Sources of carbon to suspended particulate organic matter in the northern Gulf of Mexico

The introduction of fossil hydrocarbon-derived material, whether by anthropogenic inputs or natural seepage, provides a unique source of carbon source to the sea. In most areas of the Atlantic and Pacific oceans, surface primary production is the main carbon source to the deep ocean. Typically, only about 1% of the carbon fixed at the surface reaches the deep seafloor. Along the way, most of the organic matter is consumed and degraded, exchanging between the different carbon pools in the water column — dissolved inorganic carbon (DIC), dissolved organic carbon (DOC), and sinking and suspended particulate organic carbon (POC_sink, POC_susp), before the residual amount finally reaches the seafloor.

Of these pools, POC_susp is uniquely capable of providing insights into the sources of carbon to the water column and those fueling the microbial loop. The small particle size and relatively short residence time that characterize POC_susp make it more sensitive than other carbon pools to recording variations in inputs. The different carbon pools can be defined operationally by size. The smallest end of this size continuum is DOC, which we define as organic carbon that passes through a filter of 0.7-µm pore size. POC_susp is any organic matter collected on the 0.7-µm filter, while POC_sink is comprised of those particles typically larger than 50 µm (Deuser, 1986). Due to the small difference in size, DOC and POC_susp are more similar to each other chemically than to POC_sink (Druffel et al., 1996). As DOC is partially controlled by microbial processes, the DOC-POC_susp connection provides a link between microbial processes and the larger particles that can move carbon up the food chain (Chanton et al., 2012; Cherrier et al., 2014). Linkage with the microbial loop has also been observed in the microbial uptake of dissolved inorganic nitrogen and possible methanodiazotrophy through δ¹⁵N isotope analysis (Montoya et al., 1990; Fernandez et al., 2016).

The residence times of DOC, POC_sink, and POC_susp range widely, resulting in different degrees of sensitivity of each carbon pool to different inputs. DOC is the second largest carbon reservoir in the ocean, amounting to about 650 Pg of mostly recalcitrant carbon, with a residence time of 1000–6000 years (Williams and Druffel, 1987). The concentrations of POC_susp and POC_sink in the water are much lower than DOC, but the flux of POC_sink through the water column is much greater than POC_susp, with a residence time of about a month (Deuser, 1986). POC_sink is determined primarily by surface phytoplankton production (Chanton et al., 2018), which draws on the DIC pool, the largest carbon reservoir in the ocean at 38,000 Pg (Hansell and Carlson, 2014). POC_susp floats in the water for 5–10 years (Bacon and Anderson, 1982). Its low concentrations in the open ocean, typically from less than 10 µM C in surface waters to about 1 µM C at depths below 500 m (McNichol and Aluwihare, 2007), and short residence times increase the sensitivity of this pool to other carbon sources, although the DOC and POC_sink pools have also been observed to reflect variations in carbon input associated with the Gulf of Mexico (GOM) Deepwater Horizon (DWH) spill (Yan et al., 2016; Walker et al., 2017; Chanton et al., 2018; Geiring et al., 2018).

Studies in the late 1980s and early 1990s used stable and radiocarbon isotope analysis of POC to determine sources of oceanic particulates. Druffel et al. (1992, 1996, 2003) analyzed stable and radiocarbon isotopes of particulates from the Sargasso Sea and central North Pacific which indicated that POC_sink and POC_susp are derived primarily from surface phytoplankton production. POC_susp is enriched in radiocarbon from being formed at the surface and becomes more depleted with depth. In the GOM, δ¹³C values for surface production range from –20 to –22‰ (Chanton and Lewis, 2002), and, at the time of the oil spill, Δ¹⁴C values for surface production ranged from 39 to 41‰ (Chanton et al., 2012, 2018).

In marginal seas, the sources of POC_susp can vary. Bauer et al. (2002) found highly depleted δ¹³C and Δ¹⁴C values for POC_susp at depth along the Mid-Atlantic Bight (MAB). The radiocarbon depletion near the seafloor of POC_susp has been attributed to resuspension of old sediment or organic matter and adsorption of DOC onto POC_susp (Druffel et al., 1992, 1996, 2003; Bauer and Druffel, 1998; Bauer et al., 2002). Flocculating particles due to heterotrophic activity could also play a role in depleting PO¹⁴C_susp (Druffel et al., 1992). Bauer et al. (2002) suggested that deep shelf PO¹⁴C from the MAB could be caused by natural hydrocarbon seepage; however, at the time of collection there was no evidence of such seepage, leading Bauer et al. (2002) to conclude that the depletion was due most likely to resuspended sediment. The similar correlation of δ¹³C and Δ¹⁴C observed between particulates from the MAB (Bauer et al., 2002) and the Desoto Canyon of the Gulf of Mexico (Figure 1) (Cherrier et al., 2014) suggests similar sources of carbon. In 2014, Skarke et al. (2014) reported the discovery of a major hydrocarbon seep field in the same area of the MAB that Bauer et al. (2002) had sampled. The hydrocarbons from this seep field could well be the cause of the correlation between the depleted δ¹³C and Δ¹⁴C of POC from the MAB (Figure 1).

