The sources of the tremendous amount of Cenozoic siliciclastic sediment deposited in the Gulf of Mexico region remain debated because of a lack of definitive provenance-identifying characteristics. In an effort to build on prior provenance analysis, we present 101–160 single-grain detrital zircon U-Pb ages for each of 10 outcrop samples from Upper Paleocene to Upper Miocene sandstones from a ∼10,000 km2 swath of central Louisiana corresponding to the ancient Mississippi River Delta, the largest Cenozoic depocenter in the northern Gulf of Mexico region. Sample depositional age control is derived from biostratigraphy and/or regional lithostratigraphic correlation. U-Pb ages in each of the samples range from Cenozoic to Archean, and correspond to the ages of various geologic terranes that underlie the modern Mississippi River drainage basin. However, the prominence of various age distributions changes systematically through the Cenozoic stratigraphy, and pronounced shifts in the abundance of certain age distributions between stratal packages appear to be correlated to shifts in heavy mineral assemblages recorded across the northern Gulf of Mexico coastal plain. Comparison of coastal plain detrital zircon age distributions to age distributions from North American sedimentary cover and the ages of major North American crystalline basement rocks, aided by a sediment mixing model, illuminates the provenance of each of the stratal packages, and suggests that (1) the Mississippi River catchment has resembled its present configuration, at least in the east-west dimension, for much, if not all, of the Cenozoic, and (2) depositional episodes on the Louisiana coastal plain characterized by high sediment supply also corresponded to high proportions of sediment sourcing from the Sevier-Laramide region of the interior western United States. Sediment supply to the Louisiana coastal plain by the paleo–Mississippi River has generally been high during the Cenozoic, except for an anomalous low during the Middle Eocene, when the abundance of sediment derived from the Rocky Mountain region decreased dramatically relative to sediment derived from the Appalachian region.

The Gulf of Mexico Basin contains one of the most voluminous accumulations of Cenozoic sediment in the world. Here, sedimentary cover mantles attenuated transitional continental crust along the southern North American continental margin and extends southward to a region underlain by oceanic crust in the Gulf of Mexico (e.g., Buffler and Sawyer, 1985; Sawyer et al., 1991). Along the coastal plain, numerous Cenozoic stratal packages offlap the continental margin (Fig. 1) and thicken basinward to an aggregate thickness of as much as 15 km (Sawyer et al., 1991; Culotta et al., 1992; Peel et al., 1995). Regional paleogeographic reconstructions and isopach maps for each of these stratal packages show that the Cenozoic sediment accumulated in several deltaic depocenters distributed along the strike of the basin. In the northern U.S. Gulf Coast region, the largest depocenters are approximately colocated with the lower reaches of the modern Brazos, Mississippi, and Red Rivers, and there is an additional, eastern depocenter in the southern Mississippi-Alabama border region that is thought to be linked to an ancestral Tennessee River (Fig. 1; e.g., Storm, 1945; Murray, 1955; Williamson, 1959; Brown, 1967; Winker, 1982; Salvador, 1991; Galloway et al., 2000, 2011, and references therein). Despite an extensive history of basin analysis around the Gulf of Mexico, many of the interpretations of sediment provenance at major depocenters remain equivocal and controversial, and there is a need to augment existing work with newly emerging geochemical provenance-identifying measurements (e.g., Galloway et al., 2011). Given the tremendous volumes of sediment in the Gulf Coast region, this issue is important for understanding the erosion of the North American continental interior, from the Rocky Mountains to the Appalachians (e.g., Boettcher and Milliken, 1994; Pazzaglia and Kelley, 1998; Pederson et al., 2002; McMillan et al., 2006; Roy et al., 2009). In addition, understanding sediment provenance and the orientation of sediment dispersal systems may refine definitions of stratal packages across the basin and aid assessments of hydrocarbon reservoir potential, particularly for downdip strata that are covered by extensive allochthonous salt canopies (e.g., Meyer et al., 2005, McDonnell et al., 2008).

Previously, geologists have used the composition of non-opaque heavy mineral assemblages in modern and ancient Gulf Coastal plain sediments to interpret the provenance of stratal packages (e.g., Bornhauser, 1940; Cogen, 1940; Goldstein, 1942; Hsu, 1960; Mange and Otvos, 2005). Early work demonstrated the presence of several heavy mineral provinces in modern rivers and beaches across the region, with kyanite and staurolite being relatively prevalent on the Mississippi and Alabama coastal plain, and garnet being relatively prevalent on the Louisiana and Texas coastal plain (e.g., Goldstein, 1942; Hsu, 1960). Due to the fact that these minerals are commonly found in orogenic belts in the eastern and western United States, respectively, the pattern suggests that heavy mineral assemblages are potentially a useful tool for discriminating between stratal packages with eastern versus western source regions. However, interpreting provenance based on heavy mineral assemblages has at least two critical shortcomings, and a basin-wide framework for this type of analysis in ancient sediments has never been established. First, diagenetic processes alter primary heavy mineral assemblages and obscure provenance information, beginning even at modest burial depths of ∼1 km (e.g., Milliken, 1988, 2007). Second, the provenance information contained in heavy mineral assemblages, including the kyanite-staurolite– and garnet-rich assemblages on the Gulf Coast, is often non-unique, such that regions on opposing sides of the continental interior may be permissible sources for a given stratal package (e.g., Todd and Folk, 1957; McCarley, 1981).

Most recent attempts to understand sediment provenance in the Gulf Coast region have relied primarily on integrating analysis of the architecture, size, and age of deltaic depositional systems with geologic observations from upstream hinterland regions and downplayed sandstone compositional observations (e.g., Winker, 1982; Galloway, 2005). To refine these drainage basin reconstructions, some workers have applied detrital zircon U-Pb age spectra provenance analysis to Upper Paleocene–Lower Eocene Wilcox Group strata on the Texas coastal plain (Hutto et al., 2009; Mackey et al., 2012; see also Lawton et al., 2009). Zircon crystallization ages can be correlated to basement provinces around North America (e.g., Dickinson and Gehrels, 2009a) and because the mineral is highly refractory, provenance information is generally not lost during diagenesis (e.g., Harrison et al., 2002). Studies of detrital zircon age spectra in the Texas Wilcox Group argued that the unit derived from the Laramide uplifts of the Rocky Mountains (e.g., Mackey et al., 2012). The northern limits of the paleo–Wilcox Group catchment, however, remain difficult to define because the crystalline rocks that appear to be the sediment source cover an expansive region, extending ∼1000 km along the strike of the interior Rocky Mountains, from northern Mexico to southern Wyoming (Karlstrom and Humphreys, 1998). Furthermore, North American continental interior drainage evolution through the later Cenozoic has not been evaluated with detrital zircon provenance analysis of sandstones in the Gulf of Mexico region.

In this contribution, we aim to reconstruct the sediment sources for Upper Paleocene to Upper Miocene sands deposited within the Mississippi delta region, using U-Pb ages of detrital zircons. We integrate our measurements with existing detrital zircon age spectra from (1) modern river sediments, (2) possible sediment sources, and (3) correlative stratal packages along the strike of the Gulf of Mexico coastal plain in order to better understand drainage patterns across large swaths of the North American continental interior during the Cenozoic. We also integrate our observations with recently published measurements of sediment flux to the coastal plain (Galloway, 2008; Galloway et al., 2011) in order to constrain Cenozoic rates of sediment export from the Sevier and Laramide regions of the western United States as well as the Appalachian region of the eastern United States.

Louisiana occupies the southwestern flank of a broad basement low on the Gulf Coastal Plain, called the Mississippi embayment, the central axis of which nearly corresponds to the course of the modern Mississippi River (Fig. 1). In our study area in central Louisiana, the Mississippi embayment is bound to the northwest by a broad basement high, called the Sabine Uplift, and bound to the northeast by another basement high, the Monroe arch (Fig. 1; Murray and Thomas, 1945; Sawyer et al., 1991). Although these basement highs have undergone several episodes of differential uplift and subsidence since the Mesozoic (e.g., Moody, 1931; Jackson and Laubach, 1988), the most recent episode appears to have been in the Late Paleocene and Early Eocene (see Jackson and Laubach, 1988; Laubach and Jackson, 1990; Tikoff and Maxson, 2001). Projection of Wilcox Group strata across the Sabine Arch culmination shows a maximum of ∼180–300 m of total erosion due to the combined effects of structural relief development and eustatic fall, such that Tertiary structural relief on this arch is modest (Murray and Thomas, 1945; Jackson and Laubach, 1988).

