Yam Seep at Four-Way Closure Ridge: a prominent active gas seep system at the accretionary wedge SW offshore Taiwan

Hydroacoustic surveys at the accretionary wedge southwest of Taiwan reveal a confined active hydrocarbon seepage area of ~ 49,000 m2 in ~ 1350 m water depth on the northern crest of the Four-Way Closure Ridge, which we call Yam Seep. In this study, multibeam and side-scan sonar data acquired during surveys with an autonomous underwater vehicle during an expedition with R/V Ocean Researcher I in 2017 showed that the area is characterized by rough topography and high seafloor backscatter. Seafloor observations with a video sled and sediment sampling with gravity corers and the MeBo seafloor drill rig during an expedition with R/V SONNE in 2018 revealed that the area is almost entirely covered by intensely fractured methane-derived carbonates, which indicate that seepage has been ongoing for thousands of years. Hydroacoustic anomalies (‘flares’) in the water column indicated the presence of several gas bubble emission sites mostly at the center and eastern flank of the area in 2019. Drilling through massive carbonates in the northwestern part of Yam Seep induced free gas escape from a depth of ~ 5.1 m. This suggests the presence of gas hydrates in the subsurface as the seep area is located well within the gas hydrate stability zone. The inter-disciplinary investigations of the Yam Seep demonstrate that upward migration of light hydrocarbons and seafloor discharge has a considerable influence on the seabed properties.


Introduction
Fluid flow, dewatering, and the ascent of gassy hydrocarbons through ocean sediments are fundamental processes of the Earth's carbon cycle. Such processes not only affect geochemical budgets of the ocean, but migration of fluids and volatiles also plays an essential role in triggering geological processes like submarine landslides, earthquakes, and mud volcanism (Kastner et al. 2014). Subsurface fluid circulation systems can result in focused emission of fluids and/or gaseous compounds from the seafloor into the ocean as cold seeps. These sites are geological windows that can provide important information on the transfer processes and reactions in the geosphere below (Suess 2014). Chemosynthesis-based communities that utilize the chemical energy of reduced compounds such as CH 4 , H 2 S and other light hydrocarbons are frequently associated with fluid outflow at seeps (Rubin-Blum et al. 2017;Sahling et al. 2002;Wang et al. 2022). Similar to hot vents, cold seeps support the growth of tubeworms, vesicomyid, and mytilid bivalves, as well as sulfide-oxidizing bacteria (Sibuet and Olu 1998). Biomass production by these communities can be several orders of magnitude higher than typical densities at non-seep environments on the nearby ocean floor (MacAvoy et al. 2002). Authigenic mineral precipitates are further obvious seafloor manifestations of a cold seep. Hydrocarbon-seep deposits are typically dominated by carbonates, but a variety of other precipitates including barite, iron sulfides and gas hydrates are also common (e.g., Bohrmann et al. 1998;Fan et al. 2018;Tong et al. 2013;Torres et al. 1996). Mineral authigenesis is linked to biogeochemical cycling of carbon and sulfur (Claypool and Kaplan 1974;Berner and Raiswell 1983;Berner 1990). Carbonate precipitation is induced by the anaerobic oxidation of methane (AOM), and typically occurs at the base of the sulfate reduction zone, which, at methane seeps is often situated close to the seafloor (Paull et al. 1992;Boetius et al. 2000). Due to highly dynamic processes at seeps, the morphology of the seafloor may change during the seepage processes and distinct bottom features are created, like pockmarks, bulges or very irregular surfaces instead of otherwise plain seafloor (Judd and Hovland 2009;Paull et al. 2015).
Cold seeps occur along all continental margins. At active margins, plate convergence and strike-slip faulting are driving tectonic fluid expulsion by dewatering of sediments, which is generating seep fluids from different depths (Scholl and von Huene 2007). Cold seeps at active plate boundaries were reported from the Aleutian Trench , the Cascadia margin (Suess et al. 1985), Costa Rica (Bohrmann et al. 2002), the Hikurangi margin (Greinert et al. 2010), the Makran Trench (Römer et al. 2012), the Nankai Trough (Sato et al. 2012) and many other oceanic to continental plate convergence zones. Fluid expulsion at passive margins is frequently driven by sediment loading, differences in compaction, saline fluids resulting from contact with evaporitic rocks, and migrating hydrocarbons. These seep sites have been documented along the North Atlantic margins (Paull et al. 1995;Skarke et al. 2014), in the Gulf of Mexico (MacDonald et al. 1990), the Black Sea (Klaucke et al. 2006), the Congo deep-sea fan (Sahling et al. 2008a, b;Pape and Bohrmann 2022), and many other passive margins.
The area studied here is located in the South China Sea. Convergence of the Philippine Sea Plate and the Eurasian Plate formed an active margin in the east and rifting formed the passive margin in the west of the South China Sea (Bowin et al. 1978). Seep research was first focused on the passive margin of the South China Sea, where a Chinese-German research cruise in 2004 discovered seep carbonates exposed in a large area named 'Jiulong methane reef' (Han et al. 2008). Similar methane-derived carbonates, dead and living chemosynthetic bivalves and other seep manifestations were found along the northern and southern continental margins of the South China Sea (Feng et al. 2018b). Detailed investigations using diving platforms like submersibles and ROVs have been performed at active seep sites on Formosa Ridge Fujikura et al. 2007;Machiyama et al. 2007) on the northeastern slope and at 'Haima' seep on the southwestern slope of the South China Sea. Both sites are well known for their active gas emissions, methanederived carbonate precipitates, specific biomarkers, and chemosynthetic bivalves (Feng et al. 2018a;Liang et al. 2017;Wei et al. 2020).
In contrast to the passive margins, seeps along the active margins of the South China Sea are less well investigated although the seafloor sediments and the lowermost water column commonly show elevated concentration of methane Chuang et al. 2010Chuang et al. , 2013Chuang et al. , 2019. Geophysical indications for the widespread presence of gas hydrates also occur within the accretionary prism southwest of Taiwan (Schnürle et al. 2011). Klaucke et al. (2016) investigated seepage at ~ 1500 m below sea level (mbsl) at the so-called Four-Way Closure Ridge (FWCR) by analyzing high-resolution bathymetry, deep-towed side-scan sonar imagery, high-resolution seismic profiling and towed video observations. A prominent seep area identified on top of the northern FWCR is characterized by patches of high seafloor backscatter. Seafloor video observations showed active gas emissions, the presence of extensive authigenic carbonate slabs, and chemosynthetic bivalves (Klaucke et al. 2016).
During a drilling cruise in 2018 using the remotely operated drill rig MARUM-MeBo200 onboard R/V SONNE we re-visited the main seep area described by Klaucke et al. (2016). Since the shape of the seep area resembles the root vegetable 'yam', we call the area "Yam Seep". The objective of this work is to constrain the geological processes that have formed the Yam Seep and constrain how they affect seafloor properties and gas seepage. We present new and comprehensive results with a high-resolution video survey from the Yam Seep together with micro-bathymetry data collected with an autonomous underwater vehicle during a previous cruise.

