Methane release from the southern Brazilian margin during the last glacial

Seafloor methane release can significantly affect the global carbon cycle and climate. Appreciable quantities of methane are stored in continental margin sediments as shallow gas and hydrate deposits, and changes in pressure, temperature and/or bottom-currents can liberate significant amounts of this greenhouse gas. Understanding the spatial and temporal dynamics of marine methane deposits and their relationships to environmental change are critical for assessing past and future carbon cycle and climate change. Here we present foraminiferal stable carbon isotope and sediment mineralogy records suggesting for the first time that seafloor methane release occurred along the southern Brazilian margin during the last glacial period (40–20 cal ka BP). Our results show that shallow gas deposits on the southern Brazilian margin responded to glacial−interglacial paleoceanographic changes releasing methane due to the synergy of sea level lowstand, warmer bottom waters and vigorous bottom currents during the last glacial period. High sea level during the Holocene resulted in an upslope shift of the Brazil Current, cooling the bottom waters and reducing bottom current strength, reducing methane emissions from the southern Brazilian margin.

these species can live in a methane-laden environment and that the δ 13 C of calcareous benthic foraminiferal tests are influenced by the δ 13 C signature of methane 17 . Thus, anomalously negative δ 13 C excursions in fossil foraminiferal carbonate from sedimentary records are used to assess the geologic history of methane seepage 6,7,[18][19][20] . Additionally, tests of planktonic foraminifera deposited in sites influenced by methane seepage may also show negative δ 13 C excursions due to carbonate overgrowth 21 .
The southern Brazilian margin (SBM) is a passive hydrocarbon basin where several circular and elliptical depressions of collapsed sediments (i.e. large pockmark fields) have been associated with methane release from hydrate reservoirs and shallow free gas layers 5,22,23 . Multi-channel seismic data revealed complex subsurface faults, fractures and channels connecting deep and shallow gas hydrates as well as shallow subsurface free-gas deposits to the pockmarks 5,23 . The presence of chemosynthetic communities 24,25 and high biogenic methane concentrations 23 are suggestive of modern active cold methane seeps on the SBM, while the presence of authigenic carbonate nodules with δ 13 C values between −16‰ and −30‰ likely resulted from past methane release 26,27 . However, only a few studies have examined methane seepage history in the Brazilian margin 26,27 .
To determine the temporal evolution of past seafloor methane seepage at the SBM, we examined the δ 13 C signatures of benthic (Cibicides spp. and Uvigerina spp., ≥250 μm) and planktonic (Globigerinoides ruber and Globigerinoides sacculifer, ≥250 μm) foraminifera from marine sediment core GeoB6201-5 (26°40′S, 46°26′W, 475 m water depth, 235 cm long) collected at a funnel-shaped depression (i.e. pockmark) located on the SBM ( Fig. 1 and supplementary Figure S1) 28 . The pockmark is placed at the boundary between two southern sourced water masses, the denser Antarctic Intermediated Water (AAIW) and the less dense South Atlantic Central Water (SACW) that are separated by the 27.1 isopycnal at ca. 500 m water depth 29 (Fig. 1B; see supplementary information for further details). We compare δ 13 C records from core GeoB6201-5 to δ 13 C records from nearby cores collected outside of the pockmark, namely GeoB2107-3 (27°17′S, 46°45′W, 1048 m water depth, 780 cm long) and 14GGC (26°68′S, 46°50′W, 441 m water depth, 200 cm long) 30 (Fig. 1A and supplementary information). The δ 13 C records from these two sediment cores provide the regional background conditions not affected by methane seepage. The presence of authigenic carbonate precipitation derived by the anaerobic oxidation of methane (AOM) in core GeoB6201-5 was determined by the elemental composition of individual benthic foraminiferal tests (Cibicides wuellerstorfi) as well as the mineralogical composition of sediments (see Material and methods).
