Application of detrital zircon U-Pb geochronology to surface and subsurface correlations of provenance, paleodrainage, and tectonics of the Middle Magdalena Valley Basin of Colombia

Although advances in detrital zircon U-Pb geochronology have fueled a resurgence in sediment provenance studies, subsurface and intrabasinal applications remain limited. Rapid data acquisition through laser-ablation–inductively coupled plasma–mass spectrometry (LA-ICP-MS) has propelled U-Pb analyses of sand-sized zircon grains to the forefront of provenance studies seeking to discriminate among potential source regions (e.g., Dickinson and Gehrels, 2008; Nie et al., 2012; Gehrels, 2014). Further applications for sedimentary basin analysis include assessments of: (1) the relative volumetric contributions from known sources in modern watersheds (Saylor et al., 2013); (2) the complex mixing and downstream dilution of source signals within modern rivers (Amidon et al., 2005; Link et al., 2005; Zhang et al., 2012; He et al., 2014); (3) large-scale paleodrainage patterns, including the onset and evolution of major paleorivers (Davis et al., 2010; Dickinson et al., 2012; Mackey et al., 2012; Blum and Pecha, 2014); and (4) absolute ages of stratigraphic units (Fildani et al., 2003; DeCelles et al., 2007; Dickinson and Gehrels, 2009; Horton et al., 2015). Although accurate ages for syndepositional volcanic zircons clearly improve chronostratigraphic correlations, U-Pb geochronology may provide a viable correlation tool on the basis of comparable age distributions (e.g., Rainbird et al., 2007; Lawton et al., 2010; Beranek et al., 2013; Lewis and Sircombe, 2013), an approach that could considerably enhance subsurface investigations.

Despite the wide range of applications, complexities over multiple scales highlight various difficulties in pinpointing sediment source regions, reconstructing paleodrainage patterns, and correlating stratigraphic units. Whereas most basin-scale studies focus on temporal provenance shifts registered within key stratigraphic sections, uncertainties persist over the effectiveness of signal transmission from source to sink. These issues can include (1) uneven contributions from source areas, (2) storage, buffering, or recycling within drainage systems, and (3) potential sequestration due to intrabasinal variations in hydrodynamics or depositional environments (e.g., Métivier and Gaudemer, 1999; DeGraaff-Surpless et al., 2003; Allen, 2008; Lawrence et al., 2011; Saylor et al., 2013; Latrubesse, 2015). In river-dominated basins, many of these potential problems can be minimized by sampling a range of depositional subenvironments over a sufficiently large region (e.g., Potter, 1978; Ingersoll, 1990; Ingersoll et al., 1993). Nevertheless, few investigations have explored both the temporal and spatial variations in U-Pb provenance signatures for both surface and subsurface deposits within a single sedimentary basin.

In this study, we explore an ∼150 m.y. provenance history for surface and subsurface samples from a single elongate basin exhibiting important along-strike, cross-strike, and proximal-to-distal variations. The Middle Magdalena Valley Basin of Colombia is a narrow intermontane basin spanning an area of 30,000 km² within the northernmost Andes (Fig. 1). Depositional systems in the north-trending basin are dictated locally by transverse sources derived from the two major flanking mountain ranges, the Central Cordillera and Eastern Cordillera, and regionally by the longitudinal Magdalena River, which today flows ∼1500 km along the basin axis to the Caribbean coast. The Middle Magdalena Valley Basin is a prolific hydrocarbon basin, with ∼100 yr of exploration efforts. Nevertheless, despite hundreds of wells, problems of stratigraphic correlation are highlighted by complex spatial variations in Mesozoic–Cenozoic depositional systems (Morales, 1958; Van Houten and Travis, 1968; Van Houten, 1976; Butler and Schamel, 1988; Schamel, 1991; Cooper et al., 1995; Ramon and Rosero, 2006; Naranjo-Vesga et al., 2013).

These complexities are reflected in competing models for paleodrainage and basin evolution in the Middle Magdalena Valley Basin. The Paleogene Middle Magdalena Valley Basin has been envisioned as the locus of either northward axial paleodrainages or eastward transverse paleorivers (Nie et al., 2010, 2012; Moreno et al., 2011; Saylor et al., 2011; Caballero et al., 2013a, 2013b; Silva et al., 2013; Bayona et al., 2013; Reyes-Harker et al., 2015). Alternative scenarios propose a north-flowing proto–Magdalena River developing in middle to late Miocene time (Hoorn et al., 1995, 2010). Evolution of the Middle Magdalena Valley Basin drainage systems and establishment of a through-going Magdalena River also directly affected sediment accumulation and offshore hydrocarbon prospectivity of the Magdalena delta in the Caribbean Sea.

Further uncertainties center on potential basin responses to the important regional phases of Mesozoic extension and Cenozoic shortening in the northernmost Andes, as well as critical transitions involving focused basement uplift and punctuated magmatism (e.g., Colletta et al., 1990; Dengo and Covey, 1993; Gómez et al., 2003, 2005a, 2005b; Horton et al., 2010a; Bayona et al., 2012; Parra et al., 2012; Nie et al., 2012; Saylor et al., 2012a, 2012b; Caballero et al., 2013a, 2013b; Reyes-Harker et al., 2015). Here, we present U-Pb ages of detrital zircons for 21 subsurface samples of Mesozoic–Cenozoic basin fill, and we integrate these data with results from 83 surface outcrop samples and seven modern river samples to help correlate the provenance and tectonic histories of surface and subsurface basin fill.

The Middle Magdalena Valley Basin is a long-lived sedimentary basin that has recorded uplift and exhumation of the major ranges and block uplifts of the northernmost Andes at 4°N–7°N (Fig. 1). Modern sediment is delivered to the narrow intermontane Middle Magdalena Valley Basin by transverse rivers and alluvial fans from the flanking Central Cordillera and Eastern Cordillera, with the Magdalena River flowing longitudinally northward along the basin axis into the Caribbean Sea. Additional sediment sources include the isolated San Lucas range and Santander Massif, which form block uplifts at the northern terminations of the Central and Eastern Cordilleras, respectively.

Although the 5–10-km-thick clastic fill of the elongate Middle Magdalena Valley Basin now defines a 20–100-km-wide by 500-km-long swath, the Mesozoic–Cenozoic succession contains a composite history involving a more-expansive early Cenozoic foreland basin (Gómez et al., 2003, 2005b; Caballero et al., 2010, 2013a, 2013b; Moreno et al., 2011) and isolated Mesozoic extensional subbasins (Cooper et al., 1995; Sarmiento-Rojas et al., 2006; Kammer and Sánchez, 2006; Mora et al., 2013; Tesón et al., 2013). To the west, the north-trending Central Cordillera is mostly composed of Mesozoic–early Cenozoic igneous rocks of the Andean magmatic arc unroofed during Mesozoic extension and Cenozoic shortening. In contrast, to the east, the north-northeast–trending Eastern Cordillera consists of the Andean retroarc fold-thrust belt, dominated by Cretaceous sedimentary rocks that, along with localized basement uplifts, were erosionally recycled as the Middle Magdalena Valley Basin transitioned from a broad foreland basin to an intermontane hinterland basin (Caballero et al., 2010, 2013a; Moreno et al., 2011; Saylor et al., 2011; Horton, 2012; Parra et al., 2012; Sánchez et al., 2012; Silva et al., 2013; Moreno et al., 2013; Tesón et al., 2013).

