Applying ramped pyrolysis radiocarbon analysis to suspended river sediments, we generate radiocarbon (14C) age spectra for particulate organic carbon (POC) from the lower Mississippi-Atchafalaya River system (MARS) to better understand a major river system's role in carbon transport. Ramped pyrolysis 14C analysis generates age distributions of bulk carbon based on thermochemical stability of different organic components. Our results indicate higher proportions of older material in the POC during higher discharge. Ages increase throughout the high-discharge age spectra, indicating that no single component of the POC is responsible for the overall age increases observed. Instead, older material is contributed across the POC age spectrum and unrelated to increased bedload suspension. In this comparison of 2 spring discharges, less than half of the POC transported during higher discharge is less than 1000 14C years in age, constraining of the role of the MARS as a flux of atmospheric CO2 toward longer-term sedimentary sinks in the Mississippi delta and the Gulf of Mexico. The results suggest that delta-building processes benefit disproportionately from high discharge events carrying larger amounts of sediment because these events involve both a higher proportion of millennially-aged carbon from floodplain exchange of POC and a potentially higher proportion of petrogenic carbon (30-530% increase). Overall, an internally consistent picture of PO14C age distributions from a major river system emerges, as differences in space and time are small compared to the range of ages of POC sources in such a large basin.
 Rivers serve as important regulators of atmospheric CO2 on geologic timescales by transferring organic carbon (OC) from petrogenic and terrestrial carbon reservoirs to marine sediments and/or the atmosphere [Garrels and Lerman, 1981; Berner, 1982; Burdige, 2005; 2007]. Because the petrogenic, biospheric and marine reservoirs are much larger than the atmospheric reservoir, even small changes in proportional fluxes between them can have significant effects on atmospheric CO2 at time scales ranging between 100 to 109 y [Garrels and Lerman, 1981; Berner, 1982]. Recent efforts have attempted to constrain these pools of OC in major river systems [Aller, 1998; Aller and Blair, 2006; Galy et al., 2007; Galy et al., 2008b; Galy et al., 2008a; Hilton et al., 2008; Galy and Eglinton, 2011; Hilton et al., 2011] . Variability in the source and composition of transported carbon as well as the seasonal and interannual storage processes of carbon between the catchment and marine basins [Hedges et al., 1986; Goñi et al., 1998; Gordon and Goñi, 2003], and uncertainty in its fate in the marine environment [Prahl et al., 1994; Hedges et al., 1997; Burdige, 2007] inhibit our knowledge of carbon transport between these reservoirs via river systems.
 Rivers transport 0.25 Pg (1Pg = 1015 g) of dissolved organic carbon (DOC) and 0.15 Pg of particulate organic carbon (POC) each year from the land to the oceans [Meybeck, 1982; Hedges and Keil, 1995]. In most cases, riverine POC is older than DOC [Raymond and Bauer, 2001; Bianchi, 2011], whereas DOC is more reactive in the marine realm. On time scales <108y, the small flux of POC into marine sediments can be considered unidirectional because of the difference in residence times of carbon in long-term sediment sinks versus relatively short-lived atmospheric, soil, and biospheric reservoirs [Berner, 1982]. In fact, sedimentary OC burial is second only to silicate weathering in removing CO2 from the atmosphere. Carbon fluxes between crustal rocks, which release petrogenic OC (OCpetro) upon weathering, and marine sediments are important to the balance of atmospheric gases only if OCpetro is metabolized to CO2. Direct flux of recently fixed biomass (OCbio) and intermediate stored soil OC can play larger roles in regulation of atmospheric CO2.
 The transfer of carbon from terrestrial and atmospheric reservoirs to marine sedimentary sinks by rivers depends both on internal river and coastal ocean processes, which both vary considerably [Hedges and Parker, 1976; Hedges et al., 1997; Blair, 2003; McKee et al., 2004; Raymond et al., 2004; Burdige, 2005; 2007]. Ocean processes alone can cause large variations in carbon storage potential. The contrast between the Ganges-Brahmaputra delta system in the Bay of Bengal and the Amazon delta and mudbank system is a primary example. The former is estimated to bury up to 85% of terrigenous OC [Galy et al., 2007], whereas the latter constantly re-exposes terrigenous OC to oxic and suboxic conditions creating the potential for substantial remineralization of OC [Aller, 1998; Aller and Blair, 2006; Blair and Aller, 2012]. Coastal margins receiving large terrigenous inputs from river systems can become net sources of CO2 to the atmosphere [Cai et al., 2003; Raymond and Oh, 2007], suggesting potential pathways for OCpetro into the biologic carbon cycle rather than passive transport from one long-term reservoir (sedimentary rocks) to another (marine sediments). River systems can both remineralize and prime fixed carbon, the latter of which is important in determining whether marine sediments will store terrigenous OC or release CO2 derived from terrigenous OC to the atmosphere [Bianchi, 2011]. Through the last few decades, work on sediment and carbon budgets from fluvial systems has shifted to small mountainous river systems (SMRs) [Milliman and Syvitski, 1992] which are deemed important in terms of the number of such systems and the mass of sediment transported by them. However, such systems likely play limited roles in transforming and mixing sources of organic carbon when compared to larger river systems [Blair and Aller, 2012].