Regional setting
Located at the convergent margin of the Eurasian Plate and the Philippine Sea Plate, southwestern offshore Taiwan is characterized by a deformation front where the South China Sea lithosphere is subducted underneath the Philippine Sea Plate. The Philippine Sea Plate is moving in a northwestern direction against the Eurasian Plate (Fig. 1a) with a convergence rate of ~ 8 cm/year (Ustaszewski et al. 2012). The collision initiated between the middle and late Miocene (Bowin et al. 1978) caused the formation of a series of accretionary ridges and thrust faults. This process created pathways for fluid and gas migration leading to the formation of mud volcanoes and cold seeps at the seafloor (Liu et al. 1997(Liu et al. , 2006. The accretionary wedge has been divided into an upper slope and a lower slope (Fig. 1a) (Klaucke et al. 2016;Liu et al. 1997Liu et al. , 2006. The upper slope domain in relatively shallow water hosts a series of mud diapirs, mud volcanoes, and related active gas and mud extrusion (Chen et al. 2014(Chen et al. , 2020a. The deeper, lower slope domain includes ridges formed by active thrusting and folding and cold seep sites (Chuang et al. 2010(Chuang et al. , 2013(Chuang et al. , 2019Fan et al. 1 3 2018). Four-Way Closure Ridge (FWCR), which hosts the seep area studied here, is one of the well-developed accretionary ridges (Fig. 1b).
FWCR is a complex anticlinal elongated structure, which is approximately 10 km in length. It is orientated in NNW-SSE direction parallel to the deformation front. It is located at ~ 1500 mbsl and towers up to 350 m above the surrounding seafloor (Fig. 1c). Chen et al. (2020a, b) identified scarps associated with the western and eastern flank of FWCR and Klaucke et al. (2016) suggested the presence of slump scars on its eastern flank (Fig. 1c).
The southwestern Taiwan accretionary prism is characterized by one of the highest weathering rates in the world, and the sediment flux toward SW offshore is estimated to account for 99 Mt/year (Dadson et al. 2003). These sediments are transported into the South China Sea through submarine canyons, including the Penghu canyon, which controls the erosion of slope sediments along the submarine ridges, and are prominent erosive conduits on the active margin. Penghu canyon (Fig. 1b) was probably once located close to FWCR, as revealed by seismic data showing similarities of deposits at the eastern flank of FWCR, suggesting removal of paleo seafloor (Klaucke et al. 2016;Kunath et al. 2020).
The Yam Seep studied herein is located at the northern crest of FWCR. It is situated on top of an extensional fault network with the main thrust fault structures on its western limb (Kunath et al. 2020). Fluids ascend along the thrust faults and extensional fault systems at shallow depth (Kunath et al. 2020). Bubble emission activity at the seafloor of the Yam Seep has been reported by Klaucke et al. (2016). Mau et al. (2020) reported enrichments of dissolved methane in near-bottom water samples collected in 2018  (Klaucke et al. 2016;Liu et al. 1997;Yu and Chang 2002). c 1-m scale bathymetry map based on dives with an autonomous underwater vehicle (AUV) and showing the northern sector of Four-Way Closure Ridge (FWCR) and Yam Seep. Several fossil seep sites, scarps, and slump scars in the area had been reported by Klaucke et al. (2016) and Chen et al. (2020a, b) 1 3 and showed that the Yam Seep is generally bathed in high salinity deep water. Several fossil seep sites inferred from backscatter anomalies were reported from a close-by area to the northwest of the Yam Seep (Klaucke et al. 2016;Chen et al. 2020a, b) (Fig. 1c).

Materials and methods
This study is based on data obtained during research cruises with R/V Ocean Researcher I (ORI-1163) conducted in May 2017 and SO266/1 on R/V SONNE conducted in October/November 2018 (Bohrmann et al. 2019). Multibeam bathymetry, seafloor backscatter and side-scan sonar data were also aquired with an autonomous underwater vehicle (AUV) during cruise ORI-1163 in May 2017. Multibeam bathymetry, seafloor and water column backscatter via the ship-borne system, seafloor observation image, sediment cores, and seawater samples were collected during cruise SO266/1. The geo-information system (GIS) software ESRI ArcGIS™ ver. 10.4 was used to create maps and sustainable spatial management of the data obtained during the cruises. The correlated bathymetry and backscatter intensity profiles along the studied track lines were plotted using ArcMap 10.7 3D analyst tool.