The Jurassic to Neogene succession in the Middle Magdalena Valley Basin (Fig. 2) recorded relatively continuous accumulation of 5–10 km of clastic sediment in marine and nonmarine settings. Age control is provided by marine fossil assemblages for the lower succession (Etayo-Serna et al., 1983; Etayo-Serna, 1985) and a combination of pollen, freshwater mollusks, and volcanic ash geochronology for the upper succession (Gómez et al., 2003, 2005b; Jaramillo et al., 2011). Initial basin accumulation during Middle–Late Jurassic (Girón Formation) to Early Cretaceous time (Los Santos, Rosa Blanca, and Paja Formations) involved proximal deposition of clastic facies in alluvial-fan, fluvial, and shallow-marine settings. Subsequent shallow- to deep-marine deposition of mixed sand, mud, and limited carbonate persisted throughout late Early Cretaceous to early Late Cretaceous time (Tablazo, Simití, and Simijaca Formations). The latest Cretaceous–Paleocene marked a shift to sandier sedimentation, with significant progradation of fluviodeltaic systems (Umir, Lisama, Seca, and Hoyón Formations). The Eocene through Pliocene history was governed by fluvial deposition involving alternating phases of channel and overbank accumulation, as well as alluvial-fan and possible fluvial megafan sedimentation (La Paz, Cantagallo, San Juan de Rio Seco, Esmeraldas, Mugrosa, Colorado, Real, and Mesa Formations).

Mesozoic–Cenozoic stratigraphic units of the Middle Magdalena Valley Basin show complex spatial variations, leading to ambiguous stratigraphic correlations (Morales, 1958; Sarmiento Rojas, 2001; Sarmiento-Rojas et al., 2006). There are no field localities with complete exposures of the Jurassic through Neogene succession, and many units are defined only in the subsurface. The lack of uninterrupted exposures and the subsurface presence of intrabasinal structural highs (e.g., the subsurface Infantas, La Cira, and Cáchira highs) and associated unconformities (Morales, 1958; Gómez et al., 2003, 2005b; Caballero et al., 2013a) have precluded delineation of systematic cross-strike (east-west) trends in lithofacies and associated depositional systems. Along-strike (north-south) variations in the aforementioned stratigraphic units (Fig. 2) reveal more straightforward lithostratigraphic trends along the basin axis, likely due to more-continuous north-trending structures within the regional structural framework.

For the purpose of this study, we systematically consider results from an ∼300 × 50 km swath of the Middle Magdalena Valley Basin (Fig. 1), including three wells from the western to far northeastern basin (Cocuyo, Guane, and Cagui wells), three surface sites from the eastern basin (Nuevo Mundo syncline west limb, Nuevo Mundo syncline east limb, and La Salina footwall), and three surface sites from the central-southern basin (Opón syncline, Rio Ermitaño syncline, and Guaduas syncline).

Detrital zircon U-Pb geochronological results for samples of Mesozoic-Cenozoic sandstones and modern river sands (Supplemental Tables 1¹ and 2²) were acquired through LA-ICP-MS, principally at the University of Arizona LaserChron Center following procedures established for the facility (Gehrels, 2000, 2014; Gehrels et al., 2008). New U-Pb results, consisting of a single U-Pb date for each analyzed zircon grain, are reported for 21 subsurface samples acquired from three wells. Random selections of >100 inclusion-free zircon grains were analyzed for each sample, along with known standards to correct for inter- and intra-element fractionation (Gehrels et al., 2008), resulting in a 1–2% (2σ) age uncertainty for each analysis. Common Pb corrections were employed using measured ²⁰⁴Pb and an assumed initial Pb composition (Stacey and Kramers, 1975). Errors in determining ²⁰⁶Pb/²³⁸U, ²⁰⁶Pb/²⁰⁷Pb, and ²⁰⁶Pb/²⁰⁴Pb ratios yielded total measurement errors of ∼1–2% (2σ) for each preferred age, which is reported as either the ²⁰⁶Pb/²³⁸U age (grains younger than 900 Ma) or ²⁰⁶Pb/²⁰⁷Pb age (grains older than 900 Ma). Age discordance filters were employed following previous studies in the region, with all single-grain analyses exhibiting >10% uncertainty and all ²⁰⁶Pb/²³⁸U ages displaying >20% discordance or >5% reverse discordance omitted from consideration. Given the potential for minor discordance, all provenance and chronostratigraphic interpretations were developed on the basis of age populations defined by results for three or more zircon grains. U-Pb data are plotted as age probability distribution functions and age histograms. Significant age populations are generally reported as age ranges rather than individual age peaks (i.e., the probability maxima of distribution functions), in order to emphasize and correlate age trends across multiple samples.

Previously published U-Pb data are incorporated into our regional analysis, including results for 83 samples from six exposed stratigraphic sections and seven samples of modern river sands (Nie et al., 2010, 2012; Ibañez-Mejia et al., 2011; Bayona et al., 2012, 2013; Caballero et al., 2013a, 2013b). These data were acquired by LA-ICP-MS in the same or similar laboratory settings (in laboratories at the University of Arizona, University of Texas at Austin, and Washington State University) using comparable methods, as detailed in the original studies (Nie et al., 2010, 2012; Ibañez-Mejia et al., 2011; Bayona et al., 2012, 2013; Caballero et al., 2013a, 2013b). To enable reasonable comparisons across these recent data sets from different researchers, we employed the aforementioned filters for age errors and age discordance across all sample results.

Detrital zircon U-Pb age distributions for modern river sands in the northernmost Andes (Nie et al., 2012; Bande et al., 2012; Saylor et al., 2013) demonstrate the distinctive age signatures for several competing sediment sources: principally two contiguous ranges, the Central Cordillera and Eastern Cordillera, but also two localized block uplifts, the Santander Massif and San Lucas range (Fig. 1).

(1) Along the western flank of the Middle Magdalena Valley Basin, the Central Cordillera forms a contiguous north-trending barrier constructed of upper Paleozoic to lower Cenozoic igneous and metamorphic rocks (Fig. 1). Modern river sands collected from the mouths of the two largest drainage networks, the Rio Nare and Rio Nechi (Fig. 1), are both governed by zircon U-Pb age peaks of 95–70 Ma (Figs. 3A and 3B), consistent with considerable exposure of the Upper Cretaceous Antioquia (Antioqueño) batholith (88–83 Ma) over roughly half the cumulative drainage area in the Central Cordillera. Additional sources represented in the river sand U-Pb age spectra include Paleogene (65–50 Ma) and Permian–Triassic (300–200 Ma) granites, and Paleozoic metasedimentary and meta-igneous rocks of the Cajamarca complex, defined chiefly by 300–250 Ma metamorphic ages with minor Proterozoic to early Paleozoic (roughly 1000–400 Ma) inheritance (McCourt et al., 1984; Aspden et al., 1987; Aleman and Ramos, 2000; Cordani et al., 2005; Ordóñez-Carmona et al., 2006; Vinasco et al., 2006; Cardona et al., 2010; Villagómez et al., 2011).