The depositional history of the Mississippi embayment is generally straightforward. Following continental rift-related sedimentation during the Late Triassic breakup of Pangaea and opening of the Atlantic Ocean, the Jurassic and Cretaceous history of the region was dominated by carbonate deposition across a broad, passive continental margin, with brief episodes of siliciclastic sediment influx (e.g., Salvador, 1991; Galloway, 2008). Predominantly carbonate and fine-grained siliciclastic sedimentation, with only minor sand sedimentation, persisted into the Late Cretaceous and the Early Paleocene, but the Middle Paleocene marked the start of a relatively continuous episode of coarse-grained siliciclastic deltaic sedimentation, initially sourced by the paleo–Mississippi River, that has persisted to the present (e.g., Moody, 1931; Murray and Thomas, 1945; Salvador, 1991; Galloway, 2008). On northern Louisiana portions of the coastal plain, Cenozoic sedimentation has generally occurred in delta plain, deltaic, and shore zone depositional settings (e.g., Fisher and McGowen, 1967, 1969; Galloway et al., 2000; Galloway, 2008). The locus of Mississippi Delta sedimentation has generally been focused over Louisiana, but it has shifted southward through time (Galloway, 2008). Throughout the Cenozoic, sediment loading caused flexural subsidence of the continental margin, tilting sediments in northern Louisiana into a south-dipping panel (Fig. 1; Nunn, 1985) and creating belts of rocks that trend approximately east-west across the state.

The Cenozoic depositional history of the region has been subdivided into 14 depositional episodes (excluding the Early Paleocene), bound by either basin-margin unconformities or maximum-flooding disconformities that broadly correspond to key lithostratigraphic units of the region at the formation level (Fig. 2; Galloway et al., 2011). Isopach maps for each of these depositional episodes have been used to reconstruct the rates sediment flux into the basin through time (Fig. 2; e.g., Galloway and Williams, 1991; Galloway et al., 2000, 2011; Galloway, 2008). In general terms, sediment influx was rapid during deposition of the Upper Paleocene Wilcox Group. During the deposition of the Lower Eocene Wilcox Group, sediment influx began to wane, and there was a distinct minimum in sediment influx during deposition of the Middle Eocene Claiborne Group. In general, the remainder of the Tertiary was characterized by relatively high, but temporally variable, sediment influxes (Fig. 2). Because these sediment flux reconstructions integrate across multiple depocenters, we sought to pair them with a proxy for Mississippi delta sediment flux for this study. Following the approach employed by Carvajal et al. (2009), we measured the maximum amount of continental shelf progradation in southern Louisiana, orthogonal to the continental shelf margins, based on shelf edge reconstructions for each of the Cenozoic depositional episodes defined by Galloway (2008) (Fig. 2). The maximum continental shelf progradation in front of the Mississippi delta in Louisiana during the Cenozoic generally mimics the regional sediment flux history (Fig. 2; e.g., Winker, 1982; Galloway et al., 2011), suggesting that Mississippi River sediment flux constitutes a large portion of the overall Cenozoic sediment budget in the region.

The patchwork of crystalline basement rocks underlying the North American continental interior is likely the ultimate source of detrital zircons on the Louisiana coastal plain (Fig. 3), although most zircons were probably recycled through sedimentary cover prior to deposition on the continental margin. Several recent studies provide excellent reviews of North American basement terranes (e.g., Whitmeyer and Karlstrom, 2007), including several detrital zircon provenance studies (e.g., Dickinson and Gehrels, 2009a; Park et al., 2010), and readers are referred to those studies for additional information on this subject. The basement terrane compilation in this study (from Whitmeyer and Karlstrom, 2007; Rankin et al., 1989; Tollo et al., 2004; Miller et al., 2006; Dickinson and Gehrels, 2009a) differs slightly from compilations used in many recent North American detrital zircon provenance studies, which treat the Yavapai and Mazatzal provinces as a composite, extending from the Cheyenne belt in southern Wyoming to northeastern Mexico (Fig. 3). The region is actually underlain by three distinct southwest-northeast elongate Paleoproterozoic and Mesoproterozoic geologic provinces (Bennett and DePaolo, 1987; Karlstrom and Bowring, 1988; Whitmeyer and Karlstrom, 2007, and references therein), which include the 1.80–1.72 Ga Yavapai province, the 1.72–1.65 Ga Mazatzal province, and the 1.55–1.35 Ga Granite-Rhyolite province, all of which were intruded by a suite of 1.45–1.35 Ga granitoids (Bickford et al., 1986; Windley, 1989; Anderson and Cullers, 1999; Whitmeyer and Karlstrom, 2007) (Fig. 3). The three provinces generally exhibit a southward decrease in the ages of basement terranes, from Archean in the north, to Mesoproterozoic in south (Fig. 3A).

The regions of sedimentary cover that are the most probable sources of detrital zircons to the Louisiana coastal plain are those parts of the continental interior that were deeply eroded during the Cenozoic, including the eastern, interior part of the Rocky Mountains that is covered by a broad Jurassic–Cretaceous foreland basin related to the Sevier orogeny (e.g., DeCelles, 2004) and a superimposed network of intramontane basins that formed when the Sevier foreland was partitioned during the Laramide orogeny (Fig. 3) (Dickinson et al., 1988; Pazzaglia and Kelley, 1998; Crowley et al., 2002; DeCelles, 2004). On the opposing side of the continent, numerous studies point to persistent erosion of the Appalachian hinterland and foreland, locally at rates as high as 30 m/m.y., since the Cretaceous (Boettcher and Milliken, 1994; Granger et al., 2001; Matmon et al., 2003; Reed et al., 2005), and growing evidence suggests an acceleration in erosion of the Appalachian region since the Miocene (Boettcher and Milliken, 1994; Gallen et al., 2013; Miller et al., 2013). Recent drainage reconstructions suggested that these regions in the western and eastern continental interior were both integrated and isolated from the paleo–Mississippi River at various times in the Cenozoic (e.g., Galloway et al., 2011); we assess these reconstructions through analysis of detrital zircon age spectra.

Sand or sandstone was sampled from 10 natural outcrops and quarries in north-central Louisiana (Fig. 1; Table 1). Sampling focused on the most coarse-grained material at a site, and in general, the grain size of samples ranges from fine- to coarse-grained sand. Limited exposure and deep weathering of outcrops made detailed stratigraphic descriptions at many sites impractical. However, the sample sites share a few general stratigraphic characteristics, including lenticular bedding (generally laterally continuous over meters to tens of meters), which at many sites could be observed to grade laterally into finer grained deposits, and a lack of marine fossils. At many sites, channel scours and/or trough or planar cross-stratification were observed, and coarse-grained strata at many sites are interbedded with lignites (Wilcox Group; see Glawe, 1989; Catahoula Formation; see Wrenn et al., 2003). The stratigraphic observations at each site suggest that all samples were deposited in channels or crevasse splays in a delta plain to delta front setting, along the coastal depositional profile (Elliot, 1974; Reading and Collinson, 1996). This interpretation is in good agreement with several paleogeographic reconstructions based on evaluation of regional outcrop and subsurface data, which suggest a lower delta plain or delta front depositional setting for the various sample sites (Murray, 1947; Galloway et al., 2000; Galloway, 2008). We collected ∼10–20 kg of sand or sandstone at each site. Each sample represents a composite of material from several beds, distributed over the maximum exposed stratigraphic thickness (generally 1–5 m). This sampling strategy was implemented in an effort to obtain well-mixed samples, in light of the fact that detrital zircon populations may be poorly mixed in fluvial deposits (DeGraaff-Surpless et al., 2003; Lawrence et al., 2011; cf. Saylor et al., 2013).