In April of 2010, the DWH Blowout released 717–789 million liters (4.5–4.9 million barrels) of oil and 500,000 t of gaseous hydrocarbons into the northern central Gulf of Mexico (Lehr et al., 2010; Joye et al., 2011). An estimated 30% of the released hydrocarbons formed a deep-water hydrocarbon plume between 1000-m and 1200-m depths (Valentine et al., 2010; Ryerson et al., 2012). The bulk of the gaseous hydrocarbons were primarily methane (Joye et al., 2011), but less than 0.01% of the gases reached the surface (Kessler et al., 2011; Yvon-Lewis et al., 2011). Crespo-Medina et al. (2014) measured methane oxidation rates in the water column following the DWH event. At the depth of the deep-water hydrocarbon plume, concentrations of methane and the gene methane monooxygenase (pmoA) were elevated, as were methane-oxidation rates. Cherrier et al. (2014) presented evidence that this DWH-derived CH₄ was found in the POC_susp of the Gulf in 2011–2012 (Figure 1). Assimilation of methane by methanotrophs has been found to be very efficient in other systems, e.g., converting 63–85% of methane into biomass at landfill sites (Börjesson et al., 1998, 2001). Du and Kessler (2012) estimated, using theoretical calculations of oxygen usage from the dissolved oxygen anomaly present in the deep-water hydrocarbon plume, that hydrocarbon degradation generated 0.36 ± 0.11 mg biomass per mg hydrocarbon. They also estimated that 0.10 ± 0.11 Tg of hydrocarbons, primarily methane, were converted into microbial biomass within the deep-water hydrocarbon plume. This biomass, including after cell senescence and viral lysis, would be an input to the POC_susp pool.

Methane can also be generated microbially in the oceanic water column under aerobic conditions (Karl et al., 2008). This process occurs in the Gulf of Mexico (Rakowski et al., 2015), as evidenced by CH₄ concentrations and relative microbial abundances co-varying significantly from the seafloor to the euphotic zone. Thus, our characterization of the sources contributing to POC_susp included characterizing the isotopic composition of biogenically produced methane.

Several studies following the DWH event analyzed POC and plankton in the GOM, finding depleted δ¹³C and Δ¹⁴C signatures (Graham et al., 2010; Chanton et al., 2012; Cherrier et al., 2014). Chanton et al. (2012) and Cherrier et al. (2014) found a linear relationship between δ¹³C and Δ¹⁴C signatures, from modern photosynthetic production to a hydrocarbon endmember, with both the plankton and POC falling along the line, indicating the incorporation of material originally sourced from hydrocarbons, as well as the movement of this material up the food web (Wilson et al., 2016).

The purpose of this study was to characterize δ¹³C and Δ¹⁴C signatures of POC_susp in the GOM following the oil spill and as the system recovered from it, determining a new post-spill baseline for δ¹³C and Δ¹⁴C signatures of POC_susp in the GOM in the process. (For parallel work on POC_sink, see Chanton et al., 2018.) In the Atlantic and Pacific Ocean basins, the baseline derives from a single dominant source, modern photosynthetic production. No other carbon source contributes more depleted ¹³C or ¹⁴C to the pool of POC_susp, so that co-variation between the isotopes (as in Figure 1) is not observed. In the GOM, however, the presence of two distinct carbon sources, one associated at times with anthropogenic activity, drives the depletion of both δ¹³C and Δ¹⁴C of POC, resulting in the co-variations depicted in Figure 1. The continuous input of hydrocarbon-derived material, petrocarbon, from natural seeps could also cause baseline signatures of the GOM to be more depleted than those from the Atlantic or Pacific oceans where input of hydrocarbon-derived material is quantitatively unimportant.

In this study we pursued three goals. First, we tested the hypothesis that, at sites with one-time (DWH) inputs of petrocarbon derived from methane and oil, initial depletion of the carbon isotopic signatures of POC_susp would be followed by recovery, shifting the depleted values observed by Cherrier et al. (2014) towards more enriched baseline-like signatures. Assessing baseline signatures of POC_susp in the Gulf prior to the DWH spill in 2010 has been difficult, as no PO¹⁴C data were collected in the GOM prior to that time. Second, we addressed the question, to what extent are the carbon isotopic values of POC_susp in the GOM affected by the seep sites that provide a continuous source of hydrocarbon-derived material to the Gulf as opposed to the one-time input from the DWH? Third, we aimed to determine the relative importance of modern surface marine production and riverine inputs to the different carbon sources in the Gulf, as reflected in POC_susp. In this study, we analyzed the stable carbon and radiocarbon isotopes of POC_susp collected from across the northern GOM to determine the carbon sources to these particulates. For sources we considered surface marine primary production, riverine input, sediment, two sources of biodegraded methane, and biodegraded oil.