(2) At the northern termination of the Central Cordillera, the San Lucas range (Fig. 1) forms an isolated block uplift composed of Jurassic volcaniclastic rocks capping Triassic–Jurassic granodiorite and Grenvillian-age metamorphic basement developed on a Mesoproterozoic protolith (Gómez Tapias et al., 2007; Clavijo et al., 2008; Cuadros et al., 2014; Blanco-Quintero et al., 2014). U-Pb results from modern river sands collected at the mouth of the Rio Santo Domingo along the northwestern Middle Magdalena Valley Basin margin exhibit a dominant Permian–Triassic (300–200 Ma) population within a broad population of Carboniferous–Jurassic (350–150 Ma) ages, along with a minor early Paleozoic (500–400 Ma) signal (Fig. 3C).

(3) In the Santander Massif, along the northern margin of the Eastern Cordillera, two modern drainage networks encompass Triassic–Jurassic granites and subordinate Ordovician–Silurian metasedimentary rocks and Precambrian basement (Fig. 1). These distributions are faithfully represented by U-Pb age spectra for modern river sands from the Rio Umbala and Rio Manco (Figs. 3D and 3E). The rivers are governed by 300–150 Ma ages, with clustering of 210–190 Ma ages, consistent with Late Triassic–Early Jurassic emplacement ages for the Santa Barbara granodiorite, Pescadero granite, and associated intrusions (Goldsmith et al., 1971; Irving, 1975; Forero-Suarez, 1990; Dörr et al., 1995; Mantilla Figueroa et al., 2013). Populations of 500–400 Ma and 1200–550 Ma grains record input from less-extensive igneous and metamorphic rocks of Ordovician–Silurian and Proterozoic age, respectively.

(4) To the east, the Middle Magdalena Valley Basin is flanked by the broad Eastern Cordillera, a north-northeast–trending fold-thrust system composed mainly of Cretaceous clastic deposits (Fig. 1) originally derived from source areas in the South American craton (Guyana Shield) to the east (Horton et al., 2010b). U-Pb results from the Rio Cravo Sur and Rio Cusiana show primarily Proterozoic grains (1800–900 Ma), with peak ages of 1100–900, 1600–1500, 1850–1750, and 1250–1200 Ma, and secondary early Paleozoic ages (500–400 Ma; Figs. 3F and 3G), which are clear indicators of recycling from exhumed Cretaceous sedimentary rocks (Bande et al., 2012; Saylor et al., 2013).

Although there is modest overlap in ages among these different tectonic provinces, several essential distinctions are recognized in the U-Pb age distributions for modern rivers (Fig. 3H). First, the Eastern Cordillera is unambiguously dominated by 1800–900 Ma zircons, which are diagnostic of voluminous Cretaceous clastic deposits originally derived from Precambrian cratonic sources to the east in the Guyana Shield and possible subsurface basement sources of the Llanos Basin (Figs. 3F, 3G, and 3H). Second, the Central Cordillera is uniquely characterized by Late Cretaceous–early Eocene (100–50 Ma) zircons, with a limited population of Permian–Triassic (300–200 Ma) zircons (Figs. 3A, 3B, and 3H). Third, intrabasinal Middle Magdalena Valley and marginal Middle Magdalena Valley uplifts such as the San Lucas range and Santander Massif (and by inference, additional subsurface structural highs such as the Infantas, La Cira, and Cáchira highs) share the 300–200 Ma signature but are differentiated by Jurassic (200–150 Ma), Carboniferous (350–300 Ma), and early Paleozoic (500–400 Ma) populations (Figs. 3C–3E and 3H).

Detrital zircon U-Pb ages for 21 sandstone samples from three wells define several key populations and provenance shifts for subsurface Jurassic through Neogene clastic deposits of the Middle Magdalena Valley Basin (Supplemental Tables 1 and 2; Figs. 4 and 5). Whereas the Guane and Cocuyo wells are closer to the western margin, the Cagui well is situated near the far northeastern basin margin (Figs. 1 and 2).

The oldest stratigraphic units of the Middle Magdalena Valley Basin, which are not exposed at the surface, consist of coarse-grained deposits of Middle–Late Jurassic to earliest Cretaceous age. At the base of the succession, U-Pb analyses of two subsurface samples of the coarse-grained nonmarine Girón Formation yield similarly unimodal age populations composed of Jurassic grains dated at 200–165 Ma (Figs. 4 and 5). Over 70% of analyzed zircon grains fall in the 190–170 Ma age range, with strong Early to Middle Jurassic peaks at 185–175 Ma in the southwest (Guane well) and 180–170 Ma in the northeast (Cagui well). Results from the three youngest grains define a population at ca. 166 Ma, providing a maximum depositional (stratigraphic) age constraint for the subsurface Girón Formation. The only other age signature involves a limited occurrence of Precambrian (principally 1800–900 Ma) grains, including an 1100–1000 Ma signal.

For the relatively finer-grained marine deposits of the Lower Cretaceous succession, U-Pb ages for three samples from the Los Santos, Paja, and Rosa Blanca Formations (Guane and Cagui wells) show the continuation of the prominent Jurassic (200–150 Ma) age signature, with an Early Jurassic concentration at 190–175 Ma (Figs. 4 and 5). The establishment of a significant population of Precambrian (1800–900 Ma) zircons, composing roughly half of the total analyzed grains, is expressed in subpopulations of late Mesoproterozoic–early Neoproterozoic (1300–900 Ma), early Mesoproterozoic (1600–1500 Ma), and Paleoproterozoic (1800–1700 Ma) age. The U-Pb results also reveal the first appearance of several minor signals in the Middle Magdalena Valley Basin, including late Paleozoic (300–250 Ma) ages in the northeast (Cagui well) and early Paleozoic (400–500 Ma) ages in the southwest (Guane well).

U-Pb age spectra for two sandstone samples from the uppermost Cretaceous Umir Formation and Eocene Cantagallo Formation indicate an up-section rise in the late Paleozoic (300–250 Ma) population and, within the major Jurassic (200–150 Ma) signal, a shift toward younger Middle to Late Jurassic (170–150 Ma) ages (Figs. 4 and 5). The continued broad distribution of Precambrian ages (principally 1400–900 Ma), which accounts for roughly half of analyzed grains, includes a substantial Mesoproterozoic component with notable 1400–1300 Ma and 1250–1150 Ma peaks.

Three samples of the uppermost Eocene–lowermost Oligocene Esmeraldas Formation, collected across the three wells, point to spatial and temporal variations among several detrital age populations within the Middle Magdalena Valley Basin. In the southwest (Guane and Cocuyo wells), an up-section reduction of the Jurassic (200–150 Ma) signal is accompanied by increased late Paleozoic (300–250 Ma) and continued Precambrian (1800–900 Ma) populations, including Mesoproterozoic subpopulations matching those of subjacent units (Fig. 4). In the northeast, the Cagui well displays the persistence of a strong Jurassic (200–150 Ma) signal and a broader distribution of Precambrian (1800–900 Ma) ages, including a sizeable Paleoproterozoic–early Mesoproterozoic (1800–1400 Ma) population (Fig. 5). In addition, early Paleozoic (500–400 Ma) and Eocene (40–30 Ma) peaks occur at the expense of the late Paleozoic (300–250 Ma) population.

