and 3981 m levels; Fig. 4) were crushed, and zircon grains were isolated using standard heavy mineral separation techniques. Zircons of variable size, shape, and quality were selected randomly, mounted into epoxy pucks, polished, and then analyzed at the Arizona LaserChron Center of the University of Arizona by laser ablation–multicollector–inductively coupled plasma–mass spectrometry (LA-MC-ICP-MS) following established procedures (Gehrels, 2000, 2012; Gehrels et al., 2008). Measurements were calibrated and fractionation and common lead corrections were applied using a Sri Lankan age standard (563.5 ± 3.2 Ma isotope dilution–thermal ionization mass spectrometry [ID-TIMS] age) measured after every five unknown grains. Gehrels et al. (2008) provided additional details regarding corrections and laboratory methods.

Zircon U-Pb ages were calculated and are reported (Table DR3 [see ^{footnote 1}]) on the basis of ²⁰⁶Pb/²³⁸U ratios for ages younger than 1000 Ma and ²⁰⁶Pb/²⁰⁷Pb ratios for ages older than 1000 Ma, in order to minimize age uncertainties. Using these results and corrections based on the age standard, the age uncertainty is 1%–2% for most grains. Data filtering involved the omission of grains with >10% uncertainty, >30% discordance, or >5% reverse discordance. The LA-MC-ICP-MS results are plotted as age distribution histograms and normalized relative age probability plots (Fig. 9). Whereas the age distribution histograms show the number of zircons that fall within each 50 m.y. bin, the relative probability plots combine ages and uncertainties into a single curve depicting relative age distributions.

Seven major detrital zircon age populations were identified collectively in analyzed samples (Fig. 9A), and are discussed from youngest to oldest.

A 25–50 Ma age peak, the youngest and largest population, represents Cenozoic grains from the Andean magmatic arc, approximated by the present Western Cordillera (Mamani et al., 2010) west of the study area. Because very limited igneous rocks of this age are present in the Eastern Cordillera, the dominance of arc-derived zircons is considered a strong indicator of a western source throughout the succession.

A 50–150 Ma population principally represents zircons from the Late Cretaceous–early Eocene magmatic arc, an extinct arc (the Toquepala arc) situated farther west within the present forearc (Mamani et al., 2010). This signal indicates input from east-directed drainages tapping the entire western arc complex, rather than the Western Cordillera exclusively.

A distributed 100–350 Ma population represents a long-lived Carboniferous–Early Cretaceous magmatic arc (the Chocolate arc) immediately west, and partially overlapping, the Cretaceous–early Eocene (Toquepala) arc (Mamani et al., 2010; Decou et al., 2011). Within this interval, however, there is a 250–350 Ma signal of Carboniferous–earliest Triassic magmatism representing the Mitu rift system, a zone of backarc igneous activity and volcano-sedimentary extensional basins spanning the Altiplano and Eastern Cordillera (Sempere et al., 2002).

Additional age peaks at 400–500 and 500–650 Ma represent grains originally from Paleozoic strata, including widespread Ordovician siliciclastic rocks derived from Famatinian and Pampean sources characteristic of the Andean orogen and westernmost foreland (Chew et al., 2007; Ramos, 2008; Ibanez-Mejia et al., 2011; Siks and Horton, 2011). In southern Peru and Bolivia, this Ordovician basin was concentrated in the Eastern Cordillera but persisted from the western Amazon foreland to eastern Altiplano (Bahlburg et al., 2011).

The oldest populations of 1000–1200 and 1500–2800 Ma are consistent with several sources. Late Mesoproterozoic ages may derive from Arequipa basement rocks WSW of the basin (Sempere et al., 2002; Loewy et al., 2004; Ramos, 2008; Bahlburg et al., 2011) or direct input from the Sunsás Province of the Amazonian craton. Alternatively, this 1000–1200 Ma signal reflects recycling of nearby sediments from extensive Cretaceous or older strata ultimately derived from Arequipa or Sunsás sources (Bahlburg et al., 2011; Michalak, 2013). Similarly, the principally Paleoproterozoic (>1500 Ma) grains must be ultimately derived from the Amazonian craton, but were likely delivered directly from Cretaceous or older sedimentary units of the Western Cordillera.

