Andean topographic growth and basement uplift in southern Colombia: Implications for the evolution of the Magdalena, Orinoco, and Amazon river systems

The northern Andes form the chief orographic barrier separating the modern Orinoco, Amazon, and Magdalena River watersheds (Fig. 1). Although these river systems have collectively drained the northern half of South America, governed significant Andean erosion, and influenced Caribbean and Atlantic ocean chemistry, their genesis remains highly debated (e.g., Figueiredo et al., 2009; Hoorn et al., 2010; Sacek, 2014). During Paleogene time, most of northern South America drained northward into the Caribbean Sea, forming a large delta in the Maracaibo Basin of Venezuela (Díaz de Gamero, 1996; Escalona and Mann, 2006; Mann et al., 2006). Today, the Orinoco River empties into the equatorial Atlantic >1000 km farther east, and the Magdalena River of Colombia is now the largest single contributor of sediment to the Caribbean Sea (Fig. 1). The disruption of the original northward drainage configuration and the establishment of independent Orinoco, Amazon, and Magdalena systems are critically tied to basement uplift in the Eastern Cordillera fold-thrust belt and proximal Andean foreland of Colombia.

Paleocurrents, palynological assemblages, and mammal fossils suggest that Amazon River capture of the former southern Orinoco drainage system and isolation from the inter-Andean Magdalena drainage system (Fig. 1) was driven primarily by Neogene uplift of the Eastern Cordillera and Mérida Andes (Hoorn, 1994; Hoorn et al., 1995; Díaz de Gamero, 1996). In contrast, some reconstructions of the northern Andean and southern Caribbean margin propose that an independent Magdalena River was already established by 40–30 Ma (Gómez et al., 2005; Escalona and Mann, 2006). Although uplift of the Mérida Andes in Venezuela (Fig. 1) was long considered the driving mechanism behind the delineation of the Orinoco and Magdalena systems, more recent work recognizes the fundamental influence of foreland basement arches, thrust systems, and sediment accumulation on drainage evolution (Mora et al., 2010a; Roddaz et al., 2010; Silva et al., 2013; Caballero et al., 2013; Horton et al., 2015a). Understanding the evolving drainage configuration of northern South America has fueled a vigorous debate about the formation of the modern Amazon River, with estimates ranging from a middle Miocene to a Pleistocene onset of transcontinental drainage (e.g., Potter, 1997; Campbell et al., 2006; Campbell, 2010; Latrubesse et al., 2010). New constraints on the timing, mechanics, and geomorphology of the transition from mid-Cenozoic to modern drainage configurations are critical to reconstructions of past rivers and their influence on the geological and biological dynamics of South America (e.g., Hoorn et al., 2010; Ribas et al., 2012; Baker et al., 2014).

In this study we utilize a multiproxy approach to constrain topographic growth of the Eastern Cordillera in southern Colombia and the resulting establishment of an isolated Magdalena River system. We focus our study on the Garzón Massif, the 2500–3000-m-high basement uplift that bounds the Andean headwaters of the Magdalena watershed, and whose foreland counterpart, the Macarena high and broader Vaupes arch, forms the Amazon-Orinoco drainage divide (Fig. 1). We constrain the Neogene exhumation age of this massif by studying clastic deposits of the Neiva hinterland basin in the adjacent Upper Magdalena Valley (Fig. 2A) using sediment provenance techniques to track the influx of basement-derived material, and apatite fission track (AFT) thermochronometry to constrain the timing and pace of exhumation. In addition, we use two paleosol-based paleoclimatic proxies to place rough bounds on initial development of the modern topographic barrier and its corresponding orographic rain shadow.

The northern Andes contain the Cenozoic structural and sedimentary record of advancing fold-thrust systems in the Central and Eastern Cordilleras. In Colombia, the Central Cordillera demarcates the latest Cretaceous–Paleogene thrust front, whereas the Eastern Cordillera represents a younger bivergent contractional belt that reactivated a Mesozoic rift system in a combination of thin-skinned ramp-flat thrust systems and thick-skinned basement-involved structures (Colletta et al., 1990; Dengo and Covey, 1993; Casero et al., 1997; Ramon and Rosero, 2006; Mora et al., 2006; Parra et al., 2009, 2012; Mora et al., 2013; Wolaver et al., 2015). The north- to northeast-trending Central and Eastern Cordilleras are separated by the Magdalena Valley, which narrows southward toward the Magdalena River headwaters in southern Colombia, where the two range systems intersect (Fig. 1).

The Upper Magdalena Valley is bounded to the west by the Chusma fault system (Fig. 2A), an Eocene–Oligocene thrust system defining the eastern front of the Central Cordillera (Butler and Schamel, 1988; Mojica and Franco, 1990; van der Wiel and van den Bergh, 1992; Sarmiento and Rangel, 2004). The eastern boundary of the Upper Magdalena Valley is defined by the west- to northwest-verging Algeciras fault system (Fig. 2A), a complex series of thrust, transpressional, and right-lateral strike-slip faults, including the alternatively named Garzón, Suaza, Pitalito, and Altamira faults (Chorowicz et al., 1996; Casero et al., 1997; Montes et al., 2005; Velandia et al., 2005), which bound crystalline basement rocks of the northeast-trending Garzón Massif at the southern tip of the Eastern Cordillera (Fig. 1). This complex fault system separates Jurassic granitic rocks (Algeciras monzogranite) from late Mesoproterozoic–early Neoproterozoic high-grade metamorphic rocks (gneiss, schist, and migmatites), marks the southern expression of major active transpression in Colombia, and has been dominated by wrench tectonics throughout the Quaternary (Chorowicz et al., 1996; Velandia et al., 2005; Bustamante et al., 2010). However, the pronounced structural relief along the >150-km-long Algeciras fault zone and low-angle orientation (∼15° southeast dip) requires dip-slip displacement for a significant portion of its history (e.g., Bakioglu, 2014). The temporal transition from shortening to principally strike-slip deformation is considered to have occurred during latest Miocene–Pliocene time, but the precise timing remains unclear (Velandia et al., 2005; Egbue and Kellogg, 2010; Egbue et al., 2014). As a result of this varied deformation history, the Garzón Massif defines a structural high with an overall north- to northeast-plunging configuration in which regional-scale exposures of the prevailing Precambrian metamorphic basement and flanking Jurassic granitic rocks plunge beneath a northern carapace of younger geologic units (Fig. 2).

The burial and exhumation histories of the Garzón Massif are not well constrained. Eocene–Oligocene strata have been mapped in nonconformable contact upon Proterozoic metamorphic rocks on the eastern flank of the Garzón Massif, suggesting that parts of the basement massif were exposed at the surface prior to large-scale burial beneath this clastic sedimentary succession (Rodríguez et al., 2003; Bakioglu, 2014; Wolaver et al., 2015). The full massif was exhumed at some point after this phase of Eocene–Oligocene sedimentation, although precise constraints on uplift timing remain elusive. Previously reported AFT ages span from 13.9 ± 2.3 to 9.2 ± 2.0 Ma (van der Wiel, 1991), while other estimates based on structural and stratigraphic relationships range from 12.9 to 6.4 Ma (Guerrero, 1993, 1997; Butler and Schamel, 1988; van der Wiel et al., 1992; Wolaver et al., 2015). The range of interpreted ages for the onset of Neogene basement uplift may reflect true spatial variations or the diversity of approaches taken by the previous efforts. We aim to refine these estimates by integrating new results capable of providing a more complete picture of the Garzón Massif’s topographic history and the influence of basement block uplifts on large drainage systems.