 Riverine OC is heterogeneous in nature, consisting of carbon of various origins and reactivities; from aged, refractory to recently fixed, more reactive forms of carbon [Hedges and Parker, 1976]. Riverine processes, such as in situ photosynthesis, microbial metabolism, photochemistry, POC exchange with floodplain sediments, and adsorption/desorption to mineral surfaces, can also affect OC composition before the OC is transported into the coastal ocean [Blair et al., 2004; McCallister et al., 2004; Raymond et al., 2004; Bianchi, 2011]. Recently fixed OC sources are favorable for saprophagic microbial metabolism and consumption at higher trophic levels [Thorp and Delong, 2002; Mayorga et al., 2005], however there is evidence that microbes can and do utilize aged OC that is subsequently transferred up trophic levels or released as CO2 to the atmosphere [McCallister et al., 2004; Caraco et al., 2010]. Additionally, storage of sediments within river channels, which occurs over periods of months to decades, may alter the character of POC during exchange with the suspended load [Sutula et al., 2004]. Bedload sediment can store OC at such time scales [Nittrouer et al., 2008; Nittrouer et al., 2011a], however processes of alteration would differ.
 Such pre-conditioning of OC suggests that discharge is an important variable in deciding a river's role in carbon cycling [Mitra et al., 2002; Gordon and Goñi, 2003; Bianchi, 2011; Hatten et al., 2012]. For instance, high discharge events are also large factors in delta building [Roberts, 1997], and OC burial rates are proportional to lithogenic flux onto continental margins that are adjacent to river systems [Burdige, 2005; Allison et al., 2007]. Thus, it follows that extreme high water events would carry a disproportionate amount of POC to potentially long-term carbon sinks. Here, we employ ramped pyrolysis 14C analysis [Rosenheim et al., 2008] to compare, in time and space, PO14C age spectra from two different spring discharge events on the Mississippi-Atchafalaya River System (MARS), sampled throughout the lower reaches of the MARS. As with other river systems, the river channel and floodplain within the MARS can act as reservoirs for POC, allowing for a range of chemical transformations to occur including both the degradation [Sutula et al., 2004] and storage of OC (which can constitute a transformation of the 14C). Conversely, during higher discharge, the river channel and floodplain exchange POC with the suspended load, effectively changing the signature of the POC [Demas and Curwick, 1988; Galler and Allison, 2008; Nittrouer et al., 2008; Allison et al., 2012]. This POC will have potentially been altered by partial degradation, differentially priming it for storage and/or remineralization during riverine transport and/or deltaic and marine sedimentation. The goals of this comparison are to determine to what extent different discharge events will elicit differences in the age spectra, and to what extent PO14C age spectra from a major river system vary when sampled over different locations at different times.
2.1 Study Site and Sampling
 The MARS generally receives its annual maximum and minimum discharge during the spring and autumn, respectively (Figure 1). During a high discharge event in April-May, 2008 (Figure 1), 105 km3 of water was discharged from the Mississippi River. Peak discharge reached 41,229 m3 · s-1 at Tarbert Landing, which currently is the fourth highest discharge recorded at Tarbert Landing since 1961 [USACE, 2011] and represents the eighth time since the building of the Bonnet Carré diversion and spillway (Figure 1) in 1931 that the river was diverted immediately upstream of New Orleans, Louisiana. Thus, the study areas in the Mississippi channel and outflow (Figure 1) saw the highest possible flow that the lower river is permitted to carry during the spring 2008 discharge event. We compare these measurements to samples taken during an above-average, but more moderate spring discharge in 2009 (Figure 1). Table 1 compares the fluxes during both hydrographic events, as defined by the dates that the river discharge at Tarbert Landing was above average during those years. Daily loadings of water, sediment, and POC for each sampling day are summarized in Table 2.
Table 1. Mass Flux Comparisons Between the Spring Discharges of 2008 and 2009a
1σ Mass POC
All river data are from Tarbert Landing (Figure 1). The proportion row is simply the 2008 mass quantity divided by the 2009 mass quantity, where 1 values (italicized and bold) are the uncertainties of the mass quantities propagated into their quotient.
Table 2. Daily Fluxes for Water, Sediment, and POC During Actual MARS Sampling Daysa
Total Carbon Flux
Daily fluxes of different components for each 24 hour period containing the sampling day are also shown. Calculations are not performed for sampling days at river outlets because water and sediment discharge data have little meaning at those sites. Multiplying by a factor of 1.5741 × 10-5 converts these figures into fluxes of mass or volume per second.