U-Pb results from four samples of the Oligocene Mugrosa Formation signify a considerable reduction in the formerly dominant Jurassic (200–150 Ma) signal across the three wells, accounting for only ∼5% of analyzed zircon grains (Figs. 4 and 5). This drop is accompanied by the sustained presence of a regionally significant late Paleozoic (300–250 Ma) population along with an increased Precambrian (1800–900 Ma) signal. Both margins of the Middle Magdalena Valley Basin exhibit high amounts of Precambrian detritus, constituting roughly 60–80% of total grains analyzed. For the western basin margin, the Oligocene Mugrosa Formation represents the greatest proportion of Precambrian grains for any part of the Middle Magdalena Valley succession within the Guane and Cocuyo wells (Fig. 4).

For the uppermost basin fill, U-Pb results for seven subsurface samples of the Miocene–Pliocene Colorado and Real Formations define key provenance shifts and a sharp spatial contrast across the Middle Magdalena Valley Basin. In the northeast, five samples from the Cagui well display nearly exclusively (∼85–95%) Precambrian ages, with prominent age peaks of 1700–1300 and 1100–900 Ma, along with subordinate Triassic–Jurassic (250–150 Ma) and early Paleozoic (500–400 Ma) populations (Fig. 5). In contrast, in the southwest, samples from the Guane and Cocuyo wells record much fewer (<50%) Precambrian ages, and a relative strengthening of the Jurassic (200–150 Ma) signature (Fig. 4). The stratigraphically highest sample, from the upper Miocene Real Formation in the Guane well (Fig. 4), shows not only the loss of a late Paleozoic (300–250 Ma) signal, but also the emergence of the most substantial proportion (up to 10%) of 100–0 Ma grains among all stratigraphic units along the western basin margin; these pre-Miocene (30–100 Ma) grains likely reflect contributions from Cretaceous–Paleogene igneous sources.

A survey of detrital zircon U-Pb ages for outcrop samples enables spatial comparisons and correlations among surface and subsurface deposits. Cumulative data sets are presented for the eastern, central, and southern segments of the Middle Magdalena Valley Basin (Figs. 1 and 2), including the type sections of several stratigraphic units, following the analytical and data treatment procedures of the original studies (Nie et al., 2010, 2012; Ibañez-Mejia et al., 2011; Bayona et al., 2012, 2013; Caballero et al., 2013a, 2013b; Silva et al., 2013; Reyes-Harker et al., 2015). The eastern Middle Magdalena Valley data set includes cumulative results for 40 samples from the La Salina locality (footwall of the La Salina thrust), Nuevo Mundo syncline west limb, and Nuevo Mundo syncline east limb (Fig. 6). The central-southern Middle Magdalena Valley data set consists of results from 43 samples of the Opón syncline, Rio Ermitaño syncline, and Guaduas syncline (Fig. 7).

Exposures of Mesozoic strata are extremely limited within the basin. A single mid-Cretaceous sample from the Simijaca Formation in the southern Middle Magdalena Valley Basin (Guaduas syncline) yields a U-Pb age signature (Fig. 7) governed by Precambrian ages (1800–1450, 1350–1250, and 1150–1000 Ma). This broad Proterozoic age distribution can be correlated with the aforementioned detrital zircon age spectra for subsurface Cretaceous strata, and it can be readily discriminated from the unimodal Jurassic (200–150 Ma) age signature for Middle–Upper Jurassic deposits at the base of the basin fill succession (Fig. 4).

The strong Proterozoic signal is compatible with crystalline basement sources such as the expansive Precambrian exposures of the South American craton (Guyana Shield) to the east. However, given the similarity of U-Pb age distributions, it is likely that localized highs composed of basement and Paleozoic cover strata within the Eastern Cordillera and/or Llanos Basin also provided Proterozoic detritus. Such local sources, however, were largely eliminated by late Early Cretaceous time, when marine depositional conditions characterized the Eastern Cordillera and flanking Llanos and Magdalena regions (Cooper et al., 1995; Sarmiento-Rojas et al., 2006; Mora et al., 2009).

The absence of the 200–150 Ma signal characterizing subsurface Cretaceous equivalents farther north (Fig. 4) suggests that the preferential influence of isolated uplifts such as the San Lucas range and Santander Massif in the northern Middle Magdalena Valley Basin does not persist for the central-southern Middle Magdalena Valley Basin.

Variations in U-Pb age populations are expressed in 10 samples from the Upper Cretaceous–lower Paleocene Umir, Seca, Labor-Tierna, lower Lisama, and lower Hoyón Formations. Most distinctive is the first appearance of Late Cretaceous–Paleocene (90–60 Ma) zircons, with 90–70 Ma grains delineating the central-southern Middle Magdalena Valley Basin (Fig. 7) and a small 80–60 Ma population defined in the eastern Middle Magdalena Valley Basin (Fig. 6). These signatures are not detected in correlative subsurface strata farther north. However, similar to the age distributions of older Cretaceous units (Figs. 4 and 5), Precambrian zircons (primarily 1800–1200 Ma) are present and remain dominant in the eastern Middle Magdalena Valley Basin, with subordinate populations of Paleozoic (500–400 and 300–250 Ma) ages.

Collectively, these results are consistent with continued clastic contributions from distal cratonic sources to the east, but also a localized introduction of 90–60 Ma zircons along the western margin of the southern Middle Magdalena Valley Basin (as particularly well expressed in the Guaduas syncline; Figs. 1 and 2). This western source can be linked to earliest uplift of Cretaceous–Paleocene magmatic arc rocks of the Central Cordillera, but it is restricted to the southern segments of the Middle Magdalena Valley Basin.

Widespread changes in U-Pb age spectra are expressed in a sizeable set of 34 upper Paleocene–lowermost Oligocene samples from the upper Lisama, upper Hoyón, La Paz (including Toro Shale), lower Esmeraldas, and lower San Juan de Rio Seco Formations. The most pronounced shift is represented by the regional mid-Paleocene emergence of a Late Cretaceous–Paleocene (90–60 Ma) age population (Figs. 6 and 7), with a subset of probable syndepositional zircon grains concentrated at 65–55 Ma (Figs. 6 and 7). Additional populations include a broad Precambrian (1800–900 Ma) distribution, a Jurassic (200–150 Ma) signal, and minor clusters of early Paleozoic (500–400 Ma) and Permian–Triassic (300–200 Ma) ages.

The Late Cretaceous–Paleocene population, which dominates the entire Paleogene succession of the southern Middle Magdalena Valley Basin (Fig. 7) and progressively replaces an extensive Precambrian component in the eastern Middle Magdalena Valley Basin (Fig. 6), is absent in nearly all subsurface Paleogene deposits of the northern Middle Magdalena Valley Basin (Figs. 4 and 5). These results may point to progressive northward growth of an integrated western source, with local concentrations of these 90–60 Ma grains suggesting temporally or spatially focused pulses of sediment influx from Cretaceous–Paleocene igneous rocks of the Central Cordillera. Such northward propagation of Central Cordillera uplift matches paleogeographic reconstructions (Gómez et al., 2003, 2005b; Bayona et al., 2011; Ayala et al., 2012).