Various age populations are expressed in the 4-km-thick San Jerónimo succession (Fig. 4). Whereas lower levels show a wide range of populations, upper levels are dominated by Cenozoic ages (Fig. 9A). Cenozoic (younger than 50 Ma) zircons constitute a major component throughout all 11 samples, underscoring the importance of arc-derived sediment from the Western Cordillera. Additional western sources include 50–150 Ma grains from Cretaceous–Paleogene arc rocks, and 1000–1200 Ma grains recycled from Cretaceous cover strata or directly eroded from possible exposures of Arequipa basement. Although the 400–650 Ma and >1500 Ma populations are nonunique, when combined with paleocurrent data, it appears that recycling of Cretaceous and older strata from western regions offers the most reasonable option. These considerations suggest that Paleogene basin fill was dominantly sourced from the broader Western Cordillera, including Cenozoic arc rocks, Cretaceous sedimentary and arc rocks, and possible exposures of deeper Triassic, Paleozoic, and late Mesoproterozoic Arequipa basement.

U-Pb ages for the 11 aforementioned samples, including nine sandstones and two reworked tuffs, yield stratigraphic age constraints for the San Jerónimo (Puno) Group. Although often applied strictly for provenance determinations, detrital zircon geochronology also proves useful in constraining maximum depositional ages on the basis of the youngest age population within individual samples (e.g., Bahlburg et al., 2011). In this study, the proximity and large volume of volcanic material from the active arc suggest that the youngest age population may be considered identical to the depositional age of the host sample (e.g., Fildani et al., 2003; DeCelles et al., 2007; Leier et al., 2010). Any lag time between volcanic zircon crystallization and sedimentary deposition would be considered minimal, likely <0.5–1 m.y. A key test of this inference would involve a systematic up-section reduction in the ages of the youngest zircon populations among successive samples.

Consideration of the 0–50 Ma age population for the 11 samples (Table DR3 [see ^{footnote 1}]) is accomplished through plots of the age distribution histograms and relative probability functions, which combine individual age analyses and uncertainties (Fig. 9B). The reported U-Pb ages for the youngest zircon populations in these samples represent weighted mean ages and their corresponding errors at the 95% confidence interval (2σ).

The results constrain the maximum depositional ages of the 4000 m succession from 36.9 ± 0.8 Ma to 28.2 ± 1.0 Ma (Figs. 4 and 9). In this study, the sandstones are as useful as reworked volcanic units in defining depositional ages. The nine sandstones (0–3106 m level) all yield young age populations, ranging from 36.9 ± 0.8 Ma to 28.6 ± 0.8 Ma, reflective of abundant syndepositional volcanic zircons. The two reworked tuffs (3147 and 3981 m levels) have comparable unimodal age peaks with the youngest age populations at 28.9 ± 1.7 Ma and 28.2 ± 1.0 Ma (Figs. 4 and 9), suggesting rapid accumulation of upper stratigraphic levels.

The weighted mean ages for the youngest zircon populations in several adjacent samples overlap within error, consistent with rapid accumulation. The results also show a general but systematic up-section reduction in the age of the youngest populations (Fig. 9B), suggesting active volcanogenic contributions throughout late Eocene–Oligocene accumulation, consistent with regional syntheses of continuous arc magmatism (e.g., Sandeman et al., 1995; Mamani et al., 2010). For this reason, the youngest populations are considered to approximate true depositional ages, thus providing benchmarks for the magnetostratigraphic analyses.

To further constrain depositional ages, 155 magnetostratigraphic sites were sampled at ∼10 m spacing over a 1605 m stratigraphic interval of the San Jerónimo Group in the upper portion of the northern section (Figs. 2 and 4). At least three oriented core samples were collected at each site using a hand-operated, gasoline-powered drill. At the University of Texas at Austin Paleomagnetics Laboratory, thermal demagnetization experiments were performed employing incremental heating steps of 0, 200, 300, 350, 400, 450, 500, 525, 550, 575, 600, and 650 °C. Between steps, starting with a natural remanent magnetization (NRM) measurement at 0 °C, the magnetization of each core was measured (in both the up and down orientations) using a 2-G cryogenic magnetometer and an automated sample handler system (e.g., Kirschvink et al., 2008) inside a magnetically shielded room.