POC_susp samples were collected during fourteen cruises over six years from 2010 to 2017 (except 2011), from a total of 43 separate sites across the northern Gulf of Mexico (Figure 2). At the time of collection, sites were classified as seep or non-seep, with seep samples collected directly over areas of seafloor seepage, while non-seep samples (designated north central Gulf, NCG) were collected in areas not directly influenced by natural seepage. These determinations were based on map data from MacDonald et al. (2015) and shipboard acoustics used to detect hard bottoms or bubbly streams indicative of a seep site. Over the course of sampling we collected particles from 13 seep and 30 NCG sites.

Water column samples were collected by CTD-Rosette, filtering 1–20 L of water through pre-combusted 47-µm 0.7 GF/F borosilicate filters in plastic housings under gentle pressure filtration (5–10 psi). Filters were stored frozen in combusted aluminum foil and brought back to the lab for acidification in a filter housing unit with dilute (1N) HCl (Fernández-Carrera et al. 2016). Stable carbon isotope ratios of POC_susp were analyzed on subsections of the filters using a Carlo-Erba elemental analyzer connected to a Finnegan MAT delta Plus XP Stable Isotope Ratio Mass Spectrometer (EA-IRMS) at the National High Magnetic Field Laboratory. The results were converted into δ¹³C with respect to VPDB (Vienna Pee Dee Belemnite). Samples analyzed after 2014 had a small split, about 10% of the total sample, of CO₂ removed prior to graphitization to be analyzed for δ¹³C. After δ¹³C analysis, the remaining filter was combusted, and the resultant CO₂ was purified cryogenically using the methods of Peterson et al. (1994). The purified CO₂ was flame-sealed in a 6-mm ampule and sent to Woods Hole National Ocean Sciences Accelerator Mass Spectrometry (NOSAMS), University of Georgia Center for Applied Isotope Studies (UGA) or Lawrence Livermore National Laboratory Center for Acceleration Mass Spectrometry (LLNL CAMS) for analysis of natural abundance of radiocarbon. The radiocarbon signatures are reported in the Δ¹⁴C notation as described in Stuiver and Pollach (1977). The blank correction as described in Fernández-Carrera et al. (2016) was applied to both δ¹³C and Δ¹⁴C values. Forty coal samples, representing fossil ¹⁴C dead carbon, were run to access our procedural blank over the course of this study. The average Δ¹⁴C value was –995 ± 7‰. We also ran 25 azalea leaf standards collected in Tallahassee, Florida, in 2013. The average Δ¹⁴C value was 31 ± 8‰. There was no variation between AMS labs in these samples or the coal blanks.

To estimate the DOC blank adsorbed onto the filters, we attached a second GF/F filter below the filter collecting the POC_susp. The top filter would collect the POC, leaving the second filter to adsorb only DOC (if that were occurring). From this process, we estimated the DOC blank on the filters to be 0.12 µmoles CO₂, representing 0.3 to 2% of the amount of carbon on the filters.

For water-column CH₄ and its δ¹³C isotopic composition, water samples were collected by CTD-Rosette and dispersed to glass bottles in June and July of 2013 at two sites in the northern GOM at 28.669°N; 88.3584°W, and at 28.32554°N and 88.3865°W (Figure 2). Methane concentrations were determined by the methods detailed in Magen et al. (2014). The stable isotopic composition of water-column methane was determined from 4-L bottles preserved with KOH as described in Magen et al. (2014). A headspace was introduced into the bottles and flushed into cryogenic trapping system, cryo-focused and run on a Thermo Finnegan Delta V Isotope Ratio Mass Spectrometer at Florida State University. The results were converted into δ¹³C with respect to VPDB (Vienna Pee Dee Belemnite).

The weighted average of the NCG POC_susp per collection year for both Δ¹⁴C and δ¹³C was calculated to estimate the inventory of carbon and their isotopic signatures in the water column. The weighted average of radiocarbon was calculated by multiplying the total µmoles CO₂ sent for radiocarbon analysis for each year by the individual sample size of µmoles CO₂. This proportion was then multiplied by the Δ¹⁴C for each sample and summed for the overall weighted average for that year. For the δ¹³C samples, the value for mg C per subsample used for δ¹³C analysis was calculated from the regression of the %C and δ¹³C of known standards. This value was extrapolated to the whole filter, assuming the sample was spread evenly. The value for total mg C for each sample was converted to µmoles and then summed following the radiocarbon method above.