 Samples were taken from several locations on the river (Figure 1) in collaboration with ongoing research employing the R/V Lake Itasca (University of Texas, Mississippi distributary) and the R/V Whiskey Pass (LUMCON, Atchafalaya distributary), as well as a dedicated cruise aboard the R/V Acadiana (LUMCON, Mississippi River outlet). Samples within both rivers were taken using a 5 L Niskin bottle, and were decanted into 4-5 1 L Nalgene sample bottles while continuously agitated to ensure suspended sediment remained in suspension. On the Mississippi, samples were taken at 3 proportional depths – 0.1, 0.5, and 0.9, alternatively referred to as surface, intermediate (inter.), and bottom depths, respectively, in the text and figures - relative to the deepest cross-sectional bottom depth as measured by an onboard multibeam bathymetry system. We interpret the deepest cross-sectional bottom depth as the thalweg and targeted it for all samples taken aboard the R/V Lake Itasca. Casts were made while the vessel drifted over the sampling site to minimize wire angle. On the Atchafalaya, samples were taken near the surface (1-3 m) by hand-casting the Niskin bottle. At offshore sites of the Atchafalaya outflow, samples were taken above and below a depth defined by a halocline (approximately 1.5 m). Offshore samples were taken with a Rosette sampler at surface, intermediate, and bottom depths with a single cast. Although samples from different hydrographic years are not necessarily from the same site, we assume that the distribution of 14C in the POC was similar at different points on the lower MARS between New Orleans and Southwest Pass because we targeted the thalweg where sediments generally have the highest likelihood to stay in suspension. During the lowest of our sampling discharges (15,000 m3 · s-1) and using a cross-sectional area of 16,000 m2 for this part of the river [Nittrouer et al., 2011b], the furthest sampling sites were 48-60 h apart by discharge rate, and most comparisons below are between locations that were < 24 h apart at actual discharge during sampling. Neither of these time spans are significant relative to processes affecting POC age spectra described below.
2.2 7Be Sediment Deposition Rate Analysis
 To determine sedimentation rates of flood-derived sediments in wetlands receiving water and sediment from the 2008 high discharge event, we collected samples from the Dupre's Landing wetland in Pass-a-Loutre Wildlife Management Area (Figure 1, star, and Figure 2, upper left). Dupre's Landing wetland is a teardrop shaped island that is dominated by fresh water marsh, including Typha latifolia, Panicum hemitomon, and Sagittaria lancifolia. The marsh elevation grades from a high point at the front (riverside) edge of the marsh that is it typically flooded by only by storm surges tides and large river floods to a low spot near the rear of the marsh that is typically flooded for extended periods of time. Cores were collected using a 10.16 cm (I.D.) Hargis-style coring device equipped with a blade at the bottom, which was able to retrieve a largely undisturbed and uncompacted marsh sediment core. Distinct surface layers of different thicknesses were observed that were comprised largely of poorly consolidated sediments devoid of roots, suggesting a rapid, young flood deposit. To test whether this layer represented high-discharge overbank deposition, we cut one core into 2 to 4 cm intervals, dried and ground the sediment, and packed ~15 g per increment in sealed plastic petri-dishes. These samples were then analyzed with a low-energy gamma detector, and the amount of 7Be (t½ = 53.1 days) was quantified from the 477KeV photopeak (Figure 2, upper right). Beryllium-7 is a naturally occurring particle-reactive radionuclide that is commonly used to quantify episodic sediment deposition rates in coastal systems and elsewhere [Sommerfield et al., 1999; Allison et al., 2005].
2.3 Sample Pre-treatment and Ramped Pyrolysis 14C Analysis
 Suspended matter was filtered from 4-5 L water samples onto pre-combusted quartz-fiber filters with a nominal pore size of 0.7 µm. Filters and sediment were dried (60 °C, 24 h), weighed, and then acid treated with 10% reagent-grade HCl to dissolve mineral carbonates prior to drying (60 °C, 24 h) and re-weighing. Entire filters and their sediment were then processed using the ramped pyrolysis technique [Rosenheim et al., 2008], and a spectrum of 14C ages was generated for each sample by 14C dating of five individual aliquots of CO2 (i.e., ~20 µmol aliquots containing 240 µg C each) that evolved over a 5 K · min-1 temperature ramp. Although more extensive details of the method are given in Rosenheim et al. , in brief, the approach involves volatilizing POC during a slow, stable temperature ramp (5 K · min-1) from ambient temperature to 1000 °C. Volatiles form from more reactive, less diagentically stable components at lower temperatures, and more refractory forms of carbon are broken down at higher temperatures. Volatiles are rapidly oxidized downstream (away from the solid sample which sees no O2), and the resulting CO2 is trapped with liquid N2 and transferred to a vacuum line for purification. Clean CO2 is then radiocarbon dated using standard accelerator mass spectrometer (AMS) techniques. Radiocarbon dating for all samples was performed at the National Ocean Sciences AMS facility at the Woods Hole Oceanographic Institution, and splits (approximately 10%) of each gas sample were analyzed for stable isotope composition. Stable isotope measurements were performed on a GV Optima isotope ratio mass spectrometer using ratios of CO2 isotopologues mass ratios of 45/44 and 46/44, corrected [Craig, 1957, adapted for a triple collector instrument] to δ13C and δ18O ratios. In this study, we employ the full spectrum of ages generated rather than only the youngest age that is of interest in 14C chronology applications of the ramped pyrolysis radiocarbon system [as in Rosenheim et al., 2008].