Multibeam echosounder data acquisition and processing
A Dorado class AUV developed and operated by the Monterey Bay Aquarium Research Institute (MBARI) (Caress et al. 2008) was used to collect multibeam bathymetry and side-scan sonar backscatter of the FWCR, including the Yam seep area. The AUV was equipped with a Reson 7125 400 kHz multibeam and Edgetech 110 kHz chirp side-scan sonar, both yielding mapping data with a lateral resolution of 1 m. The survey was flown at a 50-m altitude with a 150-m survey line spacing. The AUV survey starting position was initialized using tracking from the GPS navigated ship using an ultra short baseline (USBL) sonar. The realtime survey navigation was derived from a Kearfott inertial navigation system that limits navigation drift to less than 0.05% of distance traveled when aided by velocity estimates from a 300-kHz Doppler velocity log.
Seafloor bathymetry, seafloor and water column backscatter were also obtained using the hull-mounted Kongsberg EM122 multibeam echosounder during cruise SO266/1. The sonar frequency of the EM122 is 12 kHz with an angular coverage sector of up to 150° and 288 beams per ping.
The open source software package MB-System (vers. 5.5.2213 and5.5.2289;Caress and Chayes 1996) was used for post-processing of multibeam bathymetry and AUV side-scan data. The AUV navigation was adjusted so that features match in overlapping and crossing swathes using the MB-System tool MBnavadjust. Water column data were displayed and edited to extract hydroacoustic anomalies attributed to rising gas bubbles ('flares') and allowed for manual flare identification and geo-picking of flare sources at the seafloor using the QPS software Fledermaus (vers. 7.7.2) and the FMMidwater tool box. As the seafloor footprint of a flare in the multibeam records was measured at ~ 50 m in diameter, we define this distance as the approximate error of the geo-picked flare sources. Flare sources identified during different crossings and located within the seafloor footprint were grouped in flare source areas. The 3D projection of extracted anomalies was edited and displayed using the QPS software Fledermaus. The position of AUV multibeam grids was adjusted to fit with the GPS navigated ship-based bathymetry using the Global Mapper ® (Blue Marble Geographics) application software. The water column flare data sets and detailed bathymetric data were displayed using the 3D Fledermaus software (QPS).

Coring and drilling
During cruise SO266/1, sediment and carbonate cores were drilled at two sites in the Yam Seep using the seafloor drill rig MARUM-MeBo200 (Table 1, Fig. 2b). Sediment core MeBo-28 (GeoB23234-1) was drilled at the eastern flank of the Yam Seep at 1348 mbsl ( Fig. 2b). At this station, a drilling depth of 59.9 m below seafloor (mbsf) with a core recovery of 50.5 m was achieved. Sediment core MeBo-24 (GeoB23225-1) was drilled in the northwestern part of the seep area at 1347 mbsl, down to ~ 5.1 mbsf. At this station, cemented carbonates were drilled with a core recovery of ~ 4.9 m and the drilling was aborted when significant amounts of free gas started to escape from the borehole.

Stable carbon isotope analysis of authigenic carbonates
Powdered carbonate sub-samples were reacted to carbon dioxide (CO 2 ; addition of 100% phosphoric acid at 75 °C) and stable carbon isotope ratios ( 13 C/ 12 C reported as δ 13 C in ‰ vs. V-PDB) of the released CO 2 were analyzed with a Finnigan MAT 251 mass spectrometer at the Stable Isotope Laboratory at MARUM. For details of the procedure see Loher et al. (2018).

Visual seafloor observation
The towed Ocean Floor Observation System (OFOS) on R/V SONNE was used for visual seafloor documentation of the Yam Seep during two surveys (OFOS-1 (GeoB23202-1); OFOS-2 (GeoB23206-1); Fig. 2a). The OFOS was equipped with a HD dynamic color video camera (Micro Sea Cam 2002) and a high resolution still frame camera (Canon EOS 5D Mark IV). Positioning of the OFOS was tracked by the IXSEA USBL system. Scale at the seafloor was provided by three laser pointers (Micro-Sea-Laser) of known distance between two laser points (20 cm).

Spatial distribution of dissolved methane in the near-seafloor water body
Four vertical hydrocasts were conducted along a 220 m-long transect across the Yam Seep area on October 20, 2018 (GeoB23208-1, -2, -3, and -4, Table 2). The concentration of dissolved methane in near-bottom waters was previously reported by Mau et al. (2020) in a study of methane over a larger region. In this study, differences in concentrations of dissolved methane are discussed with respect to relative positions above the seep structure.