Thermal demagnetization results are depicted on vector component (orthogonal projection) and equal-area (stereographic projection) diagrams to identify and distinguish temperature steps representing the declination and inclination of the primary magnetization vector. For each sample, least-squares regression lines were fit to demagnetization trajectories to define primary magnetization directions (Kirschvink, 1980; Butler, 1992). The resulting geographic (in situ) declinations were corrected for bedding dip, and maximum angular deviation (MAD) from the least-squares fit was calculated as a measure of uncertainty.

Data quality was assessed using the MAD of the least-squares fit for each sample and the internal variability of primary magnetization directions for samples from the same site. For each site, data quality was categorized into three classes: (A) MAD < 15° (for all samples) and maximum deviation of declinations < 30°, (B) MAD < 15° and maximum deviation of declinations < 45°, and (C) MAD < 20° and maximum deviation of declinations < 45°. Although magnetic reversals could be defined by various combinations of A, B, and C quality sites, U-Pb age control allows us to focus on the highest-quality data from A sites. The reversal chronology defined for the sampled interval was then correlated to the geomagnetic polarity time scale (GPTS; Gradstein et al., 2004).

Progressive thermal demagnetization resulted in isolation of the primary magnetization vector for each sample (Fig. 10). In total, 61 magnetostratigraphic sites (41 class A, 11 class B, and 9 class C), selected on the basis of the MAD of the least-squares fit and intrasite variations in declination (Table DR4 [see ^{footnote 1}]), were classified as normal or reverse polarity, based on the site-averaged, tilt-corrected magnetic declination.

Vector component (orthogonal projection) and equal-area (stereographic projection) diagrams show the thermal demagnetization results for representative samples (Fig. 10), allowing discrimination of declination and inclination values for the primary magnetization vector over a range of temperature steps. These samples include a class A normal polarity site (Figs. 10A and 10B), a class A reverse polarity site (Figs. 10C and 10D), and a class B normal polarity site (Figs. 10E and 10F).

Magnetic reversals are defined by the stratigraphic boundaries between normal and reverse intervals. By focusing on the highest-quality class A sites, we defined four individual intervals of uniform polarity. This pattern of magnetic reversals, and the depositional age constraints provided by U-Pb geochronological results within and below the sampled section (Figs. 4 and 9) allow for two potential magnetostratigraphic correlations to the GPTS (Gradstein et al., 2004). Correlation 1 matches the observed polarity intervals to magnetic chrons C11n–C10r (Fig. 11), yielding a 1.5 m.y. age range, from 30.2 to 28.7 Ma. Correlation 2 matches the observed polarity intervals to magnetic chrons C10n–C9r (Fig. 11), yielding a 0.9 m.y. age range from 28.7 to 27.8 Ma. Although both correlations appear reasonable, correlation 2 provides a consistent overlap with the youngest U-Pb age populations for 11 successive samples (as plotted in Fig. 12). Although both options yield an average accumulation rate in excess of 1000 m/m.y., the preferred correlation suggests a very rapid rate of ∼1800 m/m.y. (Fig. 12).

Integration of new age control from U-Pb geochronology and magnetostratigraphy provides a continuous long-term accumulation history (Fig. 12), with a generalized rate of 440 m/m.y. (4020 m from 36.9 to 27.8 Ma). For the lower ∼1600 m, very rough calculations based on U-Pb ages, without corresponding errors, yield a wide range of potential accumulation rates from ca. 37 to 30 Ma, with a moderate long-term average of ∼230 m/m.y. Although by no means definitive, the age constraints are permissive of an initial rapid phase at ca. 37–35 Ma (possibly ∼1500 m/m.y.), followed by slower accumulation (possibly ∼130 m/m.y.) from ca. 35 to 30 Ma. In contrast, the upper ∼2400 m interval is precisely constrained by the magnetostratigraphic GPTS correlation and supporting U-Pb ages to define a very rapid accumulation rate of ∼1100 m/m.y. from ca. 30 to 27.8 Ma. The significant increase in accumulation at ca. 30 Ma, as recorded at the ∼1600 m level, approximately coincides with the stratigraphic boundary between the lower sandy unit and upper conglomeratic unit.