Complex shifts in U-Pb age distributions characterize 30 samples of Oligocene–middle Miocene clastic fill, including the upper Esmeraldas, Mugrosa, Colorado, and Santa Teresa Formations. The most distinct change involves an up-section reduction of the Cretaceous–Paleogene (100–50 Ma) signature that dominated the underlying Eocene section. The progressive relative decline of this population in the central-southern Middle Magdalena Valley Basin (Fig. 7), and its early–middle Miocene disappearance in the eastern Middle Magdalena Valley Basin (Fig. 6), is coupled with the joint introduction of Jurassic (200–150 Ma) and late Paleozoic (300–250 Ma) signals. These trends are accompanied by Precambrian age distributions (1800–900 Ma) that define an up-section increase in the central-southern Middle Magdalena Valley Basin but a complex increase and then subsequent decrease in the eastern Middle Magdalena Valley Basin. The complex fluctuation of Precambrian detritus in the eastern Middle Magdalena Valley Basin can be correlated with comparable shifts in the subsurface record of the Guane and Cocuyo wells (Fig. 4). However, none of the three wells shows the Late Cretaceous–early Paleogene signal, and the Cagui well is uniquely distinguished by the Miocene dominance (up to 90%) of Precambrian zircon populations (Fig. 5).

The results signify a complex interplay among source areas. Phases of relatively diminished influx from Cretaceous–Paleogene igneous sources (such as the 100–50 Ma signal) in the Central Cordillera broadly coincide with amplified contributions of Paleozoic and Precambrian ages. Although Precambrian age distributions are consistent with direct input from a cratonic source, these grains were most likely recycled from uplifted Cretaceous sedimentary rocks during mid-Cenozoic growth of the Andean fold-thrust belt encompassing the Eastern Cordillera and associated basement uplifts.

U-Pb results from uppermost levels of the Middle Magdalena Valley Basin show a distinctively cosmopolitan assemblage distinguished by the abrupt arrival of young zircons of chiefly Cenozoic age. Results from 10 samples of the upper Miocene–Pliocene Real and Mesa Formations show a pronounced introduction of grains younger than 100 Ma, with significant age probability peaks at 88–84, 81–78, 43–40, 14–11, and 9–6 Ma (Figs. 6 and 7). The youngest ages cluster in the 12–6 Ma range and reflect contributions from syndepositional magmatic sources that approximate the depositional ages for the southern-central Middle Magdalena Valley Basin (Opón syncline). A broad spread of subordinate age populations includes Precambrian (1800–900 Ma), early Paleozoic (500–400 Ma), Permian–Triassic (300–200 Ma), and Jurassic (200–150 Ma) ages comparable to results for the underlying Oligocene–middle Miocene deposits.

The overall age distributions correlate with subsurface age spectra for the Guane well (Fig. 4), matching both the appearance of young grains and the multiple modes of older pre-Cenozoic populations. Neither data set, however, correlates with the Precambrian-dominated age signatures for Miocene–Pliocene strata of the Cagui well (Fig. 5). The internal consistencies among some localities yet sharp contrasts among others underscore the continued competition among disparate sediment sources in the Central Cordillera and Eastern Cordillera. The wholesale emergence of 100–0 Ma zircons indicates that late Miocene paleodrainage systems of the Middle Magdalena Valley Basin gained access to a new sediment source, either through bedrock unroofing or drainage modifications. This late Miocene shift is particularly well expressed in a comparison of composite U-Pb age distributions for Cenozoic basin fill (Fig. 8), which shows the introduction of Late Cretaceous–Cenozoic zircons, consistent with the birth of a proto–Magdalena River.

Key provenance trends summarized from the U-Pb results (Figs. 3–8) facilitate spatial comparisons and correlations of Mesozoic–Cenozoic stratigraphic and tectonic histories for surface and subsurface deposits within the Middle Magdalena Valley Basin. Broader consideration of statistical similarities among U-Pb age spectra is enabled by calculation of cross-correlation coefficients between probability density plots of all sample pairs (Saylor et al., 2012b, 2013; Saylor and Sundell, 2014). Following recent studies (e.g., Chapman et al., 2012, 2015; Vermeesch, 2013), these correlation values are displayed graphically in multidimensional scaling (MDS) plots (Fig. 9) in which greater similarity is shown as closer proximity. The MDS plots for 104 Jurassic through Pliocene samples (Figs. 9A–9D) and seven modern river sand samples (Fig. 9E) reveal clear quantitative distinctions among samples derived from the major source regions.

The Eastern Cordillera, San Lucas and Santander block uplifts, and Central Cordillera all define diagnostic U-Pb age signatures (Fig. 9). Whereas the Eastern Cordillera is dominated by Cretaceous sedimentary units originally derived from the Precambrian craton, the Central Cordillera is composed largely of Cretaceous–Paleogene igneous rocks of the Andean magmatic arc, and the San Lucas and Santander block uplifts are characterized mostly by Jurassic igneous rocks and their sedimentary derivatives. These age distinctions, as expressed in modern rivers (Figs. 3 and 9E), serve as critical markers in the sedimentary record that allow discrimination of evolving sources during Mesozoic–Cenozoic basin evolution, thus providing the paleogeographic framework for tectonic reconstructions of the region (Fig. 10).

Coarse-grained Upper Jurassic–lowermost Cretaceous units of the lowermost Middle Magdalena Valley Basin are dominated by bedrock signatures of localized uplifts. The unimodal 200–150 Ma U-Pb age populations (Figs. 4 and 5) are consistent with local exposures of Lower to Middle Jurassic granitoid intrusions and correlative volcanic rocks (Cáceres et al., 2003; Gómez Tapias et al., 2007; Bayona et al., 2011). Unimodal age signatures also define the modern rivers draining intrabasinal and marginal Middle Magdalena Valley uplifts such as the San Lucas range and Santander Massif (Figs. 3C–3E, 3H, 9A, 9E, and 10A). A similar Jurassic age signature is observed in selected Upper Jurassic strata (Horton et al., 2010b), suggesting the possibility of a correlation tool between the Middle Magdalena Valley Basin and Eastern Cordillera (e.g., Fabre, 1987; Cáceres et al., 2003).

Local sources of sediment are consistent with interpretations of a Jurassic–earliest Cretaceous extensional setting containing localized coarse-grained depocenters (Cooper et al., 1995; Sarmiento-Rojas et al., 2006). In this regional extensional system, segregated fault-bounded subbasins (e.g., Mora et al., 2013; Tesón et al., 2013) were fed sediment by small drainage networks eroding uplifted ranges dominated by Triassic–Jurassic igneous rocks (Fig. 10A). These isolated ranges were likely developed in the structural zones presently occupied by marginal block uplifts such as the San Lucas range and Santander Massif (Fig. 1).

Cretaceous deposits record a mix of distal cratonic sources and diminished basement uplifts. U-Pb results are most distinguished by the introduction of a substantial Precambrian (1800–900 Ma) age population (Figs. 4, 5, and 7). Although continental crystalline basement underlies the Eastern Cordillera and Middle Magdalena Valley Basin, and small populations of inherited Precambrian grains are detected in Paleozoic–Mesozoic rocks of the Central Cordillera (Figs. 3A and 9E; Cordani et al., 2005; Vinasco et al., 2006), the sizeable fraction and broad spread of dated Precambrian grains in the absence of major western signatures suggest most basement contributions were derived from the South American craton (Guyana Shield) to the east (Figs. 9A and 10B). However, the coupled reduction in the formerly dominant Jurassic (200–150 Ma) populations and emergence of subordinate Paleozoic (300–250 Ma and 500–400 Ma) signals attest to deeper unroofing of local intrabasinal Middle Magdalena Valley uplifts and probable extrabasinal uplifts, principally in the Eastern Cordillera, but also conceivably in isolated zones of the Central Cordillera. The absence of syndepositional Cretaceous zircons underscores the lack of significant contributions from the Andean magmatic arc to the west, possibly due to transport distance or a regional west-dipping topographic gradient.

