Gas hydrate dissociation linked to contemporary ocean warming in the southern hemisphere

Ocean warming related to climate change has been proposed to cause the dissociation of gas hydrate deposits and methane leakage on the seafloor. This process occurs in places where the edge of the gas hydrate stability zone in sediments meets the overlying warmer oceans in upper slope settings. Here we present new evidence based on the analysis of a large multi-disciplinary and multi-scale dataset from such a location in the western South Atlantic, which records massive gas release to the ocean. The results provide a unique opportunity to examine ocean-hydrate interactions over millennial and decadal scales, and the first evidence from the southern hemisphere for the effects of contemporary ocean warming on gas hydrate stability. Widespread hydrate dissociation results in a highly focused advective methane flux that is not fully accessible to anaerobic oxidation, challenging the assumption that it is mostly consumed by sulfate reduction before reaching the seafloor.

S ubmarine gas hydrates on continental margins store large quantities of methane (e.g., 0.5-12.7 × 10 21 g 1,2 ), mainly produced by microbial degradation of organic matter, and argued to form a large capacitor able to regulate the Earth's climate [3][4][5][6] . The global gas hydrate reservoir is particularly vulnerable to dissociation at its feather edge, i.e. where the top and base of the gas hydrate stability zone (GHSZ) intercept the seafloor as a consequence of decreasing pressure (depth) and increasing bottom water temperature. The feather edge typically lies in upper slope depths (300-600 m), where ocean warming and sea-level lowering are capable of reducing the volume of the GHSZ to drive sediment degassing 5 . Field observations [7][8][9][10][11][12] and numerical modelling [13][14][15][16] suggest that anthropogenic-related ocean warming may be causing hydrate dissociation in upper slope settings. However, the mass of methane reaching the atmosphere is forecast to have a minor impact on climate during this century 6,14,17,18 , in part owing to sulfate reduction in sediments 19 and in part to the dissolution and oxidation of methane in the water column 20,21 . Clearly, it is important to improve our understanding of the long-to shortterm dynamics of the upper limits of gas hydrate systems on upper continental slopes in relation to contemporary climate change, and the effectiveness of the sulfate reduction filter in preventing methane to reach the oceans.
Our multi-disciplinary and multi-scale investigation of a bottom simulating reflector (BSR) outcrop on the southern Brazilian margin allows an investigation of gas hydrate dynamics and ocean interactions over long-(millennial) to short-(decadal) scales and provides the first robust evidence from the southern hemisphere of hydrate destabilization related to contemporary climate change. Geochemical and geophysical data, including the first autonomous underwater vehicle (AUV)-borne sub-bottom profiles of a BSR outcrop, allow us to document a massive advective flux of methane through the feather edge of the GHSZ, resulting in the formation of an elongate pockmark field associated with hundreds of water column gas flares. The pockmarks record long-term degassing, possibly in response to stable postglacial water temperatures, while the observed BSR outcrop is in thermodynamic disequilibrium with bottom water temperatures and the present-day edge of the GHSZ, consistent with ocean warming over several decades. Our results add to growing evidence that gas hydrate dissociation and sediment degassing related to contemporary ocean warming is a global phenomenon. The advective flux of methane through the feather edge of the GHSZ is three orders of magnitude greater than background diffusive flux and cannot be entirely consumed by anaerobic oxidation in the sediment, challenging the assumption that the sulfate filter prevents methane from reaching the seafloor. Nonetheless, gas bubbles are inferred to dissolve within 50 m of seafloor, consistent with methane oxidisation in the water column before reaching the atmosphere. Estimated methane leakage rates at the edge of the GHSZ on the Brazilian margin are lower than those in the northern hemisphere, and indicate that hydrate dissociation may be an important process in the global carbon cycle and the Earth's climate in a long-term (e.g. 10 3 years) perspective.

Results and discussion
Degassing along the edge of gas hydrate stability. The study area is located on the Rio Grande Cone (RGC), which consists of a 250,000 km 2 and 7.5-km-thick, Barremian-Recent 22,23 sediment depocenter in the southern Brazilian margin (Fig. 1a). A gas hydrate province was first recognised from a continuous, slopecrossing BSR observed over an area of at least 45,000 km 2 in water depths of 500-3500 m 24,25 . Seafloor investigations have subsequently yielded samples of gas hydrates, authigenic carbonates, and chemosynthetic ecosystems indicative of gas venting from pockmarks fields in two areas, on the mid-and upper slopes [26][27][28] . Ship-borne multibeam bathymetric imagery show that the upper slope pockmark field lies in water depths of 520-660 m, near the upper limit of the GHSZ, and is elongated parallel to at least 12 km of the slope 26,29 (Fig. 1b). Here we present additional evidence that this pockmark field is associated with gas venting to the water column (Fig. 1c, d), and show that it lies immediately downslope of the regional BSR outcrop on the seafloor (Fig. 1e, f).