The age model of GeoB6201-5 is based on six AMS 14 C ages performed on planktonic foraminifera G. ruber (supplementary Table S1), regional planktonic foraminiferal biostratigraphy and stable oxygen isotope (δ 18 O) Figure 1. Location of core GeoB6201-5 (this study) and adjacent control cores GeoB2107-3 64 Fig. S2). The chronology of core GeoB6201-5 is additionally supported by planktonic foraminifera biostratigraphy 34,35 , where the presence of Globorotalia menardii and the low abundance of Globorotalia inflata indicate Biozone Z (Holocene), while the absence of G. menardii and the high abundance of G,inflata characterize the glacial Biozone Y (last glacial) (supplementary Fig. S2b). The chronology of core GeoB2107-3 was previously published in Heil 36 and Gu 37 .
The absence of anomalously negative δ 13 C values in contemporaneous foraminifera from the adjacent cores (GeoB2107-3 and 14GGC 30 ; Fig. 2B-C) rules out the possibility of an enhanced primary productivity event and associated increased carbon flux to the seafloor (the so called Mackensen effect) 40 as well as the influence of a bottom water mass with significantly lighter δ 13 C 30,39 as the main sources for the observed negative δ 13 C excursions recorded in core GeoB6201-5. The anomalously negative benthic foraminiferal δ 13 C values in our records are consistent with those previously reported from modern and past cold-methane seeps (supplementary Table S2) 13,14,19,[41][42][43] . For example, living epifaunal specimens of Cibicidoides wuellerstorfi from a cold seep in Monterey Bay off California show mean δ 13 C values as low as −3.3‰ 14 while the same species from our core GeoB6201-5 had δ 13 C values down to −3.2‰ (Fig. 3B). The infaunal species U. peregrina, from a Blake Ridge seep diapir in the NE-Atlantic had a mean δ 13 C of −4.2‰ for the late Holocene 43 while U. peregrina specimens from our core GeoB6201-5 had values down to −4.6‰ (Fig. 3B). This suggests that the carbonate tests of foraminifera were precipitated under the influence of low δ 13 C values of dissolved inorganic carbon (DIC) due to sulfate-dependent AOM [44][45][46] .
Depleted-δ 13 C signals in foraminifera shells can be also related, in part, to authigenic carbonate overgrowth induced by AOM 20,21,41,47 . Our specimens of planktonic foraminifera also show anomalous negative δ 13 C values down to −6‰ (Fig. 2C), which are unlikely to have been caused by AOM in the water column, as suggested in previous studies 6,19 . Instead, gas methane is rapidly consumed by the methanogenic bacterial consortium in the water column and/or is deflected laterally by bottom currents 48 . Therefore, we suggest that the anomalously negative δ 13 C values in the foraminiferal shells of core GeoB6201-5 were partially affected by post-depositional authigenic carbonate overgrowth. Elemental analyses of C. wuellerstorfi from core GeoB6201-5 reveals an elevated content of magnesium (Mg) and sulfur (S) between 140 and 235 cm core depth (supplementary Table S3), which is supported by bulk sediment mineralogy that shows the presence of high-Mg calcite (HMC) at the same interval (Fig. 2). Elevated Mg and S as well as the presence of HMC closely matches the interval of depleted-δ 13 C in foraminifera ( Fig. 2 and Table S3), confirming that authigenic precipitation affected the original δ 13 C of foraminiferal shells. However, not all samples containing elevated Mg, S and HMC were associated with depleted δ 13 C. For example, shells analyzed at 195 cm core depth show δ 13 C values of −1.39‰ and an absence of Mg and S (supplementary Table S3), suggesting that the anomalously negative δ 13 C value represents the original test composition. On the other hand, mineralogical analyses indicated HMC in sediments at 180 cm core depth ( Fig. 2 and   Table S3). Our results suggest that AOM-derived authigenic carbonate precipitation may have added negative δ 13 C to some of the original foraminiferal shells, but this process was not responsible for all depleted δ 13 C values. Thus, the depleted-δ 13 C signals in core GeoB6201-5 evidence methane outgassing from the SBM, that occurred during the last glacial.