Depositional age control for the outcrops generally is from biostratigraphy and regional lithostratigraphic correlation (Table 2), and some of the sample sites are considered to be type localities of various stratigraphic units in the state of Louisiana. Depositional ages range from Late Paleocene to Late Miocene. The Pliocene or Quaternary strata were not sampled because exposure is generally poor across the low-lying southern Louisiana coastal plain, and outcrops are difficult to distinguish from the multiple generations of Quaternary terrace fill that are inset into the coastal plain (Rosen, 1969; Blum and Törnquist, 2000; Mange and Otvos, 2005; Otvos, 2005). There is some controversy over the age of the Sicily Island Catahoula sample (sample CG11-5; Table 2). A trace amount of Late Miocene pollen taxa has been reported at this site (Wrenn et al., 2003); however, we assign an Oligocene age for the Sicily Island site on the basis of the distinctive lithostratigraphic character and regional biostratigraphy suggesting an Oligocene age (e.g., Berry, 1916) and the regional map pattern (Fig. 1).

U-Pb Geochronology

Zircons were separated from bulk sediment samples, and >100 single-grain U-Pb ages for each sample were measured using laser ablation–inductively coupled plasma–mass spectrometry (University of California at Santa Barbara) following the conventional protocols described by Cottle et al. (2009, 2012) (all U-Th-Pb data are presented in the Supplemental File1). Due to changing precision and accuracy of the various U-Pb and Pb-Pb geochronologic systems for different windows of geologic time, for 206Pb/238U ages younger than 800 Ma we display 206Pb/238U ages, and for 206Pb/238U ages older than 800 Ma, we display 206Pb/ 207Pb ages (Fig. 3; Gehrels, 2012). Data exhibiting ≥15% normal discordance or ≥5% reverse discordance, calculated from the ratio of the 206Pb/238U age to the 207Pb/235U age, were filtered. Our approach to filtering eliminated ∼2%–10% of grains from any given sample and was designed to strike a balance between the need to discard grains with highly discordant ages with the need to preserve very young (Cenozoic) or very old (Archean) grains, which may exhibit high discordance, but nevertheless contain critical provenance information. Generally, no 204Pb could be detected within the zircons, to within the instrumental uncertainty (see  Appendix 1) (<5% of total analyses had significant 204Pb); therefore, rather than attempting to correct ages for common Pb, we discarded the few analyses that were discordant due to significant 204Pb concentrations.

Analysis of Detrital Zircon Data

In light of the complex nature of age spectra on the Gulf of Mexico coastal plain, which likely integrate zircons from multiple source regions (e.g., Mackey et al., 2012), we employ two statistical calculations in the consideration of our data (adapted from Andersen, 2005, also see Vermeesch, 2004). Within these calculations, n represents the number of randomly selected detrital zircons in a sample from a sedimentary host rock, which contains N total detrital zircons. The detrital zircon sample is intended to quantify the abundance of the ith population, or Xi, for all age populations in the host sediment. However, only a finite number of detrital zircon ages can be measured (N >> n), so it is important to understand the detection limit of a given data set. Herein, a detection limit (XL) is defined as the abundance of the largest age population (Xi) of zircons likely to be unrepresented by a sample age distribution (xi) after n single-grain age measurements (after Dodson et al., 1988; Andersen, 2005):
where pL is the probability level assigned to the detection limit.
A sample distribution with n single-grain ages would be considered representative under the condition that xi = Xi/N for all age populations present in the host sediment; however, it is important to understand the uncertainty for xi when comparing values of xi among different sample distributions. Following Andersen (2005), and assuming that xi = Xi, 2σ relative error (2σi) for xi can be estimated as:

In order to deconvolve zircon contributions to downstream sediment sinks from various upstream sources containing complex age spectra (e.g., Amidon et al., 2005a, 2005b; Saylor et al., 2013), we also employ a mixing model that (1) creates synthetic age spectra based on all possible mixing proportions of end-member sediment source zircon populations, and then (2) calculates the similarity between the various synthetic age spectra and measured age distributions (modified from Amidon et al., 2005a, 2005b). Similarity between synthetic and observed age spectra is measured by calculating the areal mismatch between normalized relative probability plots, and we use a 5 m.y. smoothing window (e.g., Amidon et al., 2005a, 2005b; Saylor et al., 2013). In an effort to construct sample distributions with sufficiently large values of n to make robust characterizations of xi for these calculations, age spectra for key stratigraphic groups or formations are combined into composites, each consisting of several hundred single-grain age measurements from multiple sample localities. We refer to composite age distributions as detrital zircon chronofacies (e.g., Lawton et al., 2010; LaMaskin, 2012; May et al., 2013). We also perform mixing calculations on a detrital zircon age distribution from modern Mississippi River sediment in Louisiana (Iizuka et al., 2005) to evaluate sediment source mixing at a time in geologic history when the Mississippi River catchment boundaries are known (the present) (Fig. 3).

Detrital zircon ages range from Archean to Cenozoic (Figs. 4 and 5) in every sample (for complete U-Pb zircon geochronology data and concordia diagrams, see Supplemental File [footnote 1]). Five age distributions encompass nearly all (>95%) of the measured zircons ages, which are 540–0 Ma (D1, Phanerozoic), 770–540 Ma (D2, late Neoproterozoic), 1550–950 Ma (D3, Mesoproterozoic), 2050–1600 Ma (D4, late Paleoproterozoic), and 3050–2450 Ma (D5, Archean). D1, D3, and D4 can be further divided into subdistributions (see Table 3 for a full list of subdistributions). In general, the most prominent age probability peaks on the relatively probability plots (and the largest proportions of zircons) correspond to a few (sub)distributions, i.e., (1) 540–0 Ma (D1), (2) 1350–950 Ma (D3a), and (3) 1820–1600 Ma (D4a); we place special emphasis on these subdistributions herein. In the following, we describe zircon age distributions within each lithostratigraphic group or formation. Generally, we require ≥3 overlapping single-grain ages in order to define an age probability peak on the relative probability plots that we consider to be robust (Gehrels, 2012). In addition to presenting age spectra from the various stratigraphic units, we summarize results of Kolmogorov-Smirnov (K-S) statistical tests that evaluate the null hypothesis that age distributions are drawn from the same original parent population, in our case at the 95% confidence level (see Supplemental File [footnote 1]; Berry et al., 2001; Gehrels, 2012).

For samples CG11-2, CG11-3, and CG11-9, >3 of the youngest single-grain age measurements overlap with the depositional age control derived from biostratigraphy and regional lithostratigraphic correlation, to within 2 standard errors, suggesting that the existing depositional age control is robust (Fig. 6; Table 2). For the remaining samples, although regional biostratigraphy provides depositional age control, fewer than 3 single grain ages overlap with biostratigraphically determined depositional ages. Importantly, no zircons in any of the samples are younger than biostratigraphic depositional ages (see Table 2).

Wilcox Group Samples

Both Wilcox Group samples (CG11-3 and CG11-10) are dominated by age distributions that date to the late Paleoproterozoic, with 33%–36% of single-grain ages within the 1820–1600 Ma subdistribution (D4a, see Fig. 4; Table 3). The samples also contain a high proportion of 540–0 Ma (D1) single-grain ages, accounting for 31%–40% of the overall sample distributions (Fig. 5; Table 3). The most prominent Phanerozoic age peaks are within the 85–55 Ma (D1c) and the 120–85 Ma (D1d) subdistributions, although there are also minor age peaks at 184 and 436 Ma (Fig. 5). Although 1550–950 Ma grains (D3a and D3b) account for no more than 10%–21% of the overall age distributions, one key contrast between the two Wilcox Group samples is that CG11-10 has several robust Mesoproterozoic age peaks (within D3a and D3b in Fig. 4), whereas CG11-3 has only one, which is defined by only 3 overlapping single-grain ages. No more than 2%–7% of single grains are within the 770–540 Ma (D2), the 2050–1820 Ma (D4b), or the 3050–2450 Ma (D5) distributions (Fig. 4; Table 3). There are very few robust age probability peaks in these (sub)distributions, but CG11-3 has age probability peaks at 2517 and 2724 Ma within the Archean (D5). At a 95% confidence level, a difference between the parent age populations for these two samples cannot be resolved (see Supplemental File [footnote 1]).