Seafloor bathymetry and general composition of deposits
The Yam Seep is located on the summit line of the FWCR at ~ 1347 mbsl (Figs. 1c and 2a). The seep area is identified as an area with distinctly rougher seafloor, compared with the relatively smooth seafloor in the surrounding area. The seafloor is also characterized by anomalously high seafloor backscatter (Fig. 2b). Together, these seafloor properties characterize a triangular shaped area of ~ 400 m maximum length and ~ 175 m maximum width (Fig. 2b). The total seafloor area characterized by rugged bathymetry is ~ 49,000 m 2 , whereas the area showing relatively higher seafloor backscatter accounts for ~ 37,000 m 2 . Both the rough seafloor and high backscatter are considered to reflect seepage-related features, such as high density of authigenic carbonate crusts.
In order to distinguish relative depths within the seep area, a reference plane was created assuming an undisturbed seafloor (Fig. 2a). Elevation difference between this plane and the surveyed seafloor shows two elevated areas as follows: the northern elevated area and the southern elevated area (Fig. 2a). The southern elevated area is an elliptically shaped mound that peaks at a water depth of 1337 mbsl (Fig. 2a). The northern elevated area is characterized by an elongated structure, which reveals two peaks located in its central and western parts. The SW-NE trending topographic profile (profile A1-A2, Fig. 2d) crossing two local elevated areas shows an overall slope gradient of ~ 2.8° and a total water depth variation of 19 m.
The seep area shows an inhomogeneous, reticular appearing backscatter pattern, which is mainly N-S to NW-SE directed. Highest backscatter intensities were found along the rim of the seep area, with the largest high-backscatter area at the western rim (Fig. 2b). N-S-oriented high backscatter branch-like patterns correlate well with the location of seafloor fractures illustrated in the micro-bathymetry map ( Fig. 2a and b). Three separate high seafloor backscatter patches of ~ 7 m in diameter were identified within circular seafloor depressions a few meters to the west and south of the main seep structure ( Fig. 2a and b).
The failure of two gravity corer deployments (GC-1 and GC-2) to recover sediments (Figs. 2a and 3) are assumed to result from the presence of carbonates in near-surface sediments, which prevented penetration. Visual inspection during landing of the MeBo system at station MeBo-24 showed a flat sediment covered surface. Drilling at this station had to be aborted at a coring depth of ~ 5.1 mbsl (Fig. 3), because   of virulent escape of bubble-forming gas that apparently was ascending from below this depth. Core MeBo-24 revealed about 4.9 m of carbonates consisting of whitish botryoidal and acicular aragonite, grayish micrite, and cemented shells, which likely have sealed the pathway to rising free gas from the deeper accretionary structure. Stable carbon isotope signatures of the carbonates (δ 13 C-CaCO 3 ) ranging from − 37.9 to − 58.3‰ V-PDB suggest that they represent methanederived authigenic carbonates (Naehr et al. 2007) with the carbon incorporated primarily originating from the AOM. Hemipelagic sediments were recovered at stations GC-4 and GC-8 close to the eastern rim of the Yam Seep and at station MeBo-28 further east of the seep area (see Fig. 2b for locations). Concentrations of sulfate dissolved in pore waters extracted from these cores showed only slight depletions with depth (Bohrmann et al. 2019). This indicates that the zone of the sulfate-methane-interface (SMI) was situated well below 4.8 mbsf (GC-4) and 6.9 mbsf (GC-8), respectively. These cores penetrated into the sediment and indicate that upward flux of methane is low at these sites at present (see, e.g., Borowski et al. 1996). Core MeBo-28 was located further east of the seep area (Fig. 2b) and recovered sediments down to 59.8 mbsf. The pore water sulfate concentrations indicate a modern SMI located at ~ 14-20 mbsf (Bohrmann et al. 2019;Fig. 3). Carbonate precipitates were present at ~ 4.7 m and ~ 4.8 mbsf depth in core GC-4, and at ~ 5.1 mbsf in core GC-8 (Fig. 3).

Spatial distribution of seepage sites and temporal variability of gas ebullition
During cruise SO266/1 in October 2018, hydroacoustic anomalies in the water column indicative of gas bubble emissions, called flares, were recorded during eight crossings of the Yam Seep. Vertical seafloor projections of the flares, called flare source points, suggested seafloor gas emissions from 41 seafloor sources (Fig. 4b). However, due to echosounder footprint and the corresponding uncertainties of the flare source point locations, we postulate a minimum of eight flare source areas (Fig. 4a) including repeated sightings of flares from the same source areas. Flare source points assigned in 2018 were mainly concentrated along the northeastern rim and the center of the seep area. Such flares could be traced from the seafloor to about 880 m above seafloor (i.e., 550 mbsl) (Fig. 4c). The AUV side-scan survey conducted in spring 2017 revealed the presence of two bubble emission sites located at the northern center of the seep area (Fig. 4a). While one site was located inside the flare source area interpreted from the ship-based flare detection in 2018, no gas emission was observed for the other site in 2018.
3D flare extractions enabled characterization of the variation of bubble emission sites, bubble rise, and intensity. Multiple flares were sometimes observed to occur simultaneously, with three simultaneous flares on 10/20/2018 with maximum heights of 550 m (Fig. 4b). The highest single flare observed during a crossing on 10/31/2018 reached a height of ~ 880 m above seafloor (Fig. 4c). These results indicate that while gas bubble seepage has persisted for at least several years within the Yam Seep, the particular venting sites varied spatially and temporally. Our flare analysis revealed that gas bubble emission in 2018 was sourced from at least eight areas along the northeastern rim and from central areas of the Yam Seep. However, the bubble ebullition was spatially and temporally variable during different years and even different days of observation.
The variations in the dissolved methane concentrations in discrete water samples collected over the Yam Seep in 2018 (Mau et al. 2020) match results from hydroacoustic investigations. Maximum concentrations of dissolved methane (18.2 nmol/L) were measured in a water sample collected ~ 5 m above seafloor (1399 mbsl) at the western rim of the seep area during station CTD-2-1 (Fig. 5). Elevated methane concentrations of up to 10.3 nmol/L were also measured in samples taken at the eastern rim of the seep area at ~ 5 and 10 m above seafloor (1340-1345 mbsl) during CTD-2-3. Concentrations of dissolved methane in samples from above the central seep area (cast CTD-2-2) ranged between ~ 5.5 and ~ 6.4 nmol/L, with highest relative Although bottom water currents have not been investigated in this study, the relatively low CH 4 concentrations measured in water samples collected at casts CTD-2-2 and CTD-2-4 support the observation that most of seafloor gas emissions at the Yam Seep in 2018 occurred from its rim. However, seepage of methane in the dissolved phase from the northern central sector of the seep area cannot be excluded. According to seawater temperature and salinity, methane concentration in-situ atmospheric equilibrium would be 3.6 nM, but all measured concentrations were much higher. Therefore, seepage of methane in the dissolved phase is probably focused at spatially limited areas like fractures, might be temporally more variable and, thus, less significant. The cause(s) for the methane enrichments in water samples collected from the western rim of the Yam Seep (cast CTD-2-1), where no flares were observed, may be explained by (i) the presence of diffusive gas seepage sites, (ii) temporal variability of gas seepage, and (iii) lateral transport of water masses enriched in methane. It should be noted, however, that data from a single station conducted are not representative for the entire seep area, as visual observations revealed the heterogeneity of recently active areas and carbonate precipitates partly preventing and focusing fluid flow.