Seismic reflection profiles across the RGC upper slope (Fig. 2a,  b) show that the regional BSR intercepts the seafloor at ca. 515-520 m water depth, at the upper limit of the pockmark field. The BSR outcrop is observed both on conventional multichannel seismic profiles (peak frequencies 10 1 Hz) and, to our knowledge for the first time, on sub-bottom profiles (SBP) acquired 40 m from seafloor using an AUV (peak frequencies 10 3 Hz). The AUV profiles show strong acoustic blanking, both at depths corresponding to the BSR, and as vertical columns rising above the BSR (Fig. 2c, d). Blanking on high frequency data is attributable to attenuation of the acoustic signal by reverberation and scattering due to the presence of gas bubbles in sediment pores [30][31][32] . The blanking features on SBP imagery are therefore consistent with the presence of free gas both beneath the BSR and rising through the thin (10s m) GHSZ, beneath the pockmark field. At the seafloor, AUV-borne multibeam imagery (Fig. 1b) reveals hundreds of high backscatter depressions, up to 600 m wide and 3-8 m deep, and thousands of smaller unit pockmarks 33 . High backscatter signatures are consistent with the presence of authigenic carbonates, as observed at the seafloor over wide areas using a remotely operated vehicle (ROV; Fig. 3a). Raw side-scan sonar images reveal 394 gas flares rising up to 50 m into the water column, from seafloor depths of 525-540 m, which can also be observed in ship-borne multibeam imagery (Fig. 1c, d). The flares are, therefore, located between the BSR outcrop at 515-520 m and the edge of GHSZ at 550-585 m calculated using water column temperature measurements obtained locally in the pockmark field (Figs. 2b and 4; Supplementary Data 1) and the equilibrium equation for pure methane hydrate in seawater 34 .
Nine giant Calypso-type piston cores (20-40 m long) were obtained within the pockmark field and along a transect crossing it in water depths of 397-1588 m (Fig. 1a) and show the presence of massive and laminated, bioturbated, dark olive and dark greenish grey, rarely dark grey, muddy sediments (Supplementary Fig. 1) containing authigenic carbonate nodules (centimetres in diameter). Gas hydrate was recovered in three piston cores within the pockmark field (PC95, PC66, and PC97) in water depths of 550-585 m, as mm−cm-thick laminae oriented parallel to sediment layering, at depths of up to 8 mbsf (Fig. 3b).
Advective vs diffusive gas fluxes. Evidence from ROV observations and cores allow us to quantify both the advective flux of methane to the ocean through the edge of the GHSZ and the broader diffusive flux of methane towards seafloor across the upper slope. Within the area of gas flares, ROV seafloor observations at two sites ( Fig. 1f) indicate that acoustically observed flares are formed by 5-6 individual bubble streams, which flow intermittently, varying from virtually zero to up to three bubbles per second in a time span of less than a minute (Fig. 3a, Supplementary Movie 1). Bubbles are estimated to be ca. 1 cm in diameter based on observations during sampling of a single bubble stream at each site (see "Methods"), but may show a high variability in diameter, even in the same cluster 35 , and commonly between 0.2 and 0.5 cm as observed in other seep sites around the world [35][36][37] . Observations of the two sites in the RGC over longer periods (up to 1 h) show a lower bubbling rate of 0.023-0.106 bubbles per second (44-200 cm 3 h −1 ) and mass transfer rates of 0.5 × 10 -3 g s −1 and 2.3 × 10 -3 g s −1 (for bubble of 1 cm in diameter; see Supplementary Note 1 for mass transfer rates and methane fluxes calculated for different bubble sizes and bubbling rates). Assuming these rates apply to all 394 flares mapped in the area, and that each flare contains five bubble streams (an assumption that is in agreement with observations of other flares offshore Svalbard 38 , Pakistan 35 , and Cascadia margin 39 ), the total methane advective mass flux is 25.2-115.9 mmol cm −2 year −1 , and the total methane mass transfer rate from sediments to the ocean is 31.3-144 Mg year −1 .