We studied the Neogene succession at two field localities in the Neiva Basin within the Upper Magdalena Valley of southern Colombia and obtained samples along six field transects spanning depositional ages of 13.8–6.4 Ma (Figs. 2 and 3; Supplemental Table 1¹). The chronostratigraphic framework for studied basin fill was established by isotopic dating of interbedded ash beds and previous magnetostratigraphic studies, as described in the next section. This middle to upper Miocene stratigraphic interval is exposed in two main areas (Fig. 2A), where the basal contact is commonly buried, or the succession unconformably overlies Jurassic, Cretaceous, or Eocene–Oligocene units. The lower Miocene Honda Group is exposed in the northern basin (La Venta site; Fig. 2B) and the upper Miocene Gigante Formation crops out along the southern basin margin (Gigante site; Fig. 2C). Due to variable exposure and accessibility, all Honda and Gigante samples were taken from these localities. Previous efforts have established a stratigraphic framework for Neogene clastic fill in the Upper Magdalena Valley, resulting in a variety of naming schemes for comparable stratigraphic sections (Van Houten and Travis, 1968; Wellman, 1970; Van Houten, 1976; Butler and Schamel, 1988; Guerrero, 1997). Here we employ the stratigraphic and geochronological framework for the Miocene Honda Group detailed in the La Venta paleontological, magnetostratigraphic, and ⁴⁰Ar/³⁹Ar geochronological studies of the northern Neiva Basin (e.g., Guerrero, 1993, 1997; Flynn et al., 1997). For the upper Miocene Gigante Formation, exposed primarily in the southern Neiva Basin, we use the nomenclature of van der Wiel et al. (1992), although their Honda divisions differ slightly from those of Guerrero (1997).

In the north, at the La Venta site (Fig. 2B), the middle Miocene Honda Group is divided into the La Victoria and Villavieja Formations (Fig. 3). The units are separated by a 5–10-m-thick clast-supported pebble conglomerate, the Cerbatana Conglomerate, which is exposed across the northern Neiva Basin (Guerrero, 1997). The older La Victoria Formation is principally composed of bioturbated mudstones, with decimeter-scale bands of gray, purple, and green. Many intervals contain carbonate nodules, with well-developed rhizoliths near the top of the section. These claystones and siltstones are periodically interrupted by meter-scale, trough cross-bedded sandstones with erosive bases. Although most pinch out laterally over 20–30 m, several thicker sandstone intervals are continuous across the 2–4 km basin width and form important marker beds (Guerrero, 1997). The La Victoria section is ∼500 m thick in the type section, thickens southward to ∼1000 m near Gigante, and has been interpreted as a meandering river system with significant soil development in fine-grained overbank zones (Wellman, 1970; Guerrero, 1997).

The Villavieja Formation is 580 m thick in its type section and conformably caps the 5–10-m-thick Cerbatana Conglomerate. Whereas the base contains fossiliferous green to gray mudstones and paleosols similar to the La Victoria Formation, the upper 350 m is characterized by brilliant red paleosols and mudstones referred to as the Polonia red beds. Soil carbonate nodules are less common and not as well developed as in the red beds, but still occur in discrete horizons. Identification and classification of paleosols within these mudstones is based on descriptions provided by Guerrero (1993). Trough cross-bedded sandstones occur frequently in this formation and show an upsection increase in their abundance relative to overbank deposits. In contrast to the multistory channel belts of the La Victoria Formation, the Villavieja sandstones occur as restricted 1–5-m-thick channelized units that persist no more than 10–20 m along strike. The contact between the Villavieja Formation (upper Honda Group) and overlying Gigante Formation (Huila Group of Guerrero, 1997) is marked by a low-angle angular unconformity at La Venta, but is conformable to the south near Gigante (van der Wiel et al., 1992; Guerrero, 1997). Here the Villavieja Formation is thinner (∼250 m) and has a greater proportion of fluvial channel to overbank facies than the La Venta site (van der Wiel and van den Bergh, 1992).

The chronology of the middle Miocene Honda Group (La Victoria and Villavieja Formations) was established previously through magnetic polarity stratigraphy of the basin-fill succession and hornblende and plagioclase ⁴⁰Ar/³⁹Ar geochronology of 8 samples of interbedded tuffs. The Honda Group spans from 13.8 to 11.6 Ma; the La Victoria–Villavieja contact is estimated at 12.5 Ma (Guerrero, 1993; Flynn et al., 1997) (Fig. 3). On the basis of these age constraints, it appears that the calculated average sediment accumulation rates increased from the La Victoria Formation (∼400 m/m.y.) to the overlying Villavieja Formation (∼500 m/m.y.). This finding is consistent with a pronounced change in sedimentary architecture, from the broad, amalgamated channel belts of the La Victoria Formation to the narrow, isolated channel bodies of the Villavieja Formation (e.g., Leeder, 1977; Mackey and Bridge, 1995; Heller and Paola, 1996).

In addition, paleocurrent orientations were measured in trough cross-bedded sandstones throughout the Honda Group using measurements of at least 10 right and left limbs at each site (e.g., DeCelles et al., 1983). In lower stratigraphic levels, the La Victoria Formation records predominantly east-directed sediment dispersal. The intermediate section, including the Cerbatana Conglomerate, appears to alternate between eastward and westward paleoflow. These deposits are capped by an interval of north-directed paleoflow recorded in the upper Villavieja Formation (Supplemental Table 2²). The paleocurrent changes and contemporaneous shifts in sedimentation rate and channel architecture are considered indicative of the transition from an unconfined, east-directed meandering fluvial system to an axial braided fluvial system confined by the development of topography in the Garzón Massif and broader Eastern Cordillera directly east of the La Venta region (e.g., Guerrero, 1997).

In the southern Neiva Basin (Fig. 2C), near Gigante, the type section of the upper Miocene Gigante Formation (composed of the Neiva, Los Altares, and Garzón Members) (Fig. 3) is exposed at Quebrada Guandinosita (van der Wiel, 1991). At the base, the ∼150-m-thick Neiva Member conformably caps the upper Honda Group (Villavieja Formation) and represents a considerable shift in depositional environments, with the lower 50 m predominantly composed of pebble conglomerates and trough cross-bedded coarse-grained sandstones with lenticular channel-fill features. The intermediate 50 m of the Neiva Member is a poorly exposed, mudstone-dominated interval capped by 50 m of multistory clast-supported cobble-boulder conglomerates interpreted to represent a north-flowing longitudinal braided fluvial system resulting from a post–Honda Group increase in sediment supply (van der Wiel, 1991).