3.1 Sedimentation Rates
 Peaks in sediment load were short lived and only occurred during the initial stages of high discharge [Figure 1; USGS, 2010]. Both overall sediment load and carbon loading from two sampling days in the Mississippi River were relatively similar compared to river discharge rates (Table 2). High suspended sand and bedload transport rates were observed, however, throughout the high discharge event of spring 2008 [Nittrouer et al., 2008; Nittrouer et al., 2011a]. The corresponding high water levels delivered sediment to overbank areas where flow velocities were slow enough to allow sediment deposition and exchange of POC with these deposits. Sediment accumulation rates from Dupre's Landing in Pass-a-Loutre (Figure 1, star) indicate that the undisturbed surface sediment was indeed a high discharge event deposit (Figure 2, upper right and lower left). Detectable 7Be was present to a depth of 7 cm in core A and a least squares linear regression fit through a plot of natural logarithm-transformed activity against depth (Figure 2, lower left) was used to determine the sediment accumulation rate following modification of established methods [Sommerfield et al., 1999; Allison et al., 2005; Kolker et al., 2009]. These results indicate a sediment accumulation rate of 0.0084 cm · d-1 or, when averaged over a weekly timescale that more reasonably represents the scale of flood-driven processes, 0.6 cm · week-1. The flood deposit at the Dupre's Landing wetland thinned at a rate of 0.14 cm of deposit thickness per meter from the river channel (Figure 2, lower right), providing an estimate of the length scale over which sediments are delivered in wetlands in the Mississippi River delta. These rates can be compared to centennial and decadal-scale sediment accumulation rates in the lower region of the Mississippi River Delta Plain. Studies from marshes in eastern Cubits Gap, which, like Dupre's Landing, actively receive river water and sediment, had long-term 210Pb-derived accretion rates exceeding 1.59 cm · yr-1 (or 0.03 cm · week-1) [Wilson and Allison, 2008]. Marshes from West Bay, which did not receive substantial quantities of river water prior the opening of a diversion in 2003 had longer-term 137Cs-derived sediment accumulation rates of 0.7 -1.5 cm · yr-1 (or 0.01-0.03 cm · week-1 [Kolker et al., 2012]).
3.2 Radiocarbon Ages of Suspended Sediment
 Radiocarbon ages are based on measurements of 14C/12C in the sample relative to 14C/12C of 1950 wood (fraction modern, fM), which can be expressed as either the radiocarbon age (14C y) or the radiocarbon deficiency (Δ14C) [Stuiver and Polach, 1977]:
where short sample storage time (<2 y between sampling and measurement) renders δ14C equal to Δ14C relative to analytical uncertainty [Stuiver and Polach, 1977]. In some cases, samples were too small or lost during radiocarbon analysis preparation. No POC sample resulted in fewer than 4 isotope determinations.
 Data from the ramped pyrolysis radiocarbon technique consist of two main facets: first, the evolution of CO2 from oxidation of pyrolysis products during the pyrolysis ramp (5 K · min-1) and, second, the isotopic composition (δ13C and Δ14C) of separate, conterminous temperature intervals over which CO2 is integrated into a single sample (Figure 3). These data can be shown on a single plot with two y-axes – pCO2 (chemically, this estimates decomposition reaction rates () from the mixture of chemicals comprising the POC because temperature is linear with time) and 14C y – plotted on a temperature abscissa (Figure 3). In these plots, the width of the bars indicates the temperature interval over which the CO2 sample was integrated and the height corresponds to the points indicating radiocarbon age. The CO2 evolution data can be used for modeling of the components comprising different pyrolysis reactions and for assessing the overall stability of the OC in the sample using the temperature of maximum reaction rate - Tmax. The simplest model of pyrolysis reactions assumes Gaussian distribution of activation energies for different components with similar Arrhenius frequency factors [Burnham and Braun, 1999]. Such models are over-simplified in terms of any one chemical component [Yang et al., 2007] but satisfactory for complex mixtures of reactants [Tarturis, 1993; Burnham and Braun, 1999; Cramer et al., 2001; Cramer, 2004]. All of the data described below can be fit with three Gaussian curves (Figure 4). Because 5 measurements were made, the age of each Gaussian curve can be solved in an over-determined linear system of equations (Table 1 and Figure 4 - see Rosenheim et al.  for more details).
 To put age spectra in terms of other bulk radiocarbon measurements of riverine POC [Goñi et al., 1998; Raymond and Bauer, 2001; Gordon and Goñi, 2003; 2004; Raymond et al., 2004; Goñi et al., 2005; Mayorga et al., 2005; Wakeham et al., 2009; Caraco et al., 2010], we use the yields over each temperature interval to calculate a geometric mean radiocarbon content comparable to a bulk radiocarbon date. The calculated bulk isotope compositions (Δ14Ccb and δ13Ccb) and calculated bulk age (14Cycb) over n temperature intervals can be expressed as:
where fi , Δ14Ci, δ13Ci, and fM,i are the mole fraction of CO2 produced, measured permil radiocarbon depletion ((fM,i-1) × 103), measured stable carbon isotope composition, and the fraction of modern 14C over temperature interval i, respectively, and ∑ fi = 1. Large variations about mean values are due to both the ability of ramped pyrolysis 14C analysis to separate a single bulk 14C age into a spectrum of ages and the stratification of the water column during near-mean discharge (Figure 3, a, c, e). It is important to note comparison of geometric mean data (Δ14Ccb and δ13Ccb) from ramped pyrolysis analysis to bulk measurements in the literature involves different degrees of uncertainty; bulk radiocarbon values will only express analytical uncertainty, whereas the larger uncertainties on Δ14Ccb represent the geologic uncertainty of a mixture of different ages.