Clastic fill of the 4000-m-thick San Jerónimo (Puno) Group recorded accumulation from ca. 37 to 28 Ma and subsequent fragmentation into a late Oligocene–Miocene hinterland basin system (Fig. 13). The basin history can be broken into three parts.

Initial basin development was marked by accumulation of the lower ∼1600 m (including the lower sandy unit) between ca. 37 and 30 Ma (Fig. 12) in a distal braided/sheet-flow environment interpreted as part of a broad early Andean basin (Fig. 13A). A Western Cordillera sediment source is indicated by NE-directed paleocurrents and the dominance of Cretaceous–Cenozoic arc-derived detritus identified through conglomerate clast counts (high volcanic clast proportions; Fig. 4), sandstone petrography (high proportions of lithic, volcanic lithic, and feldspar grains; Fig. 8), and detrital zircon U-Pb geochronology (prominent <150 Ma age peaks; Fig. 9).

Carlotto et al. (2005) considered the Lagunillas fault (Fig. 2) as the principal source of San Jerónimo sediments (Fig. 13A), and we attribute initial accumulation at ca. 37–35 Ma to possible late Eocene activation of this structure and resulting flexural accommodation. Although the fault does not expose any rocks older than the Early Cretaceous, we infer removal of a relatively thick Cretaceous succession, consistent with regional depositional patterns for the West Peruvian Trough, which occupied most of the Western Cordillera (Mégard, 1987; Scherrenberg et al., 2012). Recycling of Cretaceous rocks and coeval input from distal western sources of pre-Cretaceous strata and Precambrian basement can account for the broad distribution of zircon age spectra for the lower sandy unit (Fig. 9), implying that significant rock uplift and exhumation had commenced in the Western Cordillera by ca. 37 Ma.

At ca. 30 Ma, a transition from sandy distal fluvial to proximal fluvial and alluvial-fan deposition (Fig. 4) for the upper ∼2400 m (including most of the upper conglomeratic unit) suggests a shift in sediment supply and/or source-area proximity (Fig. 13B). This shift approximately coincides with a transition to more feldspar-rich detritus (Fig. 8) and increasing proportions of sedimentary clast compositions (Fig. 4; Fig. DR1 [see ^{footnote 1}]). NE-directed paleocurrents, volcanic-rich sandstones and conglomerates, and significant detrital zircon ages younger than 50 Ma indicate the persistence of a western sediment source (Figs. 4, 8, and 9). Conglomeratic deposition also corresponds to a substantially higher accumulation rate (∼1100 m/m.y.) from ca. 30 to 27.8 Ma (Fig. 12), consistent with a more proximal setting undergoing rapid subsidence adjacent to a syndepositional structure.

We suggest that the new source region may represent activation of the Pucapuca-Sorapata fault system (Figs. 2 and 13B), a feature that segregates two belts of Paleogene basin fill and potentially links with other thrust faults to the north and south (Carlotto, 2013; Perez and Horton, 2014). This interpretation would explain the abrupt appearance of a late Paleozoic U-Pb age signature (Fig. 9) diagnostic of the Permian–Triassic Mitu Group, which is exposed today in the uplifted fault block. Collectively, we attribute accelerated accumulation and important shifts in facies and provenance to eastward advance of shortening within the formerly contiguous foreland basin, resulting in basin fragmentation and establishment of a new hinterland basin system (the initial Descanso-Yauri Basin) in the western Altiplano, west of the Pucapuca-Sorapata fault (Fig. 13B).