A transect of six piston cores across the edge of the GHSZ (Fig. 1a) provides an understanding of methane distribution below the seafloor and estimates of its diffusion rates,    Fig. 3 Seafloor gas bubble stream and gas hydrate in sediment. a Bubbling site and authigenic carbonate block on the seafloor in a pockmark obtained by inspection with a remotely operated vehicle at the ROV08 flare site; and b thin hydrate laminae in dark grey massive mud obtained from piston core PC66 (see Fig. 1 for location). ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-17289-z three orders of magnitude higher than the maximum diffusive flux recognised in the oceans 42 . It is interesting to note that this massive advective flux bypasses the local sulfate reduction zone without changing the local SMT depth (5.1 m at PC64; Fig. 5), indicating it is very localised. This result implies that the advective flux of methane is not entirely accessible to sulfate reduction 6,43 . The mechanisms by which a focused methane flow migrates through the sediment to bypass the sulfate reduction filter remain poorly understood, but recent numerical modelling has shown that rapid gas hydrate dissociation in sediments at shallow depths (<14 m) below seafloor may create hydraulic fractures that focus methane flow 44 , producing a flux that cannot be entirely consumed by sulfate reduction 45 . Our results are therefore consistent with a highly focused massive advective flux of methane within the pockmark field that is bypassing the sulfate reduction filter to form gas flares. This finding challenges the assumption that the majority of the methane released by hydrate dissociation in upper slope settings will be consumed anaerobically before reaching the seafloor 10,46 .
Origin of gas. Methane is the dominant gas present in all bubble, hydrate, and pore water samples (Supplementary Data 3). Carbon stable isotopic analyses reveal a wide range of δ 13 C values (Supplementary Data 4) compatible with a biogenic origin (<−50‰ 47 ). Methane sampled within the pockmark field has no detectable 14 C isotope, indicating that it is sourced from fossil carbon (>43,500 years B.P.). The similar δ 13 C values in gas bubbles and gas hydrate, together with the absence of 14   carbon is the likely source of methane in the flares. However, a noticeable enrichment in 13 C in methane in pore waters from the pockmark field relative to hydrate suggests that in situ hydrate dissociation is not the sole source of methane. Deeper biogenic methane is commonly enriched in the 13 C isotope 48 . Acoustic blanking beneath the pockmarks indicating free gas rising from beneath the BSR strongly supports the idea that biogenic methane is also being sourced from or below the base of the stability zone (Fig. 2c, d). In either case, methane may flow laterally underneath the base of the GHSZ 49 to mix with gas in sediment pores and vent to the seafloor (Fig. 6). In contrast, downslope of the flares area pore water methane is found to have highly depleted in δ 13 C values as well as modern 14 C (e.g. PC93; Supplementary Data 4), indicating in situ methanogenesis with little or no mixing with methane from advective sources.
Degassing in response to stability zone dynamics. Methane release to the water column on the upper RGC is interpreted to be taking place by advective flow through the feather edge of the GHSZ from sources at or beneath its base, most likely via a subvertical system of fractures that, over time, have facilitated the formation of a pockmark field in water depths of 520-660 m (Fig. 1b). The presence of carbonate concretions resulting from the anaerobic oxidation of methane 26 implies high gas flux over timescales of at least thousands of years at this location 29 . The striking parallelism of the pockmark field with the adjacent BSR outcrop in water depths of 515-520 m raises the question of whether gas venting could record the response of the feather edge of the GHSZ to changing oceanographic conditions since the last glacial maximum (LGM, ca. 20 ka BP). Globally, ocean warming since the LGM has counteracted the effect of rising sea levels (pressure increase) on gas hydrate stability, notably on upper continental slopes where the influx of warmer water resulted in downslope retreat of its feather edge 15,50 . Post-glacial warming is argued to have driven widespread hydrate dissociation and degassing along continental margins, leading to the formation of a gas hydrate-depleted zone on upper slopes 51 . Such a natural, long-term depletion process may have diminished the quantity of gas hydrate available for dissociation related to short-term anthropogenic ocean warming 15 . On the southern Brazilian margin, in contrast, it has been hypothesised that elevation of the permanent thermocline during sea-level rise resulted in a post-glacial cooling of the upper slope that suppressed hydrate dissociation and methane release 52 . The proposed mechanism was suggested to be relevant to the gas hydrate system of the RGC 52 , where the base of the permanent thermocline lies in depths of 500-700 m and so contains the upper limit of the GHSZ 29 . We note that long-term post-glacial cooling would have favoured gas accumulation and so could account for the presence of a well-developed BSR outcrop, making the southern Brazilian margin a potential modern analogue to LGM margins. However, constraints on deglacial temperature changes in the water column on the upper RGC slope record temperature increases of 3.5°C in bottom waters for the interval 18-10 ka BP, with a possible peak of 6.6°C at ca. 14 ka BP 53 . The latter value is comparable to present-day bottom water temperatures, suggesting a long-term (latest 14 ka) stabilisation of temperatures that would have maintained the edge of the GHSZ near its present-day depth. Long-term stability of the GHSZ could in part account for a well-developed BSR outcrop in the RGC.