Visual seafloor observations of seep indicators
Visual seafloor observation during surveys with the OFOS enabled analyzing the characteristics of seafloor deposits and distributions of hydrocarbon seep-related benthic fauna along the tracks. Constrained by the life cycle, living chemosynthetic fauna indicate recent fluid availability, while authigenic carbonates document long-lasting (paleo)-seepage.

Properties and distributions of exposed carbonates as indicators of paleo-seepage
The general rough topography of the Yam Seep is associated with exposed carbonates, which occur widespread as crusts, and/or irregular outcrops. Carbonate crusts are relatively massive and thick carbonate formations and occasionally seem to comprise several layers (e.g., Figs. 6a and 7b) or to form comparably thin layers with smooth surfaces (e.g., Fig. 9c). Irregular shaped masses of free-standing carbonates are also seen (e.g., Fig. 6b). Evaluation of seafloor pictures revealed that almost all areas along the tracks hosted dense carbonate formations. Elevated areas revealed mainly flat carbonate surfaces or irregularly-formed carbonate outcrops. The southern elevated area was found to be characterized by sediment draped carbonate (Fig. 6b). Carbonate crusts were observed over a distance of several decimeters in areas of plain seafloor. Such crusts were partially draped by sediment and colonized by clams (Fig. 9c). Exposed flat carbonate crusts were dominant in elevated areas showing thick formation, including several fields of oblique thick carbonate crusts. We did not observe a clear trend in the distribution of the different types of carbonate formation as the carbonate crusts at relatively plain seafloor are less exposed. Plain areas may also be paved by flat carbonate crusts. For example, failure of sediment recovery during station GC-2 (Fig. 2) at plain seafloor (Fig. 9b) suggests the presence of hard carbonaceous substrate beneath.

Appearance and distributions of fractures in carbonates
Intense fracturing of the carbonate structures adds to the rough topography of the area. The fractures generally have an N-S orientation (Figs. 2b and c). In the AUV-derived high-resolution backscatter map, fractures can be observed at the northern and southern elevated areas as linear zones of low backscatter (Figs. 2b and c). The southern elevated area was also visually found to be characterized by fracturing (e.g., Fig. 6b). Fractures and cracks of the carbonate paved seafloor occur with different sizes and dimensions. They appear as linear troughs with steep, even vertical walls (Fig. 6a), which are likely caused by tectonic movements within the accretionary ridge, since the N-S orientation corresponds to normal faults of the structure (Kunath et al. 2020). Fractures range from ~ 15 to ~ 100 cm in width. They are more common in the center of the seep area. At the northern elevated area, a several m-long and dm-wide, SW-NE oriented fracture was discovered (Fig. 6a). Other cracks are more irregular structures with uneven walls (Fig. 6b). Therefore, the observed cracks and build-ups are interpreted as erosive structures. Cracks with variable orientations separated carbonate slabs at the northern elevated and the southern elevated area (Fig. 6b).
Along the OFOS tracks, fractures were often found to be partly filled with sediments. These are characterized by dark stained patches suggesting near-surface anoxic conditions and are presumably caused by the presence of Fig. 7 Images of the seafloor taken with the OFOS at the Yam Seep (for locations see Fig. 8d). All scale bars are 20 cm. a Bivalve shells on a carbonate paved seafloor within a wide-opened crack in carbonate crusts. b Chemosynthesis-based communities clinging on the ver-tical wall of cracks in carbonate crusts, and dark stained (presumably anoxic) sediment on the floor of the fracture. c Mussels covered by microorganisms (white) within cracks of carbonate slabs. d Anoxic sediments and shells in an apparently poorly cemented stage H 2 S near the seafloor (Figs. 6a, b, and 7b). Such staining was not observed for sediments covering carbonate slabs (Figs. 6a and 7c). The majority of anoxic sediments were also found to host microbial mats (Fig. 8a). Microbial mats and patches of anoxic sediment occur in the entire seep area including its rim and close surroundings suggesting that fluid migration is happening in shallow sediments even outside the morphologically identified seep area. The most extensive area (> 40 m in length) characterized by the presence of dark stained sediments and microbial mats was located at the NE rim of the Yam Seep (Fig. 8a).