Within the Gigante Formation, the upsection stratigraphic transition to the Los Altares Member is marked by a decrease in grain size, with the lower 70 m characterized by alternating mudstones and channel sandstone deposits. The most significant change in this member is the appearance of pumice-rich volcaniclastic debris flow deposits separated by thin sandstone intervals with erosive bases (Van Houten, 1976; van der Wiel, 1991). Planar-laminated volcaniclastic sandstones dominate upper levels of the Los Altares Member. Clasts from the overlying Garzón Member are predominantly of metamorphic origin, suggesting a shift in sediment source from the Central Cordillera magmatic arc to basement rocks of the Garzón Massif.

The depositional age of the Gigante Formation is established by previously reported biotite and hornblende K-Ar ages of volcaniclastic rock (Van Houten, 1976; van der Wiel, 1991; van der Wiel et al., 1992). These dates roughly delimit the Los Altares Member to 8.0 Ma at the base and 6.4 Ma at the upper transition to the Garzón Member (Fig. 3). New constraints on the maximum depositional ages are provided here (see following) by the youngest U-Pb populations in two of our detrital zircon samples: 8.6 ± 0.6 Ma from the upper Neiva Member and 7.7 ± 0.7 Ma from Los Altares volcaniclastic debris flows. These U-Pb age constraints are consistent with previous geochronological results, and verify a late Miocene age for the Gigante Formation.

To determine the provenance of the Miocene La Victoria, Villavieja, and Gigante Formations, modal compositional analyses of petrographic thin sections were performed for 15 sandstone samples from the northern (La Venta) and southern (Gigante) segments of the Neiva Basin. Thin sections were stained for calcium and potassium feldspars to aid in grain identification. For each sample, at least 300 point counts of mineral grains >62.5 µm were recorded following the Gazzi-Dickinson method (Ingersoll et al., 1984). Primary classifications are listed in Supplemental Table 3³. The counts were gathered into composite metrics of quartz, feldspar, and various lithic fractions for plotting on standard ternary diagrams (Dickinson et al., 1983). Enigmatic microcrystalline grains with iron oxide staining added some uncertainty in distinguishing components of the lithic fraction, particularly in terms of chert versus microcrystalline volcanic or metamorphic fragments. We focus on Q-F-Lf (quartz–feldspar–Folk [1980] lithic fragments) and Qm-F-Lt (monocrystalline quartz–feldspar–total lithic fragments) ternary diagrams (Fig. 4) rather than Qt-F-L (total quartz–feldspar–lithic fragments) distributions in order to minimize the influence of chert on the overall quartz fraction (e.g., Folk, 1980).

Samples throughout all three formations exhibit heterogeneous compositions relatively evenly distributed among quartz, feldspar, and lithic fractions (Figs. 4 and 5; Table 1). Most samples are categorized as lithic arkose or feldspathic litharenite (Fig. 4B), with the oldest unit, the La Victoria Formation, slightly more quartzose than the overlying Villavieja and Gigante Formations. An exception is uppermost Gigante sample S15, which exhibits the largest quartz fraction at 50.2%. Total feldspar is relatively uniform throughout the Honda Group (La Victoria and Villavieja Formations), with the exception of sample S5, which is substantially richer in plagioclase and nonvolcanic lithic fragments than the rest of the Honda Group. This sample, from immediately below the Cerbatana Conglomerate (at the La Victoria–Villavieja contact), may highlight a provenance change linked to the transient shift in depositional conditions responsible for gravel input. In upper stratigraphic levels, less mature samples from the Gigante Formation (S11–S15) tend to have much higher proportions of volcanic lithic fragments and feldspars than samples from the underlying Honda Group (Figs. 4 and 5; Table 1), likely due to a greater proximity of magmatic arc rocks in the Central Cordillera.

Detrital zircon U-Pb geochronological analyses were performed on 11 sandstone samples from the Neiva Basin; 7 La Venta and four Gigante samples were crushed and zircon grains separated using standard techniques, including water, heavy liquid, and magnetic separation. A random selection of 100–130 zircons in each sample was measured via laser ablation–inductively coupled plasma–mass spectrometry (LA-ICP-MS) at the University of Arizona LaserChron Center, following standard analytical procedures (Gehrels et al., 2008; Gehrels, 2012, 2014). A known standard of Sri Lanka zircon was measured between every five unknown grains to correct for transient variations in interelement and intraelement fractionation over the course of the run. The ²⁰⁶Pb/²⁰⁷Pb age is reported for all grains with a ²⁰⁶Pb/²³⁸U age older than 900 Ma; due to the relatively low abundance of ²⁰⁷Pb in grains younger than 900 Ma, the ²⁰⁶Pb/²³⁸U age is reported for these grains. Individual measurements with errors >10% in ²⁰⁶Pb/²³⁸U and ²⁰⁶Pb/²⁰⁷Pb ratios were discarded. For grains where the ²⁰⁶Pb/²⁰⁷Pb age was reported, measurements exhibiting >30% discordance or >5% reverse discordance were also discarded (Gehrels et al., 2008). U-Pb results are presented in Supplemental Table 4⁴ and plotted as age histograms and probability functions (Fig. 5) to reveal stratigraphic variations in U-Pb age distributions. For eight samples with populations of young Cenozoic zircons (samples Z1, Z2, Z3, Z4, Z5, Z8, Z9, and Z10), the calculated mean of tightly clustered grain ages (Table 2; Fig. 5; Supplemental Fig. 1⁵) provides a robust constraint on the maximum depositional age of the host strata and may approximate the true depositional age in cases with considerable contributions from syndepositional volcanic sources. Seven of the eight samples (excluding sample Z5) yield mean ages that closely match the depositional ages defined by previous ⁴⁰Ar/³⁹Ar and magnetostratigraphic results (van der Wiel, 1991; van der Wiel et al., 1992; Guerrero, 1993, 1997; Flynn et al., 1997).

U-Pb age populations in our sample set (Fig. 5) are compatible with potential sediment source regions surrounding the Neiva Basin. These sources include: (1) late Mesoproterozoic–early Neoproterozoic metamorphic basement of the Garzón Massif; (2) Jurassic granitoids; (3) Cretaceous sedimentary rocks containing recycled Proterozoic cratonic zircons; and (4) Cenozoic volcanic rocks from the Central Cordillera magmatic arc. In the following we describe the age, distribution, and distinguishing characteristics of these different sources.

The Garzón Massif is composed of several distinct metamorphic units with both igneous and sedimentary protoliths. Grains derived from principally gneisses and schists of this basement massif display a range of Precambrian ages, including principally late Mesoproterozoic–early Neoproterozoic crystallization ages of protolith and metamorphic zircons concentrated at 1200–900 Ma, with limited older zircon grains and no grains older than 1500 Ma (Restrepo-Pace et al., 1997; Cordani et al., 2005; Cardona et al., 2010; Ibañez-Mejía et al., 2011). Because comparable metamorphic basement is not exposed farther west, identification of a sizeable 1200–900 Ma population in the detrital record would represent a robust indication of an eastern (Garzón Massif) sediment source for the Neiva Basin.