The AOM-depth in sediments at seep sites is mostly a function of fluid flow intensity; high flux displaces the AOM towards the seafloor while the AOM occurs deeper within sediments affected by low flux 49 . Authigenic carbonates containing aragonite (i.e. nodules, concretions and crusts) precipitate preferentially when the AOM occurs near the sediment-water interface due to a high methane flux 9,49,50 . No evidence of aragonite was observed in core GeoB6201-5, but the absence of aragonite in sediments can also be associated with post-depositional dissolution. Modeling experiments reveal that dissolution of aragonite in sediments starts just after the decrease in the methane flux and the halt in AOM, and ca. 2000 years would cause complete aragonite dissolution 51 . In addition, AOM-derived authigenic carbonates in methane seepage sites are generally enriched in δ 18 O 9,49,52,53 , which is not observed in our results (Fig. S6) Fig. S6). This interpretation is supported by the similarity of the post-glacial Cibicides spp. δ 18 O curve of seep core GeoB6201-5 and non-seep core 14GGC 30 , collected at a similar water depth (Fig. S6a). Interestingly, an anomalous offset (of 0.2-2.72‰) is observed after the 17.5 cal ka BP (Fig. S6b). The origin of this offset is unclear. Yet, no depleted δ 13 C values is observed at this interval. It is unlikely that appreciable thermal gradient differences existed within the top few cm, and it is also unlikely that diagenetic processes can account for the isotopic offset of taxa that experienced the same postdepositional environment. It seems more likely that the decoupling of δ 18 O signals between the epifaunal Cibicides species and the infaunal Uvigerina species results from different ambient water conditions where they lived (bottom water compared with pore water). We therefore argue that such changes in the isotopic composition of benthic foraminifera reflect bottom water conditions at the time the shells were calcified. Further analyses of benthic foraminifera assemblages from core GeoB6201-5 are needed to confirm our interpretation of changes in seafloor conditions at the pockmark during the last glacial cycle. The observed depleted δ 13 C values in core GeoB6201-5 indicate that the AOM occurred close and/or at the seafloor, suggesting methane seepage from the SBM during the glacial.
Core GeoB6201-5 was collected from a pockmark depression on the SBM (26°S), where multi-channel seismic reflection profiles reveal shallow subsurface overpressured gas deposits accumulated in stratigraphic traps and gaps between pockmark arrays sensitive to hydrodynamic processes on the seafloor 5 . These deposits are connected to paleo-pockmarks through deep-seated complex subsurface faults, fractures and channels, highlighting a genetic relationship between salt diapirism and pockmarks 5 . During the last glacial (40-20 cal ka BP), global relative sea level was ca. 60-120 m lower than modern sea level 55 (Fig. 3A). Hydrostatic depressurization over the sediment column probably contributed to expansion of the gas and the subsequent release of methane to the ocean. In addition to decreased hydrostatic pressure, the lower sea level displaced the upper water column structure downslope, bringing the core of the Brazil Current (BC) close to the core site location 56 . The BC increased local bottom temperatures, since surface and thermocline waters from the BC warmed up to 4 °C between 45 and 20 cal ka BP (Fig. 3B-C) 57,58 . Increased temperatures amplified the effect of decreased pressure on the shallow gas in subsurface sediments. The shift of the upper water column structure in relation to our core site is evident when the benthic δ 18 O record from core GeoB6201-5 (475 m water depth) is compared to the benthic δ 18 O record from adjacent core GeoB2107-3 located deeper in the water column (1084 m water depth) (Fig. S3e). The large glacial isotopic differences between benthic foraminifera from both cores (δ 18 O) suggests that core GeoB6201-5 was bathed by a warmer water mass while the core located deeper in the water column (GeoB2107-3) was exposed to a cooler water mass (Fig. 3E). In contrast, lower δ 18 O values occurring from the mid-deglaciation onwards suggest that both cores were bathed by a water mass with a similar δ 18 O signature. The downslope shift of the BC may also have intensified the bottom current velocity at site GeoB6201-5, which may have acted as a hydraulic pump 8,59 , facilitating the exhumation of gas methane from subsurface sediments to the ocean. The same mechanism was evoked to explain the pockmark formation in the Strait of Gibraltar (Mediterranean Sea) during the low sea level stand of the Last Glacial Maximum LGM) 8 .