A two-endmember mixing model based on Δ¹⁴C was used to estimate the percent carbon incorporated from photosynthesis and from all petrocarbon sources, combining methane and oil. The following equation was used to determine the percent carbon from modern surface production, with the denominator equaling the total range of radiocarbon from 39‰ (for marine production; Chanton et al., 2018) to –1000‰:

The percent from hydrocarbons is 100 – %_modern. We completed a sensitivity test for the two-endmember model by adjusting the total range of Δ¹⁴C by the standard deviation of the modern endmember. The Δ¹⁴C of the modern endmember was the average GOM plankton value from Chanton et al. (2018), 39 ± 26‰ (n = 79).

In addition to the two-endmember mixing model we also employed the Bayesian mixing model MixSIAR Version 3.1 (Stock and Semmens, 2016) to determine the contribution from other potential sources of organic carbon in the GOM. This R version uses probabilities from a Bayesian method to estimate the percent contribution from multiple sources. Using Δ¹⁴C and δ¹³C as our tracers, we assumed no fractionation between the sources and the isotopic signatures of the POC. We report the contribution of each source at the mean, including the standard deviation, resulting after 3 × 10⁶ iterations. Our model was run nesting depth within site type, including the residual * process errors with uninformative priors. Our carbon sources included (mean ± standard deviation [S.D.] for Δ¹⁴C and δ¹³C, respectively) surface productivity (38.8 ± 25.8‰, and –21 ± 2‰; Chanton et al., 2012, 2018), riverine organic carbon (–154 ± 68‰ and –26 ± 1‰; Cai et al., 2015), sedimentary organic carbon (–200 ± 29‰, and –22 ± 1‰; Chanton et al., 2015), biogenic methane in the water column (38.8 ± 25.8‰, n = 79, and –41 ± 1‰, n = 19; this study), as well as DWH methane (Δ¹⁴C = –1000‰ and δ¹³C = –57.4‰; Crespo-Medina et al., 2014) and Macondo oil (Δ¹⁴C = –1000‰ and –27‰; Graham et al., 2010).

Statistics were performed using R 3.1.3 R Core Team (2015). The data were not normally distributed, as indicated by the results of Levene’s test. Therefore, the non-parametric Mann-Whitney U Test was used to compare the different groups for both δ¹³C and Δ¹⁴C. We continued the designation of seep versus NCG (non-seep, north-central Gulf) and divided our samples further into two depth categories (euphotic <300 m and deep water >300 m). The four categories we examined were as follows: seep euphotic, seep deep, NCG euphotic, and NCG deep. We used the adjusted p-value of 0.008 to indicate significance, accounting for the multiple comparisons (α = 0.05, 6 comparisons). From the z value we calculated the effect size:

where r is the effect size, z is the z score, and N is the sample size.

Overall, we collected 277 POC_susp samples in the GOM. The δ¹³C signatures (n = 277), and Δ¹⁴C values (n = 123) are summarized in Table 1. POC_susp from the NCG euphotic zone (<300 m) had δ¹³C signatures ranging from –17.8 to –35.4‰ (n = 108), with Δ¹⁴C signatures from 54 to –515‰ (n = 55). Deep-water particles (>300 m) from the NCG had δ¹³C signatures ranging from –20.6 to –35.2‰ (n = 65), with Δ¹⁴C signatures from –48 to –603‰ (n = 39). Seep euphotic POC_susp had δ¹³C signatures ranging from –19.3 to –28.8‰ (n = 62), with Δ¹⁴C signatures from 71 to –468‰ (n = 14). Deep-water suspended particles from seep sites had δ¹³C signatures ranging from –21.4 to –34.3‰ (n = 33), with Δ¹⁴C signatures from –263 to –756‰ (n = 9). POC_susp collected from the euphotic zone in the western GOM (WGOM, west of 94° longitude) had δ¹³C signatures ranging from –22.8 to –29.2‰ (n = 8), with Δ¹⁴C signatures from 51 to –204‰ (n = 7). All data are provided in Table S1.

The δ¹³C and Δ¹⁴C signatures for all POC_susp samples, across all collection depths and sampling years, were highly variable, as seen in Figure 3 where the δ¹³C and Δ¹⁴C signatures of NCG POC_susp are color-coded by year. Suspended particles from seep sites were not color-coded by year, as the majority of the samples were collected in 2013. We observed two trends of δ¹³C and Δ¹⁴C co-variation in the POC_susp data (Figure 4). One trend, the lower limb, indicated co-variation based on blending of modern surface production and petrocarbon input, observed for both NCG and seep sites. The upper limb, observed primarily for NCG sites, reflected the addition of another carbon source to modern production, with depleted δ¹³C but more enriched Δ¹⁴C signatures, possibly associated with the biodegradation of biogenic methane produced in the water column (Rakowski et al., 2015). In this study, we used two different mixing models to characterize both the co-variation based on petrocarbon input and the variation based on all of the sources that could contribute to GOM POC_susp.