Considerable volumes of intrusive igneous rocks in the northern Andes were emplaced during several phases spanning the Jurassic, with an age range of 200–150 Ma in southern Colombia at 1°–4°N (Aspden et al., 1987; Bustamante et al., 2010; Villagómez et al., 2011). These granitoid plutons are preserved in both the Central and Eastern Cordilleras on the western and eastern flanks of the Neiva Basin (Gómez Tapias et al., 2007). Here the Jurassic granitic plutons are in intrusive and fault contact with Paleozoic strata and older metamorphic basement (Figs. 2A, 2C), including principally Mesoproterozoic rocks of the Eastern Cordillera (Garzón Massif) and Neoproterozoic–lower Paleozoic rocks of the Central Cordillera (Cajamarca complex).

Cretaceous sedimentary rocks in Colombia display substantial variation in their detrital zircon U-Pb age spectra. Lower Cretaceous stratigraphic units are often heterogeneous, containing significant zircon populations of Cambrian–Ordovician age (500–400 Ma), late Mesoproterozoic–earliest Neoproterozoic (Sunsás-Grenville) age (1250–950 Ma), and older Proterozoic grains ranging from 2060 to 1300 Ma (Horton et al., 2010, 2015a; Saylor et al., 2013). The proportion of younger grains diminishes upsection, such that uppermost Cretaceous units are commonly dominated by 2060–1300 Ma ages, likely due to bulk contributions from Paleoproterozoic–middle Mesoproterozoic cratonic sources and gradual burial of Andean sources (Horton et al., 2010, 2015a). Cretaceous strata were deposited unconformably over Jurassic sedimentary and plutonic rocks (Ramon and Rosero, 2006; Mora et al., 2010b), and are widespread on both flanks of the Neiva Basin.

The Central Cordillera has been part of a magmatic arc complex in Colombia since the Late Cretaceous (Aspden et al., 1987; Bayona et al., 2012). Compositional analyses of Neogene basin fill in the Upper Magdalena Valley have been tied to coeval magmatic sources in the adjacent Central Cordillera (Van Houten and Travis, 1968; Van Houten, 1976; Guerrero, 1993, 1997). Although minor volcanic provinces exist elsewhere, there are no mapped Cenozoic volcanic units at the southern end of the Eastern Cordillera (Fig. 2; Gómez Tapias et al., 2007). Therefore, significant populations of Cenozoic-age zircon grains can be considered clear signatures of a Central Cordillera source along the western flank of the Neiva Basin. Within the basin, young zircon grains of middle to late Miocene age may reflect direct erosion from syndepositional volcanic rocks or direct input from volcanic ash-fall deposition.

At the base of the Neiva Basin succession, samples Z1–Z4 from the middle Miocene La Victoria Formation contain predominantly Cenozoic and subordinate Jurassic age zircon grains (Fig. 5), suggesting a western source in the Central Cordillera during initial filling of the Neiva Basin. The abundance of Miocene zircon grains allows for calculation of the youngest U-Pb age populations for successive samples Z1–Z4, which yield weighted mean ages of 14.4 ± 1.9 Ma (Z1, n = 4), 13.75 ± 0.4 Ma (Z2, n = 34), 13.75 ± 0.9 Ma (Z3, n = 13), and 13.2 ± 1.3 Ma (Z4, n = 5) (Table 2; Fig. 5; Supplemental Fig. 1). These ages agree well with existing magnetostratigraphic results constraining accumulation of the La Victoria Formation between 13.8 and 12.5 Ma (Guerrero, 1993; Flynn et al., 1997). Therefore, the youngest zircon populations not only constrain the maximum depositional ages, but likely represent accurate estimates of the true depositional ages of stratigraphic units in close proximity to the Andean magmatic arc (e.g., Perez and Horton, 2014; Horton et al., 2015b). Sample Z4 from the Cerbatana Conglomerate at the La Victoria–Villavieja transition contains clasts of principally igneous rocks with subordinate quartzite, slate or schist, and chert (Van Houten and Travis, 1968; Guerrero, 1993, 1997). The chert fragments are likely derived from Cretaceous sedimentary rocks, which are preserved on both the western and eastern sides of the basin. Therefore, although the U-Pb age distributions show a sustained influx from the Central Cordillera along the western flank of the Neiva Basin, we cannot rule out possible contributions from recycled Cretaceous cover strata from the Garzón Massif to the east.

Within intermediate stratigraphic levels, the Villavieja Formation records significant variations in U-Pb age distributions reflective of exhumation of the Garzón Massif within the broader Eastern Cordillera east of the Neiva Basin. Sample Z5 shows a continued dominance of Cenozoic age distributions, including a youngest population of 13.2 ± 1.0 Ma (Z5, n = 5), but an increased proportion of Jurassic ages (Table 2; Fig. 5). Upsection, sample Z6 from the upper Villavieja Formation marks the abrupt disappearance of Cenozoic grains (Fig. 5) and wholesale replacement by a unimodal Jurassic signature, with 79 of 87 total ages clustered between 200 and 170 Ma. For sample Z7, at the top of the Villavieja Formation, the Jurassic population abruptly disappears and is replaced by a heterogeneous, multimodal mix of Precambrian ages, which span principally from 1600 to 1000 Ma (Fig. 5). Although the 1200–900 Ma grains are emblematic of the Garzón Massif, the older grains indicate input from other sources, most likely recycled from Cretaceous–Paleogene sedimentary units (e.g., Horton et al., 2010, 2015a). Overall, the conspicuous absence of Cenozoic zircon populations in samples Z6 and Z7 suggests no major western sediment source for the upper Villavieja Formation. This significant departure from underlying samples requires an extensive provenance shift ca. 12 Ma, consistent with early detrital contributions from the east, likely from the Garzón Massif and/or Cretaceous–Paleogene cover strata.

In the upper levels of the Neiva Basin, further provenance variations are expressed in detrital zircon U-Pb results for the upper Miocene Gigante Formation. Samples Z8, Z9, and Z10 from the lower to middle Gigante Formation (Neiva and Los Altares Members) are all dominated by Cenozoic and Jurassic zircons (Fig. 5), similar to lower basin fill (Z1–Z5), possibly implying a return to a western source. The youngest U-Pb age populations for successive samples Z8–Z10 yield weighted mean ages of 8.6 ± 3.4 Ma (Z8, n = 2), 8.6 ± 0.6 Ma (Z9, n = 15), and 7.7 ± 0.7 Ma (Z10, n = 23) (Table 2; Fig. 5; Supplemental Fig. 1). In contrast to these three samples, sample Z11 from the uppermost Gigante Formation (Garzón Member) contains a large population of zircon grains spanning from 1500 to 950 Ma, a minor population of Jurassic zircons, but no Cenozoic zircons (Fig. 5). This age distribution is nearly identical to reported ages from Garzón crystalline basement to the east (Restrepo-Pace et al., 1997; Cordani et al., 2005; Cardona et al., 2010; Ibañez-Mejía et al., 2011), suggesting a shift to a major sediment source region in the Garzón Massif east of the Neiva Basin.