Displacement along the SW-directed Ayaviri thrust fault at 29–26 Ma is revealed by accurately dated footwall growth strata (Fig. 3; Perez and Horton, 2014). The timing of this structure roughly coincides with a modest paleocurrent shift from NE- to N-directed flow within the upper conglomeratic unit, as recorded in both the principal and southern (Llalli) sections (Figs. 2 and 4; Fig. DR1 [see ^{footnote 1}]). This change in sediment dispersal is consistent with a broad transition from relatively transverse flow to oblique dispersal between the bounding Pucapuca-Sorapata fault to the SW and the emerging Ayaviri thrust to the NE. We attribute the shift in sedimentation pathways to slip along the Ayaviri thrust, which further fragmented the basin and generated a topographic barrier along the Altiplano–Eastern Cordillera boundary sufficient to deflect sediment dispersal systems (Fig. 13C). This final stage of basin partitioning transformed the Altiplano study region from a once-extensive early Andean foreland basin into an isolated intermontane hinterland basin (Ayaviri or Tinajani Basin; Rousse et al., 2005; Carlotto et al., 2005; Carlotto, 2013).

Sedimentologic and chronostratigraphic results show rapid and sustained accumulation in the northern segment of the central Andean Plateau during the late Eocene–Oligocene, with final filling in the Miocene (Fig. 14). Next, we evaluate (1) various structural controls on late Eocene basin initiation, (2) the extent of late Eocene–Oligocene shortening, and (3) potential mechanisms for punctuated Oligocene subsidence, with consideration of a pulse of focused shortening along the Altiplano–Eastern Cordillera boundary and the broader implications for long-term Andean mountain building.

Basin inception could be attributed to (1) extensional or strike-slip deformation, (2) postextensional or postmagmatic thermal subsidence, (3) dynamic subsidence linked to subduction or lithospheric foundering, or (4) shortening-induced flexure.

Phases of fault-induced subsidence have been proposed for forearc, intra-arc, and backarc segments of the Andean margin (Godoy et al., 1999; Burns et al., 2006; Montes et al., 2005; Folguera et al., 2010; McNulty and Farber, 2002; Schildgen et al., 2009; Giovanni et al., 2010). Variations in plate convergence could account for a dynamic history in which long-term shortening was interrupted by phases of extension (Sébrier et al., 1988; Jordan et al., 2001a; Mpodozis and Cornejo, 2012). In addition, oblique convergence along the Peruvian margin (Pardo-Casas and Molnar, 1987; Pilger, 1984) is cited as a driver of potential strike-slip deformation (Soler and Bonhomme, 1990 ; Jaillard and Soler, 1996; Jaillard et al., 2000).

For the northern Altiplano, if Paleogene deposition occurred in isolated strike-slip basins (e.g., Carlotto, 2013), there should be a record of significant local tectonic rotations. However, most rotations in this zone were focused over the past 10–15 m.y., and many appear to be regional in extent (Rousse et al., 2003, 2005), consistent with rotation due to an along-strike gradient in Neogene shortening. Although Paleogene strike-slip or transtensional faulting cannot be ruled out, there is little evidence to support proposals of major strike-slip motion along the Eastern Cordillera–Altiplano boundary (Gilder et al., 2003; Sempere et al., 2004).

Thermal subsidence has been suggested for Andean arc and adjacent retroarc regions that underwent magmatic heating (Jordan and Alonso, 1987). Postextensional cooling and associated subsidence (e.g., McKenzie, 1978) seem unlikely, given the absence of significant Paleocene–Eocene extension (Mpodozis et al., 2005; Scherrenberg et al., 2012). In contrast, postmagmatic cooling is a potential mechanism for the central Andes, where shifts in arc position over the past 40 m.y. (Haschke et al., 2002; Mamani et al., 2010) have induced abrupt spatiotemporal variations in magmatism and, by inference, subsequent cooling-induced subsidence. However, within the northern Altiplano, there was limited magmatism prior to 40 Ma (Soler and Bonhomme, 1990; Sébrier and Soler, 1991; Mamani et al., 2010). Moreover, the documented rates of subsidence far exceed predicted values for postheating lithospheric configurations (McKenzie, 1978; Turcotte and Schubert, 2002). Although we rule out thermal subsidence as the principal driver for rapid accumulation of upper Eocene–Oligocene basin fill, it may have acted as a subordinate mechanism for the Miocene accumulation history following widespread Oligocene magmatism (Sandeman et al., 1995).