Chemosynthetic animals as indicators of recent seepage
Hydrocarbon seep-related animals documented at the Yam Seep include bathymodiolinae mussels (Fig. 7a-c), which harbor symbiotic chemoautotrophic bacteria (Duperron Gigantidas platifrons (previously known as Bathymodiolus platifrons) and Gigantidas securiformis (formerly Bathymodiolus securiformis) have been reported to occur at Yam Seep (Kuo et al. 2019) and are abundant in other hydrocarbon seeps in the South China Sea as well (Wang et al. 2022). For Yam Seep, dense accumulations of mytilid mussels were discovered in the northern elevated area (Fig. 8b) at the bottom of fractures of carbonate precipitates. In the center of the seep area, mussel accumulations were also found attached to subvertical walls of fractures (Figs. 6a, 7b). Mytilid mussels were also found at the northern elevated area settling mostly within the fractures, cracks, or clinging at the steep fracture flanks. Additional dense mussel accumulations were observed at the eastern rim of the seep area (Fig. 9d).
Apart from living mussels, the massive extent of accumulations of shells from dead mussels provides evidence of past occurrences of living mussels in a certain density, which hence indicate active seepage at the Yam Seep in the past. Shell accumulations were found in fractures of carbonates at the central and western parts of the seep area. Shell accumulations were present at the rim of the seep area and frequently found peripheral to living mussel accumulations. In some areas, a transition of dense mussel communities to Fig. 9 a Seafloor backscatter map of the Yam Seep with outlines of medium and high seafloor backscatter areas. b Soft sediments characterize low backscatter intensities. c Carbonate crusts partly covered by a thin layer of sediments produce medium backscatter intensities.
Clams were found to live in the sediments. d Dense faunal assemblages dominated by mytilid mussels and crabs covering the seafloor were seen in areas of high seafloor backscatter. White scale bar in b-d is 20 cm dense shell accumulations was observed and some regions were characterized by a mixture of living mussels and shells of mussels and clams (Fig. 7a). At the western rim of the Yam Seep a field containing shells was found to extend 10 m along the track of OFOS-2 (Fig. 8d). At the eastern rim, shell debris was found partly covered and buried by sediment with dark patches and potentially slightly cemented (Figs. 7d and  8d). This observation suggests the past fluid flow with dissolved methane at specific locations that sustained chemoautotroph communities, and may indicate a regional variation of the fluid transporting dissolved methane to the seafloor.
Living vesicomyid clams were barely spotted as they live buried in soft sediments. In some areas clams were seen embedded in soft sediments within fractures (Fig. 9c), e.g., at the northern and southern parts of the seep area (Fig. 8b). The presence of vesicomyid clams implies that soft sediments fill the fractures between the authigenic carbonates. The colonization at the fractures additionally suggests supply of sulfide to support the vesicomyid clam communities. Such environments have been suggested to be the precursor stage towards the capping of sediments by carbonate precipitation (Sahling et al. 2008a, b).
The occurrence of live chemoautotroph mussels and clams is key indicator for present hydrocarbon seepage. They suggest different scenarios due to the habitat requirement and chemical fuels. Whereas mussels cling to a hard substrate and use dissolved methane to fuel their symbionts, clams burrow into soft sediments percolated by hydrogen sulfide (Dubilier et al. 2008). The different settlements of mussel accumulations at the Yam Seep (surface covering in the western part and attached on vertical outcropping carbonate walls in the central area) suggest varying dissolved methane emissions depending on the carbonate structures at these different locations. Such differences can be argued to be related to the structures of the whole carbonate paved area, in which fluids with dissolved methane only occurred at a certain level. The spatial differences might be attributed to the occurrence of gas hydrate accumulations underneath the fractured carbonate slabs.
Tubeworms (polychaetes) were widely observed on carbonates (Fig. 6a). Siboglinid tubeworms live in symbiosis with chemoautotrophic, sulfur-oxidizing bacteria, are considered as one of the primary producers in seep communities (Sahling et al. 2008a, b), and have also been reported from the South China Sea at the Haima cold seep (Wang et al. 2022). Since tubeworms are sessile and can live for more than a century, their presence indicates that seepage activity has persisted for decades (Bergquist et al. 2000;Fisher et al. 1997).
At the Yam Seep, tubeworms occurred in bushes composed of up to several tens of individuals, but also of 2-3 single tubeworms only in some places (Fig. 8c). Tubeworms generally rooted below the carbonate slabs and extended out from fractures. Tubeworm bushes were mainly found associated to shallow fractures, while single tubeworms occured on isolated carbonate masses or at the edges of carbonate slabs. At a deeply incised fracture in the central part of the seep area, tubeworm bushes were attached to vertical flanks of exposed carbonate layers (Fig. 6a). Individual tubeworms were observed to reach lengths of up to ~ 1 m. The widespread occurrence of authigenic carbonates and tubeworms again substantiates the long-lasting history of gas seepage in the study area.
Other non-chemosynthetic faunal elements, such as corals and anemones, were frequently observed attached to carbonates and along fractures. Such animals might not indicate seepage, but indicate the long-term exposition of the carbonate precipitate as hard ground for settlement and nutrients supply through water currents (Deng et al. 2019;Hovland and Risk 2003;Wang et al. 2022). Other mobile animals such as galatheoid crabs or spider crabs were widely found close to or within the fauna communities of the seep area. Such an abundance of organisms suggests a sufficient food source.