Study-wide concentrations of POC_susp followed an expected gradient with highest concentrations at the surface decreasing with depth (Figure 5). Concentrations in the euphotic zone (<300 m) ranged from 0.32 to 62.5 µM (mean ± S.D.: 4.71 ± 9.38, n = 78). The four highest concentrations at the surface came from non-seep sites that were heavily influenced by riverine input. Concentrations of POC_susp in deep water (>300 m) ranged from 0.16 to 12.02 µM C (mean ± S.D.: 1.44 ± 2, n = 72).

We tested the hypothesis that δ¹³C and Δ¹⁴C of POC_susp in the north central Gulf at non-seep sites would vary temporally, becoming increasingly enriched in heavy isotopes following their depletion associated with the 2010 injection of fossil hydrocarbons into the water column from the DWH oil spill (Cherrier et al., 2014; Fernandez et al., 2016; Weber et al., 2016). We calculated the weighted averages of the non-seep NCG particles collected each year to create an inventory that accounts for the quantity of organic carbon contributing to the δ¹³C and Δ¹⁴C signatures (Table 2). Data from Cherrier et al. (2014) were included in 2011 and 2012. Data were separated by depth, into NCG euphotic and deep water (Figure 6). The δ¹³C and Δ¹⁴C of euphotic POC_susp and δ¹³C of deep-water POC_susp (Figure 6A–C) exhibited depletion from 2010 to 2011. Afterwards the signatures increased in 2012, before decreasing again until 2014 and then increasing again and stabilizing by 2015. The Δ¹⁴C of deep-water particles (Figure 6D) showed the clearest indication of fossil carbon incorporation, over a longer time period from 2010 to 2012 with recovery from 2012 to 2014, stabilizing in the following years (2015–2017) at around Δ¹⁴C = –160‰. During the recovery period of 2012–2014, the linear regression calculated between these two years suggested a recovery rate (in ‰ per year towards more ¹⁴C enriched values) of 159‰ (Figure 6D).

Dissolved CH₄ concentrations in the water column varied from 2.6 to 11.6 nM, while the isotopic composition of methane varied from –37 to –52‰. A subsurface maximum in methane concentration was observed in both profiles within the euphotic zone at 60–75 m depth (Figure 7). The average δ¹³C value, weighting the two profiles equally, was –41.4 ± 1.0‰ (n = 19). For the mixing model, we assumed that this methane was produced from modern photosynthetic carbon, not derived from nearby seeps, and had a Δ¹⁴C value of 39 ± 26‰ (n = 79). The δ¹³C value we observed was similar to those measured in the Atlantic and Pacific oceans which varied between –43 and –45‰ (Holmes et al., 2000). Karl et al. (2008) have suggested that oceanic water column CH₄ is produced aerobically as a by-product of methylphosphonate decomposition in phosphate-stressed waters, supporting our assumption of a modern ¹⁴C value for this methane. Rakowski et al. (2015) observed depletion of phosphate in the euphotic zone at the methane maximum, where we similarly observed it. The ¹³C value is consistent with production from a methylated substrate in limited supply that is consumed quantitatively (Kelley et al., 2012; Tazaz et al., 2013).

The two-endmember mixing model indicated that the bulk of the carbon in POC_susp from the NCG euphotic (<300 m), seep euphotic and NCG deep (>300 m) sites was derived from modern surface production (Table 3). In contrast to seep euphotic, seep deep POC_susp had incorporated the most petrocarbon, averaging 57% ± 13 (n = 9), while NCG deep POC_susp averaged 29% ± 16 (n = 39) petrocarbon (Table 3). Our sensitivity test showed the greatest variation in the euphotic POC_susp, where Δ¹⁴C was close to the modern endmember. The two-endmember model estimations of the %modern for the euphotic test samples were within 2–3% of the model values, and the %petrocarbon estimations also varied by 2–3%. The deep-water source estimations varied by 0–1% for both %modern and %petrocarbon.

The MixSIAR model (Table 3) suggested that POC_susp in the north central GOM is heavily derived from riverine inputs (up to 46%), followed by modern surface production (up to 45%), and, at seep deep sites, by oil (up to 46%). All deep-water POC_susp had higher contributions from hydrocarbons than euphotic POC_susp. Euphotic POC_susp at seep sites also had higher contributions from oil-derived carbon than at NCG euphotic sites; i.e., 8.2 ± 3.6% (n = 14) compared to 2.0 ± 1.2% (n = 55). Sediment, fossil methane, and biogenic methane contributed very little to the organic carbon in POC_susp, with high standard deviations of their means.