Arc advance and retreat in southern Peru is tied to shallowing and then resteepening of the subducting Nazca slab. Periods of flat-slab subduction along western South America have been linked to large-wavelength subsidence and anomalous accumulation patterns (e.g., Dávila et al., 2010; Shephard et al., 2010). Dynamic subsidence related to coupling of Andean lithosphere to flat-slab conditions is likely to account for no more than several hundred meters of accommodation space. In contrast, active foundering of thickened and negatively buoyant segments of the lower lithosphere could drive extreme amounts of dynamic subsidence (Pysklywec and Cruden, 2004). This mechanism has been invoked for rapid middle to late Miocene accumulation within closed basins farther south in the Bolivian Altiplano and Argentina Puna Plateau (Fig. 1), where earlier shortening potentially generated a thickened orogenic root (Garzione et al., 2006, 2008; DeCelles et al., 2015). For the northern Altiplano, however, the inception of rapid subsidence does not appear to have followed any substantial period of local shortening and crustal thickening.

Shortening and flexural loading along several candidate structures could explain the observed magnitude and rate of accumulation. Although late Eocene shortening and burial heating have been proposed for the Eastern Cordillera (Cordillera de Carabaya; Farrar et al., 1988; Kontak et al., 1990), this zone would have been situated ∼100–150 km to the east (Fig. 14A). A more probable explanation involves shortening in the Western Cordillera. Although most responsible structures may now be buried beneath the Neogene magmatic arc, their existence appears to be required by Paleogene tectonic rotations along the arc and forearc (Arriagada et al., 2008; Roperch et al., 2011).

We favor the interpretation of late Eocene–Oligocene flexural subsidence linked to activation of the Lagunillas, Pucapuca-Sorapata, and additional faults now buried beneath the Western Cordillera (Figs. 13 and 14A). These faults may represent inversion structures guided by inherited extensional faults of the Permian–Triassic Mitu rift system and Cretaceous West Peruvian Trough (e.g., Mégard, 1987; Sempere et al., 2002; Scherrenberg et al., 2012). Comparable Paleogene successions elsewhere in southern Peru suggest a regionally contiguous foreland basin system, including a possible Paleocene–middle Eocene record of distal basin fill that is not preserved in the study area (Carlotto, 1998, 2013; Carlotto et al., 2005). Although modeling of meaningful flexural scenarios is precluded by insufficient constraints on the original basin geometry and dimensions of thrust loads, the duration, magnitude, and rates of sediment accumulation are comparable to other Andean flexural basins (e.g., Jordan et al., 2001b; Echavarria et al., 2003). Moreover, the interpretation is consistent with forearc provenance evidence for late Eocene growth of a major sediment source in the Western Cordillera of Peru (Decou et al., 2013).

Our data support a regionally continuous foreland basin of Paleogene age and significant thickness in the central Andean Plateau of southern Peru, Bolivia, and northern Argentina (Fig. 14). In the northern Altiplano, the Eocene–Oligocene San Jerónimo (Puno) Group commonly exceeds 5 km (Newell, 1949; Palacios et al., 1993). For the central Altiplano, a 6-km-thick succession is composed of the upper Eocene–lowermost Miocene Potoco Formation or lower Corocoro Group (Ahlfeld and Branisa, 1960; Evernden et al., 1977; Martinez, 1980; Kennan et al., 1995; Horton et al., 2001; Murray et al., 2010), which provenance data show to be derived from a zone of early Andean uplift along the Chile-Bolivia border (Horton et al., 2002; Wotzlaw et al., 2011; Charrier et al., 2013). Although it remains difficult to identify the structures responsible for regional flexural subsidence, there are viable candidates. First, basement-involved fault zones in the western Altiplano are suggested by surface and drill-hole evidence for Precambrian basement capped at shallow levels by Neogene volcanic cover (Lehmann, 1978; Martinez, 1980; Martinez et al., 1995, 1996). Second, in the southern Altiplano, seismic-reflection and low-temperature thermochronometric data indicate late Eocene–Oligocene development of upper-crustal fold-and-thrust structures involving pre-Cenozoic basement (Elger et al., 2005; Oncken et al., 2006; Ege et al., 2007).