Whereas local signatures are consistent with individualized basement uplifts, the prevalent basement ages require an integrated drainage system across a broad region linking the distal eastern craton to the Middle Magdalena Valley Basin (Figs. 10B and 10C). This region includes the present-day Eastern Cordillera and Llanos Basin, where several subbasins were active during Early Cretaceous extension (Cooper et al., 1995; Sarmiento-Rojas et al., 2006; Mora et al., 2009). The coalescence of local fault-related subbasins into larger basin systems enabled regional westward transport of basement detritus. Postextensional thermal subsidence during mid- to Late Cretaceous time induced further basin expansion and depositional overlap of formerly active normal faults and local sediment sources (Fig. 10C). Regionally, progressive stratigraphic onlap eastward across the Llanos Basin (Fig. 1) and onto the South American craton (Guyana Shield) resulted in a Late Cretaceous shift to older Precambrian sources, with lessened 1200–900 Ma and enhanced 1900–1300 Ma populations (Horton et al., 2010b; Saylor et al., 2013).

A comprehensive reversal of sedimentary polarity is recorded by the introduction of a detrital zircon U-Pb age population indicative of a new sediment source along the western margin of the Middle Magdalena Valley Basin (Figs. 10D and 10E). The Late Cretaceous–Paleocene (90–60 Ma) population (Figs. 6 and 7) is unambiguously linked to erosion of igneous rocks in the Central Cordillera (Figs. 8, 9B, and E9E), as comparable ages are not observed elsewhere. The dominance of this signal is demonstrated by the unimodal 95–70 Ma signatures for modern rivers draining the region (Figs. 3A, 3B, and 3H). Less clear is the precise timing. Although the central-southern Middle Magdalena Valley Basin shows a roughly Maastrichtian appearance, data sets from the eastern-northeastern Middle Magdalena Valley Basin show a striking mid-Paleocene shift (e.g., Nie et al., 2010, 2012; Caballero et al., 2013a, 2013b). The introduction of this key signature progressively from south to north suggests that uplift of the Central Cordillera was not a single simultaneous event, but rather progressed northward along strike during Paleogene shortening and transpressional deformation. This spatial variability also precludes usage of this provenance signal as a robust tool for stratigraphic correlation. Subsurface samples of Paleogene units from more northwestern regions of the Middle Magdalena Valley Basin show none of the Central Cordillera sources detected in the southern, central, and eastern Middle Magdalena Valley Basin. Instead, an up-section increase in the late Paleozoic (300–250 Ma) population (Figs. 4 and 5) implies derivation from intrabasinal and marginal Middle Magdalena Valley uplifts comparable to the San Lucas range and Santander Massif adjacent to the northern Middle Magdalena Valley Basin (Figs. 1, 3C–3E, and 3H). This conspicuous along-strike variability in provenance history indicates a basin dominated by transverse rather than longitudinal deposystems.

Regionally, the provenance record is consistent with proposed latest Cretaceous–Paleocene uplift and exhumation in the Central Cordillera (Figs. 10D and 10E; Gómez et al., 2003; Villagómez and Spikings, 2013). This tectonic episode appears to represent a fundamental switch from a regional postrift setting to a zone of shortening and transpression linked to active subduction and possible minor collisions along the western margin.

Another distinctive reversal in sediment dispersal is expressed by the emergence of U-Pb age populations revealing growth of a substantial new sediment source along the eastern margin of the Middle Magdalena Valley Basin. A marked Oligocene amplification of a broadly distributed Precambrian population (1800–900 Ma) coincides with increased late Paleozoic (300–250 Ma) to Jurassic (200–150 Ma) ages and an abrupt decline of the formerly dominant Cretaceous–Paleogene (100–50 Ma) signature (Figs. 8 and 9C). The ubiquitous increase in Precambrian detrital populations across all sample localities (Figs. 4–7) offers a potential correlation tool for surface and subsurface units. Although Precambrian signatures are consistent with a distal cratonic source (Figs. 3F–3H), independent thermochronological and sedimentary evidence for early uplift in the Eastern Cordillera (Parra et al., 2009a, 2009b, 2012; Mora et al., 2010; Bande et al., 2012; Saylor et al., 2012a, 2012b; Sánchez et al., 2012) requires that Precambrian grains were recycled from uplifted Cretaceous strata in the Andean fold-thrust belt (Figs. 8 and 9C). Therefore, these old zircon signatures are regarded as a high-fidelity record of early rock uplift and exhumation of the Eastern Cordillera, with a dominance of eastern over western sources (Fig. 10F).

Following this provenance reversal, complex early–middle Miocene shifts are defined by fluctuations in various age populations. Most pronounced are irregular spatial variations in Precambrian populations, with decreases in most of the northern Middle Magdalena Valley Basin (Figs. 4 and 6) contrasting with increases in the central-southern and far northeastern Middle Magdalena Valley Basin locations (Figs. 5 and 7). A sharp increase in Precambrian detritus occurs within the Cagui well (Fig. 5), indicating a punctuated point source of sediment from the Eastern Cordillera. Further spatial variability is delineated by the Cretaceous–Paleogene signature: Although this 100–50 Ma population is severely reduced to nonexistent in most locations, its local persistence in the Opón syncline (Rio Oponcito) locality (central-southern Middle Magdalena Valley Basin) signifies spatially focused input from the Central Cordillera. These complex alternations reflect continued competition among eastern and western sources (Figs. 8 and 9C–9E) and suggest a compartmentalized provenance record displaying significant spatial variations, both along-strike and across-strike (Fig. 10F). Such provenance variations attest to a Cenozoic record in the Middle Magdalena Valley Basin dominated by transverse (east-west) rather than longitudinal (north-south) deposystems, consistent with a closed topographic basin disconnected from the Caribbean Sea (Moreno et al., 2011; Caballero et al., 2013b; Moreno et al., 2013; Silva et al., 2013; Reyes-Harker et al., 2015).

Final modification then establishment of the modern axial drainage configuration in the Magdalena Valley was achieved in late Miocene–Pliocene time (Fig. 10G). U-Pb results for upper Miocene–Pliocene deposits show the most cosmopolitan provenance assemblage for the entire Middle Magdalena Valley Basin succession, as defined by multimodal Precambrian, Paleozoic, and Mesozoic populations (Figs. 4, 6, and 7). This diverse and very broad distribution (Fig. 9D) contrasts with the concentrated unimodal and bimodal distributions expressed in most underlying deposits (Figs. 9A–9C). Although minor up-section increases and decreases characterize the various modes, a substantial shift is recorded by the sharp introduction of a 100–0 Ma signal (Fig. 8). This critical signal appears across all but one of the surface and subsurface samples (Figs. 4, 6, and 7), and it is only absent from the Cagui well (Fig. 5), which shows the continued dominance of Precambrian detritus derived from a point source of sediment eroded from Cretaceous strata in the Eastern Cordillera fold-thrust belt. The Cretaceous–Cenozoic (100–0 Ma) signature requires important contributions from the Central Cordillera (e.g., McCourt et al., 1984; Aspden et al., 1987; Gómez Tapias et al., 2007; Villagómez et al., 2011; Bayona et al., 2012, and references therein), but most probable candidates include the large Antioquia batholith and satellite intrusions (broadly 95–80 Ma) adjacent to the central Middle Magdalena Valley Basin and multiple 65–45 Ma batholiths and stocks (including the Manizales, El Bosque, and Hatillo intrusions) near the transition from the Middle Magdalena Valley Basin to the Upper Magdalena Valley Basin.