To characterize POC_susp in the broader context of the GOM, we plotted the δ¹³C and Δ¹⁴C signatures for other carbon reservoirs in the GOM, including: POC_sink, non-seep sediment, seep sediment, and DIC (Figure 8). We also included signatures (as provided in Methods) for the different potential carbon sources to these pools, including modern surface production, riverine input, sediment, biogenic methane, DWH methane, and oil (Figure 8).

The data were non-normally distributed; therefore, the non-parametric Mann-Whitney U tests were used to analyze the variation of δ¹³C and Δ¹⁴C signatures of POC_susp, from 2010 to 2017. We compared the importance of depth in the water column (euphotic <300 m or deep >300 m) and site type (NCG or seep) in determining the isotopic signatures of POC_susp (Tables 4 and 5) using the adjusted p-value of 0.008 to define significance. There were significant differences in δ¹³C signatures when comparing between depths, regardless of site classification. For instance, the median values for δ¹³C of POC_susp from NCG euphotic (–24.1‰) and NCG deep (–27.0‰) were significantly different (U = 1656, p < 0.001, r = 0.43), as were those for NCG deep (–27.0‰) and seep euphotic (–24.8‰) (U = 992, p < 0.001, r = 0.42). Comparisons δ¹³C of POC_susp involving the same depth zones yielded similarities: the median value for NCG euphotic (–24.1‰) did not differ from that for seep euphotic (–24.8‰) (U = 3065, p = 0.27, r = 0.08), nor did the median for NCG deep (–27.0‰) differ from that of seep deep (–27.0‰) (U = 1034, p = 0.96, r = 0.003). For radiocarbon, comparisons between different depth zones revealed significantly different Δ¹⁴C signatures, except when comparing median values for NCG deep (–224‰) and seep euphotic (–78‰), which were similar (U= 132, p = 0.007, r = –0.37). POC_susp from the NCG euphotic and euphotic seep (medians of –60‰ and –78‰, respectively) were the same (U = 359, p = 0.56, r = 0.07), but unlike δ¹³C, there were significant differences between NCG deep (median = –224‰) and seep deep (median = –552‰) (U = 28, p < 0.001, r = 0.56) (Table 5).

The first goal of this study was to characterize any temporal trends in the δ¹³C and Δ¹⁴C signatures of suspended POC in the northern Gulf of Mexico following the DWH oil spill in 2010. Isotopically depleted values for POC_susp indicating a fossil petrocarbon source were observed following the spill (Figure 6), particularly in the ¹⁴C content of suspended particles below the euphotic zone at NCG sites. The influence of apparent DWH-derived material was greatest in 2011 and 2012, with recovery beginning thereafter and proceeding until 2014 when the isotopic composition of POC_susp reached an asymptotic value (Figure 6).

Yan et al. (2016), Chanton et al. (2018), and Geiring et al. (2018) described the effects and recovery of the POC_sink pool following the blowout. Yan et al. (2016) found elevated levels of barium, a component of drilling mud, in POC_sink from August 2010 through January 2011, when they returned to pre-spill baseline levels. Chanton et al. (2018) found that POC_sink recovered in 1–3 years depending on the tracer that was evaluated. δ³⁴S and PAH indicated an approximate 2-year recovery time, while Δ¹⁴C indicated a recovery time of ~3 years. These recovery periods are on a similar time scale to our estimate of a 4-year recovery period in the Δ¹⁴C of POC_susp in deep water at NCG sites (Figure 6).

A second goal was to determine the extent of fossil carbon influence on suspended particles. We found a wide range of natural variability in both δ¹³C and Δ¹⁴C signatures of POC_susp across the northern GOM, from seep and non-seep sites and a range of depths, from surface to 1900 m (Figure 3). Compared to suspended particles in the Sargasso Sea and Pacific Station M, the GOM exhibits more variability and greater depletion in δ¹³C and Δ¹⁴C signatures (Figure 1; Druffel et al., 1992, 1996). The primary carbon source to the Sargasso and Pacific particulates is modern photosynthetic production, which does not create a co-variation of δ¹³C with Δ¹⁴C. The depletion in Δ¹⁴C in the Sargasso and mid-Pacific was not observed to be greater than –100‰, even to depths of 4000 m. The δ¹³C value of the Pacific and Atlantic suspended particles was generally –20 to –22‰ without depth variation. The co-variation of δ¹³C and Δ¹⁴C observed in the GOM and the Mid-Atlantic Bight (as presented in the Introduction) are due to the incorporation of a second source that is depleted in both ¹⁴C and ¹³C, consistent with petrocarbon-derived material. In the Southern Gulf of Mexico from 20° to 22°N, Gonzalez-Ocampo et al. (2007) reported POC_susp depth trends in δ¹³C, with ¹³C depletion at depth reaching values as low as –23.7 ± 0.5‰. Values in surface water were –22.5 ± 0.5‰.