Farther south, the Puna Plateau of Argentina (Figs. 1 and 14A) contains a thick Paleogene record interpreted as the early Andean foreland basin (Kraemer et al., 1999; Coutand et al., 2001; Carrapa and DeCelles, 2008; DeCelles et al., 2011; Siks and Horton, 2011), consistent with proximal sedimentation and related structures of northern Chile (notably in the Cordillera Domeyko; Mpodozis et al., 2005), and with isotopic data suggesting late Eocene–Oligocene attainment of high elevations in the Puna Plateau (Canavan et al., 2014; Quade et al., 2015). When combined with our results from the northern Altiplano, the picture emerges of a contiguous belt of early Andean shortening along the entire western margin of the central Andean Plateau, from ∼15°S to 25°S.

We correlate this zone of Paleogene shortening northward with the Marañon fold-thrust belt (Fig. 14A), where thin-skinned structures are focused along the Western Cordillera of central and northern Peru (Mégard, 1978, 1984, 1987; Pfiffner and Gonzalez, 2013). The continuation of this belt as far north as ∼5°S underscores the regional continuity of early Andean shortening, which resulted in an extensive, well-organized foreland basin system prior to cratonward advance of deformation. We further emphasize that this regional configuration likely requires Eocene–Oligocene establishment of a contiguous western topographic barrier constituting at least one-third of the length of the Andean orogenic belt, from ∼5°S to 25°S (Figs. 1 and 14A). This western barrier would have been the product of not only Andean volcanic topography (e.g., Baker et al., 2014), but also significant isostatic uplift driven by shortening and crustal thickening. The barrier would have effectively isolated, by 40–30 Ma, an ∼2000 km length of the western forearc and coastal regions of Peru and northern Chile from the emerging Subandean and Amazon foreland system.

Rapid accumulation rates could be linked to variable processes. One intriguing possibility is that short pulses of rapid accommodation may be a dynamic product of lithospheric foundering. Theoretical models (Pysklywec and Cruden, 2004; Göğüş and Pysklywec, 2008) show that growth of a dense root during sustained shortening may lead to sinking of the unstable root and very rapid short-term subsidence. Such foundering may involve wholesale delamination of extensive roots or more localized drips of denser lithospheric mantle, and possibly lowermost crust (Beck and Zandt, 2002; DeCelles et al., 2009). For the study region, this process would require earlier large-scale shortening and accumulation of a thick orogenic root. Such shortening is potentially evidenced by Eocene deformation along the eastern margin of the central Andean Plateau (Cordillera de Carabaya of Peru and Cordillera Real of Bolivia; Fig. 14A), where thrust-related burial heating at ca. 45–40 Ma was followed by rapid cooling and erosional exhumation (Farrar et al., 1988; Kontak et al., 1990; Clark et al., 1990; Barnes et al., 2006; Gillis et al., 2006). For southern Peru, crustal buildup during Paleogene shortening may have preceded rapid foundering that drove the 30–27.8 Ma rapid accumulation of thick clastic fill documented in this study. Although foundering may have occurred over a broad region, it would have been most pronounced in the northern Altiplano, possibly due to a larger root or reactivation of a crustal inhomogeneity (Dorbath et al., 1993; Martinez et al., 1996; Carlier et al., 2005).