The two most abundant Wilcox Group composite chronofacies sample (sub)distributions, D4a and D1, account for an estimated 35% ± 6% and 34% ± 6% of the total zircon population, respectively (2σ uncertainties from Equation 2; Figs. 7 and 8; Table 4). We estimate that D3a accounts for 10% ± 4% of the total Wilcox Group zircon population.

Claiborne Group Samples

In contrast to the Wilcox Group, the largest group of zircons in each of the three Claiborne Group samples (CG11-4, CG11-6, and CG11-7) is within the 1300–950 Ma subdistribution (D3a); these grains represent 38%–46% of each sample (Fig. 4; Table 3). In addition, 6%–13% of single-grain ages from these samples are within the 1550–1350 Ma subdistribution (D3b) (Fig. 4; Table 3). Although the Claiborne Group samples exhibit moderate age peaks in the late Paleoproterozoic (D4a) and the Phanerozoic (D1), the abundance of single-grain ages within these age ranges is diminished relative to the Wilcox Group samples (Figs. 4 and 5; Table 3). Grains in any one of these age ranges (D4a, D1) represent 13%–19% of the overall sample distributions, compared to 31%–40% for equivalent age grains the Wilcox Group samples (Fig. 4; Table 3). Within the 540–0 Ma distribution (D1), Claiborne Group samples have a relatively high proportion of Paleozoic grains, as opposed to Cenozoic and Mesozoic grains (see Fig. 5), and samples CG11-4 and CG11-7 exhibit several age peaks between 467 and 406 Ma (Fig. 5). Like the Wilcox Group samples, (sub)distributions D2, D4b, and D5 are a minor fraction of the sample distributions, although each sample exhibits a minor Archean (D5) age probability peak, ranging from 2733 to 2595 Ma. There is no resolvable difference between the parent populations that sourced the three Claiborne Group samples (see Supplemental File [footnote 1]).

The most abundant (sub)distribution in the Claiborne Group composite chronofacies, D3a, accounts for an estimated 42% ± 5% of the total detrital zircon population (Fig. 7; Table 4). D1 and D4a account for a high proportion of the Wilcox Group and account for an estimated 19% ± 4% and 14% ± 3%, respectively, of the Claiborne Group (Figs. 7 and 8; Table 4).

Catahoula Formation Samples

Similar to the Wilcox Group, the two Catahoula Formation samples (CG11-2 and CG11-5) have high proportions of 540–0 Ma (D1) and 2050–1820 Ma (D4a) single-grain ages, with grains in any one of these (sub)distributions representing 22%–39% of the total sample distributions (Table 3). The most prominent Phanerozoic age peaks within these samples date to the Cenozoic and the Late Cretaceous, within subdistributions D1a, D1b, and D1c (Fig. 5; Table 3). K-S tests indicate that these samples were drawn from different parent populations (see Supplemental File [footnote 1]), and one of the primary differences between the two samples appears to be that 28% of single-grain ages from sample CG11-5 are within the 1300–950 Ma subdistribution (D3a) compared to only 9% of single-grain ages from CG11-2 (Fig. 4; Table 3). Generally, small proportions of single-grain ages are within the other subdistributions (Table 3).

The most abundant distributions in the Catahoula Formation composite chronofacies are D1, D4a, and D3a, which we estimate to account for 31% ± 6%, 27% ± 6%, and 19% ± 5%, respectively, of the overall population (Figs. 7 and 8; Table 4). Given the apparent dissimilarities in sediment sources among the Catahoula Formation samples, the chronofacies may not be accurately characterized.

Fleming Formation Samples

Similar to the Wilcox Group and Catahoula Formation samples, the overall Fleming Formation sample (CG11-1, CG11-8, and CG11-9) distributions are dominated by Phanerozoic (D1) and late Paleoproterozoic (D4a) zircons, with grains in any one of these age ranges generally representing 18%–44% of the total sample distribution (Figs. 4 and 5; Table 3). Sample CG11-1 appears to be an exception to this generalization, in that only 6% of the single-grain ages from this sample are within the 1820–1600 Ma subdistribution (D4a). The 540–0 Ma age distributions of these samples are characterized by peaks that are scattered through the Cenozoic and Mesozoic (D1a–D1e) (Fig. 5; Table 3). The samples contain smaller numbers of zircons that are within other age distributions. K-S tests indicate that the three Fleming Formation samples were not drawn from the same parent distribution.

The most abundant distributions in the Fleming Formation composite chronofacies are D1, D4a, and D3a, and we estimate that these account for 41% ± 5%, 17% ± 4%, and 16% ± 4%, respectively, of the overall zircon populations (Figs. 7 and 8; Table 4). We are less confident that we have accurately characterized the coastal plain chronofacies because of the apparent dissimilarity in sediment source zircon populations among the samples (as for the Catahoula Group), and we discuss this in more detail in the following.

For possible sediment source regions, we compiled 596–3107 detrital zircon U-Pb ages, collected at 7–34 sample sites, from multiple geologic formations, distributed across large swaths of sedimentary basins covering areas of ∼10,000–100,000 km2 (Figs. 7 and 8; Eriksson et al., 2004; Dickinson and Gehrels, 2008; Park et al., 2010; Fan et al., 2011; Fuentes et al., 2011; May et al., 2013). For the source area compilations, the Rocky Mountain region was divided into a southern part and a northern part, with the Cheyenne belt defining the boundary (Fig. 3). Although there are several geologic units that crop out around the eastern Rocky Mountains to the south of the Cheyenne belt, the most widespread are the Upper Cretaceous strata of the Sevier foreland basin. Our aggregate from these strata includes 596 single-grain ages collected at 7 sample sites in New Mexico, Utah, and Arizona (Dickinson and Gehrels, 2008). In general, the individual samples consist of heterogeneous mixes of detrital zircons, although the dominant age distributions are similar among the sites (Dickinson and Gehrels, 2008). We compiled 3107 single-grain age measurements from 34 samples sites located to the north of the Cheyenne belt, from geographically widespread Upper Cretaceous (Sevier foreland basin) and Paleogene (Laramide intramontane basins) strata. The individual samples contain heterogeneous mixes of zircons, and the abundance of a prominent and nearly ubiquitous ca. 100–95 Ma zircon distribution changes significantly among the samples (Fan et al., 2011; Fuentes et al., 2011; May et al., 2013). In general, however, the dominant age distributions are similar among the sample sites. The Appalachian Basin detrital zircon compilation includes 2276 single-grain ages, collected at 26 sample sites that span the Paleozoic strata and latest Neoproterozoic strata that now crop out around the basin (Eriksson et al., 2004; Park et al., 2010). The individual samples are similar among the sites and dominated by only a few age distributions, particularly the 1300–950 Ma distribution, equivalent in age to the D3a subdistribution described herein, and believed to be derived from underlying basement rocks of the Grenville Province (Eriksson et al., 2004; Park et al., 2010).

For modern Mississippi River sand, the optimal combination of the three end-member sources we defined includes a 25% contribution of sediment from Appalachian Basin Paleozoic strata, and a 54% and 21% contribution from the southern Sevier basin Upper Cretaceous and the northern Sevier and Laramide basins Upper Cretaceous and Paleogene strata, respectively (Fig. 9). For the composite Wilcox Group, Catahoula Formation, and Fleming Formation, synthetic mixes of detrital zircon sources that contain 0%–3% Appalachian Basin sediment, and ≥80% southern Sevier foreland basin Upper Cretaceous sediment best match the coastal plain chronofacies (Fig. 9). In contrast to the other coastal plain composite chronofacies, the Claiborne Group is best matched by sediment source mixes that involve 57% of recycled Appalachian Basin sediment, and relatively low proportions of Upper Cretaceous and Paleogene sediment from the Sevier and Laramide basin region. We consider the 2σ uncertainties on these calculations to be large, perhaps 20% or larger, given the assumption of uniform zircon concentrations in sediment sources and uncertainties on the various values of xi that are subsumed into these calculations.