Over shorter timescales, it is of interest to note that while the narrow (<3 km wide) pockmark field lies downslope of the BSR outcrop in water depths of ca. 515-520 m, present-day water column temperatures within the same area indicate the upper limit of GHSZ to lie at deeper waters (550-585 m; Fig. 2b). Historical water temperature measurements from the South Atlantic provide evidence of a warming of the Antarctic Intermediate Waters (AAIW). This water mass is formed near the Antarctic Circumpolar Current, flows at >500 m water depth at the RGC, and is characterized by a salinity minimum and dissolved oxygen maximum 54 . The warming of the AAIW can be traced back at least to the 1970s, at rates of 0.01-0.02°C year −1 55 , associated with contemporary climate change. These rates would position the edge of the GHSZ in the 1970s at 530-540 m water depth, within the depth range of most present-day gas flares on the RGC (Fig. 6). We suggest that recent ocean warming has unbalanced the edge of a long-term, stable gas hydrate system, which is in thermodynamic disequilibrium and undergoing dissociation downslope of the BSR outcrop (Fig. 6). Evidence of such a process has previously been recognised in the northern hemisphere on the US North Atlantic coast 8 and in the Arctic on the western Svalbard margin 9 , but this is the first time that it is identified in the southern hemisphere. It is interesting to note that the methane emission rates estimated for the 12-km-long studied RGC section (0.16 × 10 6 −0.75 × 10 6 mol year −1 km −1 ) are smaller than those of the western Svalbard margin 38,56,57 , suggesting a methane emission range of 10 5 −10 7 mol year −1 km −1 for such settings.

Methods
Acoustic surveys. 2D multichannel seismic data, acquired for hydrocarbon exploration and interpreted using IHS Kingdom Suite software, with a dominant frequency of 40 Hz, was used to map the BSR outcrop in the continental slope. AUV-borne acoustic data were obtained using a C-Surveyor II Autonomous Underwater Vehicle flying at 40 m above the seafloor and the Rig Supporter vessel. Seafloor morphology and imagery data were obtained with a Simrad EM 2000 multibeam echo sounder (operating frequency of 200 kHz) and a dual frequency (120-410 kHz) EdgeTech Side Scan Sonar (2200-M). Data were processed and visualised with IVS-3D Fledermaus and Sonar Wizz software. Sub-bottom profiler lines were acquired using an EdgeTech DW106 SBP (Chirp 1-6 kHz, central frequency 3.5 kHz), and lines were processed and visualised using the IHS Kingdom software. The major clusters of gas flares shown in Fig. 1d were located using the water column backscatter data obtained from the hull-mounted multibeam echo sounder installed onboard the Rig Supporter vessel. The water column backscatter data, which are highly sensitive to the presence of gas bubbles in water and, therefore, can be used to detect flares 58 , were integrated with the seafloor backscatter data using the IVS-3D Fledermaus software. Individual gas flares shown in Fig. 1f were located using the high-resolution water column data obtained from the AUV-mounted multibeam echo sounder and side-scan sonar systems. Two flares (marked in Fig. 1) were selected for ROV investigations. The ROV front sonar was used to reach the flares.
Water column temperature measurements. A total of nine temperature and salinity profiles were obtained in the RGC using Sea-Bird Electronics CTD (Conductivity, Temperature, Depth) models SBE 19plus, SBE 9plus, and a RBR XR-620 CTD instrument. Water masses were identified based on their thermohaline indexes obtained by CTD data.