The provenance variations reflected in samples Z1–Z12 show a middle to late Miocene shift from a western to eastern sediment source, a prolonged transition interpreted to reflect the increasing influence of the Garzón Massif (and broader Eastern Cordillera) relative to the Central Cordillera. Although Garzón-derived sediment may have appeared by ca. 12.5 Ma (sample Z4), the first clear signal appeared ca. 11.6 Ma (samples Z6 and Z7), and a fully dominant signal was not realized until ca. 6.5 Ma (sample Z11) (Fig. 5). This protracted transition included complex depositional alternations involving (1) an integrated contribution of sediments eroded from both the eastern and western basin margins in an axial (longitudinal) fluvial system, yielding a history of highly diverse age distributions, and/or (2) competing sediment influx from alluvial-fan systems originating from opposing basin margins, producing a record with abrupt appearances and disappearances of particular age populations. Given the observed stratigraphic appearance and disappearance of Cenozoic, Jurassic, and Precambrian age signatures emblematic of western versus eastern sources (Fig. 5), we favor the second option for most of the middle to late Miocene provenance record. We propose that the Neiva Basin was filled from 12.5 to 6.4 Ma by competing distributary systems sourced principally from the west (Central Cordillera), but with minor mixing of signals and progressively greater contributions from the emerging eastern topographic barrier (Garzón Massif and broader Eastern Cordillera).

AFT thermochronometry was employed to evaluate late Cenozoic exhumation of the central segment of the Garzón Massif structural domain. The abundance, length, and distribution of tracks resulting from the spontaneous fission of ²³⁸U isotopes provide constraints on the thermal history of a sample below the 100–130 °C closure temperature for apatite (e.g., Donelick et al., 2005; Ehlers, 2005). We collected 10 samples of crystalline basement rocks (granite, gneiss, and schist) along a transect across the Garzón Massif, crossing a relatively straight, northeast-striking strand of the Algeciras fault system along the western margin of the massif (Fig. 2). The six lower-elevation samples (Fig. 2C), including three samples from Jurassic granitic rocks west of the Algeciras fault (samples GM01–GM03) and three samples from Proterozoic gneisses and schists on the eastern fault block (samples GM04–GM06), were found to have sufficient apatite material for AFT analyses, and the results are reported in the following.

Rock samples were crushed and milled, then apatite grains were separated using a water table, heavy liquids, and magnetic separation. AFT analyses were performed by Apatite to Zircon Inc., following the procedures outlined by Donelick et al. (2005). Apatite grains were mounted, polished, and etched with 5.5N HNO₃ to expose fission tracks. Spontaneous tracks were counted to determine the AFT age, and uranium and other elemental concentrations were then determined via LA-ICP-MS. The samples were irradiated with a ²⁵²Cf source, and re-etched to expose more spontaneous fission tracks for measurement of track lengths. In order to distinguish multiple apatite populations having different kinetic parameters, track lengths, AFT ages, and U-Pb ages calculated for each grain were plotted against the multiple kinetic parameter r_mr0 (Ketcham et al., 1999; Carlson et al., 1999) and examined for potential dependencies via the χ² test. Thermal histories for each sample were inverse-modeled using HeFTy software (version 1.8.3; June 2014) (Supplemental Files for HeFTy⁶), with track lengths projected onto the c-axis to account for track angle (Ketcham, 2005). Each modeling run was performed with surface temperature fixed at 20 °C and the condition that the massif reached >140 °C burial temperatures at some point between 40 and 10 Ma, a loose constraint placed by geologic evidence that the massif was exposed at the surface during the Eocene, and subsequently buried by mid-Cenozoic strata prior to its final exhumation (Rodríguez et al., 2003; Wolaver et al., 2015).

AFT results for the six samples show uniform, elevation-correlated values, with pooled AFT ages ranging from 5.8 to 3.6 Ma (Fig. 6; Table 3). All samples were completely reset, displaying unimodal track length distributions, consistent with cooling-only thermal histories. Whereas the three samples west of the Algeciras fault (samples GM01–GM03) yield pooled AFT ages of 4.6–3.6 Ma and a composite average track length of 13.2 µm, those from the uplifted eastern fault block (samples GM04–GM06) exhibit 5.8–5.0 Ma pooled AFT ages and a 14.5 µm composite average track length (Table 3). Most samples are dominated by a single kinetic population; in all but one case (sample GM05), fewer than three anomalous grains were removed to obtain a single population for thermal modeling.

Our HeFTy modeling reveals that this uniform resetting requires that all of these basement samples were originally located below the zone of partial annealing (Ketcham et al., 1999; Ehlers, 2005), indicating principally post-Miocene removal of at least 3–4 km of overburden, assuming typical geothermal gradients (20–25 °C/km). The calculated AFT ages therefore represent minimum estimates for the onset of exhumation, suggesting that displacement on the main strand of the Algeciras fault system initiated prior to the AFT cooling ages. The three samples from the western side of the fault (samples GM01–GM03) recorded steady, continuous exhumation since ca. 6 Ma, with an average cooling rate of ∼20–25 °C/m.y. (Fig. 6). It is intriguing that the samples from the eastern side of the Algeciras fault (samples GM04–GM06) consistently show convex-upward cooling paths with rapid exhumation from ca. 6 to 4 Ma followed by a protracted period of slow, near-surface cooling (Fig. 6). These thermal histories reflect the extremely long tracks preserved in these samples, in that the only thermal models producing acceptable results require that the samples were exhumed rapidly and then remained near the surface where rates of track annealing were diminished (Fig. 6).

Although our sediment provenance results suggest that exhumation of parts of the Garzón Massif had commenced by 12.5–11.6 Ma, and previous AFT studies from areas farther south suggest rapid cooling of the Garzón Massif ca. 14–9 Ma (van der Wiel, 1991), our AFT ages require the greatest amounts of exhumation (at least 3–4 km) within the central segment of the Garzón Massif to have occurred after ca. 6 Ma. However, the conflicting thermal histories suggest important temporal and spatial variations in exhumation-induced cooling on the eastern and western flanks of the fault, and potential variations from north to south within the Garzón Massif. The convex-upward cooling paths (Fig. 6) displayed by the samples from the eastern fault block (samples GM04–GM06) reflect a significant reduction in exhumation rate since 4–3 Ma, which could be indicative of (1) a transition from prevailing dip-slip shortening to strike-slip displacement (with diminished generation of structural relief along the straight fault segment with a linear surface trace) and/or (2) reduced displacement due to westward propagation of the locus of upper crustal deformation in this central segment of the Garzón Massif (likely reaching the modern deformation boundary along the easternmost Neiva Basin). Although neotectonic data and current geodetic measurements show that the Algeciras fault system currently accommodates overall right-lateral motion, the significant structural relief across the fault (with basement rocks juxtaposed against Pliocene–Quaternary basin fill) and its low-angle orientation (∼15° east-southeast dip) are compatible with an earlier history involving major dip-slip displacement (Rodríguez et al., 2003; Bakioglu, 2014).