We admit some reservations about the results of the MixSIAR mixing model, as two of our sources, riverine input and sedimentary organic carbon had some surprising results. The MixSIAR model was used to constrain six carbon sources with two isotopic tracers. The riverine input (34–46%) was estimated to be greater than the input from surface production (11.6–45.1%) for all sites except NCG euphotic (riverine: 41%; surface production: 45%). For the workings of the model, the riverine endmember was located isotopically in the middle of the bulk of the POC_susp values, rather than along the boundary of our data (Figure 4). This location could cause the model to estimate a higher percent contribution to POC_susp because of the isotopic similarities between the POC_susp and riverine input. The data might well represent mixing between modern carbon and a more depleted source. On the other hand, the apparent high riverine contribution may be due to the differences in lability between the riverine and modern sources. Riverine carbon is less labile than fresher photosynthetic production, which causes it to cycle more slowly than surface production (Wang et al., 2004). The salinity at the stations where we collected POC_susp varied (Table S1), but most sites did not indicate major mixing with freshwater sources. However, Wang et al. (2004) found that δ¹³C and C:N ratios of POC_susp from the Mississippi and northern GOM exhibited non-conservative behavior when mixing with higher salinity waters. A decoupling occurs between the POC_susp and the freshwater input which allows the 2.30 Tg POC yr^–1 exported from the Mississippi (Cai et al., 2015) to mix with GOM POC_susp and accumulate over the POC_susp residence time of 5–10 years, increasing the contribution from riverine POC_susp.

Whereas the riverine contribution to POC_susp was higher than expected, the sedimentary contribution was low and consistent throughout the water column. Chanton et al. (2018) found that sinking POC, collected from traps 30 m above the seafloor, did not carry a strong signal from resuspended sediment, but POC_susp integrates over longer time scales. Diercks et al. (2018) detected both small-scale and hurricane-sized resuspension events, which could play a role in the transport of sedimented petrocarbon. The number and overall scale of resuspension events in the GOM is unknown, but they potentially introduce more than 4–6% into the POC_susp pool, especially near the seafloor. The sedimentary organic carbon endmember is similar in isotope space to riverine input (Figure 4), making the two sources difficult to separate. Employing sulfur isotopes and lithogenic silica content might better separate sedimentary and riverine sources.

Unlike the MixSIAR model, the two-endmember model is simpler and better constrained. This model indicated the local influence of seeps, particularly on deep-water POC_susp, and the importance of modern surface production on POC_susp away from seeps (Table 3). Nonetheless, petrocarbon still contributed about 30% of POC_susp in the deep GOM, even away from seep sites, a phenomenon not observed in the Atlantic or Pacific. Even though the two-endmember model does not capture the full complexity of all of the potential sources to POC_susp in the GOM, the results of the MixSIAR model, using only two isotopic measurements, can only be fully interpreted with reservations, for the aforementioned reasons. Therefore, we have greater confidence in the two-endmember approach.

We estimated the new, post-spill baseline signatures for NCG (non-seep) POC_susp in the GOM to be δ¹³C = –24.1 ± 0.4‰ (n = 15) and Δ¹⁴C = –17 ± 36‰ (n = 15) for the euphotic zone and δ¹³C = –27.5 ± 1.1‰ (n = 12) and Δ¹⁴C = –164 ± 19‰ (n = 7) for deep-water suspended particles. These values are the means of the POC_susp weighted averages from the NCG for the last three sampling years (2015–2017; Figure 6). Following the recovery of deep-water Δ¹⁴C signatures in 2014, we suggest that POC_susp reached a post-blowout baseline, as both the δ¹³C and Δ¹⁴C signatures had stabilized by these years. In comparison to baselines from the Sargasso Sea and Pacific Ocean, the baseline for POC_susp from the GOM was more depleted for δ¹³C and Δ¹⁴C. We estimated the δ¹³C and Δ¹⁴C baselines for POC_susp collected in the Sargasso and Pacific (Druffel et al., 1992, 1996, 2003) for our depth zones to be: euphotic δ¹³C = –22.1‰ (n = 41) and Δ¹⁴C = 73 (n = 42); and deep POC_susp δ¹³C = –21.2‰ (n = 82) and Δ¹⁴C = 16‰ (n = 83). The differences in δ¹³C between the GOM and the Sargasso and Pacific are due to differences in the increased relative importance of hydrocarbon and terrestrially sourced material to the GOM, while the Δ¹⁴C differences are caused by both collection years and the relative importance of differing carbon sources. The gap of 22–32 years between the collection of the Sargasso and Pacific samples and our samples results in the older samples being more enriched in ¹⁴C. This effect is due to the atmospheric nuclear testing that caused radiocarbon values to spike and mix from the atmosphere into the biosphere and hydrosphere, including the ocean. The mixing process caused a lag between atmospheric CO₂ radiocarbon signatures and DIC in the ocean (Levin and Hesshaimer, 2000; McNichol and Aluwihare, 2007).