Links Between Chronofacies and Heavy Mineral Assemblages

The composite detrital zircon chronofacies in central Louisiana show an intriguing correlation to regional shifts in heavy mineral assemblages, particularly to those preserved in the neighboring Houston embayment (Fig. 1; Todd and Folk, 1957; McCarley, 1981, and references therein). The abundances of the three largest zircon age (sub)distributions, D1, D3a, and D4a, are unique in the Claiborne Group compared to all other composite samples. At a 95% confidence level, the abundance of 1300–950 Ma subdistribution D3a is higher in the Claiborne Group than in any other Louisiana coastal plain chronofacies. Moreover, the Claiborne Group contains fewer 540–0 Ma zircons (D1) than any other chronofacies at a 95% confidence level, and within the 540–0 Ma distribution, a relatively large proportion are Paleozoic (D1f) in the Claiborne Group (Table 4). The Claiborne Group contains fewer 1820–1600 Ma (D4a) grains than any other sample, although not at a 95% confidence level when compared to the Fleming Formation or the modern Mississippi River (Table 4). Thus, the Claiborne Group appears to be a unique coastal plain chronofacies.

Although four key heavy minerals, garnet, epidote, kyanite, and staurolite, appear to be present in some proportion in most of the Cenozoic strata around the northern Gulf of Mexico coastal plain, in the Houston embayment (see Fig. 1) there is an upsection change from garnet- and epidote- to kyanite- and staurolite-enriched assemblages between the Wilcox Group and the Claiborne Group. Later Tertiary strata tend to exhibit garnet- and/or epidote-enriched heavy mineral assemblages, similar to the underlying Wilcox Group. Thus, like the Louisiana detrital zircon chronofacies stratigraphy, in the Houston embayment, the Wilcox Group differs from the Claiborne Group and resembles later Tertiary strata. Several observations also suggest that the Houston embayment heavy mineral assemblage trends may be geographically widespread in the northern Gulf of Mexico coastal plain, extending across multiple depocenters. In particular, kyanite- and staurolite-dominated heavy mineral assemblages occur in Claiborne Group strata in northwest Louisiana (Todd and Folk, 1957) and as far to the east as Mississippi (Grimm, 1936). Garnet and/or epidote tend to be the dominant heavy minerals in later Tertiary strata in central Louisiana (Cogen, 1940). The correlation between the chronofacies and heavy mineral data shows that Claiborne Group chronofacies differ significantly from Wilcox Group and younger Cenozoic chronofacies (see Figs. 4 and 7; Table 4).

Provenance of Modern Mississippi River Sediment

The areas considered to be likely sediment sources for the Louisiana coastal plain during the Cenozoic are all encompassed by the modern catchment, and the sediment mixing calculation suggests that the modern river sediment is a mix of material derived from Appalachian Basin Paleozoic strata and Upper Cretaceous and Paleogene strata from the Sevier and Laramide basins of the Rocky Mountains (Fig. 9). The relative prevalence of Sevier and Laramide basin sediment in the modern Mississippi River is not surprising, given that the Mississippi drains a large swath of the Sevier-Laramide region, and integrated across the landscape this region likely undergoes more rapid erosion than the Appalachian region (or any other area integrated into the modern Mississippi River catchment) (Granger et al., 2001; Matmon et al., 2003; McMillan et al., 2006; Hancock and Kirwan, 2007; Saylor et al., 2013).

Intuitively, the results of the mixing calculation must reflect prominent age (sub)distributions shared by the modern river sediment and various source regions, including (1) subdistributions D1e (250–145 Ma) and D4a (1820–1600 Ma), which are particularly abundant in the Upper Cretaceous Sevier foreland basin strata of the southern Rocky Mountains, and have been interpreted to be derived from the Cordilleran magmatic arc and the crystalline basement rocks of the Yavapai-Mazatzal provinces, respectively (Dickinson and Gehrels, 2008); (2) subdistribution D3a (1300–950 Ma), which is particularly abundant in Appalachian Basin Paleozoic strata, and has been interpreted to be derived from the underlying Grenville Province (Eriksson et al., 2004; Park et al., 2010); and (3) subdistributions D1c and D1d (120–55 Ma), which are particularly abundant in the Upper Cretaceous and Paleogene Sevier and Laramide basin strata in the northern Rocky Mountains, and have been interpreted to be derived from the Cordilleran magmatic arc (Fan et al., 2011; Fuentes et al., 2011; May et al., 2013).

However, the various diagnostic subdistributions cited here are found in multiple source areas (e.g., D3a, D4a), and the most abundant age (sub)distributions in the modern river sediment almost certainly reflect partial contributions from multiple sources. The northern and southern Sevier-Laramide detrital zircon chronofacies (Figs. 7 and 8) are very similar, and we also caution that the relative proportions of these sediment sources revealed by our mixing calculations could be particularly sensitive to small changes in the abundances of a few key age distributions in the two Sevier-Laramide basin chronofacies compilations. Nevertheless, it seems possible to identify major sediment sources in the southern versus northern Rocky Mountains, and to differentiate these from eastern North America sources. We propose a few age distributions that could be diagnostic of the various source areas.

Grains that date to 250–0 Ma (equivalent to D1a–D1e) and 1820–1600 Ma (D4a) are absent in the Appalachian Basin and present across the Sevier-Laramide basin region, where they are interpreted to derive from the Cordilleran arc (D1a–D1e) and the underlying crystalline basement rocks of the Yavapai-Mazatzal provinces (D4a) (Fig. 3; e.g., Dickinson and Gehrels, 2008); they are therefore diagnostic of sediment sourcing from the Sevier-Laramide basin region. Although 1300–950 Ma grains (D3a in coastal plain chronofacies) interpreted to derive from the Grenvillian rocks in the Appalachian Mountains are commonly found in all of the likely sedimentary sources for the Gulf of Mexico coastal plain (e.g., Eriksson et al., 2004; Dickinson and Gehrels, 2008; Fan et al., 2011), they are by far most abundant in the Appalachian Basin (64% versus 10%–23% in Sevier and Laramide basins), such that abundances of 1300–950 Ma zircons in excess of ∼23% may require some sediment sourcing from the Appalachian basin.

Within the 540–0 Ma (D1) and 1820–1600 Ma (D4a) (sub)distributions that are unique to the Sevier-Laramide basin region, the chronofacies compilations suggest several north to south differences that can be understood in the context of the regional geology. First, zircons dating to 100–95 Ma (equivalent to D1d) appear to be particularly prevalent in the northern Sevier-Laramide basin region relative to equivalent strata to the south (Fig. 8), although not in all basins (see Wind River Basin data in Fan et al., 2011). Second, although they have been identified in minor amounts in Laramide basin fill in Wyoming (e.g., Fan et al., 2011), zircons that date to the Jurassic and Triassic (equivalent to D1e) appear to be most prevalent in the southern Rockies, which is somewhat intuitive given that Cordilleran arc magmatism was confined to Mexico and southern California in the Triassic and generally focused on (but not confined to) the southern Cordilleran arc during the Jurassic (Armstrong and Ward, 1993; Torres et al., 1999; Fan et al., 2011). Third, although grains derived from the Yavapai and Mazatzal basement provinces appear to be common in both the southern and northern Sevier-Laramide region, and although the broad age probability peaks in our detrital zircon compilations from the two regions overlap (Fig. 7), 37% +13%/–15% of grains we assign to this subdistribution are younger than 1720 Ma in the northern Sevier-Laramide basin composite, whereas 53% +16%/–29% of grains we assign to this subdistribution are younger than 1720 Ma in the southern Sevier basin composite (Fig. 7), with uncertainties being determined based on 2 standard errors on the single-grain age. Although these abundances overlap within uncertainties, the relatively high proportion of Yavapai province–equivalent grains in the northern Sevier-Laramide basins (Fig. 3) and the relatively high proportion of Mazatzal province–equivalent grains in the southern Sevier basin (Fig. 3) is sensible given the north-south gradient in basement ages in the region (Fig. 3). Archean zircons are relatively rare in the Sevier and Laramide basins in the northern Rocky Mountains (north of the Cheyenne belt), as well as in modern Mississippi River sediment, despite the fact that the modern catchment extends northward to Montana, Wyoming, North Dakota, South Dakota, Minnesota, and Wisconsin, areas that are primarily underlain by Archean crystalline basement rock (Figs. 3 and 7). The modern data suggest that the rarity of Archean zircons in ancient stratal packages need not preclude drainage of Archean basement terranes in the northern United States.