The AFT cooling paths (Fig. 6) are considered to record a temporal transition in exhumation patterns in which the eastern (hanging wall) samples were rapidly exhumed prior to 4–3 Ma during thrust-reverse displacement along the southeast-dipping Algeciras fault, and then slowly approached the surface by erosion after horizontal transverse motion began to dominate along the fault. This is consistent with plate reconstructions that show that post-Miocene Nazca–South America plate convergence became increasingly oblique as the northern Andean block began to rigidly escape to the northeast (Freymueller et al., 1993; Kellogg and Vega, 1995; Trenkamp et al., 2002; Boschman et al., 2014). Moreover, samples on the western (footwall) side of the fault (samples GM01–GM03) show more continuous cooling paths from 4 to 3 Ma onward, consistent with propagation of transpressive deformation and westward encroachment into the Neiva Basin, possibly related to strain partitioning as the Algeciras fault began to accommodate more horizontal transverse motion. The further issue of older (broadly 14–9 Ma) AFT ages farther south in the Garzón Massif (van der Wiel, 1991) could be related to the regional north to northeast plunge of the massif, or could reflect an additional north to northeastward along-strike progression of deformation, possibly related to a long-term transition to a more oblique transpressional configuration (e.g., Acosta et al., 2004). Collectively, the AFT results suggest that the most rapid vertical exhumation of the Garzón Massif occurred in latest Miocene–early Pliocene time (ca. 6–4 Ma) with subsequently slower basement exhumation, greater transverse motion, and focused activity along different faults as a result of strain partitioning after ca. 4–3 Ma.

The preceding results help constrain the timing of exhumation for the Garzón Massif, but do not provide direct constraints on surface uplift. To detect climatic changes that could be driven by the development of surface topography, we use two paleoclimatic proxies that could potentially help define the onset of the semiarid conditions now characterizing the Upper Magdalena Valley (Tatacoa Desert) along the western flank of the Garzón Massif. Today, the 2500–3000-m-high massif forms an effective orographic barrier, intercepting moisture derived from easterly trade winds (Rozanski and Araguás, 1995; Poveda and Mesa, 1997). The modern climate of the Tatacoa Desert is characterized by an intensely evaporative environment; although it receives 1300 mm/yr of precipitation, the potential evapotranspiration is estimated to be >1600 mm/yr, resulting in some of the most arid conditions in Colombia (Instituto de Hidrología, Meteorología y Estudios Ambientales, 2012). We utilize two independent paleoclimatic proxies to differentiate between several possible scenarios for the uplift history of the range.

Typically in paleoelevation studies, oxygen isotope compositions of soil carbonates are assumed to be primarily controlled by Rayleigh distillation, and paleosol carbonates reflect the decrease in δ¹⁸O values of meteoric water at higher elevations (e.g., Garzione et al., 2006; Quade et al., 2007; Hoke et al., 2009). While this effect is also observed in the orographic rain shadow of a mountain range, other effects such as aridification resulting from the development of orographic barriers may cause meteoric waters to be ¹⁸O enriched due to evaporation. This scenario is precisely that expected for the Neiva Basin: although meteoric waters should be depleted after their westward transit over the Garzón Massif, they should also be increasingly subject to evaporation as the effect of the rain shadow intensifies and blocks further moisture from entering the basin. Thus, to properly interpret the oxygen isotope data in light of changes in the amount of precipitation, we use the weathering index of the individual paleosols hosting the analyzed carbonate nodules to estimate the effect of increasing aridity associated with surface uplift.

The CIA-K calibration has a standard error of ±146 mm/yr (Sheldon et al., 2002), and the CALMAG calibration has a standard error of ±108 mm/yr in the calibration data set (Nordt and Driese, 2010). Although the CIA-K index has been used to reconstruct precipitation in a variety of soil types from temperate climates (e.g., Sheldon et al., 2002; Hamer et al., 2007), there are scant data from Vertisols and Inceptisols, which are the predominant soil types in the Neiva Basin (Guerrero, 1993). We therefore also apply the Vertisol-specific CALMAG calibration in order to develop a model that is more representative of the Neiva Basin, and to compare with the more thoroughly studied CIA-K index.

We collected paired soil carbonate and bulk paleosol samples from horizons where soil carbonate nodules were present. Pedogenic carbonate nodules were processed then analyzed for δ¹³C and δ¹⁸O values. Nodules were first sawed in half and inspected for signs of alteration and recrystallized spar; 3–4 individual nodules showing limited alteration and recrystallization were selected from each horizon. The nodules were then drilled with an abrasive microtool, and 150–550 µg of powder from each nodule was placed into septum-capped Exetainer vials. The vials were flushed with ultrahigh-purity helium and reacted with 0.1 mL of phosphoric acid at 50 °C for 2 h. Carbon dioxide gas in the headspace of each vial was then measured using continuous flow isotope ratio mass spectrometry in the stable isotope laboratory at the University of Texas at Austin. An internal laboratory standard of powdered marble was run between every seven unknowns to correct for instrument drift, and NBS-18 and NBS-19 standards were analyzed at the beginning and end of each session to tie the unknowns to the Vienna Peedee belemnite (VPDB) scale. Measured values of the laboratory standard show a standard deviation of 0.05‰ in δ¹⁸O, and 0.1‰ in δ¹³C. The δ¹³C and δ¹⁸O values of the paleosol carbonate nodules are reported in Supplemental Table 5⁷.

Bulk paleosol samples were crushed with a hammer and homogenized into a fine powder using a mortar and pestle; all tools were cleaned thoroughly with deionized water between samples. Several grams of powder from each sample were fused into a glassy disc using Li₂B₄O₇, and were analyzed for their bulk elemental abundance using wavelength dispersive X-ray fluorescence (XRF) spectrometry at the Michigan State University XRF Lab. Measurement accuracy was determined by comparison with two laboratory standards, and XRF bulk geochemical results are reported in Supplemental Table 6⁸.

The oxygen and carbon isotope compositions for 28 sets of carbonate nodules were measured; bulk elemental abundances of paleosol samples from the same interval were measured for 24 of those samples. Average δ¹⁸O values for the carbonate nodules were remarkably consistent throughout the section, with an average value of –8.8 ‰ (VPDB) and a standard deviation of 0.7‰. The minimum measured δ¹⁸O value was –9.7‰ near the base of the La Victoria Formation, and the maximum value was –7.1‰ (Fig. 7; Table 4). The upper half of the Villavieja section displays a systematic increase in δ¹⁸O values by ∼2‰, consistent with long-term global cooling during the late Miocene (Zachos et al., 2001).