Seafloor backscatter characteristics related to the seep system
Joint evaluation of the bathymetric and seafloor backscatter data together with visual seafloor observation enabled a spatial correlation of seafloor deposit characteristics and faunal distribution (Fig. 9a-d). Sediment draped areas correlate with relatively smooth seafloor as indicated by the smooth topography and low backscatter intensities (Fig. 9b). The different backscatter intensities of carbonate covered and fractured areas seem to be caused by the density of colonization by clams as observed at a site characterized by medium intensity (Fig. 9c). High backscatter intensities were also detected in areas colonized by living mytilid mussels (Fig. 9d).
Not all mytilid mussel accumulations observed along the OFOS tracks are associated with high backscatter intensity, for instance when located within deeply incised fractures or partly covered by sediments. Numerous sites with chemosynthesis-based communities, including extensive accumulations of shells from vesicomyid clams (Fig. 7a), mytilid mussels (Fig. 7c), and mytilid mussels attached to vertical fracture walls (Fig. 7b) were observed during the seafloor visual survey (OFOS-2) in the central elevated area (Fig. 8d). At the eastern rim, an extensive area with dark stained sediments adjacent to a field of shell accumulations, which mostly correlate to high backscatter characteristics, has been observed. At the western periphery, shells were seen to be covered by sediments (Fig. 7d).
According to our observations we conclude that (i) high backscatter intensities areas are mostly related to shells and living mytilid mussels; (ii) medium backscatter intensity correspond to authigenic carbonates partly hosting chemosynthesis-based communities; (iii) low backscatter relates to soft sediments (even when covering authigenic carbonates in greater depth). Backscatter intensity of shells or fractures is strongly affected (reduced) by sediment draping.
The backscatter intensities recorded within the seep area are not homogeneous, but widely scatter and range from low to high (Fig. 9a). Most of the seep area is characterized by medium backscatter intensities (> 70%), which separates the seep area to the surrounding seafloor of homogeneously low backscatter. Low backscatter areas also occur within the seep area, but account for < 15% of the total seep area and occur as relatively small areas throughout the entire area (maximum < 500 m 2 ). About 15% of the seep area is characterized by high backscatter, with the largest coherent areas covering ~ 1000 m 2 .

Evolution of the Yam Seep
Swath-bathymetry, side-scan sonar imaging, and OFOS seafloor imagery showed that the seafloor topography of the active YAM seep system is dominated by authigenic carbonate. The carbonate precipitates, which are the result of AOM, are either exposed or covered by thin layers of sediments. The presence of intensely fractured authigenic carbonates suggest that formation and exposure of the Yam Seep on top of the FWCR anticline is strongly controlled by tectonic compression, uplifting of the accretionary ridge, gas migration and seafloor seepage, and focused seafloor erosion. Ongoing AOM results in the formation of authigenic carbonates and sediment cementation (Bayon et al. 2009;Hovland 2002). Further carbonate precipitation at newly developed gas emission pathways expands the carbonate paved area (Liebetrau et al. 2014;Loher et al. 2018). Gas bubble discharge recorded in 2013, 2017, and 2018 along with seawater methane enrichments measured in 2018 indicate that (i) the Yam Seep is an actively methane seeping area and (ii) up to meter-thick carbonate pavements result from long-term seepage. Seepage areas characterized by similar features have been documented in other active margins at accretionary ridges, i.e. Hydrate Ridge (Teichert et al. 2005;Paull et al. 2015) and offshore southern Nicaragua, both linked to underlying fault systems (Moerz et al. 2005;Sahling et al. 2008a, b).
Tectonic movements like thrusting and uplifting of the submarine ridge potentially cause the fracturing and exposure of the seep region. Erosion at slump scars was observed at FWCR, which were reported as the results of over-tilting of the uplifting ridge (Chen et al. 2020a, b;Klaucke et al. 2016). In addition to tectonic influence, we argue that seafloor erosion due to currents and biology might have been involved in shaping the Yam Seep. Stronger currents result in unconsolidated sediment being carried away, exposing the consolidated authigenic carbonates on the seafloor which in general formed hardgrounds for filter-feeder fauna. Internal waves (Alford et al. 2015) were reported to occur peripheral to FWCR at canyons adjacent to the upper slope and shelf, and especially related to records of Penghu canyon (Klaucke et al. 2016;Kunath et al. 2020;Yu and Chang 2002), which effect the region including our study area by scavenging on the surface sediments. Bioerosion is also a potential erosion force in our study area as we observed abundant chemosynthetic animals along or within the edge of cracks (Paull et al. 2005).
Beside tectonic processes affecting seepage sites, the formation of gas hydrates in shallow sediments can have an important influence on the seep evolution, especially when gas bubbles are emanating at the seafloor (Marcon et al. 2014). Since the Yam Seep is located well within the gas hydrate stability zone, gas upward migration would lead to the deposition of gas hydrate (Sloan and Koh 2007) underneath the exposed carbonates. Examples from several studies reported wide areas of updoming or collapsing of carbonate paved seafloor related to gas hydrate accumulation or dissociation (Matsumoto et al. 2009;Sahling et al. 2008a, b). A good example was described from the Calabrian arc in the Mediterranean Sea where massive gas hydrate below a thin carbonate crust was dissolved and the carbonate crust collapsed (Loher et al. 2018). Methane migrated below the cemented and sealed layer of precipitated authigenic carbonates, where it formed massive gas hydrates in contact with water. An accumulation of gas hydrate at such shallow sediment depth inhibits an increased buoyancy but also could push the overlying sediments up (Barry et al. 2012). Because the Yam Seep is situated ~ 700 m below the upper boundary of the gas hydrate stability zone (GHSZ) and above the lower boundary of GHSZ at ~ 650 mbsl marked by a bottom simulation seismic reflector (Kunath et al. 2020). Thermal and pressure effects by glacial/interglacial changes which move the seafloor and shallow sediments out of the stability zone can be neglected.
The local presence of gas hydrates is causing a heterogeneous supply of methane available for chemosynthesis-based habitats. Clams and microbes found within shallow fractures at relatively smooth areas and covered by sediment suggest that the area and close surroundings are characterized by methane supply. Tubeworms are commonly found adhering to carbonate crusts of different sizes and shapes, and prove that seepage has been around for a longer time. Mytilid mussels suggest a later stage of seepage when carbonate hard ground are available where mytilid larvae initially can settled down and benefit from the dissolved methane-rich bottom which they use for their metabolism and thrive via their microbes in the gills (Sibuet and Olu 1998). Corals colonizing carbonate build-ups were found frequently over the paved area, indicating that nutrients were sufficiently available for filter-feeders and that bottom water currents most likely exposed the hard carbonate substrates (Hovland and Risk 2003;Roberts et al. 2009).
Summarizing our results presented in this study, we established a simplified conceptual model with four evolutionary stages (Fig. 10) in order to discuss the dominant processes at the Yam Seep:

Stages I and II-initial seepage and carbonate precipitation
Initially light hydrocarbons begin to migrate along permeable fault networks to the shallow subsurface. When methane comes in contact with seawater-derived sulfate, AOM produces sulfide and initiates formation of authigenic carbonates. The sulfide-rich environment enables colonization by vesicomyid clams. Tubeworms can settle on authigenic carbonates that serve as hard substrates (Cordes et al. 2003(Cordes et al. , 2005Freytag et al. 2001;Sahling et al. 2008a, b).

Stage III-increasing precipitation and seafloor sealing
The continued supply of methane enhances the density of authigenic carbonates and the formation of carbonate slabs that increasingly seal the seafloor and deflect upward migration fluids. Gas hydrate (blue gradient shaded hexagons in Fig. 10) formation underneath the carbonate slabs may also be induced. Near-seafloor carbonates and gas hydrates hamper fluid upward migration and direct gas escape from the seafloor at Yam Seep. Instead, deflection of fluids towards the edges of the carbonate crust causes lateral growth of the seep-influenced area as suggested by the sporadic occurrence of carbonate chunks in hemipelagic sediments at the peripheral sites GC-4 and GC-8. Chemosynthetic organisms settle at the rim of the seep area, where gas migrates through higher permeability sediments.

Stage IV-erosion and re-shaping of seafloor morphology
Partial exposure of the paved carbonate area, fracturing of carbonate slabs, and fauna colonization occur. We postulate that tectonic movement causes ridge uplifting, sediment removal through marine erosional processes along tilted flanks and exposure of carbonate crusts. Thrusting and faulting affect the cemented carbonate slab and cause fracturing and opening of pathways for gas ebullition. Repeated sealing within cracks and fracturing of the lower cemented crusts shaped the (a) deeply fractured crusts. Meanwhile, (b) carbonate precipitation occurs in the exposed sediments at the more profound depth within deep fracture. (c) Free gas migrates along the extensional fault and is partly released from the sediment to the water column. However, part of the free gas gets trapped underneath the carbonate crusts and potentially forms gas hydrate, which fuels chemosynthesisbased communities. Peripheral to the carbonate area, fractured crusts buried in shallow sediment (d) and anoxic sediment (e) experience the processes of stage I to III. Because of blocking, new migration and seepage pathways evolve, which consequently extend the carbonate paved area. Anoxic sediments and the presence of vesicomyid clams living in sediments within fractures are the key features occurring at the rim and the relatively plain regions.

Conclusions
In addition to previous seismic and side-scan sonar studies since 2013, our study constrains the geological processes that control the formation of the Yam Seep and characterizes the carbonate-cemented, fractured, and exposed region. Hydro-acoustic data documented several flare sightings, defined eight source areas for recent gas bubble ebullition at the center and the eastern flank of the Yam Seep, and illustrated its spatial and temporal variability. The seafloor backscatter results reveal the seafloor characteristics with the highest backscatter intensity related to shell distribution and middle-intensity-patterns to fractures and cracks. Enrichments of dissolved methane were found at the rim of the seep area. Ample seep-related and widespread fauna was observed visually. The widespread occurrence of authigenic carbonates illustrates the long-term seep evolution while the presence of living chemosynthesis-based community suggests recent hydrocarbon seep activity.
The diverse evidence observed at the Yam Seep was used to propose an active margin seep area evolution under dynamic tectonic conditions, geochemical, and biological interaction, involving four stages: (I) Initial seafloor seepage of methane, (II) Carbonate precipitation, (III) Increasing precipitation and seafloor sealing, (IV) Erosion and seafloor morphology re-shaping. Stages I-III relate to the geochemical process of methane seepage and carbonate precipitation and its indicator fauna. Stage IV includes the exposure of the seep carbonates under tectonic faulting, uplifting, overtilting and the expansion of the seep area with repeating stages I-III at its periphery.
Acknowledgements R/V SONNE cruise SO266/1 was financed in Germany by the Ministry of Education and Science, project TaiDrill -SO266 (Ref. #03G0288A). The project was funded in Taiwan by the Ministry of Science and Technology (MOST) Project TaiGer drill (MOST107-3113-M-002-004). We are grateful to the master of the vessel, Lutz Mallon, and to Susan Mau, Tzu-Hsuan Tu, Jhen-Nien Chen, Janice Malnati for collecting and analyzing water samples during SO266/1. The AUV survey was also supported by the Ministry of Science and Technology (MOST) through the master gas hydrate program and the David and Lucile Packard Foundation. We thank Dong Feng, Manasij Santra, and Peter Kukla for their constructive comments that helped to improve the quality of the manuscript.
Funding Open Access funding enabled and organized by Projekt DEAL.
Data availability statement Relevant data are made available through the World Data Center PANGAEA® (http:// www. panga ea. de). Additional information is available to interested readers on request.

Conflict of interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to affect the work reported in this paper.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/.