Unlike POC_susp from the Sargasso Sea and Pacific Ocean, depth played a significant role in the variation of δ¹³C of POC_susp in the GOM over the period of 2010–2017. The significant differences observed between euphotic (<300 m) and deep (>300 m) reflect the relative importance of the two sources, with petrocarbon-derived material increasing in importance below 300 m (Tables 2 and 4). Suspended particles in deep water were depleted due to the hydrocarbon input potentially from seep sites and/or the DWH blowout, while POC_susp from the same depth (whether euphotic or deep) had similar δ¹³C, regardless of site type (seep or NCG).

We also found significant differences between the Δ¹⁴C of POC_susp in euphotic and deep POC_susp for both site types (seep or NCG) (Table 5). This finding suggests that suspended particles in the GOM, including at non-seep sites, are more depleted in ¹⁴C in deeper waters than suspended particles in other oceans, probably due to the natural hydrocarbon seepage in the GOM, although completely ruling out the lingering petrocarbon from the oil spill is difficult at this point. The significant differences between the Δ¹⁴C of deep-water POC_susp from NCG and seep areas were not observed for the δ¹³C results, as δ¹³C is less sensitive to variations in the input terms (Bosman et al., 2016). The more depleted Δ¹⁴C of POC_susp found at seep sites suggests that the presence of natural hydrocarbon seepage significantly affects the Δ¹⁴C of suspended particles in the Gulf.

Based on our assessment of the δ¹³C and Δ¹⁴C signatures of POC_susp, petrocarbon is an important carbon source to suspended particles in the GOM, in contrast to the situation in the Atlantic and Pacific oceans. In the north central Gulf during our study period, POC_susp was isotopically depleted relative to POC_sink or sediment (Figure 8). DIC fixed by marine primary production is the primary source of carbon to POC_sink and to the non-seep sediment. We found deep POC_susp from seep sites to be composed of about 45% oil-derived petrocarbon, while deep-water POC_susp from the NCG sites may contain as much as 15% oil-sourced petrocarbon and 3.5% methane-sourced petrocarbon. In the Gulf of Mexico, there are at least three clear and separate carbon sources (Figure 8): one driven by surface primary production, observed in POC_sink and non-seep sediments; the second attributed to the microbial cycling of methane and oil, observed to a greater extent in POC_susp and seep sediments; and the third, a riverine contribution, pervasive but more challenging to quantify unambiguously.

Data deposition: Data are publicly available through the Gulf of Mexico Research Initiative Information & Data Cooperative (GRIIDC) at https://data.gulfresearchinitiative.org, 10.7266/N7FX77C7.

AMS samples were run at Lawrence Livermore Center for Accelerator Mass Spectrometry, the University of Georgia Center for Applied Isotopic Studies, and the National Ocean Sciences Accelerator Mass Spectrometry facility (NOSAMS) at Woods Hole Oceanographic. We thank Alexander Cherinsky, Tom Guilderson, Ann McNichol, Kathryn Elder, and Mark Roberts. Samples were collected from the RV Endeavor, RV Pelican and RV Weatherbird and we thank their crews. We thank Burt Wolff and Yang Wang for use of the facilities at the National High Magnetic Field Laboratory, which is supported by National Science Foundation Cooperative Agreement No. DMR-1644779 and the State of Florida. We especially thank the editor, Dr. Jody Deming, for her thorough review of our work and her detailed attention to our manuscript. We thank associate editor Lisa Miller, reviewer Brett Walker, and an anonymous reviewer.

This research was made possible by grants from The Gulf of Mexico Research Initiative through its consortiums: Ecosystem Impacts of Oil and Gas Inputs to the Gulf (ECOGIG), The Center for the Integrated Modeling and Analysis of the Gulf Ecosystem (C-Image), and Deep Sea to Coast Connectivity in the Eastern Gulf of Mexico (Deep-C). This is ECOGIG Contribution #523.

The authors have no competing interests to declare.

• KLR wrote the paper with JPC, collected samples, analyzed samples and data

• SB edited the paper, collected samples, analyzed samples and data

• SW collected samples, analyzed samples

• CM collected samples, analyzed samples

• JPM collected samples analyzed data

• JPC co-wrote the paper, analyzed data