 The average calculated bulk radiocarbon age of Mississippi River sediment during the 2008 high discharge event (2010 ± 67 14C y) was older than average calculated bulk PO14C sampled during 2009 (860 ± 360 14C y, see supporting information for estimation of Δ14C when eq. 3 cannot be used due to ∑ fi ≠ 1 owing to the loss of one or more samples during analysis). For the Atchafalaya the 2009 calculated bulk PO14C ages (1500 ± 440 14C y, Figure 5) were older than samples taken from the Mississippi (860 ± 360 14C y) during the same hydrographic year (Figure 1). Indeed the entire spectra from the Atchafalaya distributary reflect older ages than the Mississippi distributary (Figure 5). Important changes in the distribution of ages of POC as the Mississippi and Atchafalaya Rivers flow into marine environments on the continental shelf are evident as well. Samples collected at the points of discharge were older than upstream samples (Figure 5). This difference is more pronounced in the Mississippi distributary than in the Atchafalaya distributary (Figure 5), suggesting that POC from the Mississippi undergoes a larger shift toward older ages than POC from Atchafalaya.
 Ramped pyrolysis 14C analysis generates 14C age distributions of bulk carbon based on thermochemical stability of individual OC components. The age distributions from each sample in this work can have a wider range than analytical uncertainty of a bulk 14C age determination. Because 14C is fixed into plant OC via photosynthesis, the presence of radiocarbon in POC provides a signature of the recent residence of carbon in the atmosphere and thus can be used as an integrative proxy for the period since sequestration of CO2 from the atmosphere. Additionally, we employ the radiocarbon chronometer on POC components that could contain carbon much older than the maximum measureable 14C age (~55,000 14C y, depending on the accelerator mass spectrometer used). This chronometer is still useful to separate three main pools of riverine POC – recent biospheric C (ages post-modern to 10's of years), soil-stored C (ages of 100's to 1000's of years), and petrogenic carbon (infinite radiocarbon ages; e.g. Galy and Eglinton ). Ramped pyrolysis aims to more directly measure 14C of the older pools of carbon [Rosenheim and Galy, 2012] rather than inferring them as is commonly achieved, although in most cases ages of individual ramped pyrolysis temperature intervals still represent mixtures of different components (Figure 4).
 Overall, the pyrolysis reaction curves for all samples are similar (Figures 3, 5). Three-Gaussian models are adequate to describe each sample (Figure 4). Additionally, Tmax are very similar for each sample; averaging 435 °C (±26 °C) during the high 2008 spring discharge, and 436 °C (±24 °C) during the more moderate 2009 spring discharge. The average Tmax of the entire data set was 436 °C (± 25). This type of concordance in POC reactivity would be expected from the lower reaches of a large integrative river system on a passive margin where the complexity of mixing several sources of carbon is potentially tempered by in situ reworking and exchange within the floodplain [Blair and Aller, 2012]. Comparison of calculated Δ14Ccb indicate other similarities. Previous work has reported bulk PO14C ages ranging between 1,070 and 6,770 ybp on the continental shelf near the outflow of the Mississippi and Atchafalaya distributaries [Goñi et al., 1997; Gordon and Goñi, 2004; Wakeham et al., 2009], and averaging 2,310 14C ybp in the Atchafalaya suspended sediments (autumn 1997 and spring 1998, [Gordon and Goñi, 2003]). These ages are generally older than calculated Δ14Ccb (Table 3, 1,290 14C y in the Atchafalaya and 1960 ± 60 or 840 ± 420 14C y in the Mississippi, depending on the year). A sample taken from the Atchafalaya in April 28, 1998 [Gordon and Goñi, 2003], has a similar age (1600 14C y) and geographic location to ARO-2009 in this study (Table 3), however the discharge at Simmesport was nearly a factor of 2 higher for that sample than for ARO-2009 in this study. Based on these comparisons, it would seem that discharge is not a master variable determining age spectrum of POC, and that geographic and temporal differences in sampling can also cause considerable PO14C age variation that is small relative to the 14C time scale. Analysis of the age spectra compared between different hydrographs (Figure 3) and between different locations (Figure 5) provides more insight into differences and similarities between sites and at different discharges.
Table 3. Ages and Compositional Proportions of the 3 Gaussians Decomposed From Ramped Pyrolysis Data of Each Sample Analyzed Hereina
Tmax (1 σ, °C)
Δ14Ccb (1σ, eqs. 4)
14C y (1σ, eq. 5)
δ13Ccb (1σ, eqs. 4)
C1 Age (14C y)
C1 % of total CO2
C2 Age (14C y)
C2 % of total CO2
C3 Age (14C y)
C3 % of total CO2
n.c. refers to samples where calculations could not be made due to loss of one or more temperature intervals and * refers to estimation of calculated bulk isotope quantities using the third temperature interval (Appendix A in the Supporting Information). Figure 4 shows a representative Gaussian decomposition; all samples were similar in that 3 Gaussians were sufficient to describe the production of CO2 during pyrolysis. Surface = 0.1D; Inter. = 0.5D; and Bottom = 0.9D, where D is the overall depth of the water at the sample site.