Flare inspection and gas sampling procedure. Two ROV dives at flare sites previously identified with side-scan sonar images and sub-bottom profiles were performed using a Mohican Inspection Class Remotely Operated Vehicle (ROV) capable of operating in depths of up to 2000 m. The ROV was installed onboard the research vessel Marion Dufresne and was equipped with two manipulators and three inspection-quality video cameras. The positioning of the ROV was determined by a high precision acoustic positioning system (HiPAP). The location of bubbling sites was performed using the ROV's front sonar, and bubbling gas was collected using a device consisting of a 300 mL stainless-steel bottle with vacuum, coupled to a ball valve and a transparent, acrylic funnel with markings to estimate bubble size. The device was positioned on top of the bubble stream and the time elapsed to fill a certain volume in the funnel was used to estimate the gas seepage rate. The valve was then opened and closed by the ROV's manipulators to collect and store the gas (and water) in the bottle for further analyses. Gas seepage was also quantified by counting the number of bubbles per time, and a measured bubble diameter of 1 cm. The amount of methane n (moles) contained in a single bubble was determined by the gas law: (n = PV/ZRT), where Z is the compressibility of methane (0.8702), P is pressure (5.294 MPa at 525 m of water depth), V is the bubble volume, R is the universal gas constant, and T the ambient water temperature (8.5°C). The compressibility factor of methane was calculated by the Peng −Robinson equation of state 59 . The annual mass transfer of methane for the 394 flares was calculated assuming an average of five bubble streams per flare in a total area of ca. 8000 m 2 . The calculations assume no temporal variability of advective flux on timescales longer than 1 h, no seasonal variations in the flux as observed in other seep sites 60-62 , a uniform rate of discharge at all 394 flare sites, and a single bubble diameter of 1 cm for all bubbles (see Supplementary Note 1 for mass transfer rates and methane fluxes calculated for different bubble sizes and bubbling rates).
Methane diffusive flux. The methane diffusive flux was calculated based on sulfate concentration profiles in cores. The observed linear decreasing trend in sulfate profiles with depth suggests that sulfate reduction is driven primarily by methane consumption and, therefore, downward sulfate flux can be used as a proxy to an equal upward methane flux 40 . The flux was obtained by the Fick's first law, and the effect of porosity on diffusion was considered through a logarithmic equation 63 . We assumed standard seawater salinity of 35 psu 64 and sediment porosity of 0.6 (within the range of uncompacted mud 65 ). The pressure and temperature for each piston core in the transect was obtained from bathymetric and CTD data, respectively (see Supplementary Data 2). The PC66 was not included in the flux calculation (and on the transect- Fig. 5) because its top was lost during core recovery.
Gas chemical and isotopic composition. Samples for gas analyses in sediments were collected at every 1.5 m in the cores (ca. 100 cm 3 ) and placed in gas-tight, inert jars, and kept at 4°C. The sediment filled one third of the jar, while another third was filled with distilled water, leaving the top third with air (headspace). Five drops of the diluted Zephiran Chloride bactericide were added into each jar to eliminate microbial activity. Gas composition (seeping gas from plumes and headspace gas from jars) was determined with a gas chromatograph equipped with a capillary column VP-Plot Alumina/KCl, 30 m × 0.53 mm, and a flame ionisation detector (FID). Helium was used as the carrier gas at a constant rate of 5 mL min −1 . The injecting temperature was 190°C and the FID temperature was held at 200°C. Carbon stable isotopic analyses of methane were performed using a Thermo Fisher Scientific gas chromatograph coupled to a Thermo Scientific DELTA-V Plus isotope ratio monitoring mass spectrometer via a Thermo GC IsoLink and Conflo IV interfaces (Thermo Fisher Scientific). The gas chromatograph contained a 30 m × 0.32 mm fused silica column, Carboxen Plot 1006, and was operated at a heating ramp of 70-150°C, over 30 min. The isotopic data are reported using the delta notation (δ 13 C) in parts per thousand (‰) with isotopic ratio relative to the international Vienna Pee Dee Belemnite standard (V-PDB). The 14 C analyses were performed using accelerator mass spectrometry in gas samples obtained from sediment pores, hydrate, and venting gas. The gas samples were injected in vacuum glass tubes (vacutainer) and send to Beta Analytic laboratory in Florida, U.S.A. The modern reference standard was 95% the 14 C activity of the National Institute of Standards and Technology (NIST) Oxalic Acid (SRM 4990C) and calculated using the Libby carbon half-life (5568 years).

Data availability
All relevant data are included in the Supplementary material to this article.