CIA-K and CALMAG indices were calculated from the measured elemental abundances in the bulk paleosol samples, and MAP was calculated using empirical calibrations developed from reference data sets (Equations 1 and 2; Sheldon et al., 2002; Nordt and Driese, 2010). The CIA-K and CALMAG reconstructions follow the same trend, although the CALMAG estimate tends to show 100–200 mm/yr less precipitation than the CIA-K estimate (Fig. 7; Table 5). Both show a pronounced excursion toward drier values in intermediate levels of the La Victoria Formation, with a gradual upsection increase in precipitation through the Villavieja Formation.

The relatively small changes in the isotopic composition of pedogenic carbonate nodules combined with the slight increase in precipitation reconstructed by the bulk elemental abundances in the host paleosols suggest that the modern rain shadow was not present during ca. 14–12 Ma accumulation of the Honda Group. The lack of a large orographic barrier is consistent with provenance data from the La Venta area in suggesting that the earliest faulting in the east did not occur until a period closer to the La Victoria–Villavieja transition, during the latest middle Miocene. Therefore, initial exhumation and probable surface uplift of the Garzón Massif (southernmost Eastern Cordillera) was only just under way during deposition of the uppermost Villavieja Formation, ca. 12–11 Ma (Fig. 8).

New results bearing on sediment provenance and basement exhumation in southern Colombia provide an important context for the growth of extensive Andean barriers and the evolution of the Magdalena, Orinoco, and Amazon river systems (Figs. 1 and 2). Topographic growth of major basement blocks in the northern Andes and adjacent foreland has diverted the Orinoco River eastward, revised the northern limit of the Amazon drainage, and funneled the north-flowing Magdalena River along the narrow hinterland segment between the Central Cordillera and Eastern Cordillera of Colombia. Basement uplift of the Garzón Massif, which defines the southernmost Eastern Cordillera in the Colombian Andes (Fig. 2), forms a critical element of this history, as it divided the Andean foreland basin from the intermontane Magdalena Valley (Van Houten and Travis, 1968; Gómez et al., 2005; Horton, 2012; Silva et al., 2013; Caballero et al., 2013) and separated the Magdalena River from the Orinoco and Amazon Rivers (Fig. 8). This topographic development also induced an orographic rain shadow, which is expressed in the modern semiarid Tatacoa Desert of the Upper Magdalena Valley. The expected signatures of this uplift event include (1) rapid exhumation of crystalline basement rocks composing the Garzón Massif and, in the Neiva Basin (Upper Magdalena Valley), (2) shifts in depositional pathways related to topographic diversion of sediment dispersal systems, (3) stratigraphic appearance of distinctive erosional products from the Garzón Massif, and (4) postdepositional processes reflective of increased aridity.

AFT data (Fig. 6) help constrain the Neogene exhumation history of the Garzón Massif. Our results suggest that the ca. 6–3 Ma phase of shortening-related basement exhumation and surface uplift (Fig. 8C) was succeeded by reorganization and regional partitioning of strain into focused shortening along the eastern margin of the Neiva Basin and right-lateral strike-slip motion along the Algeciras fault (Fig. 8D). Such a shift in deformation style is suggested by a significant decrease in modeled exhumation rates after 4–3 Ma, as observed for all samples on the uplifted eastern side of the fault. This reconstruction suggests that exhumation and generation of considerable structural relief in the central segment of the Garzón Massif was most rapid during latest Miocene–early Pliocene shortening, followed by slower exhumation when transpressional motion became well established after 4–3 Ma (Fig. 8). For the western side of the Algeciras fault, continuous exhumation since ca. 4.5 Ma, modeled at a more moderate cooling rate, likely represents exhumation along a transpressional structure that propagated westward into the Neiva Basin.

Sediment provenance results from the Upper Magdalena Valley provide insight into the early exhumational record of the Garzón Massif within the Eastern Cordillera, prior to the latest Miocene–Quaternary history delineated by the AFT results. From the Neogene compositional and paleocurrent data sets (Fig. 5), the earliest evidence for uplift-induced exhumation east of the Neiva Basin is expressed in the ca. 12.5 Ma influx of nonvolcanic lithic fragments, including chert, likely recycled from Cretaceous–Paleogene sedimentary units (Figs. 8A, 8B). Although this sediment source is replaced shortly thereafter, it may represent early relief generation along faults within the Garzón Massif–Neiva Basin transition and associated establishment of a drainage barrier related to basement uplift. A more pronounced provenance shift is detected in the uppermost Honda Group (ca. 12.5–11.6 Ma) of the northern Neiva Basin (La Venta site; Fig. 2B), where possible Cenozoic zircons indicative of a western source are replaced by the first appearance of Garzón-derived grains. In the southern Neiva Basin (Gigante site; Fig. 2C), a similar shift is observed in the overlying Gigante Formation, in which a significant fraction of clastic material appears to be derived principally from the magmatic arc to the west until ca. 6.4 Ma, the age of the uppermost sample, which is overwhelmingly dominated by detritus shed from the Garzón Massif. This north-south age discrepancy in the first appearance of Garzón-derived material may be related to the overall north-to-northeast plunge of the basement-involved Garzón Massif structures, with (1) spatial variations in the magnitude of deformation, (2) a spatial progression in the timing of uplift-induced exhumation in the massif, and/or (3) the greater proximity of the Central Cordillera to the southern (Gigante) site. These options suggest a paleogeographic configuration in which large-scale exhumation of deeper metamorphic basement within the central segment of the Garzón Massif was unlikely to have commenced until ca. 6.4 Ma. However, there is evidence for exhumation of part of the Garzón Massif (possibly focused to the south of the study region) from 12.5 Ma onward, as evidenced by periodic appearances of Garzón-derived material, a gradual transition from east-directed to north-directed paleocurrents, an increased sedimentation rate, and a change in sedimentary architecture (Fig. 5).

Paleoclimatic data spanning the 13.8–11.6 Ma deposition of the Honda Group suggest minimal evidence for aridification that could be linked to surface uplift of the Eastern Cordillera. Oxygen isotopic compositions of soil carbonates in the Neiva Basin display little change (Fig. 7), and weathering-index–based estimates of mean annual precipitation suggest that the basin became wetter, rather than drier, as would be expected with the development of an orographic rain shadow produced by surface uplift in the east. Therefore, although the sedimentary provenance data suggest possible early growth of eastern sediment sources in the Garzón Massif, the paleoclimatic data suggest that it is unlikely that these middle Miocene sources constituted a significant topographic barrier.

These findings are in agreement with biological data, which suggest that Magdalena Valley flora and fauna were becoming distinct from the Amazon Basin by ca. 10 Ma (Hoorn et al., 2010; Ochoa et al., 2012; Aguilera et al., 2013). The possible divergence of the Magdalena River from the proto–Orinoco River and proto–Amazon River systems at this time lends support to the hypothesis that the modern South American drainage system developed over a protracted period beginning in the late Miocene (Figueiredo et al., 2009; Hoorn et al., 2010; Shephard et al., 2010). In order to reconcile the results of our sediment provenance, paleoclimatic, and thermochronometric studies, we propose that initial segregation of these three continental-scale river systems commenced with earliest structural development of the Garzón Massif in southern Colombia ca. 12–10 Ma, but substantial topography in the southernmost Eastern Cordillera was not generated until the latest Miocene–Pliocene (ca. 6–3 Ma) phase of rapid exhumation of Garzón basement.