4.1 Age Spectra Through Time and Space in the Lower MARS – Similarities and Differences
 Ramped pyrolysis provides more information than bulk PO14C dates by allowing the comparison of entire age spectra between samples (Figure 3). Ramped pyrolysis data illustrate that all measured Δ14Ci in POC from high discharge (2008) contained less radiocarbon than those measured in 2009 during spring conditions closer to the mean river discharge. Increased age throughout the spectrum is in contrast to having one significantly older Δ14Ci skewing the Δ14Ccb of the sample, akin to addition of an older source of POC during high discharge. In such a case it is conceivable that the river system could transport more young carbon toward long-term sinks during high discharge simply due to higher mass flux of carbon, but still have an older bulk 14C age due to the presence of the older component. Instead, Figure 3 indicates that generally all POC is of older age during the high discharge, and thus pre-aged material is distributed throughout the age spectrum. We interpret this wholesale age spectrum increase without a difference in Tmax during high discharge as an indication that POC from the 2008 high discharge event was compositionally similar but older than that from 2009. Indeed, previous work supports this interpretation; Bianchi et al.  have demonstrated that there is no discernible relationship between river discharge and both the concentration and composition of POC in the lower Mississippi River. Likely reservoirs of compositionally similar but older material that could become entrained in the higher river discharge events include flood plain/riparian sediments that only interact with flowing water during bank-full flow and bedload sediment that is largely stationary during near mean flow.
 Entrainment of bedload into the suspended load does not seem to be a viable mechanism to decrease PO14C content across the age spectra. Coarse grain bedload sediments can remain undisturbed during average discharge [Demas and Curwick, 1988] but are subsequently transported with higher discharge events [Horowitz, 2006; Nittrouer et al., 2008; Nittrouer et al., 2011a; Ramirez, 2011]. Surface and intermediate depth samples showed larger age spectrum shifts between different discharges than bottom samples (Figure 3). Whereas the river shows a higher degree of particle size stratification during high discharge (Figure 6), the corresponding age spectra at all depths were nearly identical (Figure 3, left column). Conversely, during more moderate discharge, age spectra showed stratification with older ages at the bottom suggesting that bedload transport indeed has an effect on the POC age spectra. It is likely that this effect is constrained to fine-grained bedload with high average surface area to mass ratio rather than coarse-grained bedload whose presence in bottom samples during high discharge does not change the age spectra. During high discharge fine grained suspended sediment is more uniformly distributed (Figure 7).
 A combination of different ages and different δ13C signatures (Figure 7) between discharge events can be attributed to flood provenance and floodplain storage of POC. High discharge POC samples (2008) were on average 1.5‰ more depleted in δ13C than moderate discharge POC samples (Figures 3, 5. During more moderate discharge, the Mississippi River receives the majority of its sediment load from the Missouri River system and the majority of its water discharge from the Ohio River system [Meade et al., 1990]. The Ohio River basin has a high proportion of C3 plants, such as deciduous woodlands and mixed conifer-deciduous forests relative to the predominant grasslands of the Missouri River basin [Gordon and Goñi, 2003; Lee and Veizer, 2003]. Corn-planted acreage in the lower Ohio and Mississippi mainstem and wheat-planted acreage in the Missouri valley likely attenuate this difference with more reactive POC, but a large proportion of each watershed remains unfarmed [Vilsack and Clark, 2009]. The high discharge of spring 2008 was caused by increased rainfall in the Ohio River basin during March and April (Figure 8), potentially contributing a higher proportion of 13C-depleted (C3 plant-derived) POC during high flow conditions. The increased ages of this material (Figures 3, 6) is likely due to floodplain storage. During the spring discharge of 2009, the source of increased discharge was tributaries further downstream in the Lower Mississippi basin (Figure 8). Sediment in this part of the basin consists of a mixture of sediment from the Upper Mississippi, Missouri, and Ohio sub-basins with a majority from the Missouri basin [Meade et al., 1990].
 Comparison of δ13C of POC at both outflows during the 2009 spring discharge indicates the effects of marine processing of POC on age spectra and isotope compostition. First, there is no evidence for mixing with marine POC in the Mississippi plume; POC from Southwest Pass shows lower δ13C values than the POC in the channel, indicating that samples from the plume still had a C3 terrestrial composition (Figure 3). Lightening of the isotope composition of POC could indicate that C4-derived POC in the lower river channel was more reactive or denser in the marine environment, supporting older ages for the C3 material due to the increased age spectrum (Figure 5). Comparison of Atchafalaya outflow POC to channel POC in 2009 is more complex because of the significant fraction of Red River POC that was not directly measured, however the age spectrum shift toward older ages at the ouflow was more subtle than in the Mississippi. Conversely, the isotope composition shift is toward significantly more positive values indicating the possibility of incorporation of marine material. In either case, the potential effects of differential degradation and differential transport (hydrodynamic sorting) cannot be teased apart using this dataset.