Some Cenozoic zircons in modern river sediment may reflect volcanic airfall input into the Mississippi catchment, so that their exact provenance significance is difficult to unravel (e.g., Ruppert et al., 1994; Warwick et al., 1996, 1997; Guillemette and Yancey, 1996). Airfall input is a particularly likely source for zircons within the Cenozoic subdistribution that dates from 40 to 25 Ma and overlaps in age with the mid-Tertiary ignimbrite flare-up (e.g., Armstrong and Ward, 1991).

Provenance of Fleming and Catahoula Formations

Detrital zircon chronofacies from the Fleming and Catahoula Formations are similar, and appear to permit at least two drainage basin reconstructions. The first involves integrated drainage from the Sevier-Laramide basin region to the Appalachians, similar to the modern catchment, and the second involves the coalescing of distinct western (originating in the Sevier-Laramide basin region) and eastern (originating in the Appalachian Basin region) depositional systems, along the Louisiana-Mississippi border region.

In support of the first alternative, the three major age distributions (D1, D3a, D4a) within the modern chronofacies are similar to the Fleming and Catahoula Formations chronofacies, if not overlapping at a 95% confidence level (Table 4). The primary difference between the three chronofacies appears to be shifts in the youngest age probability peaks that correspond to the differing depositional ages for the three units (Fig. 8). In support of the second alternative, we reiterate that K-S statistical tests indicate that the individual samples that compose the Fleming and Catahoula Formation composites were drawn from parent populations in different host sedimentary rocks (see Supplemental File [footnote 1]). Assuming that this sample to sample variability does not simply reflect poor mixing of zircons in fluvial deposits (e.g., DeGraaff-Surpless et al., 2003; Lawrence et al., 2011; cf. Saylor et al., 2013), our sampling transect may actually span the divide between the easterly and westerly sediment dispersal systems. The two Oligocene Catahoula Formation samples, CG11-2 and CG11-5 (Fig. 1), are separated by an east-west distance of ∼200 km; although individual samples are small (n = 102, 109, respectively) and have relatively limited representativity (see Andersen, 2005), the eastern site (CG11-5) exhibits a relatively low proportion of zircons in D4a (22% versus 33% in CG11-2), and subdistribution D4a is apparently a hallmark of sediments derived from the Sevier-Laramide basin region. Furthermore, the eastern site (CG11-5) exhibits a relatively high proportion of zircons in D3a (28% versus 9%), which is by far the largest component of Appalachian Basin Paleozoic sediment (Figs. 3 and 7; Table 4). The easternmost Fleming Formation sample CG11-1 exhibits a similar trade-off between the abundances of zircons in subdistributions D3a and D4a relative to the two Fleming Formation samples (CG11-8, CG11-9) collected >100 km to the west.

For either alternative, the detrital zircon data do not clearly define the north-south dimensions of the catchment, and subdistributions discussed here are broadly similar among the modern and ancient chronofacies. We somewhat arbitrarily define the southern catchment limit based on the modern catchment and extend it northward to the crustal boundary between Archean cratons and Paleoproterozoic provinces in the northern midcontinent, noting that the catchment may have extended farther to the north (Figs. 3 and 10).

Reconstructions by Galloway et al. (2011) show that from the Middle Miocene to the Pleistocene, the Mississippi River drained portions of the Sevier-Laramide basin region, whereas an ancestral Tennessee River drained the Appalachian Basin region. These reconstructions indicate the presence of a depositional divide near the Louisiana-Mississippi border, nearly overlapping with the region where we observe changing detrital zircon sample distributions in both Miocene and Oligocene stratal packages (e.g., Fig. 4; Table 3), and this is broadly similar to the second alternative that we have proposed. This interpretation is supported by the fact that Galloway et al. (2011) noted the presence of distinct fluvial input axes along the updip margins of the coastal plain. In contrast, in Galloway et al. (2011) the Oligocene and Early Miocene paleo–Mississippi River is shown to integrate drainage of the northern Sevier-Laramide basin region and the Appalachian Basin, similar to our first alternative.

Regarding the Fleming and Catahoula Formation samples, it is also plausible that a large proportion of the later Tertiary sediment budget is dominated by recycling of coastal plain strata to the north (Fig. 1). The Oligocene Catahoula Formation sample CG11-5 cannot be discriminated from the Middle Eocene Claiborne Group strata using the K-S test (see Supplemental File [footnote 1]), and some of the other Catahoula and Fleming Formations age spectra cannot be discriminated from the Wilcox Group and/or the Claiborne Group using the K-S test (see Supplemental File [footnote 1]). In support of the idea of significant sediment recycling on the coastal plain, the large difference between minimum zircon age and depositional age for Fleming Formation samples CG11-1 and CG11-8 (Fig. 6; Table 2) suggests that the samples may contain a large proportion of material derived from older coastal plain stratal packages.

Provenance of the Claiborne Group

The high abundance of zircons in the 1300–950 Ma subdistribution (D3a) in the Claiborne Group, versus zircons in the 540–0 Ma (D1) and 1820–1600 Ma (D4a) (sub)distributions (Fig. 7), is striking because the pattern is recorded in multiple formations, at multiple sample sites, and the unique chronofacies appears to be correlated to a unique kyanite-staurolite heavy mineral assemblage, and the group was deposited during a time of anomalously low sediment influx rates across the northern Gulf of Mexico coastal plain (Figs. 2 and 7; Grimm, 1936; Todd and Folk, 1957; McCarley, 1981; Galloway, 2008; Galloway et al., 2011). The only possible sediment source that has a high abundance of zircons in D3a in association with relatively small D1 and D4a (sub)distributions, as well as abundant pelitic metamorphic minerals, appears to be the Appalachian-Ouachita orogenic belt and the associated foreland basins (Fig. 7; Todd and Folk, 1957; McCarley, 1981; Gleason et al., 2002; Park et al., 2010). This suggests that most D3a zircons in the Claiborne Group were ultimately derived from the Grenville Province and recycled through Appalachian Basin fill (e.g., Eriksson et al., 2004). However, the Claiborne Group contains 11%–17% of 1820–1600 Ma zircons (D4a) (Table 4), requiring some sediment input (including some D3a zircons) from the Sevier-Laramide basin region.

We envisage that during Claiborne Group deposition, the paleo–Mississippi River catchment extended from the Sevier-Laramide basin region to the Appalachian Basin region (Fig. 10), similar to the east-west dimensions of the modern Mississippi River catchment. It is difficult to define the north-south dimensions of this catchment, and we do so in a fashion similar to that for the Fleming and Catahoula Formations reconstruction (Fig. 10). Although the northern and western limits of our reconstruction are similar to other recent Middle Eocene drainage reconstructions (Galloway et al., 2011), the eastern limit differs starkly from recent studies, which suggest that the eastern half of the continent was not integrated in the Middle Eocene Mississippi River catchment and that the Appalachians were not a source of sediment to the Middle Eocene Mississippi Delta (Galloway et al., 2011).