From the detrital zircon U-Pb, sandstone petrographic, and AFT exhumational results, we interpret a three-stage history for the Garzón Massif in the Eastern Cordillera of southern Colombia (Fig. 8). During the earliest stage, from ca. 12.5 to 6.4 Ma, shortening was accommodated along a principally dip-slip segment of the Algeciras fault system east of the Neiva Basin, which intermittently blocked eastward drainage, but did not form a major topographic barrier at these latitudes (1°–4°N). The intermediate stage commenced during deposition of the basal Garzón Member of the Gigante Formation, with the arrival of coarse gravel bearing metamorphic clasts and zircons with U-Pb ages consistent with a Garzón Massif source. From ca. 6.4 to 4–3 Ma, this stage involved extremely rapid exhumation as upper crustal shortening was accommodated by continued thrust displacement within the Algeciras fault system. During the final stage, from 4 to 3 Ma onward, exhumation of the Garzón Massif slowed considerably as significant transpressional deformation was accommodated by strike-slip motion along the main Algeciras fault strand and dip-slip motion along thrust faults bounding the easternmost Neiva Basin. This history of uplift for the Garzón Massif is similar to the timing of sharp surface uplift for the Bogotá plateau along strike to the northeast (Hooghiemstra et al., 2006; Mora et al., 2008; Anderson et al., 2015), suggesting that the most rapid phase of regional uplift occurred between 6 and 3 Ma. The synchroneity of uplift for these two study localities over a >250 km distance along strike within the Eastern Cordillera is suggestive of a shared geodynamic driver.

Our results suggest that basement uplift of the Garzón Massif may be indicative of a significant geodynamic shift in the northern Andes. Modern convergence between the Nazca and South American plates is currently oblique along the Colombian margin, with regional tectonic escape of the northern Andean block to the northeast. Various mechanisms for the initiation of tectonic escape have been proposed. These include the collision of the Panama arc (Farris et al., 2011; Montes et al., 2012), the upper plate response to subduction of the buoyant Carnegie Ridge (Gutscher et al., 1999; Spikings et al., 2001; Egbue and Kellogg, 2010; Egbue et al., 2014), or a shift in convergence direction or strain partitioning along the obliquely converging Nazca–South American margin (Acosta et al., 2004; Jiménez et al., 2014). Our interpretation of a Pliocene transition to principally strike-slip motion along the Algeciras fault is most consistent with a shift to more oblique convergence, greater strain partitioning, and/or the potential influence of the geodynamic effects of the subducting Carnegie Ridge.

In addition, due to the shared basement configuration and close proximity of the Garzón Massif to the foreland structures forming the Macarena high and broader Vaupes arch (e.g., Casero et al., 1997; Acosta et al., 2004; Velandia et al., 2005), we speculate that their uplift may share a driving geodynamic mechanism, consistent with previous work suggesting a similar timing based on structural and stratigraphic constraints (Wesselingh et al., 2006; Mora et al., 2010a). These foreland basement arches form the barrier between the modern-day Orinoco and Amazon river systems (Figs. 1 and 2), and may play a key role in the initiation of a transcontinental Amazon drainage. This underscores potential links among plate boundary processes, topographic uplift, and regional drainage patterns in affecting the biological dynamics of South America.

We characterize Neogene topographic development of the Garzón Massif (southernmost Eastern Cordillera) of the northern Andes by analyzing the sediment provenance of clastic basin fill, AFT thermochronometry of basement rocks, and paleosol-based indicators of climatic variations in the Upper Magdalena Valley (Neiva Basin) of southern Colombia. Our reconstructions of middle Miocene–Quaternary tectonics, exhumation, and paleodrainage (Fig. 8) emphasize the following points.

1. Sediment sources for the Neiva Basin alternated between principally magmatic-arc rocks of the Central Cordillera and probable Eastern Cordillera sources from the sedimentary cover and basement rocks of the Garzón Massif. Most of the middle to upper Miocene basin fill was derived from western, igneous-dominated sources in the Central Cordillera, with an important shift to eastern sources and pronounced contributions from Garzón basement sources by at least 6.4 Ma. However, earliest exhumation of the Eastern Cordillera at these latitudes (1°–4°N) along upper crustal thrust structures defining the eastern margin of the Neiva Basin may be recorded by a focused ca. 12.5 Ma influx of nonvolcanic lithic fragments suggestive of initial stripping of Cretaceous sediment cover rocks capping the Garzón Massif.

2. Minimal paleoclimatic changes attributable to topographic uplift are detected in the 13.8–11.6 Ma interval. Evidence for the generation of an orographically induced rain shadow is lacking, in that reconstructed precipitation is shown to increase, rather than decrease, over this time interval. In combination with the shifting sediment sources inferred from provenance analyses, these data suggest the Garzón Massif did not form a substantial elevated topographic barrier during the early stages of basement uplift from 12.5 Ma until 6.4 Ma.

3. Inverse modeling results of thermal histories defined by new AFT data suggest profoundly different exhumation histories for opposing flanks of the composite Algeciras fault system. This pattern may be attributed to a regional Pliocene reorganization from principally dip-slip thrust motion to strike-slip motion along the Algeciras fault. Based on the minimum exhumational ages provided by the AFT data, as well as the thermal history modeling results, we suggest that thrust-induced rapid exhumation was focused between ca. 6.4 Ma and 4–3 Ma. Thereafter, the Algeciras fault likely accommodated strike-slip displacement, consistent with estimates for right-lateral motion since at least 2 Ma.

4. This sequence of events is consistent with the timing of major uplift and exhumation elsewhere in the Eastern Cordillera of Colombia, suggesting a shared geodynamic driving mechanism over a large segment of the northernmost Andes. We suggest that basement uplift in the Eastern Cordillera and northern Andean foreland basin forced the late Miocene isolation of the Magdalena, Orinoco, and Amazon drainage systems. By inference, the latest Miocene–Pliocene (6–3 Ma) phase of rapid and widespread exhumation across the Eastern Cordillera is hypothesized to have provided a huge pulse of clastic material into these continental-scale river systems.

Funding was provided by the National Science Foundation (grants EAR-1019857 and EAR-1338694), Ecopetrol–Instituto Colombiano del Petróleo (ICP), and student research grants from the Geological Society of America. We thank Juliana Barrientos-Mesa and Miguel Corcione for field assistance, and James Coogan, Mauricio Ibañez, Alejandro Mora, Mauricio Parra, Alexis Rosero, Timothy Shanahan, Lorena Suarez, and Gabriel Veloza for beneficial discussions. We thank Apatite to Zircon, Inc., for providing data at reduced academic rates. The manuscript was improved through the constructive reviews of Matías Ghiglione and Annia Fayon.