Technical Comments

Response to Comment on “A Persistent Oxygen Anomaly Reveals the Fate of Spilled Methane in the Deep Gulf of Mexico”

See allHide authors and affiliations

Science  27 May 2011:
Vol. 332, Issue 6033, pp. 1033
DOI: 10.1126/science.1203428


We hypothesized that methane from the Deepwater Horizon oil spill was quantitatively consumed and presented results from four tests supporting this finding. Subsequent published studies provide further support for our conclusions. We refute the criticisms by Joye et al., which are incorrect, internally contradictory, based on flow-rate estimates that exceed consensus values, and overall do not disprove our hypothesis or invalidate its underlying assumptions.

We proposed and performed four tests of the hypothesis that nearly all methane released from the 2010 Deepwater Horizon (DWH) oil spill was consumed by bacteria before September 2010 (1). Implicitly, we assert that the methane released no longer remains in the Gulf of Mexico. Joye et al. (2) levy a range of criticisms against the tests of our hypothesis performed so far (1, 36) but do not present data refuting our hypothesis.

The first two tests of our hypothesis were an exhaustive search for methane, and in its absence the quantification of the resulting dissolved oxygen (DO) anomaly, both spanning a 100,000 km2 area that included the DWH plume. The DO anomaly was identified, integrated over the spatial extent of the plume, and found to be of sufficient magnitude to account for respiration of all DWH methane. Although the predominant direction of plume propagation was to the southwest, Fig. 1 (1, 4) shows that we also sampled north and east of the plume, an area that Joye et al. indicate we excluded.

Fig. 1

(A) Google Earth image with the Mississippi Canyon highlighted in yellow. The green (n = 20) and red (n = 21) circles represent stations with and without DO loss from plume depths, respectively. (B) Profiles of stations containing no DO anomaly. (C) The time-dependent DO anomaly displays the DWH plume passing by Mississippi Canyon. The available evidence is not consistent with a stationary seep feature as asserted by Joye et al. (2). Data are from (12, 13).

Joye et al. (2) suggest that the expected DO anomaly in the plume horizon could total 7 × 1010 to 12 × 1010 moles of DO decline, roughly two- to four-fold larger than we identified [3.0 × 1010 to 3.9 × 1010 moles (1)], implying that a large portion of the hydrocarbons, including CH4, were not oxidized and were thus missed. However, their DO calculations hinge on high estimates of hydrocarbons emitted. The lower bound of oil flow rate cited by Joye et al. (2, 7) exceeds the final federal estimate of 58,000 barrels of oil per day (bopd), determined by the Flow Rate Technical Group (8). Joye et al.’s upper bound [84,000 bopd] is based on a preliminary estimate and lacks a valid scientific basis (9, 10). Even if we accept the high emission rates (7), the corresponding DO demand is not 7 × 1010 to 12 × 1010 moles but 5 × 1010 to 10 × 1010 moles (11); their revised lower bound is only 28% higher than the upper bound we estimated, not 10-fold as Joye et al. (2) claim.

Findings published after our original paper (1) provide additional evidence that we correctly identified the DWH plume and that the methane had been removed. Specifically, Kujawinski et al. (3) show that subsurface applications of dioctyl sodium sulfosuccinate (DOSS), a major component of the dispersant added at the wellhead, traced the deep plume, not mixing appreciably with DOSS added at the sea surface; contrary claims (2) are therefore not supported. Interestingly, for near-field plume samples collected in June (4), CH4 and DOSS concentrations were correlated (3), as were observed DO anomalies and DOSS concentrations in far-field samples collected in September [0.008 μg DOSS/μmol DO anomaly (n = 8; R2 = 0.62) for paired samples (1, 3) within 250 km of the wellhead]. Scaling these relationships to the average daily application rate and the total subsurface addition of DOSS (~2.9 × 108 g), respectively, yields a CH4 emission rate and a total DO deficit within the ranges of our initial estimates (1, 3) but below those of Joye et al. (2, 7).

We do not contend that the DO anomaly is homogeneous throughout the plume [figure 1B and figure S3 in (1)], but rather expect a heterogeneous distribution dependent on water mass circulation and mixing. To assess the suggestion that the heterogeneous DO anomalies we observed could be natural DO depletions in the Mississippi Canyon, we plotted 40 additional DO profiles measured by others, including Joye et al. (12, 13) (Fig. 1). In many profiles, DO depletion was undetectable, and among the others we found no temporally or spatially consistent DO anomalies, arguing strongly against the existence of a persistent natural feature selectively influencing the plume depth in this area. Moreover, except for this disaster, no authors have previously reported DO anomalies of this nature. Thus, the alternatives suggested by Joye et al. (2) are without experimental support, inconsistent with available data (Fig. 1), and highly unlikely in that it would have to rapidly, coincidentally, and selectively affect waters at 1000- to 1200-m depth.