After the LGM, the absence/reduced seepage of methane is evidenced by the more positive and less variable δ 13 C signals from core GeoB6201-5, with δ 13 C values becoming similar to those of modern foraminiferal calcite from non-seep environments (Figs 2 and 3B) (e.g. Curry and Oppo 39 and Chiessi et al. 38 ). The rising sea level increased the hydrostatic pressure on the seafloor and shifted the BC upslope decreasing bottom temperatures and current strength, thus reducing methane outgassing from the SBM (Fig. 3). The high sea level during the Holocene together with a decreased bottom current strength at our core site may have favored the deposition of pelagic fine-grained sediments, consequently forming a sealing layer that blocked the upward migration and allowed gas accumulation underneath the sediment, forming shallow subsurface deposits 5 . It is noteworthy that gas could also have migrated laterally within the sediments, not being recorded in our δ 13 C after the LGM, feeding nearby pockmarks as evidenced by current presence of chemosynthetic communities in this region 24 . This suggests that sea level changes and associated oceanographic processes played a key role on the stability of gas deposits in the SBM since ca. 40 cal ka BP.
Previously published evidence of past methane release from the Brazilian margin is limited to depleted δ 13 C values (−16 to −30‰) from authigenic carbonate nodules 26,27 . Some of these nodules from the southeastern Brazilian margin were U/Th-dated to 130-140 cal ka BP 27 , which are contemporaneous with an accumulation of warm water in the upper western South Atlantic Ocean that preceded Termination II 57,60 , as well as with a decrease in sea level 55 . Thus, available data are consistent with the notion that methane release from shallow subsurface gas deposits in the SBM occurred repeatedly over time, responding to glacial−interglacial changes in sea level and associated changes in the structure of the upper water column. The combination of lower sea level, warmer bottom waters and vigorous bottom current induced the destabilization of shallow subsurface gas deposits during the two recent glacial periods, while the higher sea level, relative cooling of bottom waters and reduced bottom current stabilized these gas deposits during the modern and last interglacials in the western South Atlantic Ocean.
In summary, we present the first foraminiferal δ 13 C records demonstrating methane outgassing in the SBM during the last glacial period. Our results show that shallow subsurface gas deposits in shallow water depths (ca. 500 m water depth) on the SBM are sensitive to changes in hydrostatic pressure, bottom water temperature and bottom current velocity over glacial-interglacial cycles. Additional research is needed to better understand the spatial-temporal paleoceanographic processes controlling methane gas dynamics and stability in the Brazilian margin. This is particularly important in the SBM where a huge hydrate deposit covering an area around 45.000 km 2 with water depths ranging from 500 to 3500 m, the so-called Rio Grande Cone 23 , may have also been sensitive to the changes postulated in this study.

Material and Methods
Marine sediment cores. We examined core GeoB6201-5 (26°40′S, 46°26′W, 475 m water depth, 235 cm long) 28 , taken from the center of a funnel-shaped depression (600 m diameter and 70 m high) and core GeoB2107-3 (27°17′S, 46°45′W, 1048 m water depth, 780 cm long) 61 collected outside of the pockmark, both from the SBM (Fig. 1). Core GeoB6201-5 is mainly composed of gray to very dark gray clay bearing nannofossil ooze, weakly to moderately bioturbated sediments with shell fragments and macrofossils 28  Chronology of the sediment cores. The chronology of sediment core GeoB6201-5 is based on six AMS 14 C measurements (supplementary Table S1), regional planktonic foraminifera biostratigraphy and oxygen isotope stratigraphy (supplementary Figs. S2, S3 and S4). The age model of GeoB2107-3 was published in Heil 36 and Gu 37 .  Table. S1). To avoid possible contamination from an old carbon source, the samples selected for 14 C analyses were chosen among the depths not affected by anomalous negative δ 13 C values, except for the basal age (235 cm). The age model and associated uncertainties were calculated using the R script BACON version 2.2 31 and the IntCal13 calibration curve 32 with a reservoir correction of 400 ± 100 years (supplementary Fig. S2). Beside the default parameters of the software, we used the following settings: mem.mean = 0.7, acc.shape = 0.8. and t.a = 8/t.b = 9. Sedimentation rates for core GeoB6201-5 are shown in supplementary Fig. S3.