The regional emergence of the 100–0 Ma signature (Fig. 8), and the consistency of multimodal pre-Cenozoic zircon populations across studied localities reflect drastic modifications in large-scale paleodrainage patterns. Although localized input of syndepositional Neogene volcanic material characterized the Opón syncline (central-southern Middle Magdalena Valley Basin), the enhanced Cretaceous–Cenozoic populations point to a drainage expansion, likely through capture of upstream catchments of the Upper Magdalena Valley, in which the evolving system gained access to new sediment sources (Fig. 10G). This episode corresponds with a transition from a series of compartmentalized transverse deposystems, each dominated by sources in either the Central Cordillera or Eastern Cordillera, into an integrated longitudinal system. This late Miocene switch constitutes the onset of a through-going proto–Magdalena River, flowing axially from south to north, and linking the Upper Magdalena Valley and Middle Magdalena Valley Basin to the Caribbean coast (Fig. 10G).

Tectonic events of the northernmost Andes (Fig. 1) are robustly expressed in both the surface and subsurface provenance record of the Middle Magdalena Valley Basin (Figs. 4–9). At the broad scale, the detrital zircon U-Pb geochronological data set faithfully records the three principal Mesozoic–Cenozoic episodes of the northernmost Andes. First, for the Jurassic–earliest Cretaceous phase of regional extension involving the growth of synrift subbasins (e.g., Colletta et al., 1990; Sarmiento Rojas, 2001; Sarmiento-Rojas et al., 2006; Kammer and Sánchez, 2006; Mora et al., 2009), detrital zircon data show unimodal age signatures emblematic of uplifted bedrock sources along localized normal faults (Fig. 10A). Second, a Cretaceous phase of regional postextensional subsidence (Fabre, 1987; Toussaint and Restrepo, 1994; Cooper et al., 1995; Roeder and Chamberlain, 1995; Toro et al., 2004) is reflected in the emergence of a distal eastern cratonic source replacing localized footwall uplifts (Figs. 10B and 10C). Third, a protracted Cenozoic period of Andean shortening (Dengo and Covey, 1993; Taboada et al., 2000; Gómez et al., 2005a; Parra et al., 2009b; Mora et al., 2010) is expressed as a complex mix of proximal sources dictated by a competition between eastern and western sources during the growth of basin-bounding ranges (Figs. 10D–10G).

At finer scales, U-Pb age distributions also provide accurate signatures of two critical transitions during Andean mountain building. First, earliest shortening-related uplift in the flanking Central Cordillera (Gómez et al., 2003, 2005b) can be tied to a latest Cretaceous–Paleocene reversal in paleodrainage, as defined by a shift in U-Pb age spectra from an eastern cratonic source to a western orogenic source (Figs. 9A, 9B, 9E, 10C, and 10D). Second, initial uplift of the Eastern Cordillera (Parra et al., 2009b; Mora et al., 2010; Bande et al., 2012; Saylor et al., 2012a; Sánchez et al., 2012) is identified by U-Pb age signatures (recycled populations of craton-derived Precambrian detritus) recording the latest Eocene–Oligocene unroofing of Mesozoic strata in the Andean fold-thrust belt (Figs. 9B, 9C, 9E, 10E, and 10F).

Of these provenance shifts, nearly all of them provide sufficient fidelity for correlation of surface and subsurface deposits. Rather than constraining precise matches among U-Pb age spectra of individual samples, such correlations are reliant on the up-section appearance or disappearance of the aforementioned tectonic signatures. Perhaps the most valuable criteria involve the identifications of important reversals of sediment polarity (e.g., Coney and Evenchick, 1994). Such reversals prove to be sufficient in identifying wholesale uplift of contiguous barriers but can be complicated by along-strike variations. As an example, along-strike propagation of uplift in the Central Cordillera (Gómez et al., 2003, 2005b; Villagómez and Spikings, 2013) may explain modest discrepancies from south to north within the flanking Middle Magdalena Valley Basin.

Despite the utility of detrital provenance archives, several key components of the geologic and tectonic history appear to be less well expressed, even missing, from the U-Pb data set. First, a well-documented phase of late Paleocene–middle Eocene basement uplift and erosional beveling in the central to northern Middle Magdalena Valley Basin (Fig. 10; Gómez et al., 2003, 2005b; Moreno et al., 2011) appears to be mostly aliased in the detrital record, as there is no corresponding major shift in U-Pb age distributions (Figs. 4–7 and 9B). Several subsurface basement highs and the associated Middle Magdalena Valley Basin angular unconformity (Fig. 2) are well imaged by seismic-reflection profiles, with supporting borehole data. Although localized, these are large features (Infantas, La Cira, and Cáchira highs) with over several kilometers of vertical structural relief (Morales, 1958; Moreno et al., 2011; Parra et al., 2012; Gómez et al., 2003, 2005b; Caballero et al., 2013a, 2013b; Mora et al., 2013; Tesón et al., 2013). Therefore, it is difficult to envision a case where the provenance signal would be volumetrically eclipsed by other source regions. Rather, we suggest that comparable U-Pb age signatures of 300–250 Ma and 200–150 Ma for both the uplifted basement and recycled sedimentary units preclude clear discrimination of the Middle Magdalena Valley Basin basement sources. Nevertheless, low-temperature thermochronological results from other deposits of this age reveal a sharp influx of material that can be attributed to Middle Magdalena Valley Basin basement uplift (Saylor et al., 2012b), highlighting the utility of complementary techniques.

Second, despite the proximity to the western magmatic arc, there is no robust Middle Magdalena Valley Basin record of a proposed Paleogene magmatic pulse, and U-Pb ages prove to be of limited utility in constraining Paleogene depositional ages. Bayona et al. (2012) suggested that the clear pulse of 65–45 Ma magmatism in the Central Cordillera, near the transition from the Middle Magdalena Valley Basin to the Upper Magdalena Valley Basin (McCourt et al., 1984; Aspden et al., 1987; Gómez Tapias et al., 2007; Villagómez et al., 2011), may also extend to undiscovered igneous units in the Eastern Cordillera. Nevertheless, the U-Pb results illustrate very limited proportions of possible syndepositional zircons within Paleocene–middle Eocene Middle Magdalena Valley Basin strata. Of the 47 samples from the uppermost Cretaceous–lowermost Oligocene section, only three samples (all restricted to the Guaduas syncline in the southernmost basin) contain statistically significant populations of grains (n > 2 ages) approaching stratigraphic ages, accounting for a cumulative total of ∼1% of all analyzed grains. Given the short transport distances (<100–200 km) between the Central Cordillera and the Middle Magdalena Valley Basin, the limited amount of Cenozoic ages cannot be attributed to preferential weathering and selective removal of these grains (e.g., Bayona et al., 2012). Therefore, erosion of the observed Paleogene magmatic centers in the Central Cordillera must have fed sediment to only a limited portion of the southernmost Middle Magdalena Valley Basin, with no measurable contributions to downstream localities in the central and northern parts of the basin (Figs. 10D–10F).