 Despite some clear differences between the sampled spring discharges, it is important to note that the results are overall internally consistent for a large data set sampled over different times and locations. To assign cause to the differences observed would involve more constraint on several variables in addition to discharge. One such variable is the Red River's contribution of water, sediment, and POC to the Atchafalaya. The distribution of water entering the Mississippi and Atchafalaya River Channels is managed at the Old River Control Structure (Figure 1) by diverting a portion of the discharge of the Mississippi main stem to the Atchafalaya channel where it mixes with the discharge of the Red River. This portion is controlled so that the outflow of the Atchafalaya/Red system (measured at Simmesport, Figure 1) is 30% of the total outflow of the MARS and that the Mississippi outflow (measured at Tarbert Landing, Figure 1) is 70% of this combination. The Atchafalaya receives 100% of the water, sediment, and POC discharged from the Red River. On average, the Red River flow comprised about 27.5% and 36.1% of the flow in the Atchafalaya basin in 2008 and 2009, respectively, but it reached maxima of 85.9% and 76% during these years, respectively (Figure 9). During our sampling of the Atchafalaya in 2009, the proportion of Red River discharge in the Atchafalaya was 35% (Figure 9), rendering it a substantial source of water and sediment to the Atchafalaya. Focus on the small differences between discharges and/or outflows of the MARS should involve more extensive sampling of the hydrograph in both space and time, as well as constraint on such variables as the Red River POC contribution.
4.2 Mass Flux of Carbon From Different Time Scales
 Internal consistency of the entire data set is best illustrated by the lack of remarkably old or young POC in any part of the spectra. This is evident in both the raw age spectra (Figures 3 and 5) and in applying Gaussian models (Table 3) to all samples, [Rosenheim et al., 2008]. With the exception of young POC in the surface and intermediate samples of near-mean Mississippi River flow in 2009 (Figure 3), no source of OCbio or OCpetro appears to dominate the age spectra near the terminus of the MARS. The large drainage basin of the Mississippi River system includes both old sources of carbon (coal deposits and old metamorphic rocks in the Ohio basin, Cretaceous carbonates in the Missouri basin) and young sources (fresh vegetation - woodland, prairie land, and agricultural land) which, measured independently, would have significantly greater age differences than the observed POC age spectra. This raises the question of why these potential endmembers are not observed in lower MARS POC. Previous work has shown that small, steep mountainous river systems have simpler binary to tertiary mixtures of POC whereas larger river systems have more complex mixtures [Blair et al., 2004; Galy et al., 2008b; Galy and Eglinton, 2011]. Upland sources of POC are either remineralized or diluted by lowland sources during transport in such large, integrative systems as the Mississippi. Thus, the decomposition of ramped pyrolysis data from the lower MARS into major Gaussian components (Figure 4) likely yields major components that can be comprised of compositionally-similar material of varying age (i.e. these components are not fully resolved endmembers but they represent geometric means of simpler mixtures than those represented by the bulk POC).
 These data are useful for assessing carbon fluxes in the MARS over both short (decadal) and longer (millennial) time scales. Over short time scales, we can use the amount of radiocarbon in a sample as a proxy for atmospheric carbon sequestered by the river system – protracted storage of fixed atmospheric carbon in floodplain soils would age the POC significantly compared to leaf litter, leaf waxes, and other recently fixed atmospheric carbon. Using this proxy, it is clear that higher river flows, although they may transport a higher mass of carbon (Table 1) into demonstrably larger sediment deposits (Figure 2), are not proportionally efficient in transporting and burying carbon recently removed from the atmosphere. This is important as deltaic deposits of a major river system are areas where rapid carbon burial can happen due to high preservation rates of POC. Overall, application of the analysis in Figure 4 to all of our data (Table 3) shows that only 20 to 40% of carbon (1.7 to 3.7 x 108 kg POC/y) that is transported down the MARS is younger than 1000 14C years. The amount of carbon that dates to less than 1000 14C y represents an insignificant fraction of the net primary productivity in the Mississippi River watershed (1.6 x 1015 g C/y), illustrating that the MARS does not represent an efficient mass flux of atmospherically derived OC [Lee and Veizer, 2003]. Moreover, of the twofold increase in both water volume and sediment mass transported through Tarbert Landing during the 2008 high water event relative to the 2009 spring discharge (Table 1), mass flux of POC attributable to the C-1 component (Figure 4) is statistically equivalent between 2008 and 2009 (Table 4). If we treat C-1 as a mixture of modern (fM = 1, Δ14C = 0‰) and petrogenic (fM = 0, Δ14C = -1000‰), thereby providing an upper bound on the amounts of OCbio and OCpetro, respectively, OCbio may be as low as 27% lower in 2008 and OCpetro may be as high as 530% greater in 2008. The means of these proportions are statistically indistinguishable from variability in average age and proportion of each Gaussian component (Figure 4, Table 3), however the larger ranges of C-3 in 2008 are telling of older ages only observed in high discharge samples. These comparisons would benefit from more complex models of pyrolysis chemistry and better constraint of the hydrograph when sampling suspended sediment.