Verifying the 14 C-based chronostratigraphy of core GeoB6201-5 with foraminiferal oxygen stable isotopes. We verified the 14 Fig. S3a). This approach supports the 14 C-based age model indicating that core GeoB6201-5 recorded the last 40 cal ka BP, corresponding to Marine Isotope Stages 1-3 (supplementary Fig. S4a).
Verifying the 14 C-based chronostratigraphy of GeoB6201-5 with planktonic foraminifera biostratigraphy. The presence/absence of planktonic foraminifera species and subspecies of Globorotalia menardii, the so-called "menardiform plexus", in marine cores has been widely applied as an indicator of paleoclimatic fluctuations of the late Quaternary, and the menardiform plexus is the main group of planktonic foraminiferal species used in the biostratigraphic zonation of this interval 34,35 . Ericson and Wollin 34 divided the Late Quaternary into 10 biozones (Q to Z) in accordance with the presence/absence of the menardiform plexus. The menardiform plexus was found to occur in the interglacial intervals (biozones R, T, V, and X), whereas its absence characterized the glacial intervals (biozones Q, S, U, W, and Y) of the late Pleistocene. As the Holocene is the latest interglacial, it was named zone Z (Fig. 1). In core GeoB6201-5, the absence of the menardiform plexus between 235 and 55 cm core depth (ca. 40-10 cal ka BP) is indicative of the glacial biozone Y of Ericson and Wollin 34 , while its presence after 55 cm core depth (ca. 10 cal ka BP) indicates the post-glacial biozone Z (Holocene) (supplementary Fig. S4b). In addition, the elevated abundance of G. inflata, a cold-transitional water species in the southwestern Atlantic Ocean 62 being regionally abundant during glacial biozone Y and virtually absent during the Holocene biozone Z 35 , further supports the 14 C-based age model of core GeoB6201-5 (supplementary Fig. S4b).
Mineralogical analyses. The bulk mineralogy of 13 samples of core GeoB6201-5 was determined by X-ray diffraction (XRD) (Bruker D8 Advance using Cu Kα radiation) at the Physics Institute/Fluminense Federal University. The samples were powdered before analyses and the diffractograms were produced with 2θ from 3° to 70°, with an increment of 0.02° and 0.1 s of reading time. Minerals were identified by comparison of the experimental pattern with tables from Brindley and Brown 63 .
Elemental composition of benthic foraminifera. Scanning electron microscope energy dispersive X-ray (SEM-EDX) analyses of 19 individual specimens of Cibicidoides wuellerstorfi (>150 μm) from 13 sediment samples of core GeoB6201-5 were carried out in the Institute of Chemistry/Fluminense Federal University. The analyzed specimens were cleaned by sonication for 20 s in methanol and subsequently in distilled water to remove the methanol. The samples were dried for 24 hours at 40 °C, put on carbon adhesive tape in a circular 25 mm diameter stub, and analyzed on a SEM Hitachi Analytical Table Top Microscope/benchtop SEM TM3000 equipped with a EDX detector by Bruker Nano GmbH. The EDX point analyses were obtained from the penultimate chamber of each specimen. For optimal results, a minimum of one minute for each target was selected after rarefaction optimization of results from 6, 4, 2 and 1 minute minimum.