Provenance of the Wilcox Group

The prominence of zircons dating to 1820–1600 Ma (D4a, 35% ± 6%), and 250–55 Ma (D1c, D1d, D1e, each ∼10%) almost certainly indicates that the bulk of the Louisiana Wilcox Group derived from the Rocky Mountain region (Table 4). Notably, the low proportion of Grenville Province–equivalent (D3a) zircon ages (10% ± 4%) in the Louisiana Wilcox Group appears to be significantly lower than any of the other coastal plain chronofacies (Table 4), lower than the southern Sevier foreland basin (23%), and similar to the northern Sevier-Laramide basins (10%) (see Fig. 7; Dickinson and Gehrels, 2008; Fan et al., 2011; Fuentes et al., 2011; May et al., 2013). A similar observation has been made about the Wilcox Group in south Texas, which contains only 7% Grenville Province–equivalent (D3a) zircons (Fig. 7; Mackey et al., 2012). The relatively low proportion of Grenville Province–equivalent (D3a) zircons in the south Texas Wilcox Group has been interpreted to reflect sediment sourcing primarily from fault-bound basement uplifts of the Laramide orogenic belt (Fig. 2; Mackey et al., 2012), rather than recycling of foreland basin material. This interpretation contradicts our assertion that most sediment on the Gulf of Mexico coastal plain has been recycled through older sedimentary cover, but it appears to be permissible, if not likely, for the Wilcox Group in Louisiana.

Contrasts in the Yavapai-Mazatzal provinces–equivalent age subdistribution (D4a) and Cordilleran arc–equivalent subdistribution (D1e) between the south Texas and Louisiana Wilcox Group appear to reveal distinct source regions (Figs. 7, 8, and 10). Only 39% +25%/–13% of the zircons we assigned to D4a in the Louisiana Wilcox Group are younger than 1720 Ma, whereas 64% +10%/–15% of D4a zircons in the south Texas Wilcox Group are younger than 1720 Ma (uncertainties on abundances reflect 2 standard errors for single-grain ages) (see Fig. 7). That is, the Louisiana Wilcox Group appears to contain a relatively high abundance of zircons with ages equivalent to Yavapai province basement rocks, rather than the Mazatzal province basement rocks (Figs. 3 and 10), although the abundances overlap, given 2 standard errors on the single-grain age measurements. Provided that the along-strike changes in the D4a subdistribution are robust, then it appears that the Louisiana Wilcox Group received more material that was ultimately derived from the older, northern parts of the Yavapai-Mazatzal provinces, located in Colorado and northern New Mexico, or possibly regions north of the Cheyenne belt. In terms of subdistribution D1e, we estimate that 18% ± 3% of grains in the Texas Wilcox Group date to the Triassic–Jurassic (D1e), in comparison to 8% ± 3% of grains in the Louisiana Wilcox Group (2σ uncertainties from Equation 2), suggesting that the Louisiana Wilcox Group sourced more northern parts of the Sevier-Laramide basin region.

From the detrital zircon data alone, no clear evidence limits the northwestern extent of the paleo–Mississippi catchment during Wilcox Group deposition. However, paleogeographic studies of the northern Laramide region indicate the presence of a shallow sea (i.e., the Cannonball embayment) across south-central North Dakota throughout the Paleocene and Early Eocene that was fringed by fluvial-deltaic systems originating in northwestern Wyoming (Chevren and Jacobs, 1985). Thus, central Wyoming may be a reasonable northwestern limit for a paleo–Mississippi River catchment (Fig. 10). Given the apparent need for the Louisiana Wilcox Group to be sourced by part of the Sevier-Laramide basin region, we tentatively define the southwestern catchment limit near the southern limit of the Yavapai province, such that the far western part of the catchment integrates portions of southern Wyoming and Colorado (Fig. 10). In light of the clear evidence for integration of the Appalachian region into the Mississippi drainage during the Middle Eocene, we suggest that the most reasonable reconstruction extends eastward into the Appalachian Basin region (Fig. 10), although we accept that a catchment confined to the western midcontinental interior is also a permissible interpretation.

The preferred reconstruction is very similar to at least one recent Late Paleocene drainage reconstruction, but differs from an Early Eocene drainage reconstruction for the North American continental interior (Galloway et al., 2011). In the Early Eocene reconstruction, the pronounced decrease in sediment flux to the Mississippi delta region during the middle of Wilcox Group deposition (Fig. 2) is thought to reflect integration of the Laramide region into a Houston embayment drainage network (Galloway et al., 2011; Fig. 1). In this interpretation, Early Eocene Wilcox Group sediment in the Mississippi delta is thought to be derived from relict high topography in the Ouachita region (Fig. 3) and possibly from recycling of lower Wilcox Group sediment during Paleogene uplift of the Sabine Arch (Fig. 1) (Galloway et al., 2011). However, the scenario of drainage of the Ouachita region during Early Eocene Wilcox Group deposition does not appear to be consistent with the detrital zircon sample age distribution in the upper Wilcox Group sample (CG11-3, see Figs. 2, 4, and 5). With the caveat that n = 101, the upper Wilcox Group sample contains the highest abundance (33%) of D4a zircons of any single sample in this study, zircons that appear to be a hallmark of sediment derived from the Sevier-Laramide basin region (Table 3). Moreover, the upper Wilcox Group sample contains the lowest abundance of D3a (6%) zircons of any single sample in this study (Table 3), and the available detrital zircon data from the Ouachita region suggest that the strata contain a vast majority of zircons that are within the D3a subdistribution (Gleason et al., 2002), similar to the Appalachian Basin (Fig. 7). It is more difficult to discount the possibility that upper Wilcox Group sediments are exclusively recycled lower Wilcox Group material from the Sabine Arch or other parts of the coastal plain, given the similarity between the two individual sample age distributions and the detection limits in our data set.

Our reconstruction also differs from a recent reconstruction for the Wilcox Group of southern Texas that extends that drainage network northward to the Cheyenne belt of southern Wyoming, to a region that overlaps with our Mississippi River catchment reconstruction (Figs. 3 and 10; Mackey et al., 2012). However, that reconstruction largely reflected high abundances of zircons interpreted to derive from the Yavapai-Mazatzal provinces (i.e., subdistribution D4a), and the Wilcox Group of Louisiana appears to be characterized by similarly high abundances of zircons ultimately derived from (northerly?) parts of the composite Yavapai-Mazatzal province. In light of this new observation and the paleogeographic constraint imposed by the Cannonball embayment, which integrated drainage of central Wyoming, we interpret that the Wilcox Group in both Texas and Louisiana must have been sourced by portions of the Sevier-Laramide region that overlap with the Yavapai and Mazatzal provinces (Fig. 10).

Another key difference between the south Texas and Louisiana Wilcox Group is the presence of a 150–114 Ma age distribution that has been used to link the south Texas Wilcox Group to a sediment source in the northeastern Mexican Laramide foreland (Figs. 3 and 8; Lawton et al., 2009; Mackey et al., 2012). This sample distribution accounts for ∼7% of the south Texas Wilcox Group, but it is absent in Louisiana. For the Louisiana Wilcox Group chronofacies, which consists of n = 252 grains, we estimate (from Equation 1) a 95% probability of detecting a population with a ∼1% abundance, and therefore even higher probabilities of populations with greater abundances. In light of this detection limit, it seems safe to assert that 150–114 Ma zircons could account for no more than a trace amount of the Louisiana Wilcox Group in comparison to equivalent strata in south Texas. This 150–114 Ma distribution further suggests that differences in Wilcox Group provenance can be resolved along the strike of the Gulf of Mexico coastal plain, and may help to discriminate between Wilcox Group stratal packages, particularly in the deep-water Gulf of Mexico, where a thick allochthonous salt canopy obscures reconstructions of regional sediment dispersal patterns based on seismic reflection data (e.g., McDonnell et al., 2008).