Table 4. Mass Flux Comparisons Between the Spring Discharges of 2008 and 2009a
C1 Total Mass
C1 OCbio Mass
1σC1 OCbio Mass
C3 Total Mass
C3 OCbio Mass
1σC3 OCbio Mass
All river data are from Tarbert Landing (Figure 1). Average values from Table 1 used to calculate proportion of carbon mass flux attributable to representative components in Figure 4, although those components were not measured frequently during either hydrograph. The proportion row is simply the 2008 quantity divided by the 2009 quantity. Mass of modern POC in C1 and mass of petrogenic POC in C3 represent maximum values.
 Mass fluxes also suggest that the amount of petrogenic carbon is small in lower MARS POC. The delta was constructed by the Mississippi River largely throughout the Holocene, and high discharge events have played a disproportionate role in shaping it [Roberts, 1997]. We can use the proportion of material in the C-3 component (Figure 4) to estimate the amount of petrogenic carbon that may be deposited in different discharges. Of the twofold increase in sediment mass transported in 2008 vs. 2009 (Table 3), there is indication of a 1.3–5.3 fold increase in OC petro transported during 2008's high discharge event (Tables 2, 4). Although pyrolysis is simply a proxy for diagenetic stability in nature, where both chemistry and physics determine what is remineralized and what is preserved, the data indicate that the deltaic deposit of the MARS acquires a net increase of small amounts of OCpetro and a net decrease in OCbio during high-discharge events that contribute most significantly to delta-building deposition.
 Comparison of high and low discharges yields differences in POC age spectra on the MARS, however the lack of sharp age contrast within the dataset illustrates emerging differences between POC age distributions on large, passive-margin rivers and SMR's [Rosenheim and Galy, 2012; Blair and Aller, 2012]. Our results show that the MARS suspended load is comprised of a mixture of POC devoid of very old and very young major carbon components. Ages are generally millennial in scale, including older ages in high discharge. High-discharge shifts in PO14C age structure affect the entire spectrum of ages indicating that no single, pyrolysis-isolatable component of OC is responsible. We suggest that this is a trait of large, integrative river systems with large floodplain sediment OC storage capacity.
 Overall, a high-discharge event in the MARS shows overall older ages of POC than a more moderate spring discharge. In the Mississippi, the ages of POC are dominated by the fine grain fraction of the suspended load, which is well-mixed by increased turbulence during high discharge. The coarse-grain bedload fraction, which is confined near the bottom of the river, does not exercise a significant effect on the age distributions of POC which remain similar at all depths. An unresolved marine process shifts POC age spectra toward older (Δ14C) composition in samples analyzed from the continental shelf adjacent to the outflows of both the Atchafalaya and Mississippi rivers, however the isotopic compositions (δ13C) of these sediments become both heavier (Atchafalaya) and lighter (Mississippi). The shift in ages is larger in the Mississippi distributary than in the Atchafalaya. Samples compared between both outflows were not from the same hydraulic conditions; therefore it is not possible to assign these shifts in the marine realm to differential degradation, differential transport, or some combination of both.
 If ramped pyrolysis results are extrapolated to the entire hydrograph of both events, the data suggest that the mass flux of atmospherically-sourced OC in the lower MARS POC is low relative to primary production in the watershed. Our results suggest that, in its present state, the MARS is not efficient in delivering atmospheric CO2 to longer-term deltaic sinks on societally-relevant timescales, and becomes less efficient during high discharge. Nearly equal masses of modern carbon were transported in both years, despite more than double the mass of POC being transported in 2008 vs. 2009. At the same time, the MARS transported between 130% and 530% the amount of OCpetro in 2008 vs. 2009. On timescales of delta formation, however, the MARS still serves as an efficient conduit for the transfer of terrestrial carbon to the marine-deltaic realm and potentially into longer-term sedimentary sinks, especially during high flow events. We conclude that the proportion of OCpetro:OC is likely small compared to smaller rivers where significant proportions of OC are near the old end of the radiocarbon age spectrum, despite the increase in mass of petrogenic carbon transported through the MARS in such events.
 This work was funded by NSF grants EAR-0929752 to BER and EAR-0832754 to ASK and BER, and partially by a grant from the Louisiana State Board of Regents Research Competitiveness subprogram (contract LEQSF(2009-12)-RD-A-20) to BER. Sampling was carried out in part with funds from Louisiana OCPRA (MAA) and Louisiana State Board of Regents Research Competitiveness subprogram award (contract LEQSF(2008-11)-RD-A-22) to BJR. KMR was supported during her Master's thesis by a Louisiana State Board of Regents Graduate Assistant Fellowship (LEQSF (2005-10)-GF-12). The authors wish to thank Torbjörn Törnqvist and Valier Galy for valuable comments during the preparation of the manuscript. BER and ASK thank Rebirth Brass band and the Easter Sunday 2008 second line parade to the uptown levees for impetus to investigate POC ages during high discharge. Logistical support and field work help from J.A. Nyman and Kristen Butcher is appreciated in sampling of Pass a Loutre. This manuscript benefited markedly from editorial comments by the associate editor, an anonymous reviewer, and N.E. Blair.