U-Pb age distributions from the Middle Magdalena Valley Basin illustrate a complex series of transverse paleodrainage alternations followed by a wholesale drainage reorganization involving the establishment of a major axial paleoriver (Fig. 10). After initial topographic growth of the Eastern Cordillera, the mid- to late Cenozoic evolution of the Middle Magdalena Valley Basin was governed by competing sediment sources along both the western and eastern margins. Rather than a simple mix of source signatures, most Oligocene–middle Miocene samples of fluvial basin fill indicate a clear dominance of either the Eastern Cordillera (e.g., Cagui well; Fig. 5) or the Central Cordillera (e.g., southern Middle Magdalena Valley Basin; Fig. 7). In fact, several episodes of pronounced sediment influx are recorded over a wide range of sites (in both the north-south and east-west directions) in which the Middle Magdalena Valley Basin was dominated by either a western source or an eastern source (Fig. 9). For example, Oligocene deposits from most of the surface and subsurface localities show the originally bimodal eastern versus western signals progressively superseded by a complex integrated signature (Figs. 8 and 9). At face value, the lengthy Paleocene to middle Miocene record could be largely construed as alternating “see-saw” episodes of shifting paleodrainage between two opposing transverse margins (Fig. 10). Although this characterization may seem overly simplistic, the provenance record does suggest that Middle Magdalena Valley Basin deposystems were highly sensitive to sediment flux and subsidence parameters controlled by processes in the flanking ranges. Regardless of the driving mechanisms, the U-Pb results require that these fluvial systems were mostly dominated by either an Eastern Cordilleran or Central Cordilleran source, without large amounts of mixing or downstream dilution. This time-space complexity likely reflects relatively short transport distances for the Middle Magdalena Valley Basin (tens to hundreds of kilometers), as compared to systems in the modern Andean foreland (thousands of kilometers). The results further suggest isolated deposystems within a compartmentalized basin system, without an integrated regional drainage network (Figs. 10D–10F).

The provenance data provide a refined understanding of the evolution of the proto–Magdalena River (Fig. 10G). Disagreement persists over the inferred onset of a regional longitudinal Magdalena River, with previous estimates centering on the late Miocene (Duque-Caro, 1978; Guerrero, 1993; Hoorn et al., 1995), Oligocene (Villamil, 1999), or middle Eocene (Escalona and Mann, 2011). Most studies propose that growth of the Eastern Cordillera as a topographic barrier would have deflected an axial Magdalena River northward into the Caribbean Sea, with the isolated Santa Marta Massif near the northern coast also playing a role in the drainage shift away from a former eastern outlet within the Maracaibo Basin (Fig. 1; Duque-Caro, 1978; Ayala et al., 2012). However, the U-Pb results show no clear integration of a large-scale fluvial system until late Miocene time (Fig. 8). In fact, an integrated paleodrainage system is precluded by the highly variable U-Pb age distributions, which show the appearance and disappearance of local signals in the pre–late Miocene succession. Furthermore, a late Miocene onset is consistent with the depositional record of the Magdalena delta and submarine fan (Flinch, 2003; Flinch et al., 2003a, 2003b; Rincón et al., 2007; Cadena and Slatt, 2013; Cadena Mendoza, 2014). This delayed integration of a through-going Magdalena River (Fig. 10G) may be due to the role of complex intrabasinal Middle Magdalena Valley structures or localized subsiding depocenters, or upstream processes farther south such as uplift along the margins of the Upper Magdalena Valley Basin (Butler and Schamel, 1988; Ramon and Rosero, 2006; Montes et al., 2005; Anderson, 2015).

Despite the complex time-space interaction of Mesozoic–Cenozoic tectonics and sediment accumulation in the Middle Magdalena Valley Basin, robust provenance trends are revealed by U-Pb age distributions from both surface and subsurface basin fill. Published U-Pb age distributions for detrital zircons collected from the mouths of modern rivers demonstrate that small to large watersheds provide clear signals representative of the competing source regions. Therefore, U-Pb ages within the Middle Magdalena Valley Basin stratigraphic succession allow for discrimination of sediment influxes from the contiguous Central Cordillera and Eastern Cordillera, as well as the localized Santander Massif and San Lucas range.

The principal Mesozoic–Cenozoic tectonic events of the northernmost Andes are well expressed in the sediment provenance record, including (1) Jurassic growth of extensional subbasins fed by local sources, (2) Cretaceous deposition in a regional postrift setting, and (3) Cenozoic crustal shortening and surface uplift of contiguous ranges and localized block uplifts. Finer-scale transitions in the Andean shortening history are also preserved. Early uplift in the flanking Central Cordillera is defined by a mid-Paleocene provenance shift reflecting the first arrival of material derived from the west, principally from magmatic arc rocks of Cretaceous age. Subsequent uplift and unroofing of the Eastern Cordillera are indicated by the late Eocene introduction of considerable cratonic Precambrian detritus recycled from Cretaceous–Paleogene strata.

The subsurface data set provides essential access to the older, pre-Eocene succession and reveals a complex Cenozoic history of transverse and axial paleodrainage. However, several components of the provenance record are poorly defined. A well-documented phase of late Paleocene–middle Eocene uplift of basement highs in the Middle Magdalena Valley Basin appears to be aliased in the U-Pb age distributions, likely due to similarities with recycled sedimentary signatures. In addition, despite the proximity of the Middle Magdalena Valley Basin to the Andean magmatic arc, strong evidence for a proposed Paleogene magmatic pulse is lacking, and U-Pb ages are of relatively limited utility in defining precise depositional ages. The time-space complexity of U-Pb age distributions during Oligocene–Miocene time shows alternating periods of sediment influx from the two opposing transverse margins, without significant source dilution, consistent with a compartmentalized basin system characterized by isolated proximal deposystems. The abrupt up-section introduction of youthful (100–0 Ma) zircons signifies the late Miocene establishment of an integrated longitudinal drainage network and activation of the proto–Magdalena River along the axis of the Middle Magdalena Valley Basin, coincident with growth of the Magdalena submarine fan along the Caribbean margin.

This research was funded by the Instituto Colombiano del Petróleo (ICP), a division of Ecopetrol, and the Jackson School of Geosciences, as part of a collaborative research agreement between ICP and the University of Texas at Austin. Additional support during the data synthesis phase of this research was provided by National Science Foundation grants EAR-1019857 and 1338694. Many Colombian researchers for the ICP-Ecopetrol project “Cronologia de la deformación en las Cuencas Subandinas” shared valuable information during this research. We thank Alejandro Bande, Todd Housh, Chris Moreno, Julian Naranjo, Juan Carlos Ramírez, Andres Reyes-Harker, Jorge Rubiano, C. Javier Sanchez, Jair Sierra, Eliseo Tesón, and Vladimir Torres for beneficial discussions. Reviews from François Roure and Raymond Russo helped improve the manuscript.