Integrating Sediment Provenance and Flux Rates

Although quantification of sediment export rates to the continental margin from geologic elements across the continental interior awaits additional provenance analysis from Gulf Coast depocenters, analysis that will refine and augment the catchment reconstructions presented herein, we combine our provenance analysis with various measurements of sediment flux to the coastal plain to discuss this issue qualitatively (Fig. 2; Galloway et al., 2011). The key temporal trend revealed by the Cenozoic Mississippi delta detrital zircon age spectra is that depositional episodes characterized by high sediment supply (i.e., all but the Claiborne Group) also corresponded to high proportions of sediment sourcing from the Sevier-Laramide basins or perhaps from underlying basement rocks. With possible exceptions during the Oligocene and/or Miocene, the paleo–Mississippi River catchment appears to have been largely stable, at least in the east-west dimension, throughout the Cenozoic, and pronounced increases and/or decreases in sediment influx do not appear to have corresponded to increases and/or decreases in catchment area (Figs. 2 and 10).

Our preferred explanation for the unique chronofacies and heavy mineral composition of the Middle Eocene Claiborne Group is that low volumes of sediment were exported from the Sevier-Laramide region during the Eocene, despite the fact that the region was integrated into the paleo–Mississippi River catchment. It is important to note that such a reconstruction does not require any temporal changes in sediment export rates from the Appalachian region during the Cenozoic (Fig. 10), simply an ∼10 m.y. hiatus in major sediment export from the Sevier-Laramide basin region. We suggest that there are at least two likely explanations for low sediment export from the Sevier-Laramide basin region during the Middle Eocene that are not mutually exclusive. First, regional stratigraphic analysis shows the presence of a large network of lakes across the central Laramide foreland in the Eocene (e.g., Dickinson et al., 1988; Smith et al., 2008). Strontium and oxygen isotope stratigraphy suggests that many of the vast lakes in Utah, Colorado, and southern Wyoming were internally drained at times during the Eocene (Davis et al., 2008, 2009a, 2009b; Chamberlain et al., 2012). This large network of periodically closed lake basins that extended across parts of Utah, Colorado, and Wyoming may have sequestered most Sevier-Laramide basin region sediment during the Middle Eocene. Alternatively, the Middle Eocene may have been a time of slow erosion and sediment production in the Sevier-Laramide basin region. In support of this idea, conglomerate clast composition has been used to argue that the aforementioned lakes were generally sediment starved in the Eocene relative to the Paleocene (Carroll et al., 2006). Either way, an extensive paleosol developed across the Great Plains during the Eocene, suggesting that the Middle Eocene was a time of low aggradation/erosion, and possibly low sediment throughput in the Great Plains region. It is important to note that the aforementioned Eocene lakes generally occupy the same western headwater region that we inferred for the Louisiana Wilcox and Claiborne Groups (Fig. 10); this overlap reinforces our paleocatchment reconstructions.

Analysis of detrital zircon age spectra from Tertiary strata deposited within the Mississippi River delta in Louisiana provides the basis for the following conclusions.

1. Detrital zircon ages within Cenozoic strata of the Louisiana coastal plain are diverse, ranging from the Archean to the Cenozoic. However, nearly all of zircon ages are within five broad age distributions (a few of which can be subdivided), i.e., 540–0 Ma, 770–540 Ma, 1550–950 Ma, 2050–1600 Ma, and 3050–2450 Ma. These (sub)distributions can be correlated to the various crystalline basement rocks that underlie the North American continental interior.

2. Throughout the Tertiary Mississippi delta plain stratigraphy, shifts in detrital zircon chronofacies appear to be correlated to previously defined shifts in heavy mineral composition around the northern Gulf of Mexico coastal plain. Heavy mineral assemblages and zircon age spectra in the Middle Eocene Claiborne Group are unique, in comparison to overlying and underlying Cenozoic strata.

3. Sediment source mixing calculations suggest that most (∼75%) modern Mississippi River sediment is sourced from the Sevier-Laramide region, rather than source areas to the east (i.e., the Appalachian region), in agreement with numerous studies of landscape evolution rates across the continental interior. A compilation of detrital zircon data from strata exposed in modern sediment source regions shows that several regions nested within the modern catchment have unique detrital zircon age distributions and/or abundances that may be detected in coastal plain chronofacies.

4. Detrital zircon age spectra in both Miocene and Oligocene stratal packages on the Louisiana coastal plain are broadly similar to those in modern river sediment and to those in likely sediment source regions in the Rocky Mountains. Integrated drainage of the continental interior from the Sevier-Laramide region to the Appalachian region or the sediment delivery to the Louisiana coastal plain by distinct sediment dispersal systems originating in the Sevier-Laramide basin region and the Appalachian Basin region appear to be equally plausible drainage reconstructions.

5. A high proportion (perhaps ∼57%, uncertainties discussed in text) of Middle Eocene Claiborne Group sediment on the Louisiana coastal plain derives from eastern sediment sources in the Appalachian Basin region. This is consistent with the heavy mineral composition of the Claiborne Group, which has abundant pelitic metamorphic minerals that are exposed in rocks in the Appalachian Mountains and foreland basin. Despite the unique chronofacies in the Claiborne Group, there is also evidence for some sediment input from the Sevier-Laramide basin region, and the most plausible paleocatchment configuration appears to extend from the Appalachians to the Rockies, similar to the east-west dimensions of the modern catchment.

6. Nearly all of the detrital zircons in the Upper Paleocene–Lower Eocene Wilcox Group appear to be derived from sedimentary basins in the Sevier-Laramide region, or perhaps the underlying basement rock in the Yavapai and Mazatzal provinces of the interior western United States. Similar to the Claiborne Group, the most plausible paleo–Mississippi River catchment configuration during central Louisiana Wilcox Group deposition seems to be one that resembles the modern catchment, although there is no definitive evidence for integration of the Appalachian Basin into this drainage system provided by the detrital zircon data.

7. Cenozoic depositional episodes on the Louisiana coastal plain that were characterized by high sediment supply (i.e., all but the Middle Eocene) also corresponded to high proportions (perhaps ≥75%) of sediment sourcing from either Sevier or Laramide basins in the Rocky Mountain region, or perhaps from underlying basement rocks. Anomalously low sediment supply to the coastal plain from the Rocky Mountains during deposition of the Middle Eocene Claiborne Group may have been related to ponding of Rocky Mountain sediment in a regional network of periodically closed lake basins, and/or low rates of erosion and/or sediment production across the Rocky Mountains during the Middle Eocene.

Craddock thanks Richard McCulloh and Paul Heinrich of the Louisiana Geological Survey for assistance in locating surface outcrops, and Sharon Swanson and Brenda Pierce for supporting this research. Kathleen Spiegelberg provided valuable assistance with field-work logistics. Jake Covault helped initiate this research effort. Reviews by Cristian Carvajal and Peter Warwick improved this manuscript. An exceptionally constructive and challenging review from Todd LaMaskin was particularly helpful. We thank Willy Amidon for access to his sediment mixing models.


In order to explore the degree to which age measurements might be affected by the presence of initial Pb, we plotted the number of mass 204 counts per second (cps) from the mass spectrometer against concordance, calculated from the ratio of the 206Pb/238U age to the 207Pb/235U age. Generally, grains contain minor amounts of mass 204 (<∼70 cps). Grains with <∼70 cps of mass 204 tend to be concordant, whereas grains with high concentrations tend to be slightly discordant. Several grains yielded negative mass 204 cps values, indicating that in some instances measured mass 204 cps was less than background mass 204 values. The most negative values of mass 204 cps are ∼–60 ± 10 cps. Collectively, these observations indicate that grains with 0 ± 70 cps of mass 204 have no measureable initial Pb to within the precision of our measurements. Single-grain measurements with mass 204 of 70 cps or more were rejected. Typically, this resulted in filtering of 0–8 grains per sample that would otherwise satisfy our discordance filter. This approach to dealing with initial Pb has the advantage of avoiding potential interference from 204Hg when making initial Pb corrections based on very low concentrations of mass 204.

1Supplemental File. PDF file containing analytical data for all U-Pb and Pb-Pb geochronology, concordia diagrams for each sample in this study, and the results of two suites Kolmogorov-Smirnov (K-S) statistical tests. If you are viewing the PDF of this paper or reading it offline, please visit or the full-text article on to view the Supplemental File.