An alternative explanation for rapid Oligocene accumulation involves flexural subsidence adjacent to thrust loading by the central Andean backthrust belt. The growth stratal relationships and age control provided here (Figs. 3, 4, and 9) require that the Ayaviri fault, which defines the western limit of the backthrust belt along the Altiplano–Eastern Cordillera boundary, was active at 29–26 Ma. These constraints on focused fault activity overlap with the 2–3 m.y. phase of rapid clastic sedimentation. Furthermore, the considerable throw on this fault is consistent with crustal loading necessary to explain the rapid accumulation rates comparable to other shortening-related basins now preserved in the Andean hinterland (e.g., Allmendinger et al., 1997; Horton et al., 2001; Horton, 2012). Controls on the localization of the Ayaviri fault remain speculative but may reflect a transition to flat-slab subduction (Sandeman et al., 1995; Mamani et al., 2010) and/or reactivation of an inherited normal fault or deeper crustal suture (e.g., Dorbath et al., 1993; Martinez et al., 1996; Carlier et al., 2005).

Although we do not rule out the possible role of lithospheric foundering, we consider flexural loading along the Ayaviri fault to be the principal mechanism of rapid subsidence in the northern Altiplano study region. Although the rates of fault displacement remain unknown, the growth strata record spans no more than ∼3 m.y. Therefore, we emphasize that rapid slip along major deep-seated faults (e.g., Martinez et al., 1995; Lamb, 2011; Charrier et al., 2013) may play a key role in controlling punctuated accumulation in Andean basins, and by inference, may lead to rapid surface uplift in the absence of widespread shortening or lithospheric removal.

Sedimentologic, stratigraphic, provenance, U-Pb geochronologic, and magnetostratigraphic results demonstrate that protracted late Eocene through Oligocene accumulation of coarse clastic sediment in the northern Altiplano Plateau of southern Peru involved principally fluvial deposition of the San Jerónimo (Puno) Group in a retroarc basin governed by upper-crustal shortening and flexural loading. The sedimentary facies, sandstone compositions, conglomerate compositions, and detrital zircon U-Pb age populations reveal a major sediment source linked to initial shortening along a complex western basin margin, where potential structures are overprinted by the magmatic arc. This zone of early Andean shortening is interpreted as the along-strike continuation of the Marañon fold-and-thrust belt of northern and central Peru, and the Chilean Precordillera and Western Cordillera of Bolivia and northern Argentina.

Major shifts in grain size, facies, paleocurrents, detrital zircon U-Pb age spectra, and accumulation rates are linked to the sequential activation of several thrust structures, including two faults along the western margin and a major NE-dipping fault (the Ayaviri thrust) along the Altiplano–Eastern Cordillera boundary that partitioned the foreland basin into a narrow hinterland basin. Growth strata demonstrate Oligocene displacement along the Ayaviri fault, which is well correlated with a 30–27.8 Ma phase of very rapid accumulation (>1100–1800 m/m.y.) defined on the basis of U-Pb geochronology (11 samples) and magnetic polarity stratigraphy. This genetic link underscores the potentially dominant role of individual structures in dictating sediment accumulation patterns, particularly within hinterland regions and, for some cases, may obviate the need for anomalous generation of accommodation space through lithospheric removal or other dynamic mechanisms.

Our research efforts in southern Peru were made possible through collaborative interactions, detailed discussions, and logistical support from Victor Carlotto of the Universidad Nacional San Antonio Abad del Cusco and affiliated geoscientists from the Instituto Geológico Minero y Metalúrgico (INGEMMET). Field assistance was provided by Boris del Castillo Herrera and John Christian Arredondo Sosa of the Universidad Nacional San Antonio Abad del Cusco, and Nataleigh Vann of the University of Texas at Austin. We appreciate constructive reviews from Gregory Hoke and Nadine McQuarrie, discussions with Ron Steel, Jack Holt and Ben Siks, assistance with U-Pb geochronological analyses from Mark Pecha and staff of the University of Arizona LaserChron Center, and assistance from undergraduate researchers Mary K. Bales, Adam Bowerman, Alejandra Eljuri, and Kelly Hansard at the University of Texas at Austin Paleomagnetics Laboratory. Financial support was provided by National Science Foundation (NSF) grants EAR-0908518 and EAR-1338694, a NSF Graduate Research Fellowship, Geological Society of America graduate student research grants, an ExxonMobil geoscience grant, and support from the Jackson School of Geosciences at the University of Texas at Austin.