Joye et al. (2) state that we presented no complementary evidence in support of the CH4-oxygen mass balance and that previous oxidation rate data (4) showed slow rates, inconsistent with the modeled high rates. This is incorrect. CH4 oxidation rates were measured on the September expedition, and the summarized data were published (1). The CH4 oxidation rates presented previously (4), although generally slow, clearly displayed the exponential increase our model estimated (1) as the more labile substrate, ethane, was drawn down [figure 4B in (4)]. A few samples approached a rate constant of k = 0.1 day−1, similar to the 0.2 day−1 maximum we predicted (1, 4). Also, CH4 isotope ratios were measured in June 2010 (4), but the vanishingly low methane concentrations in the September 2010 samples precluded such measurements. Furthermore, mass balance calculations show that the predicted isotopic perturbations of bulk carbon pools would be weak [∆δ13C-DIC ~ 0.06 per mil (‰), ∆δ13C-DOC ~ 2 to 3‰ (14)], even using Joye et al.’s high emission estimates (7) and assuming all the plume hydrocarbons were converted to either dissolved inorganic carbon (DIC) or dissolved organic carbon (DOC).

A third test of our hypothesis was to determine whether the bacterial community was anomalously enriched with CH4-oxidizing bacteria, which it was. We contend that our molecular biological approach had far greater measurement sensitivity than the chemical approach for defining impacted waters and that samples collected from outside the chemically defined plume cannot be considered “control” samples for biological measurements as Joye et al. suggest. Our argument instead depends on the comparison between samples collected from the chemical plume in May (15), June (4), and September 2010 (1). Methanotrophs and methylotrophs were not detected in May or June samples but were abundant in September, suggesting that a major change in bacterial community structure accompanied methane’s disappearance. In addition, although cultured members of the Methylophilaceae and Methylophaga spp. are known to oxidize methanol but not methane, isotope tracer experiments have consistently shown that these groups incorporate 13C from 13CH4, presumably through cross-feeding on methanol released by methanotrophs, to the extent that methylotroph sequences can outnumber Methylococcaceae sequences in 13C-labeled DNA (1618). We thus find the change in microbial community composition to be consistent with a methanotrophic bloom between June and September 2010. The simplest explanation is that this bloom fed on DWH methane. Suggestions that industrial methanol release or oil degradation fed the methylotrophic bloom (2) are inconsistent with available information and are unlikely given the mass dominance of methane.

Because no data have been published showing the far-field bifurcation of the oil-gas plume, and our plume data also argue against this possibility, a fourth test was to model only the time-dependent first-order CH4 oxidation rate constant necessary for complete consumption of the CH4 released to the plume. Although the peak rate constant produced by the model was slightly higher than previously published values (19), this small difference is as expected given the much higher initial methane concentrations in the DWH output (4, 7). Notably, a more sophisticated oil-spill model prior to our work (20) incorporated the published rate constant (19) with which our modeled rates agree most closely.

Lastly, Joye et al. (2) argue that, contra their contention that natural seepage in the Mississippi Canyon was mistaken for the DWH plume, natural emission does not mimic the DWH. However, Joye et al. have themselves acknowledged the relevance of the DWH CH4 release to the geological past (7). We agree that caution is needed when extrapolating our findings beyond the deep ocean— an extrapolation that we therefore did not make (1). Our results speak to the kinetics of the methanotrophic response to enhanced substrate concentration in the deep ocean, whether substrate injection is natural or industrial. Moreover, the reproducibility of the one previous study that found natural deepwater hydrocarbon releases to bypass the methanotrophic biofilter and enter the atmosphere (21) has also been questioned (22). Given that most CH4 released in the deep ocean appears to be removed by respiration rather than atmospheric equilibration, we expect conclusions drawn from methane oxidation after the DWH event to be applicable to natural deepwater releases.

References and Notes

  1. The flow rate attributed to Leifer (9) was taken from a preliminary report that was intended only for the purpose of aiding the response team in assessing the extent of the spilled oil for ongoing response efforts; the report provided no data to support the claimed flow rate of 84,000 bopd. This rate exceeds the report’s own consensus value of 35,000 to 45,000 bopd, as well as the report’s joint (with the Department of Energy) estimate of 35,000 to 60,000 bopd made to the National Incident Command. The propagation of this value by Joye et al. (2, 7) is therefore not justified.
  2. Using the gas stoichiometry and concentrations presented in (7), as well as a molecular weight of oil of 14 g/mol and a stoichiometric relationship of DO:oil of 1.5:1 (1, 4), their high hydrocarbon emission rates predict a smaller DO demand of 5 × 1010 to 10 × 1010 moles.
  3. Parameters used: plume area = 100,000 km2; plume thickness = 200 m; total oil and gas hydrocarbons released = 2.6 × 1010 to 4.7 × 1010 moles; [DIC]Background ≈ 2000 μM; [DOC]Background ≈ 40 μM; ∆[DIC] or ∆[DOC] = 1 to 2 μM; ∆δ13C-DIC ≈ 0.06‰; ∆δ13C-DOC ≈ 2 to 3‰.
  4. Acknowledgments: This research was supported by NSF through grants OCE 1042650 and OCE 0849246 to J.D.K. and OCE 1042097 and OCE 0961725 to D.L.V., and by U.S. Department of Energy grant DE-NT0005667 to D.L.V. The August, September, and October expeditions were funded by the National Oceanic and Atmospheric Administration through a contract with Consolidated Safety Services, Inc.
View Abstract

Stay Connected to Science

Navigate